Developments in

Healthcare Information
Systems and Technologies:
Models and Methods

Joseph Tan
McMaster University, Canada

Medical inforMation science reference

Hershey • New York
Director of Editorial Content: Kristin Klinger
Director of Book Publications: Julia Mosemann
Acquisitions Editor: Lindsay Johnston
Development Editor: Julia Mosemann
Publishing Assistant: Casey Conapitski
Typesetter: Travis Gundrum
Production Editor: Jamie Snavely
Cover Design: Lisa Tosheff

Published in the United States of America by

Medical Information Science Reference (an imprint of IGI Global)
701 E. Chocolate Avenue
Hershey PA 17033
Tel: 717-533-8845
Fax: 717-533-8661
E-mail: [email protected]
Web site: http://www.igi-global.com

Copyright © 2011 by IGI Global. All rights reserved. No part of this publication may be reproduced, stored or distributed in
any form or by any means, electronic or mechanical, including photocopying, without written permission from the publisher.
Product or company names used in this set are for identification purposes only. Inclusion of the names of the products or com-
panies does not indicate a claim of ownership by IGI Global of the trademark or registered trademark.

Library of Congress Cataloging-in-Publication Data

Developments in healthcare information systems and technologies : models and methods / Joseph Tan, editor.
p. ; cm.
Includes bibliographical references and index.
Summary: "This book presents the latest research in healthcare information systems design, development, and
deployment,investigating topics such as clinical education, electronic medical records, clinical decision support systems, and IT
adoption in healthcare"--Provided by publisher. ISBN 978-1-61692-002-9 (hardcover) -- ISBN 978-1-61692-003-6 (ebook) 1.
Medical informatics. 2. Information storage and retrieval systems--Medicine. I. Tan, Joseph K. H. [DNLM: 1. Medical
Informatics. W 26.5] R858.D48 2011
610.285--dc22
2010027162

British Cataloguing in Publication Data

A Cataloguing in Publication record for this book is available from the British Library.

All work contributed to this book is new, previously-unpublished material. The views expressed in this book are those of the
authors, but not necessarily of the publisher.
Table of Contents

Preface ................................................................................................................................................. xvi

Chapter 1
Evaluating Health Information Services: A Patient Perspective Analysis ............................................. 1
Umit Topacan, Bogazici University, Turkey
A. Nuri Basoglu, Bogazici University, Turkey
Tugrul U. Daim, Portland State University, USA

Chapter 2
Gastrointestinal Motility Online Educational Endeavor ...................................................................... 14
Shiu-Chung Au, State University of New York Upstate Medical University, USA
Amar Gupta, University of Arizona, USA

Chapter 3
Envisioning a National e-Medicine Network Architecture in a Developing Country:
A Case Study ........................................................................................................................................ 35
Fikreyohannes Lemma, Addis Ababa University, Ethiopia
Mieso K. Denko, University of Guelph, Canada
Joseph K. Tan, Wayne State University, USA
Samuel Kinde Kassegne, San Diego State University, USA

Chapter 4
Assessing Physician and Nurse Satisfaction with an Ambulatory Care EMR:
One Facility’s Approach ...................................................................................................................... 54
Karen A. Wager, Medical University of South Carolina, USA
James S. Zoller, Medical University of South Carolina, USA
David E. Soper, Medical University of South Carolina, USA
James B. Smith, Medical University of South Carolina, USA
John L. Waller, Medical University of South Carolina, USA
Frank C. Clark, Medical University of South Carolina, USA
Chapter 5
Information Technology (IT) and the Healthcare Industry: A SWOT Analysis ................................. 65
Marilyn Helms, Dalton State College, USA
Rita Moore, Dalton State College, USA
Mohammad Ahmadi, University of Tennessee at Chattanooga, USA

Chapter 6
Using a Neural Network to Predict Participation in a Maternity Care Coordination Program ............ 84
George E. Heilman, Winston-Salem State University, USA
Monica Cain, Winston-Salem State University, USA
Russell S. Morton, Winston-Salem State University, USA

Chapter 7
Can IT Act as a Catalyst for Change in Hospitals? Some New Evidence ............................................ 94
Teemu Paavola, LifeIT Plc, Finland

Chapter 8
Informatics Application Challenges for Managed Care Organizations: The Three Faces
of Population Segmentation and a Proposed Classification System .................................................. 102
Stephan Kudyba, New Jersey Institute of Technology, USA
Theodore L. Perry, Health Research Corporation, USA
Jeffrey J. Rice, Independent Scholar, USA

Chapter 9
Scrutinizing the Rule: Privacy Realization in HIPAA ....................................................................... 112
S. Al-Fedaghi, Kuwait University, Kuwait

Chapter 10
In What Ways Does Web Technology Support the Individual in Choice Reforms in Health Care?
A Comparison among Norway, Denmark, and Sweden .................................................................... 126
Agneta Ranerup, Göteborg University, Sweden

Chapter 11
Characteristics of Good Clinical Educators from Medical Students’ Perspectives:
A Qualitative Inquiry Using a Web-Based Survey System ................................................................ 145
Gary Sutkin, University of Pittsburgh School of Medicine, USA
Hansel Burley, Texas Tech University, USA
Ke Zhang, Wayne State University, USA
Neetu Arora, Texas Tech University, USA

Chapter 12
Open Source Software: A Key Component of E-Health in Developing Nations ............................... 162
David Parry, Auckland University of Technology, New Zealand
Emma Parry, National Women’s Health, Auckland District Health Board, New Zealand
Phurb Dorji, Jigme Dorji Wanchuck National Referral Hospital, Bhutan
Peter Stone, University of Auckland, New Zealand
Chapter 13
An Empirical Investigation into the Adoption of Open Source Software in Hospitals ...................... 175
Gilberto Munoz-Cornejo, University of Maryland Baltimore County, USA
Carolyn B. Seaman, University of Maryland Baltimore County, USA
A. Güneş Koru, University of Maryland Baltimore County, USA

Chapter 14
Intelligent Agent Framework for Secure Patient-Doctor Profiling and Profile Matching ................. 195
Masoud Mohammadian, University of Canberra, Australia
Ric Jentzsch, Compucat Research Pty Limited, Australia

Chapter 15
Towards a Conceptual Framework of Adopting Ubiquitous Technology in Chronic
Health Care ......................................................................................................................................... 214
Jongtae Yu, Mississippi State University, USA
Chengqi Guo, James Madison University, USA
Mincheol Kim, Jeju National University, South Korea

Chapter 16
E-Patients Empower Healthcare: Discovery of Adverse Events in Online Communities .................. 232
Roy Rada, University of Maryland Baltimore County, USA

Chapter 17
Towards Process-of-Care Aware Emergency Department Information Systems:
A Clustering Approach to Activity Views Elicitation ......................................................................... 241
Andrzej S. Ceglowski, Monash University, Australia
Leonid Churilov, The University of Melbourne, Australia

Chapter 18
Applying Dynamic Causal Mining in Health Service Management .................................................. 255
Yi Wang, Nottingham Trent University, UK

Chapter 19
Application of Unified Modelling Language (UML) to the Modelling of Health Care Systems:
An Introduction and Literature Survey ............................................................................................... 275
Christos Vasilakis, University College London, UK
Dorota Lecnzarowicz, University of Westminster, UK
Chooi Lee, Kingston Hospital, UK

Chapter 20
TreeWorks: Advances in Scalable Decision Trees .............................................................................. 288
Paul Harper, Cardiff University, UK
Evandro Leite Jr., University of Southampton, UK
Chapter 21
Ambient Intelligence and Pervasive Architecture Designed within the EPI-MEDICS
Personal ECG Monitor ........................................................................................................................ 302
Hussein Atoui, Université de Lyon and INSERM, France
David Télisson, Université de Lyon and INSERM, France
Jocelyne Fyan, Université de Lyon and INSERM, France
Paul Rubel, Université de Lyon and INSERM, France

Compilation of References .............................................................................................................. 312

About the Contributors ................................................................................................................... 343

Index ................................................................................................................................................... 348

Detailed Table of Contents

Preface ................................................................................................................................................. xvi

Chapter 1
Evaluating Health Information Services: A Patient Perspective Analysis ............................................. 1
Umit Topacan, Bogazici University, Turkey
A. Nuri Basoglu, Bogazici University, Turkey
Tugrul U. Daim, Portland State University, USA

The objective of the chapter is to explore the factors that affect users’ preferences in the health service
selection process. In the study, 4 hypothetical health services were designed by randomly selecting
levels of 16 attributes and these services was evaluated by the potential users. Analytical Hierarchy
Process (AHP), one of the decision making methods, was used to assess and select the best alternative.

Chapter 2
Gastrointestinal Motility Online Educational Endeavor ...................................................................... 14
Shiu-Chung Au, State University of New York Upstate Medical University, USA
Amar Gupta, University of Arizona, USA

Medical information has been traditionally maintained in books, journals, and specialty periodicals. A
growing subset of patients and caregivers are now turning to diverse sources on the internet to retrieve
healthcare related information. The next area of growth will be sites that serve specialty fields of medi-
cine, characterized by high quality of data culled from scholarly publications and operated by eminent
domain specialists. One such site being developed for the field of Gastrointestinal Motility provides
authoritative and current information to a diverse user base that includes patients and student doctors.
Gastrointestinal Motility Online leverages the strengths of online textbooks, which have a high degree
of organization, in conjunction with the strengths of online journal collections, which are more com-
prehensive and focused. Gastrointestinal Motility Online also utilizes existing Web technologies such
as Wiki-editing and Amazon-style commenting, to automatically assemble information from heteroge-
neous data sources.
Chapter 3
Envisioning a National e-Medicine Network Architecture in a Developing Country:
A Case Study ........................................................................................................................................ 35
Fikreyohannes Lemma, Addis Ababa University, Ethiopia
Mieso K. Denko, University of Guelph, Canada
Joseph K. Tan, Wayne State University, USA
Samuel Kinde Kassegne, San Diego State University, USA

Poor infrastructures in developing countries such as Ethiopia and much of Sub-Saharan Africa have
caused these nations to suffer from lack of efficient and effective delivery of basic and extended medi-
cal and healthcare services. Often, such limitation is further accompanied by low patient-doctor ratios,
resulting in unwarranted rationing of services. Apparently, e-medicine awareness among both govern-
mental policy makers and private health professionals is motivating the gradual adoption of techno-
logical innovations in these countries. It is argued, however, that there still is a gap between current
e-medicine efforts in developing countries and the existing connectivity infrastructure leading to faulty,
inefficient and expensive designs. The particular case of Ethiopia, one such developing country where
e-medicine continues to carry significant promises, is investigated and reported in this article.

Chapter 4
Assessing Physician and Nurse Satisfaction with an Ambulatory Care EMR:
One Facility’s Approach ...................................................................................................................... 54
Karen A. Wager, Medical University of South Carolina, USA
James S. Zoller, Medical University of South Carolina, USA
David E. Soper, Medical University of South Carolina, USA
James B. Smith, Medical University of South Carolina, USA
John L. Waller, Medical University of South Carolina, USA
Frank C. Clark, Medical University of South Carolina, USA

Evaluating clinician satisfaction with an electronic medical record (EMR) system is an important di-
mension to overall acceptance and use, yet project managers often lack the time and resources to for-
mally assess user satisfaction and solicit feedback. This article describes the methods used to assess
clinician satisfaction with an EMR and identify opportunities for improving its use at a 300-physician
academic practice setting. We administered an online survey to physicians and nurses; 244 (44%) re-
sponded. We compared physician and nurse mean ratings across 5 domains, and found physicians’
satisfactions scores were statistically lower than nurses in several areas (p<.001). Participants identify
EMR benefits and limitations, and offered specific recommendations for improving EMR use at this
facility. Methods used in this study may be particularly useful to other organizations seeking a practical
approach to evaluating EMR satisfaction and use.

Chapter 5
Information Technology (IT) and the Healthcare Industry: A SWOT Analysis ................................. 65
Marilyn Helms, Dalton State College, USA
Rita Moore, Dalton State College, USA
Mohammad Ahmadi, University of Tennessee at Chattanooga, USA
The healthcare industry is under pressure to improve patient safety, operate more efficiently, reduce
medical errors, and provide secure access to timely information while controlling costs, protecting
patient privacy, and complying with legal guidelines. Analysts, practitioners, patients and others have
concerns for the industry. Using the popular strategic analysis tool of strengths, weaknesses, oppor-
tunities, and threats analysis (SWOT), facing the healthcare industry and its adoption of information
technologies (IT) are presented. Internal strengths supporting further industry investment in IT include
improved patient safety, greater operational efficiency, and current investments in IT infrastructure.
Internal weaknesses, however, include a lack of information system integration, user resistance to new
technologies and processes, and slow adoption of IT. External opportunities including increased use of
the Internet, a favorable national environment, and a growing call for industry standards are pressured
by threats of legal compliance, loss of patient trust, and high cost of IT.

Chapter 6
Using a Neural Network to Predict Participation in a Maternity Care Coordination Program ............ 84
George E. Heilman, Winston-Salem State University, USA
Monica Cain, Winston-Salem State University, USA
Russell S. Morton, Winston-Salem State University, USA

Researchers increasingly use Artificial Neural Networks (ANNs) to predict outcomes across a broad
range of applications. They frequently find the predictive power of ANNs to be as good as or better
than conventional discrete choice models. This paper demonstrates the use of an ANN to model a con-
sumer’s choice to participate in North Carolina’s Maternity Care Coordination (MCC) program, a state
sponsored voluntary public health service initiative. Maternal and infant Medicaid claims data and birth
certificate data were collected for 59,999 births in North Carolina during the years 2000-2002. Part of
this sample was used to train and test an ANN that predicts voluntary enrollment in MCC. When tested
against a hold-out production sample, the ANN model correctly predicted 99.69% of those choosing to
participant and 100% of those choosing not to participant in the MCC program.

Chapter 7
Can IT Act as a Catalyst for Change in Hospitals? Some New Evidence ............................................ 94
Teemu Paavola, LifeIT Plc, Finland

This chapter presents a succesful reorganization of a patient care process that was carried out in a
middle sized Finnish hospital. The reorganization of the patient care process for joint replacement
surgery succeeded in achieving a 50 per cent increase in operations. This study proposes that IT may
have an indirect influence on the achievement of goals, such as productivity, as soon as the IT invest-
ment has been decided upon; in other words, IT benefits start accruing before the IT component is even
in place. This is a new feature to add to the previous definitions, because this particular benefit cannot
be logically derived from any of the features of the actual IT system. Paying enough attention to this
phenomen at the planning stage can be vital to the success of new IT system investment.
Chapter 8
Informatics Application Challenges for Managed Care Organizations: The Three Faces
of Population Segmentation and a Proposed Classification System .................................................. 102
Stephan Kudyba, New Jersey Institute of Technology, USA
Theodore L. Perry, Health Research Corporation, USA
Jeffrey J. Rice, Independent Scholar, USA

Organizations across industry sectors continue to develop data resources and utilize analytic techniques
to enhance efficiencies in their operations. One example of this is evident as Managed Care Organiza-
tions (MCOs) enhance their care and disease management initiatives through the utilization of popula-
tion segmentation techniques. This article proposes a classification system for population segmentation
techniques for care and disease management and provides an evaluation process for each. The three
proposed operational areas for Managed Care Organizations are: 1) Risk Status: early identification
of high-risk patients, 2) Treatment Status: compliance with treatment protocols, and 3) Health Status:
severity of illness or episodes of care groupings, all of which require particular analytic methodologies
to leverage data resources. By applying this classification system an MCO can improve its ability to
clarify internal goals for population segmentation, more accurately apply existing analytic methodolo-
gies, and produce more appropriate solutions.

Chapter 9
Scrutinizing the Rule: Privacy Realization in HIPAA ....................................................................... 112
S. Al-Fedaghi, Kuwait University, Kuwait

Privacy policies, laws, and guidelines have been cultivated based on overly verbose specifications.
This article claims that privacy regulations lend themselves to a firmer language based on a model of
flow of personal identifiable information. The model specifies a limited number of situations and acts
on personal identifiable information. As an application of the model, the model is applied to portions of
the Privacy Rule of Health Insurance Portability and Accountability Act (HIPAA).

Chapter 10
In What Ways Does Web Technology Support the Individual in Choice Reforms in Health Care?
A Comparison among Norway, Denmark, and Sweden .................................................................... 126
Agneta Ranerup, Göteborg University, Sweden

The aim of this article is to evaluate the provision of Web support in choice reforms in health care in
Norway, Denmark, and Sweden. Two main issues are investigated: (1) What institutional frameworks
for choice in health care exist, and how is the exercise of choice supported by Web technology in these
countries? (2) As a consequence of this, what roles of the individual are mediated by this technology?
The present study provides a critical analysis of current technologies for providing information about
health care. It is concluded that in Norway the individual is equipped to be a reasonably informed con-
sumer, customer, and citizen. A similar situation exists in Denmark, but here the consumer role is even
more prominent. In Sweden, there has been little technological support for these roles, but recently
national actors have initiated a project aimed at creating a national portal for public health care.
Chapter 11
Characteristics of Good Clinical Educators from Medical Students’ Perspectives:
A Qualitative Inquiry Using a Web-Based Survey System ................................................................ 145
Gary Sutkin, University of Pittsburgh School of Medicine, USA
Hansel Burley, Texas Tech University, USA
Ke Zhang, Wayne State University, USA
Neetu Arora, Texas Tech University, USA

Medical educators have a unique role in teaching students how to save lives and give comfort dur-
ing illness. This article reports a qualitative inquiry into medical students’ perspectives on the key
qualities which differentiate excellent and poor clinical teachers, using a Web-based questionnaire
with a purposeful sample of third- and fourth-year medical students. Thirty-seven medical students
responded with 465 characteristics and supportive anecdotes. All participants’ responses were analyzed
through reviewing, coding, member checking, recoding and content analysis, which yielded 12 codes.
Responses from 5 randomly chosen participants were recoded by two authors with an inter-rater reli-
ability coefficient of 0.72, implying agreement. Finally, 3 larger categories emerged from the data:
Content Competence, Teaching Mechanics, and Teaching Dynamics. We incorporate these codes into
a diagrammatic model of a good clinical teacher, discuss the relationships and interactions between the
codes and categories, and suggest further areas of research.

Chapter 12
Open Source Software: A Key Component of E-Health in Developing Nations ............................... 162
David Parry, Auckland University of Technology, New Zealand
Emma Parry, National Women’s Health, Auckland District Health Board, New Zealand
Phurb Dorji, Jigme Dorji Wanchuck National Referral Hospital, Bhutan
Peter Stone, University of Auckland, New Zealand

The global burden of disease falls most heavily on people in developing countries. Few resources for
healthcare, geographical and infrastructure issues, lack of trained staff, language and cultural diversity
and political instability all affect the ability of health providers to support effective and efficient health-
care. Health information systems are a key aspect of improving healthcare, but existing systems are
often expensive and unsuitable. Open source software appears to be a promising avenue for quickly
and cheaply introducing health information systems that are appropriate for developing nations. This
paper describes some aspects of open-source e-health software that are particularly relevant to develop-
ing nations, issues and problems that may arise and suggests some future areas for research and action.
Suggestions for critical success factors are included. Much of the discussion will be related to a case
study of a training and E-health project, currently running in the Himalayan kingdom of Bhutan.

Chapter 13
An Empirical Investigation into the Adoption of Open Source Software in Hospitals ...................... 175
Gilberto Munoz-Cornejo, University of Maryland Baltimore County, USA
Carolyn B. Seaman, University of Maryland Baltimore County, USA
A. Güneş Koru, University of Maryland Baltimore County, USA
Open source software (OSS) has gained considerable attention recently in healthcare. Yet, how and why
OSS is being adopted within hospitals in particular remains a poorly understood issue. This research at-
tempts to further this understanding. A mixed-method research approach was used to explore the extent
of OSS adoption in hospitals as well as the factors facilitating and inhibiting adoption. The findings
suggest a very limited adoption of OSS in hospitals. Hospitals tend to adopt general-purpose instead
of domain-specific OSS. We found that software vendors are the critical factor facilitating the adoption
of OSS in hospitals. Conversely, lack of in-house development as well as a perceived lack of security,
quality, and accountability of OSS products were factors inhibiting adoption. An empirical model is
presented to illustrate the factors facilitating and inhibiting the adoption of OSS in hospitals

Chapter 14
Intelligent Agent Framework for Secure Patient-Doctor Profiling and Profile Matching ................. 195
Masoud Mohammadian, University of Canberra, Australia
Ric Jentzsch, Compucat Research Pty Limited, Australia

Radio frequency identification (RFID) is a promising technology for improving services and reduction
of cost in health care. Accurate almost real time data acquisition and analysis of patient data and the
ability to update such a data is a way to improve patient’s care and reduce cost in health care systems.
This article employs wireless radio frequency identification technology to acquire patient data and in-
tegrates wireless technology for fast data acquisition and transmission, while maintaining the security
and privacy issues. An intelligent agent framework is proposed to assist in managing patients’ health
care data in a hospital environment. A data classification method based on fuzzy logic is proposed and
developed to improve the data security and privacy of data collected and propagated.

Chapter 15
Towards a Conceptual Framework of Adopting Ubiquitous Technology in Chronic
Health Care ......................................................................................................................................... 214
Jongtae Yu, Mississippi State University, USA
Chengqi Guo, James Madison University, USA
Mincheol Kim, Jeju National University, South Korea

In the advent of pervasive computing technologies, the ubiquitous healthcare information system, or U-
health system, has emerged as an innovative avenue for many healthcare management issues. Drawing
upon practices in healthcare industry and conceptual developments in information systems research,
this paper aims to explain the latent relationships amongst user-oriented factors that lead to individual’s
adoption of the new technology. Specifically, this study focuses on the introduction of chronic disease
U-health system. Using the Ordinary Line Square (OLS) regression analysis, we are able to discover
the insights concerning which constructs affect service subscriber’s behavioral intention of use. Based
on the data collected from over 440 respondents, empirical evidences are presented to support that fac-
tors such as medical conditions, perceived need, consumer behavior, and effort expectancy significantly
influence the formation of usage intention.
Chapter 16
E-Patients Empower Healthcare: Discovery of Adverse Events in Online Communities .................. 232
Roy Rada, University of Maryland Baltimore County, USA

E-patients can empower themselves and improve healthcare. In online communities, patients may
discuss adverse events that are inadequately addressed in the literature. The author as a patient joined
various online patient discussion groups and identified several such adverse events. For each such ad-
verse event, the patient findings, the medical literature, and the implications are noted. Extracts from
the literature that were provided to the patients were welcomed by the patients. Possible approaches to
financially supporting such activities are sketched.

Chapter 17
Towards Process-of-Care Aware Emergency Department Information Systems:
A Clustering Approach to Activity Views Elicitation ......................................................................... 241
Andrzej S. Ceglowski, Monash University, Australia
Leonid Churilov, The University of Melbourne, Australia

The critical role of emergency departments (EDs) as the first point of contact for ill and injured patients
has presented significant challenges for the elicitation of detailed process models. Patient complexity
has limited the ability of ED information systems (EDIS) in prediction of patient treatment and patient
movement. This article formulates a novel approach to building EDIS Activity Views that paves the
way for EDIS that can predict patient workflow. The resulting Activity View pertains to “what is being
done,” rather than “what experts think is being done.” The approach is based on analysis of data that
is routinely recorded during patient treatment. The practical significance of the proposed approach is
clinically acceptable, verifiable, and statistically valid process-oriented clusters of ED activities that
can be used for targeted process elicitation, thus informing the design of EDIS. Its theoretical signifi-
cance is in providing the new “middle ground” between existing “soft” and “computational” process
elicitation methods.

Chapter 18
Applying Dynamic Causal Mining in Health Service Management .................................................. 255
Yi Wang, Nottingham Trent University, UK

This article describes an application that illustrates the role of data mining technology in identifying
hidden causal knoledge from health and medical data repositories. Across the health care and medical
enterprises, a wide variety of data is being generated at a rapid rate. Current information technolo-
gies tends to focus on a more statical side of causal knowledge and do not address the dynamic causal
knowledge. This article shows that the dynamic causal relation data can be captured for treatment,
payment, operations purposes and administrative directed insights. Accessing this currently unreal-
ized knowledge potential would enable the delivery of actionable knowledge to medical practitioners,
healthcare system managers, policy planners and even patients to make a significant difference in over-
all healthcare.
Chapter 19
Application of Unified Modelling Language (UML) to the Modelling of Health Care Systems:
An Introduction and Literature Survey ............................................................................................... 275
Christos Vasilakis, University College London, UK
Dorota Lecnzarowicz, University of Westminster, UK
Chooi Lee, Kingston Hospital, UK

The unified modelling language (UML) comprises a set of tools for documenting the analysis of a sys-
tem. Although UML is generally used to describe and evaluate the functioning of complex systems, the
extent of its application to the health care domain is unknown. The purpose of this article is to survey
the literature on the application of UML tools to the analysis and modelling of health care systems. We
first introduce four of the most common UML diagrammatic tools, namely use case, activity, state, and
class diagrams. We use a simplified surgical care service as an example to illustrate the concepts and
notation of each diagrammatic tool. We then present the results of the literature survey on the applica-
tion of UML tools in health care. The survey revealed that although UML tools have been employed
in modelling different aspects of health care systems, there is little systematic evidence of the benefits

Chapter 20
TreeWorks: Advances in Scalable Decision Trees .............................................................................. 288
Paul Harper, Cardiff University, UK
Evandro Leite Jr., University of Southampton, UK

Decision trees are hierarchical, sequential classification structures that recursively partition the set of
observations (data) and are used to represent rules underlying the observations. This article describes
the development of TreeWorks, a tool that enhances existing decision tree theory and overcomes some
of the common limitations such as scalability and the ability to handle large databases. We present a
heuristic that allows TreeWorks to cope with observation sets that contain several distinct values of
categorical data, as well as the ability to handle very large datasets by overcoming issues with computer
main memory. Furthermore, our tool incorporates a number of useful features such as the ability to
move data across terminal nodes, allowing for the construction of trees combining statistical accuracy
with expert opinion. Finally, we discuss ways that decision trees can be combined with Operational
Research health care models, for more effective and efficient planning and management of health care
processes

Chapter 21
Ambient Intelligence and Pervasive Architecture Designed within the EPI-MEDICS
Personal ECG Monitor ........................................................................................................................ 302
Hussein Atoui, Université de Lyon and INSERM, France
David Télisson, Université de Lyon and INSERM, France
Jocelyne Fyan, Université de Lyon and INSERM, France
Paul Rubel, Université de Lyon and INSERM, France

Recent years have witnessed a growing interest in developing personalized and nonhospital based care
systems to improve the management of cardiac care. The EPI-MEDICS project has designed an intel-
ligent, portable Personal ECG Monitor (PEM) embedding an advanced decision making system. We
present two of the ambient intelligence models embedded in the PEM: the neural-network based isch-
emia detection module and the Bayesian-network risk stratification module. Ischemia detection was
expanded to take into account the patient ECG, clinical data, and medical history. The neural-network
ECG interpretation module and the Bayesian-network risk factors module collaborate through a fuzzy-
logic-based layer. We also present two telemedicine solutions that we have designed and in which the
PEM is integrated. The first telemedical architecture was created to allow the collection of medical
data and their transmission between healthcare providers to get an expert opinion. The second one is
intended for improving healthcare in old people’s homes.

Compilation of References .............................................................................................................. 312

About the Contributors ................................................................................................................... 343

Index ................................................................................................................................................... 348

xvi

Preface

A State-of-the-Art Review of Developments in

Health Information System & Technology Models
and Methods:
The MEDIA Paradigm

IntroductIon

For years, misguided health organizational information technology (IT) leadership, and the lack of
available expertise and skills in the application of health IT (HIT) models and methods have somehow
predisposed administrators of large-scale health maintenance organizations (HMOs) to become gener-
ally reluctant to migrate away from legacy health information systems (HISs). For many health provider
organizations, the more recent decline in economic activities over the last few years has also further led
to new debates on allocating limited public and private resources so gravely needed to build and sustain
large-scale interoperable systems infrastructure in order to achieve such a system migration efficiently
and effectively. Not surprisingly, progress towards adopting new health IT initiatives such as innova-
tive e-technologies for the health care industry in the United States (US) has been slower than most
other industries. Inadvertently, a major gap in the strategic and opportunistic use of emerging health
IT models and methods now exists to significantly transform the apparently fragmented nature of US
health services delivery system.
Aside from the fear of costly systems failure, many key stakeholders in the US health care system
have, admittedly, been resistant to invest in new and contemporary enterprisewide systems due to the
lack of a well-focused national health IT vision and strategy. On the one hand, attempts to successfully
diffuse interoperable, integrative HIT applications throughout the US health care system must now de-
pend on the strength of the health IT leadership shouldered by the current Administration. On the other
hand, progress in health IT implementation and diffusion will also depend on how quickly many of these
health key stakeholder groups who have been technological laggards for one reason or another can be
motivated, challenged, and appropriately enticed to design, develop, and deploy large-scale, complex
and interoperable computer-based enterprisewide systems. Such enterprisewide systems include, but are
not limited to, those that are designed to incorporate integrative health data management models, new
biomedical informatic methods, web-based semantic search capabilities, and emerging clinical decision
support methodologies. In the context of today’s complex and largely fragmented US health services
 xvii

systems, specific applications of enterprisewide, interoperable systems can range from patient-centric
records and information services systems such as electronic medical records (EMRs), electronic health
records (EHRs), personal health records (PHRs), payor-based health records (PBHRs) and computer-based
physician order entry (CPOE) systems, to health administrative-aided transaction and health informa-
tion exchange (HIE) systems such as e-prescribing systems (EPS), supply chain management (SCM),
customer relationship management (CRM), enterprise resource planning (ERP), and e-payment systems.
Briefly, key reasons why the diffusion of interoperable, integrative HIT models and methods is needed
in the US health care services sector entail:

a. an urgency, as a whole, to contain escalating health care cost - the growing health care cost has
become an increasingly unsustainable burden on US taxpayers over the years;
b. the growing complexities and uncertainties in the health information processing and exchange en-
vironment of large, medium, and even smaller health organizations populating the US health care
system - the expansion of stakeholder groups and increased federal, state, and municipal regulatory
oversight mechanisms surrounding health services delivery have and will continue to add to the
already intricate health information management (HIM) and services delivery system; and
c. the potential of interoperable, integrative HIT models and methods to aid complex data analysis
and semi-structured decision making - such analysis will not only empower care providers, but
also enable critical information sharing to occur among referring physicians in consultation with
patients, and especially when specialists and a team of caregivers are involved in key administra-
tive and clinical decision making.

With increasing attention paid to the critical role that HIT models and methods can and will play in
reforming the US health care system, we are finally seeing a trend in increased projections on health
IT spending, which is currently anticipated to exceed 15 billion dollars annually for the US (Lipowicz,
2009). According to the National Coalition on Health Care, in 2008 alone, the US has spent well over
17% of its Gross Domestic Product (GDP) on health care - a percentage that clearly exceeded those
spent by many other OCED countries (National Coalition on Health Care, 2009). Sadly, the fact that the
US has outspent almost every other country on health care has not translated into better health or even
more convenient, accessible, available or affordable health care services delivery for Americans. In fact,
findings abstracted from 2007-2008 data in a study championed by the consumer health advocacy group
Families USA has revealed that one out of every three Americans under 65 may still have to live, at one
point in time or another, without health insurance coverage (Parisi and Bailey, 2009).
Clearly, many of the issues raised here have now taken central stage in the debate raised by the Obama
Administration for championing health care reform in the US. As well, various solutions have been
considered, chief among them, using and adopting interoperable, integrative HIT applications. Such a
strategy is not without merit, as various forms of health technologies have indeed risen over the years
to similar challenges; for instance, online claims processing and e-prescribing have been successfully
deployed by various HIT vendors to serve as tools to combat rapidly escalating health administrative
costs while simultaneously leveraged to reduce wastes, increase efficiencies, eliminate redundancies,
and improve the overall quality of clinical care and services.
Raghupathi and Tan (2002, 2008) noted that various strategic applications of HIT models and meth-
ods can evidently improve the efficiency and effectiveness of US health care services delivery, adding
value to existing, legacy-based HISs, and helping to integrate the islands of health services management
xviii

systems. Their arguments not only cited the power of e-technologies to streamline increasingly complex
routine HIM processes that may require multi-provider, cross-organizational collaboration, but also the
ability of interoperable, integrative HIT capabilities to augment enterprisewide efficiencies and care
provider network connectivity.
Accordingly, the key underlying argument for migrating from legacy HISs is that both data and pro-
cesses linked to diverse functioning information systems must now be shared on an increasingly real-time
basis between both on-site and off-site caregivers if we want to improve the quality of patient care. These
interoperable, integrative systems can also streamline complex administrative and clinical workload,
easing the communication needs among health care administrators, HIT personnel, health engineers,
health informatic researchers, and HIT consultants, all of whom need to work closely together in today’s
health services delivery systems if major systems bottlenecks are to be effectively managed over time.
Figure 1 depicts the Model and Evidence Driven Integrated Analysis (MEDIA) paradigm, an integrated
model management framework that is applied in this review to provide an integrative conceptualization
of the evolving, state-of-the-art developments in HIT models and methods.
Against this background, existing HIT models and gathered evidence will not simply aggregate
but will be seen as complementing each other alongside the MEDIA integration and analysis process.
Take the case of the influenza propagation - a disease model - and imagine how it can intermingle with
a patient care process model, such as that of a primary clinic. Given the separate evidence about the
flu outbreak and patient arrival pattern, running both models in parallel within the MEDIA framework
produces either a more realistic health care supply-demand scenario in the context of a real-world com-
munity health setting, and/or offers insights into other related and potential issues needed to be addressed
when challenged with health hazard preparation such as when an apparent discrepancy is observed to
be operative within the overall system. All such information analysis and its integration can be further
quantified and addressed consistently. Such multi-attribute analytic capability is crucial to the success
of any future applications and developments of HIT models and methodologies.
The rest of this review on health IT models and methods is organized as follows. In Section 2, a
high-level systems perspective of HIS information and process flow in the context of applying HIT
models and methods is outlined. Fundamentally, the classical information system input-process-output

Figure 1. The model and evidence integrated analysis (MEDIA) paradigm

Model Integration

Ontology modeling
Evidence Fusion

Service Assurance
Probabilistic dependency modeling a. Delivery improvement
b. Risk management

Hybrid computational analysis

 xix

triad that underlies all systems engineering conceptualization is revisited. In Section 3, the focus will
shift to the complexity-uncertainty challenge encountered in the current US health care sector. Follow-
ing this, in Section 4, the emerging engineering approaches in health services applications at two levels
are highlighted. These levels include: (1) the general health engineering level; and (2) the more focused
health services modeling level. In Section 5, the MEDIA paradigm, including its three key facets are
briefly overviewed: (a) domain ontology modeling and management that lays the foundation for model
representation and infrastructural connectivity; (b) hybrid probabilistic modeling that computationally
integrates systems and subsystems models; and (c) adaptive knowledge fusion to generate quality as-
surance that support active information collection and continuing operations management. Essentially,
the MEDIA approach is based on a linked series of health engineering concepts, which deviates funda-
mentally from the traditional health care best practices in which quality is heavily dependent upon the
hands-on skills of expert clinical practitioners as well as the deployment of progressively specialized
medical instrumentation. The application of the MEDIA paradigm is also illustrated in a case study on
risk management. Finally, in Section 6, we conclude this review by speculating on future research direc-
tions and practical implications in the field of health IT models and methods.

VIewIng HIt Models & MetHods In tHe classIcal HIs Flow cycle
Model

Owing to the need to deal with increasingly complex health data processing routines and the growing
number of participating stakeholders who may need to share information and provide their particular
interpretation of such data taken from the very same electronic health databases, an exciting playground
for health researchers and practitioners now exists to incorporate efficient medical information processing
models and emerging health decision support systems (HDSS) methodologies into interoperable systems
linked to existing health databases, model management bases, and knowledge management bases.
Despite recent investments in medical information packages such as the Vista medical information
system and the GoogleHealth patient record management system, the lack of system interoperability
has made it difficult, if not impossible, to integrate isolated health records stored at varying functional
levels. Moreover, a key challenge is the integration of HIT models and methods through a knowledge
fusion process so as to enhance the capability of viewing and analyzing any chosen set of discrete and
continuous events in a health services delivery system such as relating patient-physician encounters and
the progression of different disease stages from an enterprisewide and trajectory perspective, for example,
the interactions and optimal interventional strategies for a particular patient cohort within a HMO, built
up from its connected elements, including the affiliated hospitals, clinics, practicing physicians, nurses,
information and personnel resources, labs, and departments linked to its health services delivery system.
Figure 2 depicts a generic health information flow process model in which health administrative,
clinical, and service delivery decision and policymaking must necessarily transpire at an enterprisewide
level while aided with the use of relevant and applicable HIT models and methods within the tradition-
ally defined input-process-output system triad.In the initial data collection, and information-knowledge
gathering stage, a huge amount of raw data, guided protocols, and knowledge elements are often sourced
from various input sensors and devices as well as recorded answers to questions asked of the individual
patient and/or groups of patients, including underlying reasons for these patients to seek care. All of the
xx

Figure 2. A general health information flow process model with Its input-process-output system triad

Data
Collection

Intelligent
Information
Processing

Health IT Models & Methods

Knowledge
Formation
& Feedback
Health Care Services
Decisions &
Policymaking

collected information is now ready to be pre-processed into some form of meaningful and intelligent
datasets.
These datasets are then stored appropriately either in a centrally located or in multiple electronic
locations, typically via databases and data warehouses, waiting for further processing to serve a variety
of clinical, research and administrative goals and purposes. Such purposes could range from clinical
testing and follow up, to research, administrative billing, and/or mere reporting for managerial decision
making. As a case in point, imagine the gathering of demographical data as well as specific vaccination
records for a population of young adult patients to guide the clinical management of an ongoing epi-
demic, such as the “swine flu”. The data gathered should not only be accurately coded but they should
also be securely stored in a more or less structured format to be communicated to the attending physi-
cians for further clinical evaluations, testing, and/or diagnosis. In fact, the same information may be
aggregated with other data for research analysis or it may be used to establish insurance co-payments to
the various care provider groups. Finally, the information may also be critical to plan health vaccination
programming by the different workplace or educational institutional settings tied in with the population
of young adult patients.
At the data manipulation-information mining-task processing stage, embedded patterns are often
sought and hidden knowledge uncovered within the captured datasets to be translated intelligently into
meaningful clusters and to provide additional advice and insightful guidance to the end-users, in this
illustrative case, the various caregivers. In other words, the same data-information-knowledge collected
of the young adult patients may now be further fused with other high-level information, such as expert
clinical knowledge about the state of the patient immunization types and dates as well as the state of
ongoing epidemics or pandemics, if any. The combined information can then be used to determine the
health status of the individual young adults following the immunization, for example, their resistance to
prevailing illnesses, their showing signs of some sort of common allergies, and/or their susceptibility to
some other atypical reactions. In fact, it is here that the HIT models and methods may be most relevant
to aid in the follow-up of crucial clinical and administrative decisions.
 xxi

In operationally organizing, aggregating, mixing, dicing, and mining the stored patient data, various
HIT models and methods may be combined in a complex fashion to generate a meaningful and “best”
fit for clustering and/or grouping the stored data based on some known statistical patterns. The resulting
data pattern classification are in turn routed back, on the one hand, to the clinicians to guide them in
making prognostic, diagnostic and other therapeutic decisions, and, on the other hand, to the adminis-
trative staff to assist them in determining patient diagnostic codes, and thereby, generate accurate and
appropriate third-party billings.
At the end of the data processing-data mining stage, feedback will also typically occur in terms of
second opinions, reviews and/or expert evaluations on model validity to further enhance systems fine-
tuning and performance outcomes. Such enhancements recognize the possibility for more intelligent
decision making in the face of missing or incomplete data and/or the fact that inadequate tools were
available to appropriately process such data so that adjustments to these decisions could still be made
to further improve accuracy or timeliness of key decisions and outcomes. In our case example, the nurs-
ing unit may, at this point, requests for feedback from the attending physicians as to who among these
young adults would be due for specific and further vaccination visits, be sent for more clinical testing
and evaluations, such as generating referrals to other departments or physician specialists, or they should
simply be discharged. This commonly encountered health information flow cycle is characteristic of an
open health information processing system that follows the input-process-output triad.
Hence, despite the complexity of HIT models and methods, the potential for these methodologies
to contribute significantly to future health services management is evident and apparent. While their
applications represent mostly a part of the wider and far reaching field of HIS, it should not and can-
not exist in and of itself. It must go hand in hand with the many different types of decisions on health
services management that have to be executed routinely alongside the processing of initially gathered
data, stored information, and captured knowledge. With a basic understanding on the flow of information
and processes through a generic HIS flow cycle model, it is clear that multiple disciplines – including
computer and library sciences, medical and/or health informatics, sub-fields in nursing informatics,
teleradiology, telemedicine, telehome care, telematics, biomedical informatics, as well as domains of
health IT, health engineering, and health operations research (OR) and management science, HDSS, and
even the integration of cognitive sciences, information sciences, and health sciences – must all come
together to enrich our ability in using HIT models and methods to meet a diversity and variety of clinical
and health administrative decision needs.
Ultimately, in order to achieve high quality health services management and related decision mak-
ing, not only will we need medical informatics tools such as computers, clinical protocols, formal and
informal medical terminologies, but also the integration of various informatic resources, tools, models,
devices, techniques, methods, decision aids, and methodologies to optimize the collection, storage, re-
trieval, analysis, design, and use of health information in health services delivery (Tan and Payton, 2010).

ratIonale For IntegratIng HIt Models & MetHods: tHe us

HealtH care coMplexIty-uncertaInty landscape

While the interplay of multiple disciplines and sub-disciplines in managing a growing body of health
services information and processes makes the health care system one of the most complex systems to
study, the growing intricacies of the US health services sector may also be seen in many real-world events.
xxii

The 2007 drug-resistant tuberculosis case, for example, shows the high degree of interdependency that
exists in the US health care infrastructure, and uncovers the many hidden flaws, previously ignored, in
the management of patients, hospitals, the center for disease control and prevention (CDC), as well as
the complex sociopolitical forces that link the US government with other national governments (U.S.
House of Representatives, 2007)

Key challenges & risk Management in the us Health services sector

For the US health care sector to achieve greater accessibility, affordability, and accountability, two key
determinants for its continuing growth and future developmental success include: (1) ensuring a 24/7
service availability; and, (2) having a focus on service quality.
In the coming years, as the US health care sector experiences sweeping reform, daunting risk and un-
predicted challenges will likely arise. Hence, dealing with risk and uncertainty is a critical skill in running
any modern-day health services enterprise, and this is precisely where HIT models and methods can and
will play a central role. The major threats and challenges facing the US health care sector are depicted
in Figure 3. As shown, these challenges include systems complexity, event uncertainty, and the need to
share information. Owing to the fact that such threats and challenges will often emerge unpredictably
and their diverse manifestations will also typically blur the line of most standard classifications, both
health administrators and clinicians will want specialized tools to aid them in their routine policymaking
and decision execution, as we moved forward with the US health care reform.
First, with regard to systems complexity, there is the overwhelming intricacy of diverse evolving
structures alongside the rapidly expanding health services supply networks. With hospital expansions and
alliances formed among various stakeholders, increasing complexities have also followed. Again, with
stakeholders trying to leverage on increasingly competitive global health supply chains, the resulting
system relations are even more complex. Moreover, stakeholder collaboration across diverse industrial
sectors may often follow different policy paths and practices. These factors, and the fact that most such
collaboration is developed independently, compound the intricacy. Hence, there is the need for integrated
HIT models and methods to guide future decisions.
Second, in regard to event uncertainty, there is the need for systems configuration flexibility. A health
service is subject to risk whenever and wherever the stakeholder cannot predict changes to the typical

Figure 3. Major threats & challenges facing the US health care sector industry

Challenges
Complexity Uncertainty Information Sharing
• Healthcare alliance; • New, rare, and epidemic • Incompatibility;
• Global supply chains; diseases; • Inoperability;
• Regional, national, and • Natural disaster and • Integrity and accuracy;
international coordination; man-made emergency; • Cyber attack;
• … • Terrorist attacks; • Privacy regulation;
• … • …

Service Availability and Quality

Staff Resources Information/Knowledge
 xxiii

environmental situation. As event uncertainty can arise, without prior warning, from diverse occurrences,
such as (a) variations in staff and patient flow, (b) the onset of new, rare or epidemic diseases, and (c)
emergencies, whether natural and/or man-made, all such uncertain events will require appropriate and
timely adaptation in health services management. These challenges are further convoluted by evolving
structures, novel technologies, the constant threats of natural disasters and/or man-made events, includ-
ing the unpredictable growing number of participating stakeholder groups.
Traditionally, stakeholders including patients, care providers, insurance companies, and the govern-
ment do not see eye-to-eye. As many health administrators and clinicians tend to focus largely on the
details, and likely also to stubbornly stick to their personal heuristic business decision-making biases,
challenges of prevailing complexity and uncertainty will become even more evident over time. In light
of this, it is not uncommon to have conflicting views shared among the multiple stakeholders. Moreover,
health care debates often hinge on highly sensitive issues, and it is unlikely that any lopsided discussions
or outcomes will contribute to the overall satisfaction of certain key stakeholder groups. Therefore, given
that current management practices rely largely on personal heuristics, qualitative comparisons, and sub-
jective guidelines, without precise accounting and science-based quantitative analysis, the significance
of HIT models and methods cannot be overly emphasized.
As today’s health services delivery systems often encompass huge numbers of interacting elements,
significant interdependency, and enormous uncertainty and risk, it is very difficult to expect quality
and outcomes of the decisions to be appropriately calibrated and/or justified quantitatively. Hence, the
attempts to incorporate the latest science and methodological advances in HIT models are some of the
ways to ensure better health decision outcomes and policymaking.
Third, in regard to information sharing, it is in the interest of health provider organizations, at the
minimum, to ensure that their patients receive prompt and accurate medical care services by increasing
the accessibility and availability of health information services on the one side, and ensuring the secured,
confidential, and private storage of personal medical records for sharing among affiliated caregivers on
the other side. Technologically speaking, effective information sharing has to do chiefly with systems
interoperability and the availability of an interoperable enterprisewide infrastructure. This is why a move
towards integrative, interoperable HIT models and methods is key to achieving US health care reform.
In summary, the challenges of complexities, uncertainties, and sharing of health data have deterred
key stakeholders within the US health care industry sector to work collaboratively, effectively, and pro-
ductively. But it has also highlighted the importance of having a national health IT strategy and vision
as well as building a cumulative effort to apply and diffuse integrative, interoperable HIT models and
methods.

Inadequacy of Health services Modeling practice & research

Despite the wide range of quantitative models that has emerged to address various performance optimiza-
tion risks with regards to different parts and/or aspects of health services systems over the years, there
is still the need for a system-wide perspective on how these different HIT models and methods may be
combined and integrated to aid health care decisions and policymaking.
Roberts (2007), for instance, emphasized models to represent the progression of certain disease cat-
egories as well as to predict best treatment timing and cost. Denton, Fowler, Schaefer, Batun, Erdogan,
and Gul (2009) detailed scheduling systems that model patient arrival and how they may be queued
efficiently and effectively to be served by doctors in operating rooms. Finally, Shehad, Bertino, and
xxiv

Ghafoor (2005) discussed the application of computational models to quantify information leakage and
availability based on the analysis of user communication via a computer infrastructure.
Notwithstanding, each of these isolated attempts represented a series of uncoordinated efforts to
address only a small, specific set of health services elements, without adjusting to benefit the greater
whole – that is, the limited impact of the individual studies on the total system do not account for the
intricate interdependencies among the systems elements. In other words, confined within a limited scope
of a system component or sub-system, the performance measures adopted by these studies are typically
defined in terms of cost, quality, time, or other specific but narrowly defined criterion, which, without
exception, would still impact, one way or another, on other systems elements and components on an
unknown basis and scale. Thus, isolated models and methodologies, working in and by themselves,
simply cannot address the larger challenges presented in the US health services management system.
With health engineering as an emergent field, researchers interested in HIT models and methods
are beginning to redefine the boundary of traditional health care management philosophy. The model-
ing capability to guide flexible reconfiguration of health care structures, such as new investment and/
or rearrangement of provider organizations and facilities, is critical if we are to handle the complexity
of systems modeling and analysis appropriately. Arguably, the different levels of local, regional, and
national coordination that are actually interconnected (Schnase & Cunnius, 1995) will dictate how the
national health infrastructure is to be reconfigured to manage existing resources efficiently, effectively
and productively.
All in all, for enterprisewide systems modeling, not only should external disturbance be taken into
account, but also losses incurred due, simultaneously, to rising health care cost and the need to minimize
poor quality medical (patient) care (Hick, Hanfling, Burstein, DeAtley, Barbisch, Bogdan, & Cantril,
2004) In this sense, many industry leaders and practicing engineers have now recognized the significance
of integrating the operation, audit and control management of health supply systems (Barbera & Ma-
cintyre, 2002). Unfortunately, most, if not all, of these initiatives are still in their infancy. For instance,
many health services institutions still lack the required level of modeling capability and organizational
preparedness for dealing with external interfering events.
Indeed, as we attempt to introduce the latest enterprise paradigms such as the system-of-systems
and the reconfigurable enterprises across institutions where a large number of participants interact and
evolve on a continuous basis, the lag in health services reengineering practices and health performance
management research becomes clearly evident. As early as 2003, Woolhandler, Campbell, and Him-
melstein (2003) have noted that, by 1999, 31% of the US health expenditures and 16.7% of the same
expenditures in Canada were largely health administrative overhead expenses; clearly, this points to
the high potential for HIT models and methods to achieve a significant system performance improve-
ment largely through a reduction in health administrative cost. Moreover, just as we set to data mine
and understand the interactive behavior of a complex adaptive system (CAS) (Kiel, and Elliott, 1996;
Woolhandler et al, 2003) emerging states can also be generated under various systems configuration to
be studied so as to achieve across-the-system improvements.

comparison to service enterprise risk Management

From a service supply chain management perspective, the dynamics of the US health care system may be
conceived as comprising, with special structures and performance requirements, service supply networks
of suppliers on the one hand in terms of caregivers and medical resources, and patient demands on the
 xxv

other hand in terms of patient needs and services. As we know, health systems are, without exception,
multi-layered and closely interconnected; these systems must necessarily provide high value to their
customers in terms of patient benefits and services; as well as the availability of services, security, and
privacy of captured and stored patient information. Many of these factors entail the highest priority in
risk management.
Whereas traditional OR techniques such as game theory (Shubik, 2005) 18 derived from artificial
intelligence (AI) and economics, and optimization models 18 taken from quantitative analysis have been
extensively used for enterprise modeling, current research focuses largely on core strategic and opera-
tional issues such as location configuration, demand response analysis, dynamic pricing and contract
management, as well as e-commerce challenges (Graves, Kletter, & Hetzel, 1998; Lambert, Cooper, &
Pagh, 1998).19,20 Risk assessment and risk management cover essentially many forms of risks, includ-
ing, but not limited to, business risk, financial risk, technological risk, and physical risk (Anupindi &
Akella, 1993; Argrawal & Nahmias, 1997; Kouvelis & Milner, 2002; Simchi-Levi, Kaminsky, a&nd
Simchi-Levi, 2003;) 22, 23, 24, 21.
Lately, research into the applications of HIT models and methods has also taken on a broader per-
spective. Thomas (2002), for example, analyzed the reliability of a supply chain under contingency
when it is being impacted by unexpected disasters. Using Bayesian networks, Pai, Kallepalli, Caudill,
and Zhou (2003) provided a conceptual business risk assessment framework. Based on industry prac-
titioners’ input, Blackhurst, Craighead, Elkins, and Handfield (2005) summarized the common themes
and issues surrounding supply chain disruption. Hale and Moberg (2005) presented a set cover location
model that is used in disaster preparation for identifying the minimum number and possible locations
of off-site storage facilities for supplies. Lee and Whang (2005) compared the inspection of hazardous
goods, such as for transporting explosives, with total quality management (TQM) in manufacturing. A
RAND Corporation (2004) report focused on the impact of terrorist attacks on global container supply
chain performance and advocated the importance of fault-tolerance or resilience.
Despite the stream of well-funded research in the area of service supply chain and enterprise man-
agement, the emerging body of knowledge has not adequately addressed the growing complexity and
uncertainty-risk management dimensions of the US health care sector. Little, if any, of the past scat-
tered research has provided a comprehensive analysis or given enough attention to systems and systems
modeling integration, which has resulted in the following significant challenges:

a. Current research is sparse and isolated, lacking a cumulative agenda. Even so, common terminolo-
gies or protocols are not often shared to ease communications among future researchers. Ultimately,
there is the lack of a bridging theme to aid the systemic interfacing among heterogeneous models
and documented evidence;
b. Past researchers generally hold a narrow view, focusing solely on a small subset of traditional risks.
In this sense, most studies deal primarily with individual system components or aspects thereof,
without an overall context to account for all interlaced challenges and subsystem interactions; and
c. Enterprisewide modeling and analysis is still in its infancy. Moreover, past studies tend to be mostly
conceptual and qualitative, with limited applicability to modern enterprises in addressing current
challenges.

Put together, the need for a systemic health services management framework, a representation that
would exhibit three fundamental characteristics. First, it would encapsulate a family of heterogeneous
xxvi

models at different levels of detail and for different subsystems, scalable also to the growth of the en-
terprise and its captured knowledge cannot be overly emphasized. Second, such a representation would
bridge high-level qualitative knowledge with quantitative computation by automatic conversion and
instantiation, a capability that is crucial to practical deployments of HIT models and methods. Finally,
we would also expect such a representation to incorporate quality assurance with confidence and clarity,
measured to handle uncertainty in analysis and to guide decision making in all facets of health services
management.
The MEDIA paradigm, which we will discuss in Section 5, is such an integrative paradigm exhibiting
all of the three fundamental characteristics noted above.

state-oF-tHe-art HIt Models & MetHods

Before overviewing the MEDIA paradigm, however, we will first review the state-of-the-art HIT models
and methods by categorizing and highlighting the various bodies of literature with respect to the emerg-
ing engineering approaches in health services applications.
Improving a health services delivery system can be achieved through one of two means: (1) relying
on medical equipment, technological and/or procedural advances; and, (2) focusing on improved medi-
cal, business, clinical and administrative processes. Modern health engineering technologies such as
nano-scale materials, genetic informatics, robotics and applications of virtual health have gone a long
way to provide effective and reliable prognoses and treatments to a variety of diseases. Such scientific
progression foresees an accelerating pattern in light of technological advances. Yet, management theories
and practices have taught us that the quality of health services delivery is also dependent on achieving
better health information flow and system processes that best fit the type of organizational structure and
culture inherent to particular health institutions.
In other words, a solid foundation for health care reform may be achieved only through the appropriate
balancing of the health resource configuration and process control. In light of this, a large body of models
to address medical and health services delivery issues in health system management has evolved through
best practices over the years. This approach, which concentrates on the applications of health IT models
and methods, is precisely the focus of this review. It has, in the last decade, gained increasing attention
among researchers interested in contributing to enhancing the performance of the US health care sector.

Health engineering

Today, a new body of health engineering models has emerged. These models are applicable at different
system levels, including, for example, at the low level of disease progression of individual patients,
that of a clinical session process, or at a higher level involving a health facility of interconnected sub-
systems and components, or even at the level of a complete life cycle of health services delivery of a
particular region or country. In the following section, we survey the extant literature on the new “health
engineering” paradigm, following which we shift focus then to the more specific health care modeling
and simulation domains.
Evidently, health engineering methods and IT models will not only help hospital personnel to reduce
medical errors and risk, but aid also in reducing health care costs, improving health services timeliness,
and increasing patient satisfaction. In some cases, it is of course possible that the traditional culture and
 xxvii

strict legislation imposed on the health care sector, which uniquely distinguishing it from many other
industries, may limit the transferability and diffusion of new concepts and methods.
Today, opportunities and challenges in the health care sector exist for testing and adopting concepts,
best practices, and tools from diverse engineering domains to improve the efficiency of systems processes
and the effectiveness of health services delivery in terms of quality, safety, and productivity. In this sense,
both the goals and challenges of the health care sector match those of other industries where knowledge
engineering has provided a long-term basis for resolving bottleneck issues. Hence, the combination of
knowledge engineering and health IT applications promises to provide many more reengineering op-
portunities for health services delivery, for example, the installation of bedside terminals, the adoption
of new informatic methods for infectious disease control, and therapies, and the use of innovative data
and DSS technologies in medical and clinical settings, including hospital laboratories and pharmacies.
This section reviews emerging operational systems engineering (OSE) research categorized into sev-
eral different dimensions. Specifically, the focus is on OSE methods abstracted from a mix of engineer-
ing disciplines, including human factors, factory and product design, and security engineering. As the
managerial processes and services in the health industry are often similar to those of other manufacturing
and servicing industries, we argue that the field of health services management can be enhanced through
the intelligent applications of traditional health IT OR models, emerging engineering philosophy and
methodology, and mature OSE methods.

Traditional HIT OR Models

Pierskalla and Brailer (1994) discussed the application of OR models and methods in a broad range of
health services management tasks. Major applications of health IT OR models include:

1. Demand Forecasting – A fundamental input to many other analyses in health engineering, demand
forecasting, such as predicting daily census for the different types of resource needed, is important
to improve the efficiency of health resources allocation.
2. Capacity Planning - The allocation of bed capacity within the hospital is a critical factor in opera-
tional efficiency. Thus, health services capacity planning typically focuses on total bed capacity,
bed capacity allocation to different services, surgical system capacity, capital equipment capacity,
and ancillary service capacity. Discrete event simulations and semi-Markov process models have
been used to examine bed allocation and related capacity questions.
3. Patient Screening - Screening of patients for particular disease can improve medical diagnosis on
the one hand and disease detection on the other. It may also be applied individually (individual
screening) and/or population-wise (mass screening). Specifically, for individuals, the objective is
often to prolong a patient’s life, whereas, in mass screening, the objective may be to minimize the
cost at the societal level, thereby lowering the prevalence of a specific contagious disease. Clearly,
when attempting to achieve any such objective function in OR modeling, there still may be resource
constraints and compliance levels to be factored into the solution.
4. Patient Scheduling - Scheduling is critical for matching demand with the supply of available but
limited resources. Most scheduling systems attempt to optimize the combined objectives of patient
and worker satisfaction, as well as the utilization of facilities.
5. Clinical Decision Making - Clinical decisions can be aided through OR models that incorporate
mathematics and structural analysis. Not only can such analytic models assist in the formulation
xxviii

of health care policies, but they can also be applied to the structuring of critical medical decisions,
and the fine-tuning of health systems performance.
6. Workforce Planning and Scheduling - Human resources management is one of the most costly and
intensely unpredictable activities that is to be managed intelligently across health organizations.
For example, Hershey, Pierskalla, and Wandel (1981) conceptualized the nurse staffing process as
that of a hierarchy of three decision levels over different time horizons and with different preci-
sion – that of corrective allocations, shift scheduling, and workforce planning.
7. Cost Cutting – Kumar, Ozdamar, and Zhang (2008) have developed several reengineering concep-
tual and simulation models, which were used for cost containment within the Singapore’s health
industry in the domain of supply chain management process reengineering.

Emerging Engineering Philosophy and Methodology

Over the years, various streams of engineering philosophy and methodology have been applied to improve
health systems performance. Process orientation and patient focus are the essential concepts embedded
in these methodological philosophies.
Key approaches that have been discussed in the extant literature include: (1) Lean Thinking; (2)
Six Sigma; and (3) Theory of Constraints. Young (2005) argues how a clinical session for a patient’s
treatment would serve as a good analogy for explaining how these different streams of engineering
philosophy and methodology can be combined to address various system bottlenecks encountered in
the health services delivery industry.

1. Lean thinking – Kollberg, Dahlgaard, and Brehmer (2007) believed that the idea of lean thinking
is applicable specifically to health care systems in a number of ways. For example, just-in-time
(JIT), level scheduling, and multi-skilled teams are generic techniques that can create a smooth
operation process flow through the matching of the supply-demand level of health care resources.
When applying lean thinking to health care, a measurement framework for lean initiatives that
reflects both efficiency and effectiveness of health systems performance, such as patient satisfac-
tion, referral management, process mapping, and fulfillment of targets and policies, is needed in
order to fully capture lean changes.
2. Six Sigma - This methodology involves standardized data collection and informed reporting proto-
cols based upon a well-controlled quality feedback cycle to minimize variations and quantitatively
align production or service quality to a predetermined standard. “Bridges to Excellence” is an ex-
ample of a national case initiative based on Six Sigma quality feedback methodology launched to
improve clinical care quality. This initiative targeted on physicians and their practices to enhance
patient care quality (Brantes, Galvin, & Lee, 2003) and had further been incorporated as part of
a collaborative product commerce (CPC) approach to health supply chain purchasing (Ford &
Hughes, 2007), as highlighted in the next section.
3. Theory of Constraints (ToC) – Similar to Six Sigma, ToC applies the root cause thinking processes
for analyzing system bottlenecks. Unlike Six Sigma, however, ToC attempts to deal with managing
constraints in CAS not from a technical limitation perspective, but from a “qualitative” and philo-
sophical perspective. The methodology is first applied to identify the most vulnerable constraint,
then exploiting and increasing flow through that constraint, working from the weakest link upward
to other links between that constraint and the overall system.
 xxix

Mature OSE Discipline

A maturing OSE discipline focuses on examining, analyzing, and further understanding the operating
elements and systems dynamic processes in complex systems to achieve efficient and effective systems
performance. This greatly supports the view of health care as a CAS. In this sense, OSE tools and
techniques may be applied to achieve a balance for meeting multiple goals, for example, quality patient
safety, accessibility, availability, comprehensiveness, and affordability of care.

1. Supply chain management (SCM) – In recent years, SCM topics have gained significance as health
services organizations vie to lower the cost and improve the ease of accessibility and delivery of
health services and their associated resource supplies (Brantes et al, 2003; Ford & Hughes, 2007).
38,39 .
Ford & Hughes (2007) identified potential barriers that health services organizations must
overcome in order to apply SCM principles successfully within the health care sector. In their
study, they inferred that physician services are the primary channels in a health care supply chain
to provide the relevant expertise and services to group practices, hospitals and pharmacies. These
practices, in turn, behave as secondary channels, acting as refineries and production facilities to
serve health insurance distributors and purchasing programs in the supply chain. Therefore, the
starting point for cost containment of health services in SCM will be determined by how physician
services are being managed. A specific form of SCM used by US employers is the collaborative
product commerce (CPC), which is discussed next.
2. Collaborative product commerce (CPC) – CPC attempts to extend the limited boundaries of enter-
prise collaboration for product design by leveraging innovative e-technologies to engage members
from internal as well as external constituencies. CPC approach differs from other SCM tools in
two significant aspects (Swinehart & Smith, 2005): (1) it permits inter-organizational collaboration
feeding on a common supply chain or meeting similar consumer product or services needs; and
(2) its processes are transparent to all stakeholders. CPC models, therefore, allow health provider
organizations and/or third-party payers, who are located in different health care markets, to share
in innovative product life cycle designs.
3. Business process reengineering (BPR) – BPR is a process-driven technique to improve the efficiency
and effectiveness of a business through meaningful process redesign, change management, and
system reorganization. As an example, Kumar, Swanson, and Tran (2009) conducted BPR using
simulation modeling on the complex operating theatre (OT) system in a Singapore Hospital. His
case simulation produces two recommendations: (1) a need to redesign the OT process in order to
maximize its productivity without altering the current workload of surgeons and anaesthetists; and
(2) reviewing the OT utilization data periodically so as to derive a meaningful productivity index
and accurately gauge its utilization.
4. Health management information systems (HMIS) – HMIS, which has to do with all aspects of
business information systems functions in health care, plays an integral part in any modern health
services system.43 As noted earlier, a national health IT strategy to support and ensure systems
interoperability to link health services networks throughout America is believed to be a neces-
sary step towards realizing US health care reform (IEEE-USA, 2005). Today, many traditional
HMIS functions can be easily and logically augmented to encompass database, model-based and
knowledge-based HDSS technologies when trying to build OR models to aid in the management
xxx

of health services systems, or more specifically, to improve administrative productivity, increase

clinical decision-making responsiveness, and enhance patient care quality.

Health services Modeling

Serving as the foundation for our proposed MEDIA framework to be discussed later, the extant literature
in health services models may be further subdivided along the following three dimensions:

1. the scale (level) of health system problems being studied;

2. the goal being sought or performance measure being evaluated; and
3. the modeling method being applied.

The intelligent applications of health IT models and methods depend largely on understanding how
each of these dimensions will impact on a specific health services modeling study.
Many variations and types of models exist such as policy models, procedural models, intervention
models, graphical models, deterministic as well as stochastic models. As a case in point, Eldabi and Young
(2007) indicated that quantitative models, methods or tools can be studied across different organizational
levels, from the physical and mechanical design level through the services and policy design level.

The Scale (Level) of Health System Problems

This dimension focuses mainly on the extent and/or scale of health services problems to be modeled.
Four sub-levels include: (a) Individual patient disease models; (b) Operational process models; (c) Or-
ganizational system-level models; and (d) National sociopolitical-level models
Individual patient disease models focus mainly on the biological disease processes occurring in
individual patients, that is, the infection processes among either healthy or infected person/population.
These models can range from microbiological or cellular, to organ, as well as person-to-person trans-
mission level.
O’Leary (2004), for instance, simulated person-to-person infection by building mathematical models
to predict the epidemic path of disease transmission, to assess possible infectious outcomes of events
over time, and to test the effectiveness of different intervention measures such as vaccination strategies
and quarantine policy. In general, such disease progression models are applicable to evaluating clinical
effectiveness or cost effectiveness of different interventions to particular disease (Brailsford, 2007).
The operational process models are devoted to observing and simulating individual patients residing
in a ward, a clinic, or a hospital department such as the Emergency Department (ED) or Intensive Care
Unit (ICU). Often, such models are used to facilitate business process reengineering, resource allocation,
capacity planning, staffing, and scheduling challenges. Tan, Gubaras, and Phojanamongkolkij (2002)
for instance, employed a discrete event simulation model for studying capacity planning, staffing, and
scheduling of Dreyer Urgent Care Center.
Organizational system-level models combine different departments within a large institution or en-
terprise, and attempt to study the interactions among the different departments. Typically, such models
address longer-term and broader issues, similar to the conceptualization of enterprisewide models. Sys-
tem dynamic (SD) is a common example of organizational system-level model that is often applied at a
 xxxi

strategic level where the stakeholder is interested to look at the forest more than just the trees (Brailsford,
Lattimer, Tarnaras, & Turnbull, 2004).
Finally, national sociopolitical-level models encompass and address a wide range of very high-level
issues. Having a view at a national sociopolitical level on the state of emergency, for instance, will aid
government policymakers and decision makers design a reliable health services delivery policy that will
coordinate hospitals, police, and emergency units for a rapid and more readied response. For example,
the BioSense Real-Time Clinical Connection Program of the United States (Laudon& Laudon, 2007)
builds a national surveillance system model to continuously summarize and analyze the disease and
health information by source, day, and syndrome for each ZIP code, state, and metropolitan area. This
model is designed to improve the national capabilities for disease detection and monitoring, as well as
awareness of real-time health situations.

Goal Sought & Performance Measures

In health systems, goals sought are often different from many other for-profit industries, for example,
appropriateness of care, safety, and patient satisfaction are critical relative to cost and resource utiliza-
tion, not just profits. These goals are usually measured by a set of key performance indicators such as
patient throughput as measured in system wait time and wait line; length of stay (LOS); system capac-
ity; utilization of staff, equipment, and space; cost; and various other related measures such as service
quality in terms of error and patient satisfaction.
Major groupings of studies on various health systems performance measures include the following:

1. Length of Stay (LOS) and Patient Throughput – Ramis, Palma, Estrada, and Coscolla, (2002) created
a generic simulator within a network of clinics to reduce patient LOS in the system. The simulator
also facilitated the appropriation of resources and reallocated these resources based on the number
of patients and how quickly they go through the system. Bosire, Wang, Gandi, and Srihari, (2007)
modeled a computer tomography (CT) scan facility in a hospital to study how patient wait time
can be minimized and how staffs can be used efficiently to increase patient satisfaction.
2. Resource Allocation and Capacity Planning – Rico, Salari and Centeno (2007) built a model us-
ing ARENA simulation software and OptQuest heuristic optimization to increase system capacity,
improving nursing staffing allocation, and augmenting the utilization of equipment and space. They
suggested multiple ways in which the number of nurses needed for health services delivery during
a pandemic influenza outbreak may be combined. Cahill and Render (1999) created a model to
assess ICU bed availability. Their model was applicable for excess capacity rebalancing that would
otherwise lead increasingly and unjustifiably to wasting limited health resources such as bed space
and personnel in the Cincinnati VA Medical Center.
3. Cost Containment - In the Cahill-Render (1999) study cited previously, an additional application
of their model was to contain costs by drawing from outside resources in meeting patient needs.
Stahl, Roberts, and Gazelle (2003) also investigated the strategy for setting appropriate preceptor-
to-trainee ratio in the context of a teaching ambulatory care clinic. The key purpose of their study
was to achieve an optimal financial feasibility and to cut system operational cost.
4. Policymaking - Policymaking is fueling a resurgence of interest in modeling and simulation to
improve health services delivery performance. This is especially applicable in countries such as
Canada and the UK, where the system is based largely on a single payer (the government). More
xxxii

specifically, these systems employ a governmental body of agencies to manage the entire system
through a set of national performance measures. A specific example would be the star rating used for
measuring wait time targets for emergencies such as how long it takes a patient to be served upon
admission into the emergency department (ED) of a hospital. Gunal and Pidd (2005) provided an
example in their policy-oriented simulation study that aimed to improve the accuracy of the rating
system in the UK National Health Service (NHS) through uncovering performance irregularities.

Modeling Method

As well, a number of modeling methods and techniques have been identified for health care services
implementation. Among these, the most influential, useful and widely applied methods are simulation
techniques, which include the mainstream methods as well as simulation algorithms based on artificial
intelligence as discussed in Cooper, Brailsford, and Davies (2007) and Kuljis, Paul, and Stergioulas
(2007) `s study. Table 1 summarizes the current prevalent modeling technologies and simulation meth-
odologies, as well as their potential application domains in health services analysis.

1. Discrete-Event Simulation (DES) - Among the most widely used simulation techniques in health
care as evidenced by many previously cited studies, DES appears to be tailor-made for hospital
systems to study queuing behaviors of patients waiting for appointments, investigations and treat-
ments. In particular, DES allows the modeler to construct more complex, dynamic and interactive
systems. Nonetheless, as Cooper et al. (2007) pointed out in their choice of modeling technique for
evaluating health care interventions, it may take more time and money to develop DES models.
2. System Dynamics (SD) – To resolve systems bottlenecks and understand emerging systems states,
SD modeled patient flow behaviors in complex health services systems by capturing the feedback
loops and inventory control rules for patient arrival, discharge and follow-up visits. This is similar
to studying how water flows through a heating system. SD has gained popularity in recent years
and some researchers who first applied the DES had switched to SD to deal with the dynamic
changes in patient flow. For example, after using DES to redesign the phlebotomy and specimen
collection centers in Calgary Laboratory Services (Alberta, Canada), Rohleder, Bischak, and Baskin
(2007) have later on decided also to implement a SD model to handle the unexpected performance
discrepancies found due to the dynamic interactions within these service centers.
3. Markov process models - Markov models can describe changes in the state of patient health over
time such as the case of benign vs. malignant tumors. In this sense, Cooper et al. (2007) argued
that Markov models are suitable for chronic disease interventions. Accordingly, they employed a
Markov model to evaluate the effectiveness of statins, one of the cholesterol lowering drugs, over
time for at risk patients of coronary heart disease (CHD), based on a so-called Southampton CHD
model (Cooper, 2005).
4. Monte Carlo Simulation – Essentially, Monte Carlo uses repetitive sampling process to make esti-
mates about key performance variables of interests under uncertainty conditions just like throwing
a dice repeatedly to predict the chance of picking a winning stock in the Dow Jones Industrial
market. For example, Jacobson, Lindberg, Lindberg, Segerstad, Wallgren, Fellstrom, Hulten, and
Jensen-Waern (2001) used the Monte Carlo methodology to sample various potential vaccine price
distributions determined by the mix of care providers vis-à-vis parent-guardian acceptance so as
to assess the economic values of variously combined vaccines for pediatric immunization.
 xxxiii

Table 1. Current modeling and simulation techniques and their applications in health

Technique Potential Applications

Discrete-Event Simulation Process reengineering, ward layout design, patient pathway design, scheduling, queuing
(DES) management

Strategic and operations management, alteration management, resource and asset management,
System Dynamics (SD)
patient pathway management

Continuous Simulation Physical/ biological laboratory processes control

Monte Carlo Simulation Decision making under uncertainty conditions, risk analysis in the long run

Agent-Based Simulation Demand and supply management, health economics, risk management

Decision tree Acute interventions but not for disease recurrence modeling

Markov process Cohorts of patients between health states over time, chronic disease intervention

Operations research Patient arrival patterns, LOS management, waiting time management, cost management

Human factors and

Workload analysis, safety monitoring, productivity increasing, error reducing
ergonomic models

5. Continuous Simulation - Continuous simulation is primarily employed to accommodate continuous

systems variables and therefore has limitation in its applicability for health care studies. Specifi-
cally, it is limited predominantly to physical or biological laboratory processes control simulation
such as modeling the trajectory of a missile launch. In health care, it can, for example, be used
to assess the design of equipment to further enhance the production volumes and manufacturing
process efficiencies of pharmaceutical products (Kuljis et al., 2007) 67
6. Decision tree – Based a hierarchical tree-like structure, decision trees aid decision making through
an assessment of various probabilities in terms of possible consequences corresponding to different
options and alternatives. Cooper et al (2007) showed that a decision tree can facilitate comparative
decisions on mean life expectancy of CHD patients. Here, the current response states for CHD
patients are compared to more efficient ambulance delivery services and thrombolysis (“clot bust-
ing” drug) intake that may result in the death and/or the survival of the patients.

In summary, whenever a health system problem is encountered, a critical issue then is how to make
choose intelligently among various available modeling and simulation techniques. Schriber and Brunner
(2007) proposed to explore the nature and logical foundations in every method and software and thus
gain a detailed understanding of “how simulation works”. Cooper et al. (2007) stated in their study that
the choice of modeling technique depends on several aspects including modeling technique acceptance,
model appropriateness, dimensionality, and ease and speed of model development. Generally a decision
can be made based on the complexity and dynamics of the system to be modeled in terms of interaction
xxxiv

of systems elements, model size or resource constraints. The introduction of OSE philosophies also
helps making a better combination of choices. For instance, Young (2005) studied the philosophies
driving changes in health care services delivery such as Lean Thinking and the Theory of Constraints.
He proposed that a strategic agenda could be created out of a three-way fusion of health care delivery,
industrial process and simulation capacity, on which basis, an appropriate modeling and simulation
techniques can then be selected.

tHe Model and eVIdence drIVen Integrated analysIs (MedIa)

paradIgM & Its applIcatIon

The MEDIA paradigm, which has previously been introduced at the beginning of this review in Figure
1, offers a handle to unraveling today’s highly complex and environmentally uncertain health services
enterprises by focusing on the effective integration of operational models and evidence. Three key phases
underline the MEDIA paradigm: (a) Referential ontology modeling; (b) Hybrid probabilistic model
integration; and (c) Adaptive knowledge fusion for quality assurance.
Referential ontology modeling relates to the qualitative representation of model structure that is gen-
erally constrained by the health services systems complexity and uncertainty. Nodes, representative of
crucial system artifacts, and the interdependencies among nodes, that represent their relations, typically
encapsulate the model structure. Figure 4, for example, shows the nodes and the interdependency links
among these nodes for a partial risk assessment model of operational and information risk representation
in the context of a generic model structure.
As the purpose of referential ontology modeling is essentially to support the generation of hybrid
probabilistic models, probabilistic modeling requirements should first be examined in terms of the domain
of enterprise services and elements, environment risks, and their relations (dependency). The emergent
systems complexity is then represented by referential knowledge in formal ontology based on descrip-
tion logic domain and the uncertainty among systems relations is often addressed through probabilistic
representation for ontology modeling such as applying the Bayes approach. Accordingly, a critical step
is the conversion of ontology to probabilistic models.
The next phase, hybrid probabilistic model integration, entails bridging the qualitative ontology
representations with computational analysis to quantify the relations among referential nodes and to
achieve an integration of heterogeneous models. This quantification process comprises fundamentally
an attempt to assign probabilistic distribution to nodes and/or conditional probability to relations among
the nodes. It is important to note that boundary nodes are those nodes shared by multiple models and the
same quantification process is applied to these nodes to connect among the models. Special procedures
such as those algorithms dealing with virtual and soft evidence may have to be instantiated to adjust
the associated probabilities to satisfy all of the probabilistic constraints across each of the model to be
integrated (Kim, Valtorta, & Vomlel, 2004; Xiang, 2002). In this second phase, the core migration pro-
cedures can utilize semantic web techniques such as ByesOWL framework (Ding, Peng, & Pan, 2006)
to convert the ontology to probabilistic networks.
The final phase, knowledge fusion, aims at improving the accuracy and confidence of hypothetical
estimations about key systems variables based on discrete evidence gathered from multiple locations
throughout the systems being studied. Furthermore, adaptive knowledge fusion attempts to achieve a
high quality assurance of information collection and operations management decisions in the knowledge
 xxxv

Figure 4. A model structure represented by nodes and interdependencies among nodes

vaccination Markov Cancer Model (2)

backup power
surge
epidemic
HMM Power Model (1)
computer
staff patient OR

Bayesian Parameter Model (3)

Staff A Staff B
Staff C
OR
Patient arrival Patient exit
Computer network topology

staff
Probabilistic Communication Model (5)
(4)
OR Simulation Model

Decision Model (6)

node evidence Resource utility resource

allocation
dependency control information flow link

fusion process. Appropriate quality assurance strategies are used iteratively to guide the knowledge fu-
sion process and are typically aided by dynamic quality scores or indices calculated from information
entropy of evidence and effectiveness of different operations. Essentially, these strategies will attempt
to provide insights to key questions regarding the knowledge fusion process, including: (1) how good
the resulting diagnosis or prediction may be; (2) when to engage and/or stop the knowledge integration
procedures; (3) where and what information is to be collected; and (4) finally, what resource allocation
decisions could be best implemented (Li & Chandra, 2006; Li & Ji, 2005).

MedIa procedural Framework

Figure 5 overviews a MEDIA procedural framework that can be used to guide the implementation of
the different phases involved in the construction of relevant health IT models and methods for a generic
risk management system problem. In the context of the MEDIA paradigm, any kind of a management
task such as a risk assessment scenario for a specific system like a hospital in a chosen ontology domain,
specifically, health care, will therefore pass through a set of generic procedures.

• Phase 1a: Physical constructs and security components are modeled in ontology bases for the
chosen domain. These class templates have common attributes and relations sufficient to support
the modeling of most management tasks;
xxxvi

Figure 5. Flowchart of the model and evidence driven integrated analysis (MEDIA)

Domain Specific System and Task Specific

Existing
Physical Security Relations System Security Relation
Models
Constructs Components Information Tasks Instances

Model ontology

Ontology Instantiate system models Identify boundary interfaces

Referential
Dependency Models

Translate ontology and fit parameters Link models

modify Hybrid Probabilistic

Computation Models
System Specific

Evidence Adaptive knowledge integration

• Phase 1b: Aided by the domain templates above, two different types of inputs are needed for
referential dependency models to be instantiated for the specific system and the risk management
task: (i) for an existing model, boundary nodes and their dependency, and (ii) for a new model,
complete knowledge about the model structures of nodes and dependencies;
• Phase 2: Probabilistic computational models are then generated through ontology translation and
parameter fitting to capture the interdependency and their associated uncertainty. This is helped by
additional descriptions specific to the system and task, for example, hypothesis nodes of interests
and available evidence collection positions designated for risk management;
• Phase 3: When evidence can be found, probability distributions for nodes connected within a
network as well as quality scores are updated accordingly through probabilistic inferences. The
probabilistic inferences that are carried out here should be specific to the different types of models,
for example, Markovian process, Bayesian, or DES models.

If a model needs to be modified in terms of both its structure and parameters due to changes in the
underlying systems dynamics, or a new model needs to be introduced, the process goes back to Phase
1b to repeat or iterate on the instantiation process (Cooper & Herskovits, 1992; Heckerman, Geiger, &
Chickering, 1995).
The rest of this section will turn to discussing the application of MEDIA paradigm in a multi-tier
capacity planning system.
 xxxvii

MedIa application for a Multi-tier capacity planning system

In the past, mathematical models for the optimal timing prediction of individual disease treatments such
as cancers and for the best health services scheduling such as emergency and operation rooms schedul-
ing have been developed respectively. Yet it is important for these models to work together as may be
necessary for setting up new global optimization strategies.
In the real world, for example, the service appointment and operation room scheduling do not always
guarantee that the patient gets the screening test or the surgery at the best chosen timing as dictated by
the individual disease progression model. Suboptimal timing for treatment may therefore have to be
chosen and analyzed again for potentially undesired consequences to patients. As this might further de-
lay the treatments for patients and jeopardize their life quality while increasing medical cost, these two
types of models may have to be reconnected and integrated through their interdependency to optimize
the two processes of the overall system.
In addition, the challenging issue of uncertainty arises when we consider the dynamic changes such
as an epidemic or more disturbing natural and man-made emergencies. In a normal operating environ-
ment, the discussed risk in the abovementioned case may not be very serious. However, if epidemic
cases such as SARS or events such as the 9/11 terrorist attack occurring simultaneously, for example,
excessive demand on medical aids in responding to these events may easily deprive regular patients of
“optimized” treatments. The strategy here then is to factor in the uncertainty by integrating and incor-
porating additional appropriate models.
Indeed, this integration can even be expanded either horizontally or vertically into as many channels
of the health care system as needed such as the sub-units within the same health care enterprise, or the
various business processes and institutional structures of a regional health alliance. Naturally, instead of
building a new, but highly sophisticated model from scratch by applying the same algorithmic approach
or method, applicable existing models and/or the most suitable type of model for each of these processes
can now be reused or variously combined to satisfy the overall systems model optimization requirements.
Such an application shows the dominant complexity in terms of a large number of participating systems,
subsystems, and basic units, and more significantly, their interdependency.
Figure 4, which was also discussed earlier, illustrates a partial risk assessment model of operational
and information risk representation in the patient care process. This hybrid probabilistic modeling solu-
tion employs a variety of models. A hidden Markov process model (1) predicts the state of power supply
for potential power surge. A heterogeneous Markov process model (2) captures the progression of a
given type of cancer and estimates the best treatment timings. The predictions from these models enter
a Bayesian parameter model (3) that estimates the states of corresponding variables of staff, patient,
operation room, and computer network, with the help of other available evidence. Then a simulation
model (4) can run to collect statistics on current operating room capacity, using these parameters. Another
model in the form of a probabilistic network graph (5) accounts for computer network topology and user
configuration. It calculates the risk of information leaking and availability for the different collaboration
scenarios based on staffing and computer network situations. Actually this availability prediction can
be feedback to the operating room simulation model, if applicable. Moreover, a computational decision
model (6) will enable the computation on the utility based on performance information for different
allocation actions for given resource constraints. This decision model can also be distributed to more
nodes across the system, if and as needed.
xxxviii

Imagine how a comprehensive model of a medical center can be generated through the application of
the MEDIA framework. Various sources of evidence can be introduced into this model on a continuing
basis, in terms of errors, delays, and clinic starvation alarms from physical, software or human sensory
channels, signaling emerging or potential problems. The administrators or doctors, facing an uncertain
and complex operational environment, have to ask first, given the evidence, what the big picture really is
and where the weakest links may be situated. Then the best decision (where and how) to invest resources
(capital and manpower) to avoid further deterioration of the system state and to mitigate the problems
can be made intelligently. Moreover, this process will be carried out, not haphazardly and without
knowledge, but in an active and timely manner to handle existing incomplete and uncertain information.
Imagine also that this model can further be updated with real-time system status by monitoring fa-
cilities/sensors in an “online” fashion. The key decision makers and policymakers are not asked to just
(always) rely on the “average” profile about user/disease from collected historical data. For example,
patient arrival may fluctuate from day to day; different treatment results of a special disease may re-
quire different actions to be taken; and different caregivers as well as health administrators may also
be motivated to stick stubbornly to their personal information processing styles and biases in making
certain decisions. Therefore the latest evidence, collected on a continuing basis to update information
relating to individual “special” patient, at a prescheduled pace and/or for significant events, will all be
made available as feedback to the existing model structure. Over time, the model structure or parameters
may, of course, be no longer appropriate, relevant or accurate enough for current or future predictions
(so called “concept drift”). Then, model learning and knowledge fusion will again be executed at these
times as well.
Put together, the MEDIA paradigm is clearly a useful and practical framework that can be applied
intelligently across any type of a real-world scenario to enable the mixture and integration of various
health IT models and methods to facilitate the making of short-term key health services delivery and
management decisions as well as to aid longer-term health systems planning and policymaking.

conclusIon

Essentially, our review on existing bodies of knowledge about the state-of-the-art health IT models
and methods has led us to become increasingly aware of the need for and significance of having an
integrative platform for leveraging existing models and methodologies intelligently such as the MEDIA
paradigm. This paradigm aims to offer an appropriate framework for conjugating heterogeneous proba-
bilistic models including the Bayesian network (Pearl, 1998), Markov process models (Doob, 1953),
and Monte-Carlo simulation (Fishman, 2001), through virtual or soft evidence update of distributions
over boundary nodes. The rationale for achieving such an integration is that: (1) inherent probabilistic
representation and inference is able to deal with uncertainty; (2) many such probabilistic models have
built-in mechanisms, such as in dynamic Bayesian networks and Markov process models to deal with
the temporal dependency; and (3) many of these models support localization of information processing
and can be easily implemented into agent-based architectures, aided by the interdependency present in
enterprise constructs and model components.
More specifically, just as languages spoken by people, we have seen that ontology representation
is necessary for computer models to communicate and interact with each other. Ontology defines the
concepts and the relations of generic elements in a domain and provides a common “language” shared
 xxxix

by components and subsystems of MEDIA. It is more than just a classification or a dictionary because
dynamic behaviors of the systems elements are also captured using description logic and language such
as RDF and OWL (Baader, Calvanese, McGuinness, Nardi, & Patel-Schneider, 2003; Barwise, 1997). We
have also seen how the integration of health IT models and methods can benefit through the application
of such a shared “language” for accurate and practical representation of various constructs in information
theoretic terms to represent risk and uncertainty in complex systems environments, thereby permitting
the accurate and rapid assessments of emergency situations and allowing efficient and effective health
services to be delivered productively.
Currently, there is a dire shortage of major studies in the applications of interoperable, integrative
health IT models and methods. As well, inadequate knowledge has yet to be derived from other industrial
and system engineering sectors such as manufacturing enterprises into aiding comparative processes in
health services delivery and management. For instance, the surgery department is always at the core of
many hospitals. Managing operation rooms requires a variety of expensive resources including surgical
space and facility, equipment and material supply, nurse, technician and doctor, to provide timely ap-
pointments to patients. It is a challenging task in and by itself.
Yet, the people manning the surgery department has still to interact with all the inpatient-outpatient
departments and clinics, as well as to be linked to a huge number of medically interconnected compo-
nents that will affect their own behaviors and dynamics at any moment. Reports and requests are largely
the only primary information artifacts interfacing between the operation rooms and other care providers
in the same hospital, and mostly for discrete interactions updated at a fixed time interval. The surgeons
and surgical administrator will often not be familiar or have knowledge of the status of other systems
and organizational components, and without the aids provided to them from an integrated platform of
health IT models and methods, they cannot possibly make the best decisions, including the operation
room assignments.
Ironically, most hospital departments and clinics nowadays have management models, policies, and
supporting software packages in place, but all running individually, to “efficiently” schedule patient
appointments and make medical decisions in an isolated manner. With the MEDIA framework applica-
tion, decisions on such localized “best” operations can now be optimized for the entire system. Feasible
assignments satisfying all of the required health policies for the various connected health provider
institutions can then be assured.
Our case example may now be further generalized to many other equally complex scenarios that
commonly occur in health services delivery. For example, even for a specialty clinic such as the urology
or cardiology unit located within a hospital, it has to deal with many other elements including primary
care, inpatient, outpatient, emergency, surgery, lab testing, logistics, and many other areas in terms of the
different nodes and types of information flow, including patient, provider and material flows. Similarly, an
insurance company has to navigate through patients, medical providers, pharmaceutical companies, and
other vendors or suppliers in order to keep routine financial claim transactions in order. For all of these
health provider organizations, if we do not attempt to empower the different health services managers
with the necessary, interoperable health IT capabilities, there will be no assurance that a well-coordinated
effort in patient care delivery will be sustained.
The need for such a well-coordinated effort and enterprisewide collaboration across the spectrum
of health services delivery is even greater when we consider health alliances, which may be formed
at varying scales, levels and sizes. Frequently emerging out of the cooperation among multiple health
provider institutions of primary clinics, hospitals, research institutes, insurance management organiza-
xl

tions, and even government health agencies, health alliances are dependent on the separate units to work
together in order to achieve high quality patient care. Piece-by-piece examination of related management
models and systems does not and will not offer much help when local operation decisions have to be
made based upon disparate evidence. In such networked systems and sub-systems, a medical provider
or manager needs to know far more than his or her own world in order to make better, if not the “best,”
decisions for the patients and the health alliance organization as a whole. Often, the critical need is not
more models, or even more sophisticated yet isolated models – where abundant, if still not adequate,
models have already been developed, implemented, and studied – instead, the primary need here may
just be how to integrate existing models productively and intelligently so as to allow interoperable health
IT applications to be shared. Then, and if necessary, new models and methods can be added to enhance
those aspects of decisions that may still be lacking or have yet to be addressed.
A wonderful developing story for applying the MEDIA paradigm is the 2009 case of the “swine flu”
(H1N1) pandemic that is ongoing at this point in time (Esterl, 2009). It is just impossible to have one
model or system to capture everything relating or reported about this pandemic, from the private clinics
to the US governmental health services sector plan for controlling H1N1 to the different magnitude of
H1N1 developments spreading throughout the globe. In other words, it is just not enough to harbor the
narrow perspective of a tiny component within the entire global health service system, but having an
interconnected worldview of even just how the H1N1 is affecting the different nations will significantly
improve the patient survival chance and H1N1 prevention for a particular or entire country. Even so,
the demand to integrate models and evidence for seamless analysis and management of a pandemic like
H1N1 may be attributed largely to the following facts:

1. The complexity and uncertainty are greatly increasing in the system of systems context of health
services supply chains and global health information exchange networks to which any analysis
and management systems have to appropriately respond.
2. The rapidly changing medical and health technology landscape and innovative practices in medi-
cine, health management, industry and government policy, as well as the evolving field of health IT
models and methods require the constant incorporation of new systems perspective and integrative
model approaches.
3. Unfortunately, as noted previously, vast legacy HISs have typically been built piece by piece,
resulting in the preservation only of disparate health IT models without the capability to leverage
intelligently from evidence aggregated over time from diverse sources embedded in the overall
global health care system.

Development of novel technology and unconventional business model has inspired complex enterprise
structures to evolve as in the case of the US health services delivery system. These enterprises usually
form close partnerships in order to survive and grow under increasing competition, for which decision
making and policymaking relying on existing and past discrete models for isolated system segments is no
longer sufficient. Another obstacle many of these health enterprises encountered in across-the-organization
management is the complicated systems dynamics. This needs to be dealt with by systematically increas-
ing the connectivity of existing models, resulting in the constant understanding and monitoring of the
evolving systems states comprising rapid changes in participants and configurations along both temporal
and spatial dimensions. As characterizing these different systems states efficiently can simply become
overwhelming, health services quality assurance is undoubtedly a very demanding task. Interoperable,
 xli

integrative health IT models and methods via the MEDIA paradigm aim at reducing such complexity to
meet the demands of high quality health services. The MEDIA paradigm also addresses these systems
challenges through referential ontology modeling, hybrid probabilistic model integration, and adaptive
knowledge fusion that adopt a real system-wide and process oriented perspective. Thus, the introduction
of the MEDIA framework to emerging health IT models and methods research represents a key step to
unifying our knowledge for modern health services enterprise management.
The future of health IT models and methods is dependent on the development of even more powerful
integrative frameworks beyond the MEDIA paradigm. Such a paradigm will seek to provide a consistent
approach to reusing and integrating available knowledge, dealing with complexity and uncertainty dimen-
sions in multiple ways, and encouraging the sharing and exchanging of health information privately and
securely across organizational boundaries that do not currently exist. This systematic methodological
conjugation represents a departure in traditional integrated modeling and analysis from fitting all systems
into just “one type of model” to an emphasis on knowledge interpretation, interaction and fusion among
models. This makes it much easier to achieve balance between the local and global representation of
different models. In other words, the hybrid probabilistic model integration can actually make use of
any available models and/or choose the most suitable ones for each subsystem/task, rather than having
to start the modeling process from scratch each and every time one or more new systems problem(s) is
encountered.
In closure, the MEDIA methodology or an expansion of such a methodology can further be applied
to create novel solutions to new health systems challenges faced in a wide range of practical contexts
that vie to achieve the convergence of enterprise systems and processes. The trend toward preventive
health, alternative and integrative medicine, e-health and healthy lifestyle promotion, for example, are all
virgin grounds for the applications of integrative, interoperable health IT models and methods. Imagine
how a self-monitoring community health care system that can be equipped with a regional center that
uses interoperable and integrated health IT models and methods to aid in guiding individual residents
throughout the community to cope with undue stresses arising from risky driving behaviors on a daily
basis, become quickly alerted and fully prepared for emergencies by deploying the community health
resources efficiently and effectively on an as needed basis, and channeling any and all unused health
resources to evolve the community in multiple ways to adopt a healthy, low risk and more active lifestyle
environment within the community.

Joseph Tan
McMaster University, Canada

Xiangyang Li
University of Michigan - Dearborn, USA

Yung-wen Liu
University of Michigan - Dearborn, USA
xlii

reFerences

Agrawal, N., & Nahmias, S. (1997). Rationalization of the supplier base in the presence of yield uncer-
tainty. Production and Operations Management, 6, 291–308.
Anupindi, R., & Akella, R. (1993). Diversification under supply uncertainty. Management Science, 39,
944–963.
Baader, F., Calvanese, D., McGuinness, D. L., Nardi, D., & Patel-Schneider, P. F. (2003). The Description
Logic Handbook: Theory, Implementation, Applications. Cambridge, UK: Cambridge University Press.
Barbera, J., & Macintyre, A. (2002). Medical and health incident management system: A comprehensive
functional system description for mass casualty medical and health incident management, Retrieved
from http://www.gwu.edu/~icdrm.
Barwise, J. (1997). An Introduction to First-Order Logic. Handbook of Mathematical Logic (pp. 5-46).
Amsterdam.
Blackhurst, J., Craighead, C.W., Elkins, D., & Handfield, R. B. (2005). An empirically derived agenda
of critical research issues for managing supply-chain disruptions. International Journal of Production
Research, 43(19), 4067-4081.
Bosire J., Wang, S., Gandhi, T., & Srihari, K. (2007). Comparing simulation alternatives based on quality
expectations. In Proceedings of the Winter Simulation Conference (pp. 1579-1585).
Brailsford, S.C. (2007). Tutorial: advances and challenges in healthcare simulation modeling. In Pro-
ceedings of the 39th Conference on Winter Simulation (pp. 1436-1448).
Brailsford, S., Lattimer, V., Tarnaras, P., & Turnbull, J. (2004). Emergency and on-demand health care:
modeling a large complex system. The Journal of the Operational Research Society, 55(1), 34-42.
Brantes, F.D., Galvin, R.S., & Lee, T.H. (2003). Bridges to Excellence: building a business case for
quality care, Journal of Clinical Outcomes Management, 10, 439.
Cahill, W., & Render, M. (1999). Dynamic simulation modeling of ICU bed availability. Winter Simula-
tion Conference, 2, 1573-1576.
Cooper, G.F., & Herskovits, E. (1992). A Bayesian method for the induction of probabilistic networks
from data. Machine Learning, 9, 309-347.
Cooper, K (2005). A comparative study of modeling approaches for the evaluation of healthcare interven-
tion with reference to coronary heart disease. PhD Dissertation (unpublished), University of Southampton.
Cooper, K., Brailsford, S., & Davies, R. (2007). Choice of modeling technique for evaluating health
care interventions. Journal of the Operational Research Society, 58(2), 168-176.
Denton, B., Fowler, J., Schaefer, A., Batun, S., Erdogan, E., & Gul, S. (2009). Optimization of the design
and operation of surgery delivery systems. In Proceedings of the 2009 NSF engineering research and
innovation conference.
 xliii

Ding, Z., Peng, Y., & Pan, R. (2006). BayesOWL: Uncertainty modeling in semantic web ontologies.
Soft Computing in Ontologies and Semantic Web. Springer-Verlag.
Doob, J. L. (1953). Stochastic Processes. New York: John Wiley and Sons.
Eldabi, T., & Young, T. (2007). Towards a framework for healthcare simulation. In Proceedings of the
39th Conference on Winter Simulation. 1454-1460.
Esterl, M. (2009, June 11) Swine Flu pandemic is closer, says WHO. Wall Street Journal.
Fishman, S. (2001) Discrete-Event Simulation: Modeling, Programming, and Analysis. New York:
Springer-Verlag.
Ford, E.W., & Hughes, J.A. (2007). A collaborative product commerce approach to value-based health
plan purchasing. Supply Chain Management: An International Journal, 12(1), 32-41.
Graves, S.C., Kletter, D.B., & Hetzel, W.B. (1998). A dynamic model for requirements planning with
application to supply chain optimization. Operations Research, 46(3), 35-49.
Gunal, M., & Pidd, M. (2005). Simulation modeling for performance measurement in healthcare. In
Proceedings of the 37th Conference on Winter Simulation (pp. 2663-2668).
Hale, T., & Moberg, C.R. (2005). Improving supply chain disaster preparedness: A decision process for
secure site location. International Journal of Physical Distribution & Logistics Management, 35(3),
195-207.
Heckerman, D., Geiger, D., & Chickering, D. (1995). Learning Bayesian networks: The combination of
knowledge and statistical data. Machine Learning, 20(3), 197-243.
Hershey, J., Pierskalla, W., & Wandel, S. (1981). Nurse staffing management. In D. Boldy (Eds.), Op-
erational Research Applied to Health Services, (pp. 189-220). London: Croon Helm.
Hick, J. L., Hanfling, D., Burstein, J. L., DeAtley, C., Barbisch, D., Bogdan, G. M., & Cantrill, S. (2004).
Health care facility and community strategies for patient care surge capacity. Annals of Emergency
Medicine, 44(3), 253-261.
IEEE-USA, (2005) Interoperability for the National Health Information Network, Whitepaper by the
Medical Technology Policy Committee (MTPC) of The Institute of Electrical and Electronics Engineers-
United States of America (IEEE-USA), November.
IT Governance Institute (ITGI). (2007). COBIT4.1-Control Objectives for Information and Related
Technology. Retrieved January 20, 2010, from http://www.itgi.org/
Jacobson, M., Lindberg, J. E., Lindberg, R., Segerstad, C.H., Wallgren, P., Fellstrom, C., Hulten, C., &
Jensen-Waern, M. (2001). Intestinal cannulation: model for study of the midgut of the pig. Comp Med.,
51(2), 163–170.
Kiel, L. D., & Elliott, E. (Eds.). (1996) Chaos theory in the social sciences: Foundations and applica-
tions. Ann Arbor, MI: University of Michigan Press.
xliv

Kim, Y.G., Valtorta, M., & Vomlel, J. (2004). A prototype system for soft evidential update. Applied
Intelligence, 21(1), 81-97.
Kouvelis, P., & Milner, J.M. (2002). Supply chain capacity and outsourcing decisions: The dynamic
interplay of demand and supply uncertainty. IIE Transactions, 34, 717–728.
Kollberg, B., Dahlgaard J. J., & Brehmer, P. (2007). Measuring lean initiatives in health care services:
issues and findings. International Journal of Productivity and Performance Management, 56(1), 7-24.
Kuljis, J., Paul, J., & Stergioulas, L. (2007). Can health care benefit from modeling and simulation meth-
ods in the same way as business and manufacturing has? In Proceedings of the 2007 Winter Simulation
Conference (pp. 1449-1453).
Kumar, A., & Ozdamar L. (2004). Business process reengineering at the hospitals: A case study at Sin-
gapore hospital. In Proceedings of the 18th European Simulation Multiconference (CD version).
Kumar, A., Ozdamar L., & Zhang C. (2008). Supply chain redesign in the healthcare industry of Singa-
pore. Supply Chain Management: An international Journal, 13(2), 95-103.
Kumar, S., Swanson E., & Tran, T. (2009). RFID in the healthcare supply chain: usage and application.
International Journal of Health Care Quality Assurance, 22(1), 67-81.
Lambert, D.M., Cooper, M.C., & Pagh, J.D. (1998). Supply chain management: Implementation issues
and research opportunities. The International Journal of Logistics Management, 9(2), 1–19.
Laudon, K. and Laudon, J. (2007). Can information systems help prevent a public health crisis? In
Management Information Systems (10th ed., pp. 77-78).
Lee, H. L., & Whang, S. (2005). Higher supply chain security with lower cost: Lessons from total quality
management. International Journal Production Economics, 96, 289-300.
Li, X., & Chandra, C. (2006). Efficient knowledge integration to support complex supply network man-
agement. International Journal of Manufacturing Technology and Management, 10(1), 1-18.
Li, X., & Ji, Q. (2005). Active affective state detection and user assistance with dynamic Bayesian net-
works. IEEE Transactions on Systems, Man, and Cybernetics-Part A, 35(1), 93- 105.
Lipowicz, A. (2009, October 23). Government agencies to spend $15 billion on health IT by 2014. Federal
Computer Week, Retrieved January 20, 2010, from http://fcw.com/Articles/2009/10/23/Government-
agencies-to-spend-15-billion-on-health-IT-by-2014.aspx?p=1
National Coalition on Health Care. (2009). Facts on Health Insurance Costs, Retrieved from http://www.
nchc.org/facts/cost.shtml
O’Leary, D. (2004). Models of infection: person to person, Computing in Science & Engineering, 6(1),
68-70.
Pai, R.R., Kallepalli, V.R., Caudill, R.J., & Zhou, M. (2003). Methods toward supply chain risk analysis.
In Proceedings of the IEEE International Conference on Systems, Man and Cybernetics, 5, 4560-4565.
 xlv

Parisi, L., & Bailey, K. (2009, April). Too Great a Burden: Americans Face Rising Health Care Costs.
Retrieved from http://hdl.handle.net/10207/18507.
Parlar, M., & Perry, D. (1996). Inventory models of future supply uncertainty with single and multiple
suppliers. Naval Research Logistics, 43, 191-210.
Pearl, J. (1998). Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference. San
Mateo, CA: Morgan Kaufmann Publishers.
Pierskalla, W.P., & Brailer, D.J. (1994). Applications of operations research in health care. Handbooks
in OR & MS, 6, 469-505.
Raghupathi, W., & Tan, J. (2002). Strategic IT Applications in Health Care. Communications of the
ACM, 45(12), 56-61.
Raguphati, W., & Tan, J. (2008). Information Systems and Healthcare XXX: Charting a Strategic Path
for Health Information Technology. The Communications of the Association for Information Systems,
23(1), 500-525.
Ramis, F., Palma, J., Estrada, V., & Coscolla, G. (2002). A simulator to improve patient’s service in a net-
work of clinic laboratories. In Proceedings of the Winter Simulation Conference (Vol. 2, pp. 1444-1447).
RAND Corporation (2004). Evaluating the Security of the Global Containerized Supply Chain, Techni-
cal Report, Santa Monica, CA Willis, H.H., & Ortiz, D.S.
Rico, F., Salari, E., & Centeno, G. (2007). Emergency departments nurse allocation to face a pandemic
influenza outbreak. In Proceedings of the Winter Simulation Conference (pp. 1292-1298).
Roberts, S. D. (2007). Simulation of medical decisions, IERC 2007, Nashville, TN. Retrieved January
20, 2010, from http://74.125.95.132/search?q=cache:GYvFAgiR6rIJ:ccehub.org/resources/121/download/
steve-roberts-simulationofmedicaldecisions-crc.ppt+Simulation+of+medical+decisions,+IERC+2007&
cd=1&hl=en&ct=clnk&client=safari
Rohleder, R., Bischak, P. & Baskin, B. (2007). Modeling patient service centers with simulation and
system dynamics, Health Care Management Science, 10(1), 1-12.
Schnase, J. L., & Cunnius, E. L. (Eds.). (1995). Proceedings from CSCL ‘95: The First International
Conference on Computer Support for Collaborative Learning. Mahwah, NJ: Erlbaum.
Schriber T. J. & Brunner, D. T. (2007). Inside discrete-event simulation software: how it works and why
it matters. In Proceedings of the Winter Simulation Conference (pp. 113-123).
Shehad, M., Bertino, E., & Ghafoor, A. (2005). Secure collaboration in mediator-free environment. In
Proceedings from CCS’05: ACM Conference on Computer and Communications Security, Alexandria, VA.
Shubik, M. (2002). Game theory and operations research: Some musings 50 years later. Operations
Research, 50, 192-196.
Simchi-Levi, D., Kaminsky, P., & Simchi-Levi, E. (2003). Designing and Managing the Supply Chain:
Concepts, Strategies, and Case Studies (2nd ed.). McGraw-Hill Irwin.
xlvi

Stahl, J., Roberts, M., & Gazelle, S. (2003). Optimizing management and financial performance of the
teaching ambulatory care clinic. Journal Gen Intern Med, 18(4), 266-274.
Swinehart, K. D., & Smith, A. E. (2005). Internal supply chain performance measurement: a health care
continuous improvement implementation International Journal of Health Care, 18(7), 533-542.
Tan, B., Gubaras, A., & Phojanamongkolkij, N. (2002). Simulation study of Dreyer urgent care facility.
In Proceedings of the Winter Simulation Conference (Vol. 2, pp. 1922-1927.
Tan, J., & Payton, F. C. (2010). Adaptive health management information systems: concepts, cases, &
practical applications (3rd ed.). Sudbury, MA: Jones & Bartlett Publishers, Inc.
Tan, J., Wen, J.H., & Awad, N. (2005, May). Health care and services delivery systems as complex
adaptive systems. Communications of the ACM, 48(5), 36-44
Thomas, M.U. (2002). Supply chain reliability for contingency operations In Annual Proceedings of
Annual Reliability and Maintainability Symposium (pp. 61-67).
U.S. Bureau of Economic Analysis. (2009). Manufacturing Industries Led Slowdown in 2008. Retrieved
from http://www.bea.gov/newsreleases/industry/gdpindustry/2009/gdpind08.htm.
U.S. House of Representatives, (2007). The XDR tuberculosis incident: a poorly coordinated federal
response to an incident with homeland security implications. Hearing of U.S. House of Representatives
Homeland Security Committee, June 6, 2007.
Woolhandler, S., Campbell, T., & Himmelstein, D. U. (2003). Costs of health care administration in the
United States and Canada. The New England Journal of Medicine, 349(8), 768-775.
Xiang, Y. (2002). Probabilistic Reasoning in Multiagent Systems: A Graphical Models Approach. Cam-
bridge University Press.
Young, T. (2005). An agenda for healthcare and information simulation. Health Care Management Sci-
ence, 8, 189-196.
 1

Chapter 1
Evaluating Health
Information Services:
A Patient Perspective Analysis1
Umit Topacan
Bogazici University, Turkey

Nuri Basoglu
Bogazici University, Turkey

Tugrul U. Daim
Portland State University, USA

aBstract
The objective of the chapter is to explore the factors that affect users’ preferences in the health service
selection process. In the study, 4 hypothetical health services were designed by randomly selecting levels
of 16 attributes and these services was evaluated by the potential users. Analytical Hierarchy Process
(AHP), one of the decision making methods, was used to assess and select the best alternative.

IntroductIon range including consultation (Berghout, Eminovic,

de Keizer, & Birnie, 2007), education and train-
Healthcare service providers benefit from different ing (Chen, Yang, & Tang, 2008), and home care
technologies so as to reduce cost and improve qual- (Biermann, Dietrich, Rihl, & Standl, 2002).
ity of the medical procedures (Gagnona, Godinb, Selecting the best telemedicine service among
Gagnéb, Fortina, Lamothec, Reinharza, & Clout- given ones is a complex task. The process needs
ierd, 2003). In particular, telemedicine resides on considerations of trade-offs between cost and
the center of these technologies. The American benefits of the service. Analytic Hierarchy Process
Telemedicine Association defines telemedicine as (AHP) (Saaty, 1977; Saaty, 1996) is an outstanding
“the use of medical information exchanged from method that can be used in multifactor decision-
one site to another via electronic communications making environments. It presents a structured
to improve patients’ health status” (ATA, 2009). approach to determine individual weights of
Telemedicine applications were used in a broad multiple attributes of a product or service so that

DOI: 10.4018/978-1-61692-002-9.ch001

Copyright © 2011, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 Evaluating Health Information Services

they can be compared in a simple way. Then, it There are many applications of AHP in the
simplify decision-making in the selection process. medical field. In one of the researches, AHP is
Recent technological developments enable used to develop an human resource planning
advancements in delivery of medical services, model for hospital laboratory personnel (Kwak,
appropriate healthcare at a reasonable cost, and McCarthy, & Parker, 1997). Turri applied AHP
access to quality healthcare in underserved areas approach to select a magnetic resonance imaging
in the medical sector. Telemedicine is one of vendor by using the criteria like price, service,
these developments that “enable remote medical and technology (Turri, 1988). Another applica-
procedures and examinations between patients tion of AHP was designed by Kahen and Sayers
and medical providers via telecommunication for selection of medical expert systems (Kahen
technologies like the Internet, or telephone” (Al- & Sayers, 1997). In the study, assessment of a
Qirim, 2007). Moreover, many previous researches health service prototype have done in the design
show that compared to traditional medical care, phase of the service development so as to make
telemedicine services present many benefits to the clear the factors affects users attitude toward
patients and physicians (van den Brink, Moorman, health services.
de Boer, Pruyn, Verwoerd, & van Bemmel, 2005; Usability tests of the health services (Kaufman,
Chae, Lee, Ho, Kim, Jun, & Won, 2001). Patel, Hilliman, Morin, Pevzner, Weinstock,
Diffusion of intelligent monitoring systems Goland, Shea, & Starrena, 2003), affects of
in the medical industry has gather speed with the ergonomics on the medical device development
help of recent developments in the information and (Martin, J.L., Norris, Murphy, & Crowe, 2008)
communication technologies. “Smart homes’ for and self-measurement satisfaction of hypertension
telecare by means of movement detector, oxym- patients (Bobrie, Postel-Vinay, Delonca, & Corvol,
eter, tansiometer and various other devices (Rialle, 2007) are just some of them. These studies have
Lamy, Noury, & Bajolle, 2003), a ringsensor that applied different techniques in the analyses phase
monitors patient’s blood oxygen saturation (Yang of the research. However, it is difficult to find a
& Rhee, 2000) and a web based electrocardiogram study that uses AHP method.
monitoring application facilitating collect, analyze Topacan et al (2008) identified 37 different
and storage of patient data (Magrabi, Lovell, & criteria for health information service (HIS) adop-
Celle, 1999) were designed by researchers to tion including cost, time factor, content, language,
follow-up of patients’ health status at home. security, customizability, output quality, menu
Evaluating health technologies is a compli- items, input type, sound quality and availability
cated procedure because people face with some of face-to-face communication. The studies of
difficulties while evaluating trade-offs between Karahanna et al (1999), Simon et al (2007), Chang
alternatives. Analytic Hierarchy Process (AHP) is et al (2007) identified a number of criteria with
a potential decision making method to deal with respect to triability, medical provider and vendor
complex decisions. The aim of AHP is to qualify support that are applicable for selecting health
relative priorities for a given set of alternatives on information service.
a ratio scale. In the literature, many applications Tung et al (2008) define financial cost as “the
of AHP have published in different fields includ- extent to which a person believes that using the
ing planning, resource allocations, and selecting electronic logistics information system will cost
a best alternative, etc. (Magrabi, Lovell, & Celle, money”. It was found that financial cost has an
1999). AHP method also widely used in vendor influence on adoption of electronic logistic in-
selection problems (Nydick & Hill, 1992; Tam & formation system (Tung, Chang, & Chou, 2008).
Tummala, 2001). Chang et al also examined security protection and

2
Evaluating Health Information Services

vendor support while studying electronic signature • Security means whether the service pro-
adoption. They found that vendor support affects vides secure communication between pa-
adoption, and there is no relation between security tients and doctors, or not.
and e-signature adoption (Chang, Hwang, Hung, • Triability is the degree to which patient
Lin, & Yen, 2007). may be experimented with the service on a
limited time or functionality.
• Sound Type is the voice style of the service.
eValuatIon process It can be one of computerized, male or fe-
male voices.
evaluation Model • Input Type refers the data entry procedure
of the patient.
Based on the previous studies, 21 different health • Availability of Face-to-Face Communication
service selection criteria were selected for the means whether the service enable the pa-
research. We conducted a mini survey involving tient to visit the doctor in the hospital or
randomly selected 3 male and 3 female potential not.
users. The purpose of this mini survey is to assess • Technical Support is how to assist the user
and identify critical factors and reduce the number in order to solve the specific problems with
of criteria. In this phase, potential users ranked the service.
these criteria based on importance level of them. • User Training means the demonstration
These participants also made suggestions about the the features and functionality of the service
levels of the attributes. Finally, 16 attributes were to the potential users.
selected and used in the AHP model. Attributes • Medical Provider is the hospital that sup-
used in the study is explained below; ports the medical related operations.
• Customizability is the ability of the service
• Implementation Cost of the service is the to be changed by the user preferences.
purchase price of it. This price is paid only • Clarity of Menu Items is the degree to
once while buying the service. which the user may understand the func-
• Operating Cost is the price that was paid tion of the menu item.
by the user in every month to be able to use • Output Quality refers how well the system
the service. performs the jobs (Venkatesh & David,
• Usage Time is the amount of time required 2000).
to use the service. 10 minutes usage time
means that the user spends 10 minutes in These attributes can be grouped into three
a day while entering meal information and major categories of cost, service characteristics
other data. and service provider characteristics. The cost
• Response Time is the doctor’s response factors include implementation and operating
time to the patients’ requests. As an ex- cost. The service characteristics consist of mobile
ample, in the alternative 1, the service pro- device, triability, content, availability of face-to-
vider guaranteed that doctor could respond face communication and time. Similarly, service
the patients’ questions in 45 minutes. provider characteristics include technical provider
• Language is the set of visual and auditory and medical provider. Moreover, some of these
signs of communication. sub criteria include one more level called sub
criteria 2. Mobile device consists of input type,
customizability, clarity of menu items and sound

3
 Evaluating Health Information Services

Figure 1. AHP Model

type. Content attribute, on the other hand, includes participants profile

output quality, language and security. Time factor
contains usage time and response time. Technical 6 male and 8 female potential users from different
provider consists of technical support and user age groups were selected to conduct the research.
training. The hierarchy can be visualized as a Age of the participants was in a range between
diagram shown in Figure 1. The diagram contains 23 and 60. Average age of the male and female
an overall goal and criteria broken down into two group was 37. Table 1 contains details about the
levels of sub-criteria. profile of participants.
AHP model was applied on 4 different hypo-
thetical home based telemonitoring services. Each
of the services has 16 different attributes. Levels FIndIngs
of these attributes were proposed by the potential
users in the mini research phase of the study. Each of the participants have compared the attri-
Based on these levels, 4 different alternative butes and their levels two by two in order to incor-
services were designed by randomly selecting porate judgments about them. They responded 69
among levels. All of the alternatives and their different pair wise comparison questions. Sample
levels are shown in Appendix A. questions shown in Table 2 and Table 3. In these
tables, scale number “9” refers that the attribute
has extreme importance with respect to counter
attribute. Intensity of importance decreases while

4
Evaluating Health Information Services

Table 1. Participants Profile

Age Groups Average Age

20-30 30-40 40-50 50+ Total
Male 2 2 1 1 6 36
Female 2 2 2 2 8 38
Total 4 4 3 3 14 37

Table 2. Sample Questionnaire Item for Attributes

Implementation Cost 9 8 7 6 5 4 3 2 1 2 3 4 5 6 7 8 9 Operating Cost

Table 3. Sample Questionnaire Item for Attribute Levels

Implementation Cost
200 YTL 9 8 7 6 5 4 3 2 1 2 3 4 5 6 7 8 9 500 YTL
200 YTL 9 8 7 6 5 4 3 2 1 2 3 4 5 6 7 8 9 800 YTL
500 YTL 9 8 7 6 5 4 3 2 1 2 3 4 5 6 7 8 9 800 YTL

Table 4. Sample Pair-wise Comparison Matrix for Cost Attribute

Implementation Cost Operating Cost Weight

Implementation Cost 1 1/4 0.20
Operating Cost 4 1 0.80

moving down to number “2” in the scale. More- 3. Weight of attributes (Table 4) and priority
over, scale number “1” refers that both of the vectors of alternatives (Table 5) was calcu-
attributes has equal importance for the participant. lated for each of the participants separately.
The following steps illustrate the AHP method 4. A table that shows the preferences of a single
applied in this research. participant was produced for each of partici-
pants separately (Table 6). As an example,
1. Participants responded questionnaire form. Table 6 shows that operating cost is the most
In the first part of the questionnaire (a sample important factor for the selected participant.
pair-wise comparison question shown in 5. In order to calculate total affects of 14 partici-
Table 2), they compared attributes against pants, average of the weight and priority vector
each other to calculate weight of them. values calculated in a separate table. Tables 7,
2. In the second part of the questionnaire 8 and 9 shows the overall results that is the
(sample pair-wise comparison questions average values of 14 participants.
shown in Table 3), the participants compared
level of attributes.

5
 Evaluating Health Information Services

Table 5. Sample Pair-wise Comparison Matrix for Implementation Cost Levels

A1 (200 YTL) A2 (500 YTL) A3 (800 YTL) A4 (500 YTL) Priority Vector
A1 (200 YTL) 1 4 8 4 0.57
A2 (500 YTL) 1/4 1 6 1 0.19
A3 (800 YTL) 1/8 1/6 1 1/6 0.04
A4 (500 YTL) 1/4 1 6 1 0.19

Table 6. Selection Matrix of a Participant

Attribute Weight A1 A2 A3 A4
Implementation Cost 0.040 0.573 0.191 0.045 0.191
Operating Cost 0.161 0.050 0.138 0.138 0.673
Usage Time 0.020 0.654 0.062 0.142 0.142
Response Time 0.156 0.046 0.121 0.713 0.121
Language 0.014 0.658 0.106 0.118 0.118
Security 0.099 0.250 0.250 0.250 0.250
Triability 0.033 0.188 0.154 0.188 0.470
Sound Type 0.007 0.250 0.250 0.250 0.250
Input Type 0.087 0.538 0.174 0.115 0.174
Availability of Face-to-Face Comm 0.130 0.375 0.125 0.125 0.375
Technical Support 0.013 0.713 0.121 0.121 0.046
User Training 0.003 0.228 0.291 0.384 0.097
Hospital 0.076 0.429 0.071 0.071 0.429
Customizability 0.043 0.313 0.313 0.063 0.313
Clarity of Menu Items 0.014 0.232 0.117 0.418 0.234
Output Quality 0.104 0.083 0.417 0.083 0.417
Sum 1.000 0.246 0.185 0.226 0.343

In Table 7, the local weights represent the of the service is more important than purchasing
relative weight of the nodes within a group of cost. Also, content and availability of face-to-
siblings regarding their parent. The global weights face communication are two of the significant
are calculated by multiplying the local weights service characteristics. Medical provider affects
of the siblings by their parents’ local weights. It decision of potential user more than technical
can be seen from the table that local weights of provider. Clarity of menu items, security and
each group add up to 1 and the global weights of response time are other important factors. And,
all the 16 attributes add up 1. as a characteristic of the technical provider, users
According to final calculations, Table 7 shows attach slightly more importance to user training
that potential users pay more attention on the than technical support. Compared to usage time,
characteristics of the service like face-to-face short response time has more positive effects on
communication, content, and time in the health patients selection decision of HIS.
service selection process. Moreover, operating cost

6
Evaluating Health Information Services

Table 7. Overall Local and Global Weights of Attributes

Criteria Local Weight Sub Criteria 1 Local Weight Sub Criteria 2 Local Weight Global Weight
Service Chars 0.39 Face to Face 0.27 0.105
Comm.
Content 0.26 Security 0.48 0.049
Output Quality 0.35 0.035
Language 0.17 0.017
Time 0.22 Usage Time 0.23 0.020
Response Time 0.77 0.066
Triability 0.15 0.058
Mobile Device 0.10 Clarity of Menu 0.40 0.016
Item
Customizability 0.28 0.011
Input Type 0.25 0.010
Sound Type 0.07 0.003
Ser. Prov. Chars 0.37 Medical Provider 0.78 0.289
Technical Provider 0.22 Technical Support 0.57 0.046
User Training 0.43 0.035
Cost 0.24 Operating Costs 0.65 0.156
Implementation Costs 0.35 0.084

Table 8 presents weight of the 16 attributes

Table 8. Overall Global Weights of Attributes in the decreasing order. According to Table 8,
medical provider is the most important factor for
Attribute Weight
health service selection. Secondly, people give
Medical Provider 0.289 importance on the operating cost of the medical
Operating Cost 0.156 service. Thirdly, they choose services that make
Face-to-Face Communication 0.105 available face-to-face communication with the
Implementation Cost 0.084 physician. On the other hand, input type, custom-
Response Time 0.066 izability and sound type was found three of least
Trialability 0.058 important attributes according to overall values.
Security 0.049 Table 9 presents overall priority vector values
Technical Support 0.046 of alternatives. According to the table, people
Output Quality 0.035 significantly choose to pay least amount of money
User Training 0.035 and not spare much usage time for the health
Usage Time 0.020 information services. Moreover, alternatives that
Language 0.017 provide secure data communication and storage
Clarity of Menu Items 0.016 environment, offer one week unlimited trial period,
Customizability 0.011 and include seven days twenty-four hours technical
Input Type 0.010 support were highly preferred by the participants.
Sound Type 0.003
People prefer consulting about their health status
Sum 1.000
to private hospitals instead of public ones and

7
 Evaluating Health Information Services

Table 9. Overall Priority Vectors of Alternatives

impact on potential users’ health service selec-
Attribute A1 A2 A3 A4 tion decisions. Moreover, it provides patients to
Implementation Cost 0,60 0,17 0,06 0,17
face-to-face communication with the physician.
Operating Cost 0,05 0,17 0,17 0,61
Usage Time 0,62 0,06 0,16 0,16
conclusIon
Response Time 0,09 0,16 0,58 0,16
Language 0,31 0,10 0,30 0,30
As explained in section 1, electronic health infor-
Security 0,41 0,09 0,41 0,09
mation service selection is an important and dif-
Triability 0,07 0,25 0,07 0,61
ficult problem to a medical company and patients.
Sound Type 0,19 0,25 0,37 0,19
We first identified 16 criteria and then formulated
Input Type 0,29 0,12 0,47 0,12
an AHP-based model to select health information
Availability of Face-to-Face
Comm 0,41 0,09 0,09 0,41
service as shown in Appendix B.
Technical Support 0,66 0,15 0,15 0,05
As for HIS selection, all potential users agreed
that a service characteristic was the most important
User Training 0,21 0,46 0,28 0,05
factor. Service providers and cost followed as the
Hospital 0,36 0,14 0,14 0,36
second and third most important consideration.
Customizability 0,36 0,22 0,06 0,36
The availability of face-to-face communication of
Clarity of Menu Items 0,23 0,16 0,23 0,38
the HIS was ranked first among service factors,
Output Quality 0,12 0,38 0,12 0,38
followed by content, time, triability and mobile
device. Moreover, designers should pay more at-
tention on medical provider and cost of the service
receiving guidance conducted by professionals while developing a telemedicine service. Instead
instead of online education and manual. of public hospital, private hospital should be se-
Table 10 shows the total decision weights of lected as a medical provider, because people take
the alternatives. Total decision weight of a specific into confidence in private hospitals. Also, patients
alternative was calculated by multiplying weight prefer the services that enable doctor visits.
of the attribute with overall priority vector value The most important attributes, found in the
of the attribute for that alternative and summed up research, enable the service designers and pro-
these values calculated for 16 attributes. According viders to know the characteristics of the most
to the table, people have a tendency on selecting preferred e-health service. Furthermore, govern-
alternative 4 which is followed by alternative 1. ments, especially in developing countries, face
Alternatives 2 and 3 are the least preferred services with some difficulties while providing quality
by the potential users. healthcare in underserved and rural areas. Even if
they achieve to serve, they suffer from high cost
Total Decision Weight = Σ Global Weight of the
of the health service. With the rapid development
Attribute* Priority Vector of the Attribute
of information and communication technologies,
they benefit from internet to serve the healthcare
Alternative 4 has the finest attribute values for
service. But, at this time user adoption problems
the most important three attributes listed in Table
becomes one of the obstacles in spreading of the
8. Medical provider of the alternative is a private
electronic health services. The proposed AHP
hospital. Although its implementation cost is
framework will help the authorities to overcome
higher than alternative 1 and equals to alternative
adoption problems so as to reduce health care
2, it has cheapest operating cost that has valuable

8
Evaluating Health Information Services

Table 10. Final Decision Matrix

A1 A2 A3 A4
Total Decision Weights 0,30 0,17 0,18 0,35

costs, to improve health care and access to health Chae, Y. M., Lee, J. H., Ho, S. H., Kim, H. J.,
care, to improve patience obedience. Jun, K. H., & Won, J. U. (2001). Patient satisfac-
For a future study, a comprehensive qualitative tion with telemedicine in home health services
research with large sample size can be conducted for the elderly. International Journal of Medical
to test validity of the results. Moreover, attribute Informatics, 61, 167–173. doi:10.1016/S1386-
values of proposed alternatives can be changed 5056(01)00139-3
and examined the affect of them on the health
Chang, I. C., Hwang, H. G., Hung, M. C., Lin,
service selection process.
M. H., & Yen, D. C. (2007). Factors affecting
the adoption of electronic signature: Executives’
perspective of hospital information department.
reFerences
Decision Support Systems, 44(1), 350–359.
Al-Qirim, N. (2007). Championing telemedicine doi:10.1016/j.dss.2007.04.006
adoption and utilization healthcare organiza- Chen, I. J., Yang, K. F., & Tang, F. I. (2008). Ap-
tions in New Zealand. International Journal of plying the technology acceptance model to explore
Medical Informatics, 76, 42–54. doi:10.1016/j. public health nurses’intentions towards web-based
ijmedinf.2006.02.001 learning: A cross-sectional questionnaire survey.
ATA. (2009). Americal Telemedicine Association. International Journal of Nursing Studies, 45,
Retrieved from http://www.atmeda.org 869–878. doi:10.1016/j.ijnurstu.2006.11.011

Berghout, B. M., Eminovic, N., de Keizer, N. F., & Gagnona, M. P., Godinb, G., Gagnéb, C., Fortina,
Birnie, E. (2007). Evaluation of general practitio- J. P., Lamothec, L., Reinharza, D., & Cloutierd,
ner’s time investment during a store-and-forward A. (2003). An adaptation of the theory of inter-
teledermatology consultation. International personal behaviour to the study of telemedicine
Journal of Medical Informatics, 76, 384–391. adoption by physicians. International Journal of
doi:10.1016/j.ijmedinf.2007.04.004 Medical Informatics, 71, 103–115. doi:10.1016/
S1386-5056(03)00094-7
Biermann, E., Dietrich, W., Rihl, J., & Standl, E.
(2002). Are there time and cost savings by using Kahen, G., & Sayers, B. M. (1997). Health-care
telemanagement for patients on intensified insulin technology transfer: Expert and information
therapy: A randomised, controlled trial. Com- systems for developing countries. Methods of
puter Methods and Programs in Biomedicine, 69, Information in Medicine, 36, 69–78.
137–146. doi:10.1016/S0169-2607(02)00037-8 Karahanna, E., Straub, D. W., & Chervany, N.
Bobrie, G., Postel-Vinay, N., Delonca, J., & L. (1999). Information technology adoption
Corvol, P. (2007). Self-Measurement and Self- across time: a cross-sectional comparison of pre-
Titration in Hypertension. American Journal of adoption and post-adoption beliefs. Management
Hypertension, 20, 1314–1320. doi:10.1016/j. Information Systems Quarterly, 23, 183–213.
amjhyper.2007.08.011 doi:10.2307/249751

9
 Evaluating Health Information Services

Kaufman, D. R., Patel, V. L., Hilliman, C., Morin, Simon, J. S., Rundall, T. G., & Shortell, S. M.
P. C., Pevzner, J., & Weinstock, R. S. (2003). (2007). Adoption of Order Entry with Decision
Usability in the real world: assessing medical Support for Chronic Care by Physician Organiza-
information technologies in patients’ homes. tions. Journal of the American Medical Informat-
Journal of Biomedical Informatics, 36, 45–60. ics Association, 14, 432–439. doi:10.1197/jamia.
doi:10.1016/S1532-0464(03)00056-X M2271
Kwak, N. K., McCarthy, K. J., & Parker, G. E. Tam, M. C. Y., & Tummala, V. M. R. (2001).
(1997). A human resource planning model for hos- An Application of the AHP in vendor selection
pital/medical technologists: An analytic hierarchy of a telecommunications system. Omega, 29(2),
process approach. Journal of Medical Systems, 171–182. doi:10.1016/S0305-0483(00)00039-6
21(3), 173–187. doi:10.1023/A:1022812322966
Topacan, U., Basoglu, A. N., & Daim, T. U. “Ex-
Magrabi, F., Lovell, N. H., & Celle, B. G. (1999). ploring the Success Factors of Health Information
A web-based approach for electrocardiogram Service Adoption. In Proceedings of Portland
monitoring in the home. International Journal of International Conference for Management of
Medical Informatics, 54, 1145–1153. doi:10.1016/ Engineering and Technology’08 (pp. 2453-2461).
S1386-5056(98)00177-4
Tung, F. C., Chang, S. C., & Chou, C. M. (2008).
Martin, J. L., Norris, B. J., Murphy, E., & Crowe, An extension of trust and TAM model with IDT
J. A. (2008). Medical device development: The in the adoption of the electronic logistics infor-
challenge for ergonomics. Applied Ergonomics, mation system in HIS in the medical industry.
39, 271–283. doi:10.1016/j.apergo.2007.10.002 International Journal of Medical Informatics,
77, 324–335. doi:10.1016/j.ijmedinf.2007.06.006
Nydick, R. L., & Hill, R. P. (1992). Using the
analytic hierarchy process to structure the supplier Turri, J. J. (1988). Program eases decision making.
selection procedure. Journal of Purchasing and Health Progress (Saint Louis, Mo.), 69, 40–44.
Materials Management, 25(2), 31–36.
Vaidya, O. S., & Kumar, S. (2006). Analytic
Rialle, V., Lamy, J. B., Noury, N., & Bajolle, hierarchy process: An overview of applications.
L. (2003). Telemonitoring of patients at home: European Journal of Operational Research, 169,
a software agent approach. Computer Methods 1–29. doi:10.1016/j.ejor.2004.04.028
and Programs in Biomedicine, 72, 257–268.
van den Brink, J. L., Moorman, P. W., de Boer,
doi:10.1016/S0169-2607(02)00161-X
M. F., Pruyn, J. F. A., Verwoerd, C. D. A., & van
Saaty, T. L. (1977). A scaling method for priorities Bemmel, J. H. (2005). Involving the patient:
in hierarchical structures. Journal of Mathemati- A prospective study on use, appreciation and
cal Psychology, 15, 234–281. doi:10.1016/0022- effectiveness of an information system in head
2496(77)90033-5 and neck cancer care. International Journal of
Medical Informatics, 74, 839–849. doi:10.1016/j.
Saaty, T. L. (1996). The analytic hierarchy process.
ijmedinf.2005.03.021
Pittsburgh: RWS Publications.

10
Evaluating Health Information Services

Venkatesh, V., & David, F. D. (2000). A theo- endnote

retical extension of the technology acceptance
model- Four longitudinal field studies. Man-
1
This chapter is based on a paper presented
agement Science, 46, 186–204. doi:10.1287/ at Portland International Conference on
mnsc.46.2.186.11926 Management of Engineering and Technology
2009 in Portland Oregon USA by the same
Yang, B. H., & Rhee, S. (2000). Development authors
of the ring sensor for healthcare automation.
Robotics and Autonomous Systems, 30, 273–281.
doi:10.1016/S0921-8890(99)00092-5

11
 Evaluating Health Information Services

appendIx a

Table 11. Proposed Health Service Alternatives

Attribute Alternative 1 Alternative 2 Alternative 3 Alternative 4

Implementation Cost 200 YTL 500 YTL 800 YTL 500 YTL
Operating Cost 100 YTL / month 50 YTL / month 50 YTL / month 25 YTL / month
Usage Time 10 min / day 50 min / day 30 min / day 30 min / day
Response Time 45 min 15 min 5 min 15 min
Language Turkish English Turkish / English Turkish / English
Security Available Not Available Available Not Available
Triability Not Available 1 week limited func- Not Available 1 week unlimited func-
tionality tionality
Sound Type Computerized Male Female Computerized
Input Type Sound Text Selection among alterna- Text
tives
Availability of Face- Available Not Available Not Available Available
toFace Communcation
Technical Support 7 / 24 09:00 – 17:00 09:00 – 17:00 Not Available
User Training Operating manual is Training is conducted by Online Education Not Available
given to the user a professional
Medical Provider Private Hospital Public Hospital Public Hospital Private Hospital
Customizability Frequently used menu Users are selecting fre- Menu items are shown in Frequently used menu
items are automati- quently used menu items the same order items are automati-
cally shown on top of manually cally shown on top of
the menu the menu
Clarity of Menu Items Service does not contain Textual menu items Graphical menu items Both of the graphics and
menu text are used in the menu
items
Output Quality Meal list is given as a User preference are Meal list is given as a User preference are
output taken into consideration output taken into consideration
in the meal list in the meal list

12
Evaluating Health Information Services

appendIx B

Figure 2. AHP Model

13
14

Chapter 2
Gastrointestinal Motility Online
Educational Endeavor
Shiu-chung Au
State University of New York Upstate Medical University, USA

Amar Gupta
University of Arizona, USA

aBstract

Medical information has been traditionally maintained in books, journals, and specialty periodicals. A
growing subset of patients and caregivers are now turning to diverse sources on the internet to retrieve
healthcare related information. The next area of growth will be sites that serve specialty fields of medi-
cine, characterized by high quality of data culled from scholarly publications and operated by eminent
domain specialists. One such site being developed for the field of Gastrointestinal Motility provides
authoritative and current information to a diverse user base that includes patients and student doctors.
Gastrointestinal Motility Online leverages the strengths of online textbooks, which have a high degree
of organization, in conjunction with the strengths of online journal collections, which are more com-
prehensive and focused. Gastrointestinal Motility Online also utilizes existing Web technologies such as
Wiki-editing and Amazon-style commenting, to automatically assemble information from heterogeneous
data sources.

IntroductIon next generation of medical doctors; practitioners

in this field also frequently refer to them.
For the last several decades, Harrison’s Principles Traditionally, papers and articles in specialty
of Internal Medicine, published by McGraw Hill, medical journals supplemented the material in
has served as a major source of information in the textbooks like Harrison. The latter book would
field of Gastrointestinal Motility. This book and itself be updated periodically to reflect the state
its online presentation have been, and continue of the art in medicine and the various specialties,
to be, used by many medical colleges to train the providing a consensus opinion of the standard
of care.

Copyright © 2011, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Gastrointestinal Motility Online Educational Endeavor

The advent of computers and Internet has about 170 terabytes of information on its surface
given rise to online sources of information such alone, equivalent to seventeen times the size of
as UpToDate (http://www.uptodate.com/) and the information in the Library of Congress (Ly-
WebMD (http://www.webmd.com/). While gain- man & Varian, 2003). With this increasingly
ing tremendous following and being updated fre- information-rich society, the most precious abil-
quently, these sources of online information relate ity for students and learners is no longer to find
to the medical field as a whole and not to particular the information, but to discern the most relevant
specialties. Furthermore, the information on these pieces of information and to integrate them into
sites is generally maintained by personnel of the practice. The American Library Association
respective organizations, not by specialists in describes “information literacy” as the ability
specific disciplines of medical science. These of individuals to “recognize when information
organizations are usually set up as commercial is needed and have the ability to locate, evalu-
entities, rather than not-for-profit ones. ate, and use effectively the needed information”
The progressive transformation of informa- (American Library Association, 1989).
tion has seen many journals that were previously The medical domain version of information
in paper format opting to use new electronic literacy is evidence-based medicine.
technologies; most of them now come out both Evidence-based medicine (EBM) is the inte-
in paper and electronic formats. Searchable elec- gration of best research evidence with clinical
tronic archives, such as PubMed (http://www. expertise and patient values (Guyatt et al., 1992).
pubmedcentral.nih.gov/), now place a plethora The Centre For Evidence-Based Medicine in To-
of information into the hands of researchers and ronto, Canada, states that the origins of evidence-
physicians. However, such searches are very based medicine date back to post-revolution Paris
time consuming and often produce irrelevant or (CEBM, 2007), but that the current growth is most
poorly supported articles. Sites like Harrison’s closely attributed to the work of a group lead by
Online (http://www.accessmedicine.com/) serve Gordon Guyatt at McMaster University in Canada
as information directories that can be searched, in 1992. EBM publications, reflecting interest in
hoping to place the most suitable information on this field, have grown from a lone publication in
a medical topic in a user’s hand. 1992 to thousands in 2007.
Students have gradually come to expect in- Studies have become increasingly critical
formation in quick and readily available forms of the value of textbook sources (Antman et al.,
without having to bother about inter-library loans 1992). Didactic continuing medical information
or even hardcopy versions at all. may be ineffective at changing physician perfor-
The goal of the endeavor described in this mance (Davis et al., 1997), and clinical journals
article was to adapt emerging technologies to may lack practical application (Haynes, 1993). In
improve methods of teaching gastrointestinal addition, physicians are faced with an increasing
material to students and to serve as a more effec- burden on their time, forced to diagnose patient
tive source of relevant and accurate information findings within a matter of minutes (Sackett &
for medical practitioners and specialists. Straus, 1998), and can only afford to set aside
half an hour or less per week for general medical
evidence-Based Medicine reading (Sackett, 1997). The staggering mass of
information being discovered is also daunting:
A study from the School of Information Man- 500,000 articles are added to the commonly used
agement and Systems at UC Berkeley estimates Medline medical journal database every year, and
that, in 2003, the World Wide Web contained “if a physician read 2 articles each day, every day

15
 Gastrointestinal Motility Online Educational Endeavor

for a year, (s)he would still find herself or himself its relevance, validity, and the work required to
648 years behind” (Lindberg, 2003). As research obtain it, as specified in Equation 1 (Slawson et
increases the quantity of information available, al., 2007).
medical practitioners are compelled to find ef- Further, the increasing quantity of research
ficient methods to educate themselves. being performed by commercial enterprises, as
The Centre for Evidence Based Medicine has well as other organizations with potential conflicts
cited several examples of strategic, educational, of interest, requires information be filtered for
and technical improvements in medicine that validity before incorporation into medical the
have enabled the current explosion in interest in canon. Finally, increased effort involved in ac-
this field. These include the emergence of new cessing the relevant pieces of information reduces
strategies for evaluating information; the creation the accessibility of such information. In addition,
of systematic reviews; the growing emphasis healthcare organizations are siloed, gaining the
on continuing medical education and lifetime advantage of sub-optimizing local departments,
learning; and the advent of online journals, meta possibly at the cost of the whole (Senge, 1980);
analysis of multiple studies, and ready access to this complicates the problem further.
such resources through electronic archives. Rapid
dissemination of accurate and comprehensive Information retrieval and
compilations of research results enables medical decision-Making
practitioners to make informed decisions that are
supported by the latest research results, and not As Stephen Hawking observes in The Universe in
by outdated trials. a Nutshell, the rate of growth of new knowledge is
In light of the increasing number of medical exponential. While 9,000 articles were published
journals, especially journals that focus on special- annually in 1900 and around 90,000 in 1950, there
ties, the sheer quantity of information threatens to were 900,000 scientific articles published per
overwhelm medical practitioners. The concept of annum in 2000 (Hawking, 2001). The explosive
information mastery has been coined to describe growth of information is challenging both the
the set of skills that physicians must nurture in information repositories designed to hold it, and
approaching, analyzing and incorporating or the ability of users to access relevant information.
rejecting new information. The issues mentioned At the time of this writing, Wikipedia serves
above are not limited to the medical arena alone: as the de-facto standard for online general ency-
Former Vice-President Al Gore described the state clopedias, and is among the top ten most-visited
of information management as “resembling the Web sites (Alexa Internet, 2007a). Its open-source,
worst aspects of our agricultural policy, which left volunteer-without-accountability approach led
grain rotting in thousands of storage files while to initial concerns about information validity,
people were starving” (Gore, 1992). but these concerns have been largely addressed.
The Center for Information Mastery at the Nature magazine studied Wikipedia and Encyclo-
University of Virginia asserts that the useful- pedia Brittanica and found that the two of them
ness of medical information is dependent upon

Equation 1. Usefulness (Slawson et al., 2007)

(Relevance)(Validity)
Usefulness of Medical Information =
Work

16
Gastrointestinal Motility Online Educational Endeavor

were largely similar in accuracy (Giles, 2005). The literature classified using the descriptors known
growth of Wikipedia’s information base further as Medical Subject Headings (MeSH). A broad
enhances the quality and breadth of coverage, range of search features are offered, including
and supports the possible future use of Wikipedia combined searches, exclusions, classification by
or Wiki-style architecture as an academically type of article (original research versus review)
respectable source of reference. and related articles. Another feature of the Med-
In contrast to the indexed and contributed line database, known as MedlinePLUS, provides
semi-structured format of Wikipedia, Google generalized information on health topics, and is
relies on search-keyword phrases. The useful- aimed at the public or at practitioners outside their
ness of Internet-crawling indexers, like Google, specialty domain. At the current time, PubMed
is based upon the ability to retrieve and capture serves as the gold standard in comprehensive
information from many sites, and to retrieve medical information, despite its dated interface.
relevant pages on query. Google’s initial strength HighWire Press, a Stanford-originated en-
and rise to stardom was achieved through its su- deavor, distributes thousands of journals, and
perior PageRank algorithm, which still remains provides its own search engine. In a recent study,
a carefully guarded trade secret; this algorithm the relevance of articles retrieved from HighWire
provides an uncannily relevant list of matches was found to be greater than that of PubMed, but
to any user query, ranging from commonplace with the disadvantage of a slower retrieval Speed
query phrases to obscure esoteric trivia and even (Vanhecke et al., 2006).
misspellings. In order to make a comparative evaluation
In a small supermarket today, shoppers are between different approaches, it is appropriate to
bombarded with a selection of 285 varieties of characterize the information recall ability using
cookies and 95 varieties of chips, leading the con- three parameters: precision, recall, and fall-out.
sumer to a state of decision overload (Schwartz, Precision (Equation 2) can be defined as the
2004). There is a growing need to restructure proportion of all retrieved documents that are
data to meet the informational and management relevant.
requirements of an organization or group of people Recall (Equation 3) captures the concept of
(Carlson, 2003). complete retrieval of all relevant documents.
Fall-out (Equation 4) is a measure of the
Medical Information repositories number of documents that are retrieved but are
unrelated to the issue being searched.
For the medical arena, PubMed is the most widely Medical research, while generally emphasiz-
used information database in the world, account- ing maximal precision and minimal fall-out, oc-
ing for 1.3 million daily queries by 220,000 unique casionally requires increased recall, in the case
users (Lindberg, 2003). It is a free access search of obscure diseases, or unusual side effects of
engine, provided by the U.S. National Library of medications. Medline serves as a canonical list
Medicine as the main access point to the Med- for such purposes, but at the cost of significantly
line database, a cataloged repository of medical lower precision.

Equation 2. Precision (Wikipedia, 2007)

{relevant documents}∩ {retrieved documents}

precision =
{retrieved documents}

17
 Gastrointestinal Motility Online Educational Endeavor

Equation 3. Recall (Wikipedia, 2007)

{relevant documents}∩ {retrieved documents}

recall =
{relevant documents}

Equation 4. Fall-out (Wikipedia, 2007)

{non-relevant documents}∩ {retrieved documents}

fall-out =
{non-relevant documents}

Medical practitioners using Google for their The range of challenges and issues that char-
searches will often find themselves frustrated acterize the medical domain include:
at the large quantity of articles on obscure and
irrelevant topics. A researcher searching for a • The presence of an extensive array of syn-
pharmacologic treatment of a syndrome will turn onyms for various drugs and diseases that
up with thousands of articles dealing with vari- require semantic knowledge to be encoded
ous sub-types, biochemical-signaling processes into the search engine in order to link con-
involved, and even support groups, before finding cepts that are not lexically related;
a therapeutic treatment. Due to the nature of the • The naming of disease subtypes (often after
search engine and the storage methods, there are a major contributor or discoverer), requires
concerns about Google’s or any search engine’s that hierarchies be constructed to allow users
ability to maintain a collection of such information. looking for the subclasses of disease to find
Carlson (2003) showed that due to the relatively information on the main umbrella disease,
small collection of documents indexed by an aver- and vice versa;
age search engine, a significant amount of relevant • The growth in the understanding of gener-
information would not be returned even in the alized syndromes results in a correspond-
presence of a perfectly formulated search phrase. ing need for reclassification based on new
In order to accommodate domain-specific etiologies of disease, thereby suggesting a
areas, Google has introduced the concept of “Re- dynamic organizational structure for the
fine Your Search” (http://www.google.com/coop). online medical information systems.
Without altering its main core search methodol-
ogy, Google allows users to more quickly locate In view of the growing difficulty in locat-
the type of information desired (i.e., treatment ing desired pieces of information, individuals
or symptoms). These refinement tools, provided performing research are in increasing danger of
by vendors and other private individuals or agen- information overload. As such, the next generation
cies that are deemed authoritative, subsequently of medical information access tools must aim to
label Web sites with appropriate descriptor tags. improve the ability to retrieve the right chunks
The potential conflict of interest created by these of information quickly, with zero or minimal
corporate associations is a matter of concern, due extraneous information; this concept is termed
to omissions or maliciousness of the labeling. as increasing the signal to noise ratio in the field
of electrical engineering.

18
Gastrointestinal Motility Online Educational Endeavor

VIsIon and goals For might be interested in viewing articles from the
gastroIntestInal MotIlIty perspective of case-based, symptom-based, or
onlIne test result-based diagnosis in order to apply the
information to a particular problem at hand. In
Gastrointestinal (GI) Motility Online is an essence, the vision of Gastrointestinal Motility
example of a medical information system that Online is to present information as framed by
seeks to provide access to high quality medical the interaction with the particular user at the
information online related to a particular medi- particular point in time.
cal specialty by centralizing the information and The first step in the vision of Gastrointestinal
presenting relevant information that is customized Motility Online was to collect the information in
to the user’s information requirements. a manner consistent with the goal to acquire the
The field of Gastrointestinal Motility is com- reputation for the highest quality of knowledge.
plex and interdisciplinary, involving a variety of The information base is assembled entirely from
experts. The possible user base includes layper- material provided by internationally acknowl-
sons, patients, medical students, biomedical scien- edged experts. All chapters including synopses,
tists, physiologists, pathologists, pharmacologists, articles, and reviews are written by reputed au-
biomedical students, researchers, pharmaceutical thorities. The pool of information is envisaged to
staff, house staff, specialty fellows, internists, be shared between different types of users and
surgeons, and gastroenterologists. Each role for different purposes. The design of the system
requires a different approach to depth, scholas- emphasizes a one-stop information approach that
tic relevance, and clinical direction in terms of enables the users to derive information at various
information presentation. For example, students depths. This applies to onsite information, as well
are interested in innovative research or review as to information at offsite locations.
papers; researchers would like to know the most
recent developments, and practitioners might be details of effort
more interested in using the information for dif-
ferential diagnosis purposes. GI Motility aims to A two-phase approach is being utilized for the
serve as a collaboration of medical professionals, creation of the information system: the first in-
approaching diseases and patients from different volves full leveraging of commercial technology
angles. as it exists today, and the second involves further
In a library, a user interacts with the data in research on aspects that can be incorporated in
books very differently from the way that she or future versions of our system. In the absence of a
he interacts with data in an online presentation. better term, the term gastrointestinal knowledge
The user expects the book to be focused and to repository is used for the final system, as well as
address the topic in a linear fashion. Online, the for the initial concept-demonstration prototype
same user navigates quickly, using hyperlinks, to system.
explore secondary topics. In fact, the user expects For the first phase, the acquisition of knowl-
a different presentation and a different style of edge proceeded with the establishment of titles
information; as a result, the nature of the interac- and themes for chapters, as determined through
tions will differ even with the same content. One discussions involving the concerned authors and
of the aims of Gastrointestinal Motility Online the editors for this project (Dr. Raj Goyal of Har-
is to address these different styles and to present vard Medical School and Dr. Reza Shaker of the
the desired subset of information in the manner Medical College of Wisconsin). The creation of
that the user might expect. For example, a user the gastrointestinal motility knowledge repository

19
 Gastrointestinal Motility Online Educational Endeavor

began with calls to key gastrointestinal experts A secondary role is the development of a
inviting them to submit a chapter, in electronic community of gastrointestinal specialists and
form, for inclusion in this knowledge repository. other interested parties, who can form an online
The inputs from the contributing authors were collaborative, expand upon the information re-
reviewed by these two editors and by others, on an pository, and facilitate peer communications and
anonymous peer review basis. Under the aegis of discussions. In addition, the site aims to explore
an unrestricted grant from Novartis Corporation, and to expand the concept of online information
the two editors worked closely with the staff of repositories, especially the optimal integration of
Nature Publishing Group on a number of tasks the current generation of electronic journals and
that ultimately led to the creation of the following online textbooks.
Web site: http://gimotilityonline.com.
The creation of the site involved an automated site audience
conversion process to adapt MS-Word and rich
text documents into a Web presentable format, The primary site audience is the set of gastroin-
with special emphasis placed on images, tables, testinal specialists, associated medical staff, and
and video. The majority of the investment for staff under training. Members of this primary
development lay in formatting and typesetting; user base would initially visit the site because of
the design itself was of less concern as it followed a recommendation from a colleague or because
existing Web branding and style guidelines of Na- another site (search or advertisement) directed the
ture Publishing Group. Rights for images needed user to it. Occasional patients are expected, but are
to be obtained; further, images, tables and video unlikely to become the primary users of this site.
needed to be edited to fit a standard look and feel. Initially, the core attractions for users to this
After receiving author contributions, the project site are the quality and depth of information
required approximately 18-24 months to complete, coupled with the ease of access and the low cost;
with 9-12 months required for the editorial process these same factors will also help to retain the user
itself. The current site consists of 1,000 HTML base. Most physicians and healthcare specialists
pages, 1,000 images, 500 Powerpoint presenta- spend relatively little time online (six hours per
tions and 40 videos. Articles are cross-linked by week on average); in order to impress the user
topics, and the current volume is equivalent of base, the signal to noise ratio of the site must be
about 700 pages of text. This volume is increasing high, along with the ease of locating a desired
as this endeavor continues to progress. piece of information (through the information
architecture) (Friedman, 2000). The primary user
site purpose of the site is unlikely to demand high interactiv-
ity; instead, the interest will be on locating and
The ostensible purpose of the site is to disseminate extracting the information as quickly as possible.
information on the specialty of gastrointestinal As such, the site must be efficiently organized,
medicine, primarily to gastrointestinal medical streamlined, and equipped with powerful search
practitioners and to other interested parties. The and index functions.
ambition for the site is for it to become the central After drawing a user initially, repeat visitors
hub of information on this specialty, aggregating would use the site to browse new topics of interest,
information from many sources, authors, and as well as to entrust the editorial staff to select
journals into one central location with the objec- articles that represent innovative research in the
tive of becoming a de facto standard for online relevant field. An interacting community base
gastrointestinal information. would evolve over time and eventually cause a

20
Gastrointestinal Motility Online Educational Endeavor

change in the workflow. Specialists would then information by anatomical section or syndrome.
visit the site to explore the comments from their The timing is also important; recent material
colleagues on the topics expressed and to leave is favored over older material. Since relevance
their own authoritative inputs on different articles, of the article is also important, the name of the
eventually contributing entire articles as much journal and the name of the author are also used
as a good electronic journal aspires to do today. prominently in searches.
While relatively low in terms of being tech- Furthermore, classification schemes or hier-
nically savvy, a typical user of Gastrointestinal archies to group-related topics are essential to
Motility is likely to be very comfortable with using narrow the branching process used by the search
the Internet for retrieval of scholarly information technique. The determination of related informa-
through PubMed, online textbooks, and ready tion is a complex topic. In order to address the
references such as UpToDate. A sleek, uncluttered latter issue, medical organizations are creating
design is likely to be the most attractive, even medical ontologies, such as unified medical
though it may not support applications involving language system (UMLS), to quantify and to
high load times, such as Flash or embedded vid- impose structure on conceptual relationships.
eos. Users are comfortable reading large articles With this classification, the selection criteria can
online, but they also expect printable versions to be hierarchically built up or finely specified to
be available, on an as-needed basis. obtain desired results. A lengthier discussion of
The user base, while highly intelligent, is this topic appears later in this article.
relatively small in numeric terms, and is char- Look and Feel: The look and feel of the site
acterized by a small presence on the Web. The needs to convey the sense of scholarly authority,
process of attracting a critical mass to build and but with a sleek technological approach. Medical
to maintain a user community is a high priority personnel have high standards concerning the ac-
task; this involves contacting a high percentage curacy of articles, and a professional presentation
of all members of this specialist community. aids in supporting this perception. A site that is
gaudy or shows too many bells and whistles, or
site content involves a long load time, reflects poorly on the
content, as do garish colors or lack of color.
Gastrointestinal Motility Online is a site that pro- Images and video must be consistent and
vides information on both medical practice and should be available to download as needed for
the fundamentals of the gastrointestinal tract. The reference and for closer examination. The level
information must provide both breadth and depth of interactivity available should be low, as most
on the topic, and should ultimately serve as an gastrointestinal specialists have small online
encyclopedia of the domain-specific knowledge. presence. Natural language queries, such as those
Information Architecture: As in a library, in- used with AskJeeves (http://www.ask.com/), are
formation must be properly accessible in order to unnecessary, as long as the search and browse
have value. The architecture of the information is features are precise and efficient. Pages may be
partly determined by the methods that the users presented as either textbook or journal articles,
will use to query the evolving knowledge base. based on the preference of the user.
In a library, the name of the author and the title of Extracting Content, Metadata, and Cross
the book are important. In a journal, the age of the Referencing: One of the most powerful func-
article or the issue in which it was published may tions of the Internet is the ability for Web users
be essential. In Gastrointestinal Motility Online, to span several related articles quickly because
the most likely user scenarios involve searching for of cross-referenced links. A user interested in

21
 Gastrointestinal Motility Online Educational Endeavor

the preparation of a particular chicken recipe can to rank the importance of articles so that users
quickly reference how to sauté and with what browsing information can be directed to the most
form of pan, moving quickly to sources to buy useful and informative articles.
the appropriate cast-iron skillet or wok to benefits One difficulty with searches is the likelihood
and comparisons of different brands and retail- that a particular phrase will appear in nearly all
ers. A Wiki of only gastrointestinal specialists, documents unless the search is very specific. In
with limited control from an editor, would be cases where the gastrointestinal tract is analyzed as
appropriate for the collection and dissemination an interactive system, the phrases for anatomical
of information: Gastrointestinal Motility Online locations may occur multiple times as reference
is striving towards that goal. points, but the central theme of the article may
A particularly useful feature in Gastroin- not be easy to determine by word frequency. As
testinal Motility Online, not available in online a result, a number of semantic tools, described
journals, is its cross-referencing tool. Articles later in this article, are used to analyze articles
that are closely related or broach a topic in greater and to classify them appropriately.
depth can be quickly accessed by a cross refer- Experts of the Nature Publishing Group (NPG),
ence within the Gastrointestinal Motility Online who possess prior experience in online informa-
domain. These links are established by content tion presentation, based on the online version of
extraction tools that create metadata and relate Nature magazine, helped to develop the initial vi-
that data between different documents. Footnotes sion of the online knowledge repository. The base
are available at the bottom and allow for a broader site is hosted by Nature Publishing Group. While
topical search; however, the inline linking of the site was being developed, commercial tools
articles is particularly appealing for tracking were available to handle both the production of
particular items or syndromes of interest. electronic journals and static textbook efforts like
AccessMedicine/Harrison’s Online (http://www.
auxiliary technologies accessmedicine.com/) and WebMD. However,
the gastrointestinal knowledge repository falls
Search features are incorporated in Gastrointesti- somewhere in between these two cases; accord-
nal Motility Online, but are considered secondary ingly, few over-the-shelf tools and algorithms were
to the organization of the information. A dynamic available for immediate use. As such, a significant
keyword search for anatomical sections of the fraction of the interface and architecture had to be
gastrointestinal tract is less likely to reveal useful innovated and refined, through experimentation.
information on general function than a manual Many of the existing tools for creating elec-
perusal of the literature through the prepared tronic journals are geared towards collation of
subject browse option. The efficient organiza- articles, graphics, and layout work. These tools
tion of the information, based on the anticipated reduce the time needed by the authors and editors
needs and access patterns of users, is an essential for the processes of uploading, formatting, and
feature for building a knowledge repository. By editing. In the development of Gastrointestinal
acknowledging that specialists would be more Motility Online, the use of a software suite fa-
likely to query by anatomy or syndrome, one needs cilitated handling of images and consistency of
mechanisms for structured order, as compared look and feel. One area where tools are lacking is
to mechanisms that order by article size, author the ability to organize information into a coher-
name, or recent usage. An additional feature, ent topical fashion, as in a textbook. Searching
incorporated within the community-building by keyword is especially difficult on a physician
module, is a user rating system that allows users specialty site, where the dialect is limited and

22
Gastrointestinal Motility Online Educational Endeavor

the concepts are reused multiple times. As such, ity to integrate clinical patient presentations and
editorial staff must impose additional control create a list of possible problem diagnoses.
to prevent the site from becoming a write-only At a broader technological level, eBooks
knowledge repository. represent a technology that has been adapted to
Collections of electronic journals, such as deliver information electronically. Medical eB-
Ovid (http://www.ovid.com/site/index.jsp), have ooks can be argued to be a natural outgrowth of
been primarily targeted for libraries and research the eBook movement to electronic media: volume
centers. The primary purpose of Ovid is to serve and space requirements are reduced, key phrase
as a warehouse of information, albeit uncatego- searching can be performed, and portability is
rized. Gastrointestinal Motility Online’s current enhanced. Nevertheless, few medical texts are
state differs from that of Ovid in terms of the adapted as eBooks.
presentation of the material: the former system is Although eBooks have grown in popularity,
specifically formatted to provide an online view they have not grown as rapidly as projected by
as well as a hardcopy output. The long-term goal consultants; this could be because of the follow-
of Gastrointestinal Motility Online is to collect a ing reasons:
comprehensive knowledge on subjects (as Ovid
does), as well as to add more intelligent search 1. Most readers see no need to replace print
tools or information utilities. While Ovid does not books.
organize information except into broad categories, 2. Due to the limited screen size, limited bat-
Gastrointestinal Motility Online refines classifica- tery life, and navigation interface issues with
tions, provides responses to search queries that eBooks, many people still find paper books
are more accurate, and supports tools that use easier to handle.
the knowledge in compelling ways, such as for 3. Digital rights management causes compat-
differential diagnosis. ibility and portability problems when at-
The advantages of sites built in a textbook tempting to move the eBook from desktop
style, such as AccessMedicine, is the hierarchi- to PDA or laptop.
cal organization and ready access to information. 4. Current pricing of eBooks does not account
The evolution of online textbooks has generated for the reduced value relative to paper books.
significant activity on the sites, as teaching tools. When readers finish reading a paper book,
Gastrointestinal Motility Online uses over 40 il- they can give it to a friend or sell it to a used
lustrative videos, which are not typical of a journal, bookstore; neither is possible with most
but fit the online textbook paradigm. Online text- eBooks. (Crawford, 2006)
books are excellent repositories of information,
except that updating the sites to incorporate new Additionally, delivery of information may not
information is generally cumbersome because of be simply online, but online and to a mobile user
the level of interactivity involved. using a PDA or other portable device. The con-
“Gastrointestinal Motility Online” is a hybrid, straints involved in transmitting and displaying
adopting the best qualities of online textbooks, information on a limited display panel create a
such as Harrison’s Online, and journal collections, new set of challenges. In most markets other than
such as PubMed. Organized, Well Edited, Fre- healthcare, the primary applications for PDAs are
quently Updated and Comprehensive Information for scheduling and contact management (as a busy
are self-explanatory. Search features specifically executive might use in lieu of a pen-and-paper
refer to the ability to search for a keyword or daily planner), or as a portable browser or e-mail
phrase. Differential diagnosis refers to the abil- client (as in the case of technologists and engi-

23
 Gastrointestinal Motility Online Educational Endeavor

neers). In such cases, the application of the PDA display technologies, such as holographic displays
works within the bounds of the limited display or direct-to-eye projection technologies, the abil-
and the modern constraint of minimal bandwidth, ity for PDAs to contribute to medical reference
often serving as a surrogate cell phone of sorts. appears to be technologically limited.
However, in medicine, the PDA is often As shown in Table 1, Gastrointestinal Motil-
stretched beyond its limits. The current trend is ity Online is a hybrid that lies between the two
the delivery of detailed information pages into models of electronic journals and online textbooks,
a portable format, downloadable to PDA. Since exclusively focused on providing authoritative
medical practitioners can no longer maintain a information in an organized fashion within a spe-
complete mental catalog of all drugs and par- cific domain. Using this system, a gastrointestinal
ticularly obscure clinical symptoms or diseases, researcher can find the most recent journal articles
PDAs assist physicians in their duties without because of the frequent updates, and a gastrointes-
requiring a quick trip to a computer terminal tinal clinician can easily locate a detailed diagram
or a large paper binder archive. Harrisons and of the lower esophageal sphincter. In addition,
UpToDate have both moved rapidly in this area, topical information can be cross-linked between
and the list of available drug databases is already papers—a typical feature of textbooks and very
large. PDAs fill the role of drug lookup very well, pertinent for teaching and presentations.
as well as serve as a primer for obscure diseases.
The difficulty lies with more graphically intense
data that may not display properly, or may need to concepts and InnoVatIons
be downloaded on the fly. In such cases, medical
information systems are pushing the technological The exploration of the technological space be-
limits of PDAs. tween electronic journals and online textbooks is a
PDA sales as a whole, however, are in decline, relatively new idea. All new ideas face challenges
except as a niche application. Analysts at organi- in terms of deployment and adoption. Consider
zations such as IDC and Gartner have predicted the fax number. As the number of fax machines
downtrend trend in sales of PDA. Dell has with- increases, the value of the fax increases—this
drawn its PDA line from production (Mechaca, illustrates the fact that networks attain greater
2007). In the long term, the PDA may carefully value with larger number of users. The difficulty
constrain its niche to feature more of the portable faced by Gastrointestinal Motility Online and
planner features and less multimedia and display other specialty interest sites is in terms of initial
power, rendering it less useful to medical prac- growth and development of specific communi-
titioners. Until the advent of revolutionary new ties of interest. These sites must be aesthetically

Table 1. The online endeavor

Online Textbook Journal Collections Hybrid – GI Motility Online

Organized X X
Well Edited X X
Frequently Updated X X
Comprehensive Information X X
Search Features X X X
Differential Diagnosis X

24
Gastrointestinal Motility Online Educational Endeavor

attractive, informative, efficient, and up-to-date. movie information (approximately 900,000 titles
The case of Gastrointestinal Motility Online and 2.3 million names) (IMDB, 2007). IMDB
illustrates one form of evolution of online journals draws a significant portion of its information from
and textbooks into an active online scientific com- the participation of its user base. Beyond subjective
munity. Site loyalty is achieved and maintained reviews, users are also asked to supply cast and
by the reputation of the authors and contributors. crew lists, production details, and actor biogra-
Gastrointestinal Motility Online needs only to phies. IMDB grew from two lists that started as
achieve a critical user mass before gaining the independent projects in early 1989 by participants
benefits of Web sites like e-Bay (http://www.ebay. in the Usenet newsgroup rec.arts.movies. Each
com/) or Amazon (http://www.amazon.com/) in list was maintained by a single person, record-
terms of de facto authority and brand recognition. ing items e-mailed by newsgroup readers, and
For sites which are less commercially oriented, posting updated versions of his list from time to
the loyalty of the user base is perhaps even more time. The lists were eventually combined, and
heavily emphasized. Two notable sites, which have by late 1990, the lists included almost 10,000
grown rapidly without such a strong commercial movies and television series. As the contributions
bias, are Wikipedia and Imdb. continued to grow rapidly, the IMDB formed as
Wikipedia has been both maligned and praised an independent company in 1995 and was later
for its loyal community and efforts to create a purchased by Amazon Inc. (Wikipedia, 2007b).
free encyclopedia that is accurate and up to date The approach of the IMDB system closely
without any commercial affiliations. Wikipedia resembles the envisioned system for GI Motil-
began in 2000 as a complementary project for ity Online, with the user population submitting
Nupedia, in which articles were written by experts proposed changes, followed by an editorial
and reviewed by a formal process (Wikipedia, process. Database content is generally provided
2007). In 2001, Larry Sanger proposed on the and updated by a vast collection of volunteer
Nupedia mailing list to create a wiki as a “feeder” contributors. There are only 17 members of the
project for Nupedia (Sanger, 2001); this spawned IMDB who are dedicated to monitoring received
rapid growth. By 2001, Wikipedia contained data, although 70% of IMDB’s staff serve as edi-
approximately 20,000 articles and 18 language tors (IMDB, 2007), reviewing changes, verifying
editions. As of 2007, English Wikipedia contains the information before posting the changes, and
over 1.7 million articles, making it the largest policing the forums.
en¬cyclopedia ever assembled. The site relies Peer review is considered to be one of the most
on the goodwill of its members to write, update even-handed and least biased methods of scrutiniz-
and contest articles, and has thus far proven that ing articles for publication. The development of
the Internet community as a whole is willing to a community, as well as the ability for members
contribute towards the database, albeit haphaz- of that community to voice their opinions about
ardly (recent events are more likely to be covered professional papers is pivotal to the dissemination
in detail, while significant historical figures lan- of accurate information. The model of Amazon or
guish). Given the size and relatively stable growth eBay is to allow users to comment, and thereby to
of the project, the prospect of a peer-reviewed signify reliability and approval. Gastrointestinal
and written information repository might not be Motility Online allows users to provide feedback,
so cynically doomed. both critical and supportive, in order to enhance
The International Movie Database (IMDB), a the relevance of articles.
site in the top 20 (Alexa Internet, 2007b) U.S. Web According to Harris Interactive poll, only
sites visited, is the largest Internet compilation of one-fourth of physicians use the Internet to

25
 Gastrointestinal Motility Online Educational Endeavor

communicate with their patients (Computing the main knowledge repository is not intended to
in the Physician’s Practice, 2000). In the same block or hide information, but to provide a more
poll, although 89% of physicians use the Internet relevant source. Classification and weighting
in their practice in search of information, they are accomplished by a rules-based system that
spend only six hours per week to browse medi- can assess whether an article is clinical- or basic
cal developments. Accordingly, Gastrointestinal science-related. Login will also provide both the
Motility Online has been configured to serve as ability to contribute and comment on articles, and
an encyclopedic source, as well as a high-value to obtain a specialized view depending on the type
news feed. AccessMedicine uses a Podcast update of user. The application of user-based site layouts
model, with 10-minute broadcasts generated daily is not innovative, but is important to the domain
for use in family practice. Gastrointestinal Motility of medical informatics because physicians attach
Online caters to a smaller specialty group with such a high priority on relevance.
correspondingly less frequent pace of develop- Given the large number of articles published
ments, and is therefore less time sensitive. weekly and the difficulty in ascertaining relevance
and quality, a number of automated tools will be
automation used to optimize the updating process. Sarnikar
et al. (2005) present one technique that will assist
Automated content generation or extraction from in filtering journal results and maintaining and
other publications is not feasible, due to the strin- updating the site. Their method selects articles
gent need to maintain relevance and quality. Thus, ranked by relevance using a combination of both
deployment of a pure peer-reviewed wiki-style rule-based and content-based methods, using the
community is precluded by the need to maintain following principles:
quality. If membership of the wiki is restricted
to those users whose credentials are accepted, or 1. Profiles are modeled in the form of rules.
if changes must be approved at an editorial level, 2. The purpose of the rule-based profile is to
then a wiki would facilitate the rapid exchange identify a sub-set of documents of interest
of information as well. In the second part of the to a user.
endeavor, the goal is to enable machine-assisted 3. Each role has a set of predefined rules as-
updating of the material in the gastrointestinal sociated with it.
knowledge repository. Currently, all updates 4. Rules specify knowledge sources to access
must be initiated manually. The pressure to up- (e.g., nursing journals for Nurses).
date, but to update accurately applies to many 5. Rules can specify knowledge depth and
situations: addition of new material, editing of knowledge breadth.
existing material, and deletion of parts of existing 6. Rules can specify semantic types of primary
material as new study results become available. importance to roles.
Initial thoughts and test results are documented
in Sharma (2005). Profiles are used in the gastrointestinal motility
As the site evolves, the need to keep the site context to separate information into categories:
relevant to a particular group of specialist physi- for example, new clinical findings versus basic
cians may conflict with the preferences of another science. Articles may be assigned a category and a
significant category of users: the researchers. relevance weight, given categorization rules based
Gastrointestinal Motility Online intends to use on Unified Medical Language System (UMLS)
user-based customized layout, as presented by synonym lists and the categories sign or symptom,
Sarnikar et al. (2005). The filtering of details from diagnostic procedure therapeutic or preventive

26
Gastrointestinal Motility Online Educational Endeavor

procedure and disease or syndrome semantic medical language system: What is it and how to
types. In addition to a text search in the abstract, use it? (http://www.nlm.nih.gov/research/umls/
Sarnikar and Gupta (2007) also assign weights presentations/2004-medinfo_tut.pdf).
to the type of journal. These tools can select and UMLS is an aggregate of over 134 source vo-
filter relevant articles for presentation as an RSS cabularies, including the classifications from such
XML news feed to editors or automatically as- lists as ICD-10 and DSM: IIIR-IV. It represents a
semble relevant articles for use by the editors or hierarchy of medical phrases that can be used to
Web site administrators. While these tools will classify most medical articles and textbook entries.
aid the editor, there is no replacement for the role For example, using UMLS, the following phrases
of humans in decisively selecting and classifying are grouped similarly: Deglutition Disorders,
information. Difficulty in swallowing, Difficulty in swallowing
Ontologies and semantic networks are prereq- (context-dependent category), Dysphagia NOS,
uisites to the development and classification of Dysphagia NOS (context-dependent category),
information repositories. Ontologies serve many Can’t get food down, Cannot get food down, Diffi-
purposes (Kumar, 2005) including: culty swallowing, Difficulty swallowing (finding),
Dysphagia, Dysphagia (Disorder), Swallowing
• Reuse and sharing domain knowledge difficult, Swallowing Disorders (Aronson, 2001).
• Establishing classification schemes and The system described in Sharma (2005) uses
structure techniques of Natural Language Processing
• Making assumptions explicit (NLP) to construct a semantic understanding that
surpasses text searching. Using the automated
They further enable analysis of information integration of text documents in the medical do-
and complement the stricter terminology that is main (ATIMED) system, the content and order
used in straightforward text searches. Examples of phrases are related lexically using a concept
of ontologies in use today include the National called Word-Net. Word-Net operates on the verbs,
Library of Medicine’s Medical Subject Heading subjects and objects of the sentences, comparing
(MeSH), disease specific terminologies such as sets and subsets of subject-verb-objects collections
the National Cancer Institute’s PDQ vocabulary, in order to determine topic relatedness.
drug terminologies such as the national drug data Word-Net further uses a lexical dictionary to
file, and medical sociality vocabularies such as the determine similarity in all verb pairs and then
classification of nursing diagnoses and the current subject-verb-action pairs. Sharma (2005) uses
dental terminology. In Gastrointestinal Motility the following two sentences as examples: Dys-
Online, the ontological hierarchy will be used to phagia is a disease and defined as a sensation
distinguish between sections of the gastrointesti- of sticking or obstruction of the passage of food.
nal tract, from the stomach and esophagus to the Dysphagia is related to obstruction of passage of
large intestine and colon (Kumar, 2005). food. Since both sentences contain similar objects
A key enabler for development of automatic and subjects, and use the verb “is”, the sentences
information processing is the set of ontologies are deemed similar. However, within the current
presented in the UMLS semantic network that mechanism, the phrase, “Dysphagia relates to
rely upon the concepts built in the UMLS concept obstruction of passage of food” would result in
hierarchy. Hierarchical and clustered ontologies a poorly scored correlation or low match because
allow software to construct knowledge trees, the action verb is not similar (Sharma, 2005).
conglomerating relevant knowledge. An over- Finally, the same technique allows the creation
view of the UMLS is available at the unified of new documents by collating sentences and

27
 Gastrointestinal Motility Online Educational Endeavor

Figure 1. Integration of semantic understanding to generate new information (Adapted from Sharma, 2005)

Sentence Syntax
Definition Tables

Feature
Selection

Feature 1 Feature 2 Feature 3 Feature 4

WordNet

Concept Formulation
and Creation

paragraphs from various documents. An initial branch of information technology. One level of
method of grouping sentences uses the quantity of development is the creation of maps between
concepts expressed. This method is further refined inputs and outputs, in much the same way that
by evaluating the sentences based on the following a dictionary might map between languages. The
criteria: similar-subjects, similar-objects, similar- conversion of n sources to m outputs can grow
subjects-objects, and similar-objects-subjects. to be an exponentially difficult problem; this can
Based on the structure of English grammar, these be addressed with the use of intelligent heuris-
techniques have been reliably shown to collate tics and protocols to selectively prune the data.
relevant data into a readable format. Difficulties become particularly apparent with
A diagram of the method is shown in Figure 1. changes in any client schema that cause a cascade
A sample output, based on the use of this of changes in the mappings (Sarnikar, 2005).
technique, is provided below: Using an independently developed predefined
mediating schema would restrict the amount of
REGURGITATION is defined as the spontaneous information that could be exchanged. Apart from
appearance of gastric or esophageal contents in relational schema, a client schema could also be
the back of the throat or in the mouth. In distal specified as a hierarchical schema or as an XML-
esophageal obstruction and stasis, as in achala- based message (Sarnikar & Gupta, 2007). In the
sia or the presence of a large diverticulum, the context of Gastrointestinal Motility Online, the
regurgitated material consists of tasteless mucoid specific source contexts are innovative journal
fluid or undigested food (Sharma, 2005). papers, reviews, and textbook articles. The output
contexts are specialist clinicians, researchers, stu-
The context interchange of heterogeneous dents, and other health professionals. Developing
sources of information being collated for different a schema to accurately represent journal abstracts
classes of users leverages tools from an additional and determine the relevance of those abstracts is

28
Gastrointestinal Motility Online Educational Endeavor

another method of exchanging contexts. Innova- out of medical information. The highest-value
tion in this domain will allow Gastrointestinal information for each physician may vary based
Motility Online to maintain updated, consistent- on specialty, and user-customization of graphical
quality references without requiring an editor to widgets may enhance information value. Since GI
read every journal article published immediately. Motility Online aims to support a broad base of
users with many different roles and information
needs, reusable widgets, of the type available in
lessons learned customizable Web-portals, may potentially solve
these varying needs.
The GI Motility Online site benefited from a Workflow: As stated previously, over nine
mature development environment for Web-based months was required for simply editing the site
information retrieval. information, moving documents back and forth
Interface: In the case of GI Motility Online, between editors, experts and developers. The
the user base is a readily identifiable group, method involved several inefficient technolo-
trained in a similar fashion, with specific needs gies, such as mailing CDs and sending large files
and expectations of organization and formatting. through mail servers. The advantage of such hands
The following two aspects influenced the design on interaction and communication is the result:
process: the authors of the site are particularly pleased by
the polished appearance of the delivered product.
• Pagination is a critical issue for online di- Electronic collaboration between different
dactic materials. As people do not like the sites was difficult due to the nature of document
interface to online books (Crawford, 2006), formats, which does not produce a consistent print
the development of the site must reflect the layout on different computers, and layout formats,
reading style and needs of its users. In the which do not allow easy markup or revision to
case of GI Motility Online, there are cur- the document.
rently no page delimiters; this is encouraging The use of a standardized input format and
users to print the material to be used as a efficient conversion of the RTF-formatted docu-
reference. A solution of delivering a format- ments into Web viewable formats was crucial to the
ted print-ready PDF could be applied, but development of the Web site. Though conversion
the cost begins to climb with the number of documents is a minimally complex task, the
of different formats and delivery options production of a site that allows conversation on
supported. the fly to support concurrent editing and proof-
• Tables must be handled intelligently. Tables ing between multiple users is a challenge. Online
are used in many medical publications to publishing companies, such as Atypon, offer
rapidly and clearly present information. The suites of software to facilitate publishing work-
trend in the mid-1990s for netiquette was to flow, allowing multiple authors to upload articles,
inline the tables with the text, but this can multiple editors to revise articles, and art editors
create awkward gaps or poor formatting to manage graphics, all in an organized fashion.
choices. GI Motility Online chooses the Meta-Information: The World Wide Web
path of Harrison’s Online, in linking to the provides a tremendous asset in connecting re-
tables outside the main document to preserve lated articles seamlessly. In determining related
readability. pages, an active agent—a human, machine or
some combination thereof—who must isolate
Based on their respective training, many the crucial elements of a page and capture that as
physicians expect a particular language and lay-

29
 Gastrointestinal Motility Online Educational Endeavor

meta-information, preferably categorized. Orga- certain, but he states that initial impressions are
nized meta-information allows automated tools unanimously positive and enthusiastic. Dr. Goyal
to develop connections and links to potentially notes that the majority of the traffic to GI Motility
relevant information. Online currently arrives from Google, and indeed,
Extracting crucial elements of a page is Google’s first returned site in response to “GI
best handled by encouraging authors to define Motility” is GI Motility Online (Google, 2007).
keywords, as with scientific articles. Although GI Motility Online does not currently permit any
many automated agents have historically been advertising; if this policy is revised, the usage pat-
unsuccessful in information extraction, the terns may change. Further, the numbers for site
medical domain provides some assistance with visits are expected to increase if the site were to
standardized ontologies, which allow agents to be more supported by Nature Publishing Group.
categorize information in a framework, reducing The stakeholders are satisfied with the quality
some identification errors. of the product, and are ambitiously pursuing an
Maintenance: Many articles in mainstream extension of the product to broaden coverage to
sites, such as ESPN.com, offer the opportunity other parts of the GI tract.
for users to comment and leave feedback. This Dr. Goyal has received inquiries to publish
increases user participation and value to the site, the material in a hardcopy format. Custom-built
but raises additional issues that need to be ad- books or full-page colored slides offer additional
dressed, such as: utility to medical educators, specialists, nurses,
and students, and may also offer a revenue stream.
• Profanity or other inappropriate comments These custom-built printable books are a logical
must be censored extension of the current “print what you want”
• Sites are more open to attack and security photo fulfillment services that are popular on photo
leaks due to their increased functionality sharing sites or album hosting services. Given
• Bandwidth usage increases, and may debili- that Harrison’s Principles of Internal Medicine
tate the site contains over 2,000 colored pages in book form,
there may be a future for customized books in
In addition, a version control framework must medicine.
be established for sites that allow updates, in or- Retrieval: New algorithms for search and
der to rollback unwanted changes. Furthermore, ranking are not merely lexically-based, but also
hardware resources must be allocated to store combine closely related concepts and relation-
site changes. ships between ideas. This facilitates the creation
of a richer search language that can account for
relationships, such as causation, consequence, as-
current usage and sociation, treatment, or opposite. For example, hy-
Future dIrectIons pothermia is directly opposite hyperthermia, and
may be caused by thalamic alterations. Such basic
Anecdotal evidence provided by the lead creator of chains of relationships could be easily captured
GI Motility Online, Dr. Raj Goyal, highlights that if the language and storage of meta-information
the representatives of Nature Publishing Group about articles could contain the necessary under-
describe the site as being popular, and that site lying details.
visits to GI Motility Online are increasing. Due Presentation: Using Tufte’s principles (Tufte,
to the delicate nature of conversation with his 2001), the real estate of a screen needs to be more
peers in gastroenterology, Dr. Goyal cannot be efficiently used. Currently, most search-engine

30
Gastrointestinal Motility Online Educational Endeavor

results do not often give the sense of the relevance high-capacity and high-bandwidth connections
of the article, of the correct sections, or the tone close to Internet backbones.
of the article. The information density of the re- Collaboration: Collaboration in the domain
turn pages is low, causing users to scroll through of GI motility will drive improvements in lan-
potentially hundreds of articles to find relevant guage tools, with many GI specialists throughout
information. Graphical interfaces are likely to the world who may need to collaborate via the
be a solution, as the ability of the Internet to Internet or phone services. Currently, Altavista
handle higher bandwidth applications grows. In and Google offer Web-based text translation, but
medical research in particular, an interface, that more intelligence or domain-specific knowledge
allows users to quickly see search results of the may be required. One common example in GI
primary concept and even related concepts, may translation is translating the world “oral” to mean
dramatically affect usability. “verbal”, when in fact, the correct reference is to
Online Sharing: Bulletin board systems, the mouth cavity. Overall, however, text transla-
which flourished during pre-Internet days, have tion between languages is generally adequate for
re-emerged on the Internet as forums, and are initial communication.
a popular source of information. Corporations Oral translation is a high value direction to
with significant Internet presence, and especially pursue, but at the same time, is a very difficult
gaming companies like Nintendo (http://forums. task. Psycholinguistics research is still unravel-
nintendo.com/nintendo/), have begun adding and ing the complexities of language parsing, and the
using these forums as a method of improving ability of current artificial intelligence to under-
public relations and offering support. stand language is severely limited. Nevertheless,
Intranet and Internet file sharing systems translators in this area will prove essential and
have also flourished, as bandwidth rates increase. highly desirable for the next generation of online
The rise of Youtube is one phenomenon, but the collaborators.
comfort of using the Internet to disseminate mul- Similarly, one of the holy grails of artificial
timedia (such as GI endoscope video) appears to intelligence development is the creation of an
be more solidified. These high bandwidth applica- artificial system capable of interpreting human
tions have grown in acceptance, and despite the language. Within specialized domains with lim-
threat of viruses in downloaded files, file-sharing ited vocabularies, artificial readers become more
traffic is increasing daily. feasible, but the medical domain is particularly
Megaupload and Rapidshare, two prominent difficult, due to its large specialized vocabulary.
file sharing services, are now ranked #18 and #27 in Development in this area would provide rewards
the reliable Alexa ranking (Alexa Internet, 2007b) for medical researchers, allowing the creation of
of most visited Internet sites in the U.S. However, agents, which would allow researchers to process
these services are predominantly used by non- more information by selecting and even sum-
commercial entities. Commercial enterprises may marizing articles.
be hesitant due to slow adoption, security issues, Organizationally, scientific research in do-
bandwidth maximums, or unprofessional presen- mains such as medicine would benefit tremen-
tation. Such file sharing sites will certainly cater dously from the creation of a centralized authority
in the future to commercial entities, perhaps by to monitor, synthesize and rate research. The
providing specially developed sites, or providing current system of research funding in America
branded services. The future of the Internet, and places research at the whim of special-interest
especially the ability to deliver high bandwidth, private funding and sometimes misdirected public
will increasingly rely upon specialized sites with funding in overly popular or extremely esoteric

31
 Gastrointestinal Motility Online Educational Endeavor

areas. Regulating and directing research might reFerences

also help avoid repeating inconclusive research,
which does not get published (and thus may be Alexa Internet. (2007a). Three-month traffic statis-
repeated). tics for wikipedia.org. Retrieved from http://www.
alexa.com/data/details/main?q=&url=wikipedia.
org.
conclusIon
Alexa Internet. (2007b). Top Sites United States.
Retrieved from http://www.alexa.com/site/ds/top_
Gastrointestinal Motility Online is an evolv-
sites?cc=US&ts_mode=country&lang=none.
ing knowledge base related to Gastrointestinal
Motility disorders. The current phase of the American Library Association. (1989). Presiden-
endeavor focuses on the collection and organiza- tial Committee on Information Literacy - Final
tion of knowledge from many different sources. Report. Retrieved from http://www.ala.org/ala/
Knowledge-mining tools are being developed to acrl/acrlpubs/whitepapers/presidential.htm.
utilize this information as it becomes available
Antman, E., Lau, J., Kupelnick, B., Mosteller, F.,
to add fast, relevant access and other utilities to
& Chalmers, T. (1992). A comparison of results
the information repository.
of meta-analyses of randomized control trials and
The continuous change in standards of care and
recommendations of clinical experts. Journal of
knowledge due to rapid discoveries in the basic
the American Medical Association, 268, 240-248.
and clinical sciences prompts for a system that is
more flexible than a textbook, while demanding Aronson A. (2001). Effective mapping of bio-
thoroughness and accuracy. Knowledge mining medical text to the UMLS Metathesaurus: The
tools and other advanced technologies to aid in the MetaMap program. Proceedings of the 2001
conversion and integration of articles and research AMIA Symposium, pp. 17-21.
into the mainstream science are being integrated
Carlson, C. (2003). Information overload, retrieval
into Gastrointestinal Motility Online, and look to
strategies and Internet user empowerment. In:
impact the breadth and speed of knowledge-base
L., Haddon (Ed.), The Good, the Bad and the Ir-
upgrades. Gastrointestinal Motility Online serves
relevant (COST 269) 1(1) (pp. 169-173). Helsinki,
to balance the needs of its user base while embrac-
Finland.
ing the academic rigor in a novel application of
technology to the science of medicine. CEBM. (2007). Why the sudden interest in EBM?.
Retrieved from http://www.cebm.utoronto.ca/
intro/interest.htm.
acKnowledgMent
Computing in the Physician’s Practice. (2000).
Retrieved from http://www.harrisinteractive.com/
The authors would like to thank Dr. Raj Goyal
harris_poll/index.asp?PID=58.
for his invaluable input and contribution to GI
Motility Online and this article. The authors also Crawford, W. (2006). Why aren’t ebooks
thank Richard Martin for his helpful comments more successful?. Retrieved from http://
and proofreading of the text. www.econtentmag.com/Articles/ArticleReader.
aspx?ArticleID=18144.
Davis, D., Thomson, M., Oxman, A., & Haynes,
R. (1997). Changing physician performance: A

32
Gastrointestinal Motility Online Educational Endeavor

systematic review of the effect of continuing medi- Lyman, P. & Varian, H. (2003). How much infor-
cal education strategies. Journal of the American mation?. Retrieved from http://www2.sims.berke-
Medical Association, 274, 700-705. ley.edu/research/projects/how-much-info-2003/
index.htm.
Fridman, S. (2000). Doctors lag when it comes to
computer use - Industry trend or event. Retrieved Mechaca, L. (2007). Goodbye, Axim. Mes-
from http://www.findarticles.com/p/articles/ sage posted to http://direct2dell.com/one2one/
mi_m0HDN/is_2000_March_28/ai_60907019. archive/2007/04/11/11397.aspx.
Giles, J. (2005). Internet encyclopedias go head Sackett, D. & Straus, S. (1998). Finding and
to head. Retrieved from http://www.nature.com/ applying evidence during clinical rounds: The
news/2005/051212/full/438900a.html. evidence cart. Journal of the American Medical
Association, 280, 1336-1338.
Google. (2007). Search returned http://www.
google.com/search?hl=en&q=GI+Motility. Sackett, D. (1997). Using evidence-based medicine
to help physicians keep up-to-date. The Journal
Gore, A. (1992). Infrastructure for the global vil-
for the Serials Community, 9, 178-181.
lage. Scientific American, 265,150-153.
Sanger, L. (2001). Let’s make a wiki. Message post-
Guyatt, G. Cairns, J., Churchill, D., Cook, D.
ed to http://web.archive.org/web/20030414014355/
Haynes, B., Hirsch, J., & et al. (1992). Evidence-
http://www.nupedia.com/pipermail/nupedia-
Based Medicine Working Group: Evidence-based
l/2001-January/000676.html.
medicine. A new approach to teaching the practice
of medicine. Journal of the American Medical Sarnikar, S., Zhao, J., & Gupta, A. (2005). Medical
Association, 268, 2420-2425. information filtering using content-based and rule-
based profiles. Proceedings of the AIS Americas
Hawking, S. (2001). The universe in a nutshell.
Conference on Information Systems (AMCIS
New York, NY: Bantam Books.
2005), Omaha, NE.
Haynes, R. (1993). Where’s the meat in clini-
Sarnikar, S. & Gupta, A. (2007). A context-specific
cal journals [editorial]? ACP Journal Club, 119,
mediating schema approach for information ex-
A22-A23.
change between heterogeneous hospital systems.
IMDB. (2007). How/where do you get your in- Forthcoming in International Journal of Health-
formation? How accurate/reliable is it? Retrieved care Technology and Management.
from http://imdb.com/help/show_leaf?infosource.
Schwartz, B. (2004). The paradox of choice. New
Kasper, D., Fauci, A., Longo, D., Braunwald, York, NY: HarperCollins Publishers.
E., Hauser, S., & Jameson, J. (2005). Harrison’s
Senge, P. (1994). The fifth discipline: The art and
principles of internal medicine, 16th ed. New
practice of the learning organization. New York,
York, NY: McGraw-Hill.
NY: Doubleday/Currency.
Kumar, A. (2005). Ontology-driven access to
Sharma, R. (2005). Automatic integration of text
biomedical information (ODABI). Undergraduate
documents in the medical domain. Undergraduate
Thesis, University of Arizona.
Thesis, University of Arizona.
Lindberg, D. (2003). NIH: Moving research from
Slawson, D., Hauck, F., Strayer, S., & Rol-
the bench to the bedside. Presentation to the Sub-
lins, L. (2007). What is information master?.
committee on Health.
Retrieved from http://www.healthsystem.vir-

33
 Gastrointestinal Motility Online Educational Endeavor

ginia.edu/internet/familymed/docs/info_mastery. Wikipedia. (2007a). Information retrieval. Re-

cfm#Information. trieved from http://en.wikipedia.org/wiki/Infor-
mation_retrieval.
Tufte, E. (2001). The visual display of quantita-
tive information. Cheshire, CT: Graphics Press. Wikipedia. (2007b). Wikipedia. Retrieved from
http://en.wikipedia.org/wiki/Wikipedia.
Vanhecke, T., Barnes, M., Zimmerman, J., &
Shoichet, S. (2006). PubMed vs. HighWire Press: A
head-to-head comparison of two medical literature
search engines. [Epub ahead of print] Computers
in Biology and Medicine.

This work was previously published in International Journal of Healthcare Information Systems and Informatics, Vol. 3, Is-
sue 1, edited by J. Tan, pp. 24-43, copyright 2008 by IGI Publishing (an imprint of IGI Global).

34
 35

Chapter 3
Envisioning a National
e-Medicine Network
Architecture in a
Developing Country:
A Case Study

Fikreyohannes Lemma
Addis Ababa University, Ethiopia

Mieso K. Denko
University of Guelph, Canada

Joseph K. Tan
Wayne State University, USA

Samuel Kinde Kassegne

San Diego State University, USA

aBstract

Poor infrastructures in developing countries such as Ethiopia and much of Sub-Saharan Africa have
caused these nations to suffer from lack of efficient and effective delivery of basic and extended medical
and healthcare services. Often, such limitation is further accompanied by low patient-doctor ratios, result-
ing in unwarranted rationing of services. Apparently, e-medicine awareness among both governmental
policy makers and private health professionals is motivating the gradual adoption of technological in-
novations in these countries. It is argued, however, that there still is a gap between current e-medicine
efforts in developing countries and the existing connectivity infrastructure leading to faulty, inefficient
and expensive designs. The particular case of Ethiopia, one such developing country where e-medicine
continues to carry significant promises, is investigated and reported in this article.

Copyright © 2011, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 Envisioning a National e-Medicine Network Architecture in a Developing Country

IntroductIon within a multi-provider care management context;

(d) minimizing long distance travels among rural
Healthcare consumers in general tend to seek ac- people in need of proper medical care to urban
cess to affordable health services that will meet areas or the capital city; and (e) providing medi-
their needs. From an ethical standpoint, healthcare cal information to clinical practitioners that will
has to be available when and where consumers help them keep abreast of clinical breakthroughs
need it; physical separation between consumers as well as new technological advances.
and healthcare facilities must not pose severe For Ethiopia, a lesser developing country with
limitations on the delivery of efficient healthcare, significant challenges in meeting basic healthcare
even if patients are located in remote areas. In needs, it is argued that e-medicine development is
this sense, information and communications emerging and can be fruitfully cultivated over the
technology (ICT) has been demonstrated to of- coming years if a vision and long-term strategy
fer a competitive choice for accessing affordable for this technology can be used to help increase
and effective health services, especially when the number of citizens receiving care and decrease
access is difficult and limited (Horsch & Bal- the subsequent healthcare costs. This article lays
bach, 1999; Kirigia, Seddoh, Gatwiri, Muthuri, out such a vision and strategy for a nationwide
& Seddoh, 2005; Tan, Kifle, Mbarika, & Okoli, e-medicine infrastructure to be designed. It is
2005). More recently, with the continued maturity organized as follows. First, the background and
of network such as Integrated Services Digital various design considerations for a nationwide
Network (ISDN) and Asynchronous Transfer e-medicine network is presented. Next is an over-
Mode (ATM) networks and related technologies view of the requirements for the network design
(Perednia & Allen, 1995; Tan, 2001), e-medicine followed by a more in-depth description of the local
implementation has entered a stage where both the and wide area network (LAN/WAN) architecture
health providers and consumers can now benefit envisioned. The focus of the discussion will then
significantly. shift to the existing Broadband Multimedia Net-
IT-based horizontal and vertical communica- work (BMN) and Very Small Aperture Terminal
tions among the healthcare facilities within the (VSAT) infrastructures and how these networks
organizational structure of the healthcare system may be integrated into the nationwide e-medicine
are essential. Such communications facilitate ef- infrastructure. Finally, the article concludes with
ficient information exchange and help the delivery insights into potential future work in e-medicine
of essential health services to underserved rural for developing countries.
areas. These communications can be supported
through a nationwide e-medicine network that is
based on affordable telecommunications infra- e-MedIcIne networK
structure. The network should connect all regional desIgn consIderatIons
clinics to urban area hospitals. The benefits of
such a network include: (a) establishing reliable E-medicine refers to the electronic delivery of
horizontal-vertical communications and informa- healthcare and sharing of medical knowledge over
tion sharing among facilities, thereby driving a distance employing ICT. A national e-medicine
up quality, improving efficiency, and enhancing network allows sharing and exchanging of clini-
cost-effectiveness of services; (b) achieving cal data among physicians, administrators, even
e-health commitments and bringing healthcare patients or other participating health profession-
closer to underserved and un-served rural areas; als regardless of physical distance separation or
(c) strengthening collaboration among hospitals geographical terrain of the whereabouts of these

36
Envisioning a National e-Medicine Network Architecture in a Developing Country

network participants within the national boundar- network architecture

ies. The network also facilitates communications
among physicians and academics across diverse The design of suitable national e-medicine network
cultures, affiliated healthcare organizations, and architecture in a lesser developing country such
publicly or privately funded research institutions. as Ethiopia requires several key components to be
Since there is lack of transportation and commu- integrated into a dynamic and enterprising com-
nication infrastructure in developing countries, munication infrastructure. Major architectural
medical and clinical data exchanges can be fur- components include: (1) LAN architecture for
ther secured and facilitated through an existing local networking and sharing of health-related
e-medicine network. information; here, communications may be
In developed nations, e-medicine services established using wireless cellular or ordinary
(Wright, 1998) can benefit remote locations that fixed telephone lines; (2) WAN architecture for
may not be easily accessed due to unpredict- national networking and sharing of health-related
able or harsh weather conditions found during data; this will allow communications among local
certain times of the year, for example, parts of and national physicians, healthcare workers and
North America and Scandinavian countries are clients covering urban and rural communities; and
often heavily affected by snow and other natural (3) designing a suitable back-end database and
hazards such as avalanches, falling boulders front-end user interface applications that integrate
and closure of highways due to multi-vehicle ac- seamlessly with the (prototype) implementation
cidents or other calamities. Mountainous terrain of the proposed architecture. The overall goal of
in certain parts of North American regions such the nationwide e-medicine network architecture
as Alaska, British Columbia, Alberta and New is then to provide an affordable and a low-cost
Territories implies the need for viable distance system that facilitates uninterrupted communica-
healthcare delivery solutions. E-medicine allows tions among physicians and health professionals
health professionals around the world to establish across the country. The system bolsters connectiv-
faster communication and exchange information ity among rural clinics and urban area hospitals
with clients and regional authorities irrespective to support primarily clinical e-consultation and
of geographical locations. It may also support maintenance of stored patient records.
rural dwellers to get healthcare services delivery This network should also be cost-effective,
similar to their urban counterparts. A mobile expandable, and secure. It must support a state-
e-medicine system, for instance, provides a con- of-the-art ICT access schema and connectivity
venient platform for acquisition, transmission to rural area clinics. Existing ICT infrastructure
and delivery of health-related data to healthcare will be given priority to minimize the cost of
providers through 2G/3G-based wireless networks implementing the nationwide network. In the
(Wootton, 2001). Recognizing these benefits, the Ethiopia design, expandability is a concern. First,
International Telecommunications Union (ITU) few hospitals are built in the country while more
has set a global agenda to promote e-medicine clinics are being added every year. Moreover,
applications in developing countries. Ethiopia, there is a chance to incorporate private hospitals
one of the beneficiaries of such an initiative, has in the nationwide e-medicine network as and
commissioned some ICT projects such as School- when necessary, which will further increases
Net, WoredaNet and BMN to enable fully-fledged the number of future connected sites. As well,
connectivity to make better use of the ICT in the the area of e-medicine applications will not be
health and education sectors. limited to just some specific diseases, but will
be expected to increase in type and number over

37
 Envisioning a National e-Medicine Network Architecture in a Developing Country

the long haul. In fact, the network should also infrastructure is now considered a cost-effective
support advanced applications, which require solution. Of course, set-up costs depend on the
real-time connectivity such as videoconferencing type of WAN to be used—to ensure low instal-
capabilities for future use. lation cost, it is proposed that the network design
During e-consultation or patient referral, will incorporate an existing WAN provided by the
most of the data exchanged over the network are Ethiopia Telecommunication Corporation (ETC).
sensitive patient information. Confidentiality of In Ethiopia, most of the inter-hospital commu-
patient information must therefore be respected. nications are traditionally dependent on telephone
For secure communications, protocols such as and hand-delivered referral messages. During
Secure Socket Layer (SSL) could be used. SSL referrals patients have to travel afar to one of the
ensures secured communications over web-based urban referral hospitals, carrying the written mes-
applications and provides the ability to safely sages of the referring physician. Clinics located in
exchange patient information across the network the telephone coverage areas communicate using
(Elmasri, 2000). When doctors exchange patient telephone to exchange information about avail-
information, they could adhere to medical protocol ability of specialist(s) or bed in another hospital.
that defines the rules to be followed during this Yet, the communication needs of hospitals have
process. In addition, the network and accom- grown over the years ahead of its technological
panying servers could be protected by firewall capabilities. Geographically dispersed clinics
against hacking from external parties. Firewalls lack modern telecommunication technology ac-
are software or hardware for the sole purpose of cess. Among them are instantaneous access to
keeping digital pests such as viruses, worms, and patient information, access to electronic medical
hackers out of the network (http://www.cisco.com, records, and access to the Internet. These and
Tanenbaum, 2004). other communication needs of health providers
also require the development of e-medicine ap-
plication software backed by electronic patient
networK desIgn record systems. Design of such communication
reQuIreMents networks will also require the understanding of
organizational structure of the clinics involved
As cost must be one of the driving factors for choos- in the network.
ing among existing or emerging ICT infrastruc- Since the government/public clinics are owned
tures in the country, implementing nationwide and organized under their respective regions, the
e-medicine network infrastructure may seem at WAN design should follow the organizational
first to be more expensive than building clinics structure of the administrative regions in the
or supplying existing regional clinics with medi- country. A detailed study about the inclusion
cal personnel and equipment. Yet, a cost-benefit of various clinics, their locations relative to the
analysis comparing various IT investment ap- nearest access point to existing ICT infrastruc-
proaches will provide best directions to achieving ture, traffic load and its characteristics, security,
a lower cost solution to the problem of delivering LAN/WAN protocol, topology and bandwidth
adequate and proper healthcare and disseminat- requirements and utilization, and allocation
ing confidential health information to and from of bandwidth, among other issues, have to be
various connecting points throughout the country. considered while trying to design a nationwide
With today’s oil prices at a premium, network e-medicine network architecture. For example,
connectivity among the healthcare facilities, both issues of communicating patient information elec-
in the urban and rural areas over an existing ICT tronically may further raise question on medical

38
Envisioning a National e-Medicine Network Architecture in a Developing Country

ethics, the need for developing standard medical Physical Locations of Offices and Users
protocols, and detailing policies for use in routine to be Connected on Campus
activity via the e-medicine network.
The sample hospital (Tikur Anbassa Specialized
lan architecture Hospital) is housed in a series of five buildings
(Blocks A-E). These blocks are not physically
To design the LAN for each hospital, we consider separated. Even though precise figures were un-
the central site, Tikur Anbassa Specialized Hospi- available, these five buildings are built on roughly
tal located in the capital Addis Ababa, as a model. 8,000 to 10,000 square meters. While the main
The hospital is organized into 16 departments with offices and departments in the hospital are located
each department further divided into smaller units in one of the respective blocks, most of these offices
as necessary. For instance, the Internal Medicine are in either of the first two stories of the block
department has several units such as the Renal they belong. Having routers switches in each of
Unit, the Cardiology Unit, the Neurology Unit, the departments is ideal to design a high-speed
and other units. Physicians in these departments and expandable LAN, but it will also make the
and units need to communicate whenever a patient design expensive to install, support and maintain.
visits more than one of the units. It is proposed A more cost-effective approach is to put switches
that the LAN follows the hierarchical structure per building and then get the departments to be
of the hospital. connected into various groups by using Virtual
The decision to make the selection between LAN technology.
various LAN technologies was based on: (a)
expected application to run on the network and The Rate of Network Growth
their traffic patterns; (b) physical locations of the
offices and users to be connected in campus; (c) The rate of the hospital LAN growth depends
the rate of network growth; (d) the abundance on the level of computerization in the hospital.
of the network technology in the market; and (e) Currently in this central hospital site, there is a
simplicity of installation and maintenance. Each of LAN that connects a few offices and a computer
these criteria will now be explored in more depth. room. The network employs star topology, using
a centrally located hub and Unshielded Twisted
Expected Application to Run on the Pair (UTP) cables forming a peer-to-peer LAN.
Network and their Traffic Patterns The purpose of this LAN was to enable offices to
share printer and students to get access to research
Currently we expect a Web-based e-medicine ap- documents. In this design, it is anticipated that
plication to run on the network. The application as the use of Web-based applications becomes
will use a central database server where all the user commonplace, there will be opportunities to add
and patient information will be stored. The type more applications and connect more computers
of data to be transmitted on the network should and offices to the LAN. The switches-routers se-
accommodate both text and image formats. Since lected in this design should therefore have many
all communications are to be channeled through free ports to help cascade the growing number
the server, the traffic pattern around the center of anticipated future connections.
is expected to be heavy. Higher speed devices
should be installed at the center of the LAN where
servers will be located.

39
 Envisioning a National e-Medicine Network Architecture in a Developing Country

The Abundance of Network Technology layer (core layer); (b) second layer (distribution
in the Market layer); and (c) third layer (access layer).

Capitalizing on the abundance of emerging Core Layer

network technology in the Ethiopia market, we
gathered data from existing network technology Core layer high performance switches, capable
vendors and organizations that implement com- of switching packets as quickly as possible, are
puter networks in the capital city, Addis Ababa. to be deployed. Essentially, this layer connects
Ethernet technology is common in organizations the LAN backbone media as well as connects to
that implement computer networks, such as Ad- the outside world via a firewall through WAN. In
dis Ababa University (AAU Net). AAU Net is a this design, the devices in the core layer will be
network backed by triangular shape fiber optic placed at a central location in the hospital. The
cable connecting the three main campuses. The devices will then be connected with high-speed
topology is an extended star topology that fastens cables such as fiber optics, or fast Ethernet cables.
together fiber optic cables for vertical cabling The servers will also be connected to switches,
(backbone cabling) between buildings that house shielded by a firewall.
various faculties and departments. These back-
bone cabling provide interconnection between Distribution Layer
wiring closets and Point of Present (POP). The
zones that fall within a departmental area are Distribution layer will contain switches and rout-
served by internetworking devices such as hubs ers capable of Virtual LAN (VLAN) switching
and UTP cables. and allow defining departmental workgroups and
Interestingly, what is described so far appears multicast domains. The devices should also sup-
to be the dominant design of the small number port connectivity of different LAN technologies
of networks existing in the capital Addis Ababa. since they also serve as the demarcation point
As such, it is not surprising to find that suppliers between the backbone connections in the core
of network technology devices and support in layer and the access layer. In this hospital-based
Ethiopia are restricted to only a limited number LAN design, the distribution layer represents
of vendors. switches/routers at each building connected to
the core layer on the one end and to the access
Simplicity of Installation and layer on the other. Redundant links will be used
Maintenance for maximum availability and the departments
could be grouped forming their own Virtual LAN.
To design the LAN architecture we have therefore
selected the hierarchical model. This enables us to Access Layer
design and arrange the inter-network devices in
layers. Figure 1 depicts the hospital-based LAN Access Layer is where the end-users are allowed
architecture. into the network. This layer contains switches/
It is a model preferred by most of network hubs from which PCs in each department gain
design experts for its ease of understanding, access to the hospital-based LAN. Each depart-
expandability and improved fault isolation char- ment will have at least one switch/hub, which will
acteristics (http://www.cisco.com). The model in turn have redundant links to more than two of
encompasses the following three layers: (a) first the switches in the distribution layer.

40
Envisioning a National e-Medicine Network Architecture in a Developing Country

Figure 1. Hospital LAN design

WAN architecture To choose among these possible infrastructures

for nationwide e-medicine network, the param-
Designing the WAN architecture for a nationwide eters to be considered include the geographical
e-medicine network raises the issue of WAN ser- coverage, bandwidth, mode of communication,
vice provider. Unlike LAN, WAN connectivity rental cost of WAN connection and capacity to
depends on the availability of WAN infrastructure add more LANs. Table 1 summarizes the data
in the country. The sole WAN service provider comparing among the available ICT infrastruc-
is the Ethiopian Telecommunications Corpora- tures in Ethiopia.
tion (ETC). Based on the data, it appears that WoredaNet is
ETC provides a number of services (http:// best suited to the national e-medicine network, as
www.telecom.net.et) from which the WAN in- long as the existing infrastructures are functioning
frastructure suitable for the e-medicine network efficiently and effectively. However, as noted in
may be derived. Existing WAN services include: Table 1, there may be a tradeoff between coverage
(a) Internet Services, or, providing basic Internet and capacity, that is, when the coverage is accept-
services over dial-up or leased lines; (b) Digital able the capacity may be somewhat limited. For
Data Network (DDN), supporting dedicated Inter- example, BMN coverage is ideal as it represents
net, ISDN and frame relay services; (c) SchoolNet state-of-the-art service and higher bandwidth.
VSAT, covering services for secondary schools However, it is centered primarily in the urban
and institutes of higher learning; (d) WoredaNet areas. It is also under development and we have
VSAT, covering services for districts (Woreda) thus considered it as a potential option to be used
administrations; and (e) Broadband Multimedia when integrated with the VSAT-based networks to
Network (BMN), offering high-speed optical enhance nationwide e-medicine network. Finally,
communications to major cities. the SchoolNet needs to be upgraded to support
two-way interactivity.

41
 Envisioning a National e-Medicine Network Architecture in a Developing Country

Table 1. Summarized comparison of existing ICT infrastructure

Internet DDN SchoolNet WoredaNet BMN

About 500 schools

Telephone The capital and 571 Woredas The Capital city
covered. There are
Coverage coverage regional Urban out of 594 are and 13 regional
Woredas that do
areas only areas only covered towns.
not have schools

ADSL Services:
Maximum of Downstream/
Variable bandwidth
56k dialup Maximum of Can be upgraded to upstream
Bandwidth Downstream/ up-
and 1Mbps in 1Mbps 384k upstream 45Mbps/ 256k
stream 512k/128k
Leased line downlink
and 1024k/256k

One-way broad-
Interactivity Two-way Two-way Two-way Two-way
casting

0.11 birr/min
dialup
Free For Wore- Not yet determined,
Cost 1000 birr/ Free for schools
das under development
month leased
line

Will have more

Capacity to Not scalable than 10 ports
Not scalable Can be expanded
scale enough free at each
Woreda

Thus, one alternative approach is using a an improved WAN design will result with only one
combination of VSAT networks and terrestrial WAN connection to the urban hospitals through
BMN. VSAT-based connectivity is believed to be which the hospitals will be connected to BMN
cost-effective and in the case of WoredaNet and and the rural area clinics through the WoredaNet.
SchoolNet, it enables connectivity to the public, Figure 2b depicts the second alternative solution
even in the rural areas. In addition to serving the of the WAN design.
rural areas, it also covers urban areas. Together,
this will provide modern, convenient as well as
economical connectivity to hospitals. For improv- current etc InItIatIVes
ing state-of-the-art applications such as videocon-
ferencing, connectivity via the emerging BMN is The recent development in the ETC in providing
proposed to connect urban area hospitals in the multimedia network infrastructure is the integra-
capital city and in the regions where the network tion of the VSAT-based networks (SchoolNet and
can be easily accessible. WoredaNet) with the BMN (Tiruneh, 2006). In
Figure 2a shows the e-medicine network as a other words, these VSAT-based networks can now
first alternative. Note that this approach requires be used as a point of access to the BMN.
that urban hospitals maintain two WAN connec- As part of a longer term vision and mission
tions. Having more than one WAN connection, of broadband initiatives for socioeconomic de-
however, may become expensive in the long run. If velopment in Ethiopia, ETC has also planned an
the two WAN infrastructures could be integrated, e-health setting that tries to cover rural areas,

42
Envisioning a National e-Medicine Network Architecture in a Developing Country

Figure 2a. Logical WAN design based on BMN and VSAT networks (1st Alternative)

schools, clinics, hospitals, prisons, and nursing nationwide or worldwide; (c) infrastructure that
homes, including assisted living with several supports data, video and voice/audio (multime-
requirements: (a) high quality patient data, video dia) services; (d) high quality, secured and fast
and images to be exchanged between different delivery of medical information; and (e) high
medical institutions; (b) ICT infrastructure to speed (BW) connectivity or the deployment of
connect geographically dispersed institutions, broadband infrastructure.

Figure 2b. Logical WAN design based on BMN and VSAT networks (2nd Alternative)

43
 Envisioning a National e-Medicine Network Architecture in a Developing Country

Figure 3. ETC’s broadband-integrated infrastructure for e-health applications

Key challenges faced are the need to encom- also serving as broadband access means. Figure
pass multiple locations, to use multiple access 3 shows the two recent developments.
technologies, to deliver multiple services and to Apparently, the new developments in WAN
address multiple user markets. infrastructure support the 2nd alternative e-
medicine WAN architecture discussed here as
Current BMN Development depicted in Figure 3.

ETC has already completed building the Core

Terrestrial Broadband Infrastructure, which is ca- tHe prototype
pable of providing data, video, and voice services
with 24 Points of Presence. This infrastructure Based on the specified network requirements and
services key business sites (urban areas) and architecture, a working prototype for the national
supports Multiple Broadband Access via ADSL, e-medicine network is now presented and its
FWA, WIFI, and fiber networks. operations highlighted.
The prototype is a basic e-Medicine service
Current VSAT Development (BEMS), which provides a Web-based graphical
user interface (GUI) for health providers. BEMS
In Ethiopia, a Broadband VSAT Network platform, facilitates the information exchange between re-
which supports integrated services such as video, motely located health providers for the purpose
data, Internet, and voice on a single infrastructure, of e-consultation, as well as for maintaining
is currently in place. It has countrywide cover- electronic patient information. The traditional
age (450+ schools and 550+ woredas) as part of paper-based forms and patient cards used in the
SchoolNet and WoredaNet deployments. It is hospitals will be digitized and reproduced elec-
integrated with the core multimedia network, tronically. Web-based technology is chosen for its

44
Envisioning a National e-Medicine Network Architecture in a Developing Country

ubiquity. Using Web-based technology constitutes they are being treated; and (e) providing a list of
not only a network that can be used universally, but lab test requests to laboratory technicians and
the technology also supports system-independent allowing them to input lab test results.
platforms, thus providing access to many different
computer systems at client sites (http://java.sun. BeMs architecture
com/products/jsp). Key requirements in these cli-
ent sites are simply the availability of Web browser The BEMS architecture is built on three-tiered,
software and network connectivity. client-server architecture. The first layer is where
For a secured network, password protected the client machines run Web-browser software.
system ensures that user login is needed to access This layer is used to display the user interface
capabilities of the system. In addition, user types (Web pages) of the system and send secure HTTP
are defined so that there will be a role-based access request to the Web server in the second layer. Along
to database and system functions in BEMS. To with the Web server, application server resides in
ensure compatibility with most legacy systems, the second layer. This application server manages
a relational database is advocated for storing user the clinical business logic. The bottom layer con-
and patient information. Beyond e-mails, this tains the persistent data of the system. All data
approach allows the users to mobilize structured of patients, physicians and other communicated
information exchange among the communicating messages will be stored and maintained in this
health providers. third layer. This layer runs the database manage-
ment system (DBMS) software. Put simply, BEMS
Major Features of BeMs functions as a Web-based application connected
to a Web server to provide all the interfaces of
Basically the BEMS prototype may be conceptual- the system and that of a database server to man-
ized as a database-driven Web site with the follow- age all the knowledge and information elements
ing main features and functions: (a) providing user stored in the system. Figure 4 charts the BEMS
management services where administrator can system architecture.
register users, assigning username and password, The BEMS prototype is constructed with a
and defining user type, as well as searching and combination of open source products and freely
editing user information; (b) providing patient available software components. The Web server
management services where health providers suggested is the Apache Jakarta’s Tomcat Web
can register patients, search patients and view server (http://jakarta.apache.org) with the func-
patient information on a Rolodex-like interface,
as and when necessary; (c) providing, on the one
hand, referral systems where physicians can write Figure 4. BEMS prototype architecture
referral messages to a particular department and
Web Browser
hospital, and, on the other, a system whereby a
physician can retrieve and study the list of refer-
HTTP
rals forwarded to the department s/he is working
in and allowing the physician to write feedback Web Server Application Server
instantaneously after examining the referral
message and patient information; (d) providing
a system by which physicians can request and Mysql:JDBC
Database
schedule lab test at any hospital laboratories so
that patients can get tested in the clinic/hospital

45
 Envisioning a National e-Medicine Network Architecture in a Developing Country

tionalities as well as the mandated business rules manipulate basic entities such as users, patients,
programmed in Java (Haile-Mariam, 2002). Java and medical records.
Server Pages (JSP) is used to capture the user Each component of the medical record of a
interface and the text of Web pages (http://java. patient is an aggregation of different types of data,
sun.com/products/jsp; http://www.coreservlets. which are stored in the database. In the traditional
com). Some scripting is included on the Web pages paper-based system, the medical record of a patient
in JavaScript. JSP has a capability to import java is identified by an Out Patient Card (OPCard)
classes and run them from the Web pages when the Number, which is usually called patient record
pages are downloaded to the client machine (http:// number. OPCard is a four-page hard-paper card,
www.coreservlets.com). Unlike other server side which contains patient’s generic information, such
languages such as Active Server Pages (ASP), JSP as name, sex, age, address on the first page and
makes the system platform independent. It also a table of two columns for date and clinical note
allows users to take advantage of the full power so as to record chronologically the compilation
of java programming language which overcome of health providers’ notes. All other components
some of the limitation of other scripting languages such as laboratory test results and x-ray reports,
such as PHP (http://www.coreservlets.com). among other pieces of information, are stored
The database conceptualized is the open source inside the hard-paper card referenced by the card
MySql to back up the database driven applica- number or name of the patient. The lab test results
tion. MySql works on many different operating may contain zero or more test request forms along
system platforms and is known for its speed of with the results for Urine, Parasitology, Blood
data retrieval (http://www.mysql.com). It provides Chemistry, Hematology, Serology, Bacteriology,
Application Program Interfaces (API) for many Fine Needle Aspiration Cytology and Biopsy.
programming languages including Java. Pass- When a patient is admitted to the hospital,
words are secure because all password traffic is admission and social services information is
encrypted when connecting to the MySql server. stored. The admission data include identification
For database connectivity, we use mm.mysql information and name and address of next-of-kin,
driver, which is a Java Database Connectivity marital status, and number of siblings (children)
(JDBC) driver, from MySQL AB, implemented information, besides occupational information
in 100% native Java (http://www.mysql.com/ and other demographics, as and when provided
products/connector-j). by the patient. Subsequently, follow-up data such
as vital sign measurements, fluid balance infor-
database design Issues mation and other measures will be collected and
recorded. Order sheet, which contains a list of
BEMS needs to keep track of information about treatments to be ordered following admission, is
patient and related medical records, user’s infor- also part of the inpatient medical record. In ad-
mation, and messages for both medical referral dition to these, information about the hospitals,
requests and feedbacks. A well-designed minimal departments and laboratories are also stored and
database is needed to manage this information. A captured in respective entities.
relational database model is selected to store the To minimize connectivity cost and increase
persistent data of the system, as it could be easier system performance, a distributed database is
to manage, and provides better management for recommended. Horizontal partitioning that splits
complex query of such data (Amenssisa & Dabi, tables along rows, based on the location of patient
2003). This database is expected to maintain and and healthcare facility, is seen to be an ideal choice
in the e-medicine application that tries to create

46
Envisioning a National e-Medicine Network Architecture in a Developing Country

nationwide connectivity. Finally, to use the da- Figure 5a. Administrator’s main page within
tabase, transparent data access schemes must be BEMS
defined for applications that run over the network.

BeMs Interfaces

In this part of the discussion, the design of various

BEMS interfaces is presented. BEMS is accessed
when opening the initial Web page where user
authentication is first performed. The initial page
contains a typical login screen for specifying user-
name with authenticated password. There is no
need of menu or different buttons to be submitted
based on the user types. Since the user types are
defined in the database when the user registered,
the page corresponding to the specific user type
will automatically be opened upon successful Figure 5b. User registration page
login. Currently, administrators, physicians, and
lab technician user types are defined and all of
these user types will have their own main pages
as described below.

Administrator’s Main Page

The administrator’s main page is used for manag-

ing users. The functionalities accessible from this
page include: register new user and search user
by a combination of name, father’s name, and
user name. Figure 5a shows the administrator’s
main page whereas Figure 5b depicts the user
registration page.
The other function provided to administra- Physician’s Main Page
tors is the search user function. It is possible to
search users by keying in any combinations of The physician’s main page contains a button to
name, including father’s name and/or user name. open the patient manager page and has the capa-
Note that username is treated as “unique” in the bility of retrieving a list of referrals forwarded to
user table with the search result being quickly the department where the physician is located. To
displayed. Although not shown here, the full name open a new page, the physician is free to click on
of the search result is captured as a link. This link the manage patients button or link to one of the
then leads to a page containing the relevant user referrals. Figure 6a illustrates that the patient, Ato
information from which the administrator can Andualem Lemma, is one of Dr. Aman’s refer-
edit a particular user. rals. If Dr. Aman chooses to treat this patient in
the hospital where he is privileged, he can simply

47
 Envisioning a National e-Medicine Network Architecture in a Developing Country

Figure 6a. Physician’s main page within BEMS

Figure 6b. Patient manager page within BEMS

open the patient manager page by clicking on the registered. The patient full name is a link that
manage patients button. leads to the patient information page similar to
Figure 6b shows that the patient manager page the traditional patient card used in the hospitals.
has two options, that is, physicians can “register An example of the patient card, which opens up
new patient” or “search patient” for those who were when the full name link in the previous interface
previously registered. When selecting “register is selected, can be seen in Figure 6c.
new patient,” a patient registration form, similar The patient card contains patient’s general
to the user registration page, will be opened. In information, address information and clinical
contrast, if the physician wants to look for a patient, notes that are ordered in descending order. In
s/he can input one of the search criteria such as addition to the information displayed on the
name or record number of the patient, resulting patient card, laboratory test results and medical
in the display of a matching record number or images related to the patient are accessible by
name and the hospital where the patient was first clicking corresponding buttons from the patient

48
Envisioning a National e-Medicine Network Architecture in a Developing Country

Figure 6c. Patient card page within BEMS

card interface. The physician can add clinical paper-based environment, which will only hasten
notes, refer, and/or admit the patient. From the the processing and matching of the stored com-
physician’s main page, the other option available puterized information for them.
to the physician is to see referrals forwarded to
his or her department. This is possible by clicking Graphical User Interface
the link that opens a referral page corresponding
to the patient. The patient referral page contains Apparently, human computer interface (HCI)
the referral messages and buttons that will lead design issues are critical in determining the suc-
the user to patient information as well as a button cessful deployment and continuing use of the
that can lead to the feedback input page. BEMS prototype. As indicated, the current ap-
If the physician user wants to view the patient proach attempts to optimize the interface design
information, the view patient card button will sup- in mimicking more or less the traditional forms,
port such a function. Otherwise, if the physician documentation formats and paper-based patient
would like to give feedback to the referral using record system that the physicians have grown ac-
the feedback slip, the open feedback slip button customed to using over the years. In other words,
will serve this purpose. Basically, the feedback this ensures that the BEMS supports the habits of
slip is represented as an input form similar to a the physician users. It will also serve to preserve
traditional form for capturing feedback informa- physician user satisfaction and promote a high
tion related to the current referral. Put simply, the rate of acceptance among physician users with
idea here is to mimic within a virtual environ- the new system implementation on the one hand
ment what the physicians have already become while reducing disruption to the care processes
accustomed to routinely when working within a on the other.

49
 Envisioning a National e-Medicine Network Architecture in a Developing Country

Figure 7. Laboratory information and parasitological test request page of BEMS

Nonetheless, new system development such as scheduling information from the nursing clerk,
the BEMS typically provides new opportunities all that is needed now is a mere click of the button
for revisiting the care processes that have been put corresponding to the type of “test” required from
in place over the years. Elimination of redundant the laboratory information page. The specific lab
processes as well as the need to streamline certain test request page can be designed such as to pro-
administrative and clinical processes may be vide the physician user with a dropdown list from
warranted to improve quality, cost, efficiency and which the appropriate lab test can be immediately
effectiveness of the care provided. Online requests selected and performed as scheduled. This was
of patient and referral information and querying found to be important in order to forward the lab
of databases are expectedly translatable into more test request on a real-time basis to the other user
efficient, effective, appropriate and quality care. types called, the Lab technicians.
Use of GUI also permits substantial amount of
data to be viewed together, improving the com- Lab Technician’s Main Page
munication, exchange and sharing of patient data.
Feedback from physician users should ultimately The third type of user, laboratory technician, sees
be channeled to an even more enhanced user in- a list of laboratory requests to the department s/he
terface design. In this environment, clinical test is practicing, on the lab technician’s main page.
results and specialist reports can also be captured The list contains a link to open lab test result input
quickly and shared collaboratively among all form where the lab technician can enter his or her
relevant health providers. report following the test as shown in Figure 8.
As an example, the physician can be empow-
ered to request laboratory test results in BEMS by
viewing the patient laboratory information page, ConclusIon
which is accessible from the patient card page
by a button called “Laboratory Tests.” Figure 7 The key contribution of this effort lies in envision-
provides illustrative screenshots for the labora- ing the planning and detailing of a Web-interface
tory information page and a parasitological test for a hierarchical model-based LAN architecture
request pages. When a physician wants to request that enables the integration of the inter-network
lab tests, instead of writing a prescription to the devices in layers and a WAN architecture, both
patient and having the patient wait for further

50
Envisioning a National e-Medicine Network Architecture in a Developing Country

Figure 8. Lab technician’s main page of BEMS

fine-tuned for a developing country such as as a PC. The transceiver receives or sends a signal
Ethiopia. to a satellite transponder in the sky. The satellite
The hierarchical model adopted for the LAN is sends and receives signals from a ground station
a preferred model due to its ease of expandability computer that acts as a hub for the system. Each
and improved fault isolation characteristics. The end user is interconnected with the hub station
WAN design considers the existing VSAT-based via the satellite, forming a star topology. The
WAN infrastructure in the country, the Wore- hub controls the entire network operation. For
daNet. Even if urban areas are relatively better one end user to communicate with another, each
equipped with adequate ICT technologies such as transmission has to first go to the hub station that
Internet access and digital telephone networks, then gets retransmitted via the satellite to the other
the communication infrastructure is not well de- end user’s VSAT. VSAT can handle up to 56 Kbps.
veloped in many rural areas. These regions have More importantly, the BEMS architecture dis-
to be equipped with an access to urban areas. In cussed here is designed to integrate with a large
this context, the newly emerging state-owned, part of the existing LAN and WAN infrastructure
low-cost VSAT networks such as SchoolNet and designs. The system can then be used to facilitate
WoredaNet provide the rural areas with suitable both intra- and inter-hospital communications
means of communication with urban areas and and for all forms of information exchange. The
beyond. alternative design selected will not only improve
VSAT, an earthbound station used in satellite quality of healthcare services while protecting the
communications of data, voice and video signals, privacy, confidentiality and integrity of sensitive
excluding broadcast television, comprises two patient information, but its interfaces have been
parts: (a) a transceiver that is placed outdoors in set up to mimic the physician routines working in
direct line of sight to the satellite; and (b) a device a paper-based environment. Moreover, this will
that is placed indoors to interface the transceiver also yield opportunities for further review of the
with the end user’s communications device, such paper flow and work processes to cut down on

51
 Envisioning a National e-Medicine Network Architecture in a Developing Country

redundancies and errors while simultaneously serviced on a network such as BEMS discussed
boosting both administrative and clinical efficien- in this article (Tan, 2005).
cies and effectiveness of care.
As future work in the area of developing na-
tionwide e-medicine networks, we recommend the reFerences
following considerations: (a) the intended network
should support real-time e-consultations via video
Amenssisa, J. & Dabi, S. (2003). District-based
and audio conferencing, advocate doctor-to-
telemedicine project in Ethiopia. Ministry of
patient interactions, and facilitate remote training
Health. Addis Ababa, Ethiopia.
for health professionals; (b) it should also support
a distributed database structure, where individual Apache Jakarta Project. (2007). Retrieved from
hospitals should keep their own databases, which http://jakarta.apache.org
can be further treated as one “huge” database;
Cisco Documentation. (2007). Retrieved from
(c) the definition of standards is essential to fa-
www.cisco.com
cilitate information exchange among private and
government hospitals as well as overseas; (d) the Elmasri, R. (2000). Fundamentals of database
integration of expert systems such as case-based systems, 3rd edition. Addison Wesley.
system where doctors can query the database to get
Ethiopian Telecommunications Corporation.
experience from previously stored similar cases
(2007). www.telecom.net.et
should also be considered—such a system will aid
future physicians and residents working anywhere Haile-Mariam, A. (2002). Renaissance: Strategies
in the country to learn from past successes and/or for ICT Development in Ethiopia. M.Sc Thesis,
failures of the attending specialist(s), especially for School of Engineering Postgraduate Engineering
non-trivial and complex patient cases; and (e) the Program.
infrastructure should be independent of chosen
platform and operating systems (e.g., Windows vs. Hall, M. (2007). Core servlet and Java server
Apple) and be able to support physicians needing pages. Sun Microsystems Press, Retrieved from
to remotely monitor their patients over heteroge- http://www.coreservlets.com.
neous networks, including handheld devices in Horsch, A. & Balbach, T. (1999). Telemedicine
2G/3G mobile networks and wirelessly. information systems. IEEE Trans. Inform., Tech-
Beyond the design of a nationwide e-medicine nol. Biomed., 3, 166-175.
infrastructure, there will be a host of potential
e-medicine applications that may be supported, Java Server Pages Documentation. (2007). Re-
including, but not limited to a series of healthy life- trieved from http://java.sun.com/products/jsp.
style promotion programs such as e-consultation Kirigia, J., Seddoh, A., Gatwiri, D., Muthuri, L.,
and tracking of participation in smoking cessation, & Seddoh, J. (2005). E-health: Determinants,
weight reduction, dental health, stress reduction, opportunities, challenges and the way forward
exercise and nutritional programs and many more. for countries in the WHO African region. BMC
In this regard, one of the contributing authors is Public Health, 5, 137.
actively and precisely engaged with a growing
network of researchers in generating such a series MySQL. (2007). http://www.mysql.com.
of educational modules intended for seniors and MySQL Connector/J. (2007). Retrieved from
other population groups that will eventually be http://www.mysql.com/products/connector-j/.

52
Envisioning a National e-Medicine Network Architecture in a Developing Country

Perednia, D. & Allen, A. (1995). E-medicine tech- Tanenbaum, A. (2004). Computer networks, 4th
nology and clinical applications. Journal of the edition. Prentice Hall, Inc.
American Medical Association, 273(6), 483-488.
Tiruneh, M. (2006). ETC, broadband network
Tan, J. (2001). Health management information infrastructure for e-health. ICT-H-2006 Workshop,
systems: Methods and practical applications, 2nd Addis Ababa, Ethiopia.
edition. Gaithersburg, MD: Aspen Publishers, Inc.
Wootton, R. (2001). Recent Advances: Telemedi-
Tan, J., Kifle, M., Mbarika, V., Okoli, C. (2005). cine. British Medical Journal, 323, 557-560
E-Medicine in developed and developing coun-
Wright, D. (1998). Telemedicine and developing
tries. In J. Tan (Ed.), E-healthcare information
countries. A report of study group 2 of the ITU
systems. Jossey-Bass (A Wiley Imprint).
development sector. J Telemedicine Telecare.

This work was previously published in International Journal of Healthcare Information Systems and Informatics, Vol. 3,
Issue 1, edited by J. Tan, pp. 44-62, copyright 2008 by IGI Publishing (an imprint of IGI Global).

53
54

Chapter 4
Assessing Physician and Nurse
Satisfaction with an Ambulatory
Care EMR:
One Facility’s Approach
Karen A. Wager
Medical University of South Carolina, USA

James S. Zoller
Medical University of South Carolina, USA

David E. Soper
Medical University of South Carolina, USA

James B. Smith
Medical University of South Carolina, USA

John L. Waller
Medical University of South Carolina, USA

Frank C. Clark
Medical University of South Carolina, USA

aBstract

Evaluating clinician satisfaction with an electronic medical record (EMR) system is an important dimen-
sion to overall acceptance and use, yet project managers often lack the time and resources to formally
assess user satisfaction and solicit feedback. This article describes the methods used to assess clinician
satisfaction with an EMR and identify opportunities for improving its use at a 300-physician academic
practice setting. We administered an online survey to physicians and nurses; 244 (44%) responded.
We compared physician and nurse mean ratings across 5 domains, and found physicians' satisfactions
scores were statistically lower than nurses in several areas (p<.001). Participants identify EMR benefits
and limitations, and offered specific recommendations for improving EMR use at this facility. Methods
used in this study may be particularly useful to other organizations seeking a practical approach to
evaluating EMR satisfaction and use.

Copyright © 2011, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Assessing Physician and Nurse Satisfaction with an Ambulatory Care EMR

IntroductIon larly useful in identifying perceived problems and

making adjustments to the implementation plan
The degree of interest and momentum in further- or reallocating resources as needed (Friedman
ing the widespread adoption and use of electronic & Wyatt, 1997; Burkle, Ammenwerth, Prokosch,
medical record (EMR) (or electronic health record & Dudeck, 2001; Anderson & Aydin, 2005;
systems) is at an all time high in the United States. Brender, 2006).
Healthcare providers, purchasers, payers and Our study was designed to assess physician
suppliers are all looking to the EMR as a tool to and nurse use and satisfaction with an enterprise-
help promote quality, enhance patient safety, and wide ambulatory care EMR. We incorporated into
reduce costs. Despite this energy, recent estimates the evaluation a means of assessing physician
indicate EMR adoption rates in ambulatory care and nurses’ reactions to the human-computer
remain in the 15-20% range (Hillestad et al., interface and also solicited their input and sugges-
2005). Cost, lack of uniform interoperability tions on how to improve the system’s usefulness
standards, limited evidence showing use improves to our organization. This article summarizes the
patient outcomes and clinician acceptance are methods used, findings, and relevance to other
among the barriers to widespread EMR adoption organizations in the throes of implementing and
(Bates, 2005). Those who have overcome these evaluating EMR acceptance.
initial hurdles and made the transition from a
paper-based medical record system to an EMR
often lack the time, resources and expertise to BacKground
evaluate the system’s impact on the organization,
including clinician use and satisfaction with the The Medical University of South Carolina
system (Anderson & Aydin, 2005; Wager, Lee, (MUSC) in Charleston, South Carolina has
& Glaser, 2005). implemented an electronic medical record (EMR)
Use and satisfaction are two key measures of system, known as Practice Partner Patient Record®
the success of any information system (DeLone & in the majority of its ambulatory clinics over the
McLean, 2003), including EMR system success past few years. Although EMR use is not new to
(Anderson & Aydin, 2005). Various researchers MUSC (family medicine has used the system since
have assessed physician use and satisfaction with the early 1990s and internal medicine since the
the EMR (Sittig, Kuperman, & Fiskio, 1999; Gadd mid-1990s), it was not until February 2004 that we
& Penrod, 2001; Penrod & Gadd, 2001; Likourezos secured funding to deploy the EMR throughout
et al., 2004; Joos, Chen, Jirjis, & Johnson, 2006) the ambulatory care enterprise. When the system
and some have found that user group perspectives is fully implemented, paper medical records will
can differ even within the same institution (Wager, have been replaced, and both primary care and
Lee, White, Ward, & Ornstein, 2000; O’Connell, specialty providers will share a single electronic
Cho, Shah, Brown, & Shiffman, 2004b; Hier, medical record for their patients. This message was
Rothschild, LeMaistre, & Keeler, 2005). Assessing conveyed from the top down, starting with senior
user reaction to the human-computer interface is leadership and the dean of our medical school.
also an important dimension (Sittig, Kuperman, The EMR product itself has many of the
& Fiskio, 1999; Despont-Gros, Mueller, & Lovis, attributes of a typical EMR system, including
2005) in evaluating EMR satisfaction. Likewise, electronic health data capture, results manage-
the timing of the evaluation study is important. ment, decision-support, and electronic commu-
Conducting formative evaluation studies during nications. However, MUSC has not yet installed
enterprise EMR implementation can be particu- enterprise-wide direct order entry nor activated

55
 Assessing Physician and Nurse Satisfaction with an Ambulatory Care EMR

the preventive care reminder functions. Disease- MetHods

specific progress note templates are available for
facilitating program note entry; however, direct survey Instrument
data entry is not a requirement for using the system,
and transcribed notes can also be incorporated. We conducted preliminary interviews with
The system also includes a prescription writer, physicians and nurses regarding their views and
which allows prescriptions to be entered and experiences with the EMR, but were interested
tracked in the patient record while automatically in soliciting input from all EMR users at MUSC.
checking for interactions with other medica- Given the size of the user group, we decided to
tions and documented allergies. Patients receive adopt a survey approach. We adapted the Ques-
printed prescriptions at the time of the visit; the tionnaire for User Interaction Satisfaction (QUIS),
prescriptions are not yet sent electronically to developed by researchers at the University of
the pharmacy. Maryland, for use in this study after reviewing
We provided formal training sessions to the literature, consulting with medical informat-
physicians, nurses and administrative support; ics professionals and soliciting input from EMR
however, nurses and administrative staff received physician leaders at MUSC. The QUIS has been
at least four hours of initial training, while most tested as a valid and reliable instrument in set-
physicians only had 30-45 minutes of training. tings similar to MUSC (Chin, Diehl, & Norman,
Nurses and administrative staff were trained in 1988) and was used recently by researchers af-
small groups in a classroom setting with computer filiated with Brigham and Women’s Hospital to
workstations for each individual. Physicians were assess physician satisfaction with its internally
offered several options for training, including developed outpatient EMR (Sittig, Kuperman,
small group sessions, but the far majority opted & Fiskio, 1999). The QUIS has also validated
for one-on-one training with a clinical analyst. for online administration (Slaughter, Harper, and
Knowing that physicians’ schedules were harried, Norman, 1994). The instrument assesses user
we decided to provide more intensive training to satisfaction in five major domains, with four to
nurses and administrative staff. We felt if nurses six questions in each: (1) overall user reactions,
and administrative staff were comfortable with (2) screen design and layout, (3) terminology and
the system, they could assist physicians in their system messages, (4) learning, and (5) system
respective departments as needed. capabilities. Recognizing that participants might
The purpose of the study was to gain insight respond differently depending upon whether they
into user satisfaction and determine if there were were assessing how long it takes to open the ap-
differences in user group satisfaction, and solicit plication or navigate within it, we separated the
formal feedback on how to improve the system’s item “system speed” survey component into two
use at our organization. We defined ‘user’ as any statements—system launch speed and navigation
MUSC employee or student who had been issued speed. We asked participants to rate each question
an EMR login and password. This article describes on a scale from 0 (the lowest) to 9 (the highest)
the methods used to conduct the survey and the level of satisfaction.
results from our attending physician and nurse We included several additional items in the
user community. survey: (a) demographic data (e.g., position, age,
gender), (b) use of the EMR (whether provider
dictated notes or directly entered them into EMR),
and (c) three open-ended questions, in which par-
ticipants were asked to identify the three greatest

56
Assessing Physician and Nurse Satisfaction with an Ambulatory Care EMR

benefits or advantages and three greatest limita- results

tions or disadvantages to using the EMR and any
recommendations they have for improving the use demographics of participants and
and effectiveness of the EMR within their clinic their use of eMr system
or at MUSC. The items and scale from the QUIS
were kept intact. A draft survey was sent to a We received 244 completed surveys from attend-
pilot group of physicians, nurses and administra- ing physicians and nurses for an overall response
tive staff. Suggested changes were incorporated rate of 44%. Forty-seven (47%) of physicians and
into the final version of the survey. Institutional 38% of nurses responded. 32% of participating
Review Board approval was obtained prior to physicians and 98% nurses are women; 85% are
conducting the study. between the ages of 30-60, equally distributed
within each decade of life. Nearly 37% of physi-
user population and survey cians and 19% of nurses reported having prior
administration experience with other EMR systems, and 62%
have used the EMR at MUSC for at least one year.
We sent an e-mail message from one of the authors When physician participants were asked
(who happened to serve as chair of the Physician how often they dictate patient information that
Information Council) to all attending physicians is eventually transcribed into the EMR, 39%
and nurses listed in the EMR user database. Our reported frequently (defined as more than 50%
sampling frame included all individuals who had of the time), 4% sometimes (25-50% of the time)
been issued an EMR username and password—for and 58% reported rarely or never (defined as less
a total population of attending physicians (245) than 25% of the time) (n=109). Almost 40% of par-
and nurse EMR users (304) of 549. In the e-mail ticipants reported having paper medical records
message, participants were asked to click on pulled daily for patient visits, 14% sometimes,
an embedded hyperlink to complete the online and 44% rarely or never (n=108). 70% reported
questionnaire. We gave participants the option of that they access the EMR remotely on at least a
printing and faxing the completed survey to us weekly or daily basis.
(48 surveys were returned by fax). Two reminder
notices were sent—one week and two weeks fol- Satisfaction with EMR
lowing the initial mailing. All participants were
assured confidentiality of their responses. We assessed the internal consistency reliability of
the five domains of the QUIS part of the survey
Data Analysis (Table 1) and examined mean satisfaction scores
for attending physicians and nurses for each item
Survey responses were downloaded from the
survey provider to an electronic database and then
imported into SPSS for analysis (Windows, 2005). Table 1. Reliability of domains (Crohbach’s Alpha)
Responses to open-ended questions were
grouped and assigned to categories by the primary Domain Alpha
author, experienced in qualitative research. These Overall user satisfaction .911
were then tabulated within each of the categories Screen design and layout .859
Terms and system information .891
and discussed among the participating authors. Learning .876
System capabilities .835
Overall .961

57
 Assessing Physician and Nurse Satisfaction with an Ambulatory Care EMR

Table 2. Physician and nurse mean satisfaction scores by domain/item

Do-
Item (0-9 scale) Physician Nurse
main
Terrible…………………………………….……………………....................….Wonderful 5.07 5.92*

Difficult……………….…………………………………….……………….……….Easy 5.17 6.21*

.
Frustrating……………………………………..……………………………...…Satisfying 4.33 5.09*
Overall satisfaction

Inadequate Power............………………………………...………………..Adequate Power 4.45 5.02

Dull…………………..……………………………………………………….…Stimulating 4.65 5.39*

Rigid…………………………………………………………...……………..…….Flexible 4.42 5.41*

Overall satisfaction composite score 4.67 5.50*

Characters on the screen: Hard to read……….………………………………Easy to read 6.67 7.16

Screen Design
and Layout

Highlighting on the screen simplifies task: Not at all…………………….…....Very much 5.49 6.54*

Organization of information on screen: Confusing………….……………….…Very clear 5.45 6.17*

Sequence of screens: Confusing………………………………………………...Very clear 5.56 5.81

Use of terms throughout system: Inconsistent…………………………….……..Consistent 6.09 6.58

Computer terminology is related to the task you are doing: Never………………...Always 5.89 6.31
Terms and System

Position of messages on screen: Inconsistent………..………………….……….Consistent 6.05 6.48

Messages on screen which prompt user for input: Confusing………………….Very clear 5.42 6.14*
Information

Computer keeps you informed about what it is doing: Never……..……………….Always 4.90 5.42

Error messages: Unhelpful……………………………………………….….……...Helpful 3.22 4.19

Learning to operate the system: Difficult………………………………….………….Easy 5.23 5.98*

Exploring new features by trial and error: Difficult……………………………….….Easy 4.73 5.76*

Remembering names and use of commands: Difficult…………….……………..…..Easy 4.93 5.51

Tasks can be performed in a straight forward manner: Never………………….….Always 5.06 5.74*

Learning

Help messages on the screen: Confusing…………………………….………………Clear 4.14 5.12*

Supplemental reference materials: Confusing………………………………..………Clear 4.10 5.45*

System speed (to open or launch program): Too Slow………………………Fast Enough 2.65 2.85

System speed (to navigate within EMR, e.g. open a note): Too slow………...Fast enough 4.41 4.13
System Capabilities

System reliability: Unreliable…………………………………………………..….Reliable 4.72 5.01

System tends to be: Noisy…………………………………………..……………….Quiet 7.23 7.87*

Correcting your mistakes: Difficult…………………………………………………..Easy 5.12 4.96

Experienced and inexperienced users’ needs are taken into consideration:

Never………………………………………………………………………………Always 4.68 5.31

*significant at the p>.05 level

58
Assessing Physician and Nurse Satisfaction with an Ambulatory Care EMR

within the five domains. Overall Crohbach’s alpha perceived eMr Benefits,
was 0.961. limitations and recommendations
Attending physician and nurse mean scores are for Improved use
shown in Table 2. Items were rated on a scale of
0 to 9, with 0 being the most negative, 9 the most Participants were asked three open-ended ques-
positive. Both groups gave low ratings to system tions:
launch speed and clarity of error messages. Items
rated most positive by both groups included system • What have you found to be the three greatest
noise, ease in reading characters, consistency of benefits or advantages to using the EMR?
terms, and the clarity of messages appearing on- • What have you found to be the three great-
screen. We calculated overall satisfaction scores est limitations or disadvantages to using the
for physicians and nurses using the average of EMR?
the six items in the overall user reaction domain • What recommendations do you have for
as a proxy. Overall satisfactions scores were 4.67 improving the use and effectiveness of the
and 5.50 (t-test, p < .001) for attending physicians EMR within their clinic/area at MUSC?
and nurses, respectively. We compared attending
physician ratings of the other 22 individual items Benefits/Advantages to the EMR
with those of nurses and found that the physicians’ System
mean ratings of the EMR were statistically sig-
nificantly lower in nine of the 22 items (p<.05). Nearly 82% of physicians and 56% of nurses iden-
Using one-way ANOVA, we found no difference tified availability and accessibility of the patient’s
in overall satisfaction between experience groups, record as a major benefit or advantage to using the
less than 6 months, 6-12 months, 1-2 years, and EMR. See Table 3. Included in this category were
more than 2 years. comments related to having access to other clinic
We also ran a standard multivariate regression notes and records and having remote access to
with overall satisfaction score as the dependent the EMR. One physician commented, “Everyone
variable and the 22 remaining items as indepen- has access to the entirety of the record across all
dent variables and found two items were significant disciplines.” Another noted, “The EMR greatly
predictors of overall satisfaction—(1) task can be facilitates communication among primary care
performed in a straightforward manner (p<.001) providers and specialists by having all patient
and (2) clarity of help messages on screen (p<.01). information in one place.” Other frequently cited
benefits included (a) quality of record/documen-

Table 3.Most frequently cited benefits/advantages to using MUSC’s EMR system

Physicians Nurses
EMR Benefit
n=114 n=130
Availability and access to patient record 81.6% 56.2%
Quality of record/documentation 36.0% 14.6%
Comprehensiveness and completeness of record 33.3% 25.4%
Positive impact on efficiency 13.2% 10.0%
Continuity of care and standards of care 13.2% 10.8%
Ease of use 12.3% 15.4%
Avoidance of perils of paper 9.6% 20.0%

59
 Assessing Physician and Nurse Satisfaction with an Ambulatory Care EMR

tation (legibility, data consistency and accuracy, to edit a single note without success.” A nurse
organization of record, coding/compliance); (b) noted, “When the system is down, you are put
comprehensiveness and completeness of patient completely on hold.” Several participants also
record (e.g., medications, results from diagnostic expressed frustration with the amount of typing
tests); (c) impact on efficiency (e.g., speed, timeli- required and the time-consuming nature of en-
ness in completing notes);(d) continuity of care tering data into the system. Understanding error
and standards of care (e.g., templates, continuous messages and correcting mistakes also emerged
outpatient record); and (e) avoidance of perils as points of frustration. Comments reflecting
inherent with paper records (e.g., no lost charts, concerns including statements such as: “too much
no moving charts). trial and error”, “if you don’t already know how
to do something, you’ll never figure it out”, and
Limitations/Disadvantages to the EMR “some error messages don’t tell you how to solve
System the problem.”

Limitations or disadvantages identified by at Suggestions for Improving EMR Use at

least ten participants are listed in Table 4. The MUSC
two most-cited limitations or disadvantages
of the EMR were speed and cumbersome user Participants gave numerous written suggestions
interface. Both physicians and nurses described for improving EMR use at MUSC—73% of the
the system as being too slow. Approximately 20 physicians and 49% of nurses wrote comments.
participants mentioned that the start time to open Suggestions ran the gamut from “provide more
the application was particularly problematic. training”, “fix bugs in system”, “address speed
System downtime, including the moments when and reliability of system”, to “make it more user
the “system crashes”, “locks up or freezes” and friendly.”
requires user to “reboot” were described as frus- Several common themes emerged from the
trating by 14% of physicians and 22% of nurses. written suggestions. Some of these relate to how
One physician commented that system crashes the EMR is structured and used at MUSC and
are frustrating: “The program froze three times others relate more directly to the Practice Partner
today while I was editing the same note, losing application itself. General suggestions included
the edits each time. I spent 30 minutes trying (1) address the speed/performance issues, (2)

Table 4. Most frequently cited limitations/disadvantages to using MUSC’s EMR system

Physicians Nurses
EMR Limitation
n=114 n=130
Cumbersome user interface 38.6% 24.6%
37.3%
Speed/too slow 32.3%

Having to access multiple clinical systems or poor integration with other

17.5% 4.6%
systems
Problems with templates 11.4% 4.6%
Insufficient training and support staff 11.4% 3.8%
Downtime (including system crashes or system locking/freezing up) 14.0% 21.5%
Time-consuming to use 13.2% 14.6%

60
Assessing Physician and Nurse Satisfaction with an Ambulatory Care EMR

expand training and support personnel/resources, the EMR enabled them to finish their work much
(3) improve the user interface/templates and faster than before implementation (Likourezos et
simplify data entry, and (4) integrate more fully al., 2004); although we did not ask this specific
with other MUSC applications (e.g., scheduling, question, it may help explain why MUSC nurses
lab, radiology). generally viewed the system more positively than
physicians.
Nurses also tended to value the “avoidance of
dIscussIon the perils of paper” more often than the physicians
did, yet they did not mention availability as a major
Nearly 250 physician and nurse EMR users at benefit as often as physicians did. We categorize
MUSC participated in this study. Nurses are more these measures of availability/accessibility and
satisfied overall with the EMR system than are at- avoidance of periods with paper separately, yet
tending physicians. The greater system acceptance they are related. Nurses are responsible for ensur-
by nurses may be due to the approach MUSC took ing that the patient’s record is available at the time
in implementing the EMR enterprise-wide. Initial of the visit, thus, they experience the “grief” when
efforts were focused on nurses and administrative the record is not available. Physicians are not the
staff, because these individuals were expected to ones searching for the paper record, so they may
lead in the effort. We felt if nurses were comfort- not describe the benefit in quite the same manner.
able with the system, they could assist physicians. On the other hand, physicians were more likely
The downside of this approach is it has been to mention the benefits associated with having
time-consuming and resource-intensive to roll access to clinic notes from other physicians and
out the EMR throughout all the ambulatory care more likely to comment on the overall quality of
areas. Consequently, some physician users have the documentation. Having a more “complete”
not been fully trained on all of the system capa- picture of the patient’s care (e.g., other visit notes,
bilities—yet, they have been using the application ancillary test results) stood out as particularly im-
for two or more years. portant to physicians. The EMR benefits identified
Additionally, 58% of participating physicians by MUSC physicians and nurses are consistent
report that they rarely, if ever dictate. Earlier with earlier studies (Sittig, Kuperman, & Fiskio,
studies have shown that data entry can negatively 1999; Gadd & Penrod, 2001; Penrod & Gadd,
impact physicians’ perceptions of their time, par- 2001; Likourezos et al., 2004; Joos, Chen, Jirjis,
ticularly if physicians are not proficient typists or & Johnson, 2006).
comfortable entering information in the examina- Physicians and nurses had similar concerns
tion room with the patient (Gadd & Penrod, 2001; when asked about limitations of the system or
Penrod & Gadd, 2001; O’Connell, Cho, Shah, opportunities for improving its use at MUSC. In
Brown, & Shiffman, 2004a; Hier, Rothschild, fact, speed and performance-related concerns
LeMaistre, & Keeler, 2005; Scott, Rundall, Vogt, seemed to cast a cloud over the benefits. Physicians
& Hsu, 2005; Linder et al., 2006). and nurses identified speed and the slowness of
Researchers at Partners HealthCare System the system as a major concern—launching the
conducted a time-motion study and found no dif- system was particularly problematic. Yet, in our
ferences in primary care physician time utilization multivariate regression analysis, system speed did
before and after EMR implementation, yet the not show up as a significant predictor of overall
majority of their physicians still perceived that satisfaction. Although this finding surprised us, an
the EMR required more time than paper records earlier study at Brigham and Women’s also found
to document patient information (Pizziferri et al., that system “response time” was not correlated
2005). Nurses in an earlier study reported that with overall user satisfaction (Sittig, Kuperman,

61
 Assessing Physician and Nurse Satisfaction with an Ambulatory Care EMR

& Fiskio, 1999). Factors associated with physi- and system information (p<.01) and learning
cians’ ability to effectively use the system were (p<.001). Brigham and Women’s may have a more
more likely to be predictors of EMR satisfaction advanced training environment, and because they
than speed alone. developed the EMR in-house, they have more
In addition to their concerns with speed and flexibility in customizing the screens and layout
performance, nearly 40% of the physicians felt to accommodate their physicians’ preferences.
the user interface was cumbersome, took too No differences were found among overall user
many clicks, and made it difficult to make cor- reaction and system capabilities. The physicians
rections or changes. A common complaint was participating in the study at Brigham and Women
the “system requires too many steps or clicks to had two or more years EMR experience.
perform a simple task.” Other institutions using Our study evaluated responses from both
different EMR applications have found their relatively new and experienced users. One might
clinicians share these same concerns (Sittig, assume that due to the learning curve in moving
Kuperman, & Fiskio, 1999; Wager, Lee, White, from a paper-based system to an EMR, more
Ward, & Ornstein, 2000; Gadd & Penrod, 2001; experienced EMR users would be more satis-
Penrod & Gadd, 2001; Miller & Sim, 2004; Scott, fied with the system than less experienced users
Rundall, Vogt, & Hsu, 2005). Some of the dif- as Gamm et al. suggest (Gamm, Barsukiewicz,
ficulty or frustration at MUSC may stem from Dansky, & Vasey). We did not find that to be the
a lack of sufficient training. For example, we case in our study. EMR users with less than one
observed from the open-ended comments that year of experience were equally satisfied overall
participants are frustrated with aspects of the with the system as were those who had used the
system they may not fully understand how to use application for two or more years.
(e.g., writing prescriptions, modifying templates). Participants in this study provided a host of
Other frustrations may stem from using an EMR important suggestions for improving the EMR’s
whereby primary care providers and specialists use at MUSC. Other institutions would do well
share and record their notes in a single patient to address these issues and concerns up front to
record. Historically, each set of providers had avoid problems later.
specialty-specific records that only those in the Our study has two primary limitations. First,
specific group could view and modify. the study is limited to a single healthcare orga-
Interestingly, the results of this survey are nization and one EMR product. Thus, the results
remarkably similar to those reported by research- may not be generalizable to other healthcare
ers at Brigham and Women’s in their ambulatory organizations or those who use a different EMR
care division, despite this group’s use of an in- system. Second, response bias is a concern. We
house-developed EMR system. Using the same compared the demographics of our physician and
QUIS instrument, researchers there discovered nurse participant populations with the general
that attending physicians scored their EMR physician and nurse EMR population at MUSC
lowest in the areas of system speed (although and found no differences, yet response rate re-
they did not distinguish between launch speed mains a limitation.
and speed of navigating within the application), The real value of this study to other healthcare
helpfulness of error messages, and flexibility of organizations may not be the results per se found
system (Sittig, Kuperman, & Fiskio, 1999). The at MUSC, but rather the methods and process
pattern of responses are quite similar, although used. Using the QUIS instrument, we were able
MUSC physician reactions were less positive in to identify the aspects of our EMR system that
terms of screen design and layout (p<.01), terms stood out as problematic to clinician users—

62
Assessing Physician and Nurse Satisfaction with an Ambulatory Care EMR

launch speed, navigation speed, and the clarity in easily assessing EMR satisfaction across the
and helpfulness of error messages. Equally, if enterprise.
not more importantly, we provided all clinicians
with the opportunity to offer suggestions on how
to improve the system, and since this survey, we reFerences
have incorporated many of their suggestions into
our environment. Our leadership team has taken Anderson, J., & Aydin, C. (Eds.). (2005). Evalu-
a number of steps to address the most widespread ating the organizational impact of healthcare
concerns identified by this survey and validated information systems, 2nd ed. Springer.
through other avenues. We believe other institu-
Bates, D. (2005). Physicians and ambulatory
tions will find the methods used in this study
electronic health records. Health Affairs, 24(5),
helpful in assessing clinician satisfaction and in
1180-1189.
soliciting suggestions for improvement.
Brender, J. (2006). Handbook of evaluation meth-
ods for health informatics. Elsevier.
conclusIon
Burkle, T., Ammenwerth, E., Prokosch, H., &
Dudeck, J. (2001). Evaluation of clinical infor-
Results of this survey suggest that MUSC phy-
mation systems. What can be evaluated and
sicians and nurses recognize and value having
what cannot?. Journal of Evaluation in Clinical
access to a single electronic patient record that
Practice, 7(4), 373-385.
is shared across the ambulatory care enterprise.
However, they view the current system as less than Chin, J., Diehl, V., & Norman, K. (1988). Develop-
ideal. Speed, performance and the user interface ment of an instrument measuring user satisfaction
(e.g., need to simplify data entry) are of concern. of the human-computer interface. Paper presented
Likewise, additional training and resources are at the CHI, New Ork.
needed to more effectively support the system and
DeLone, W., & McLean, E. (2003). The DeLone
its users. Our leadership team has taken a number
and McLean model of information systems suc-
of steps to address the most widespread concerns
cess: A ten-year update. Journal of Management
identified by this survey and validated elsewhere.
Information Systems, 19(4), 9-30.
Steps to enhance system speed/performance
and reliability are being taken, and multi-modal Despont-Gros, C., Mueller, H., & Lovis, C. (2005).
programs of improved training and user support Evaluating user interactions with clinical informa-
are being implemented. We expect to observe the tion systems: A model based on human-computer
results of these initiatives and report outcomes interactions models. Journal of Biomedical Infor-
in due course. matics, 38, 244-255.
Assessing user satisfaction with the EMR is
Friedman, C., & Wyatt, J. (1997). Evaluation meth-
important in providing leadership with additional
ods in medical informatics. New York: Springer.
insight into the issues and concerns. Conducting
formal evaluation studies, however, are often not Gadd, C., & Penrod, L. (2001). Assessing physician
done because of lack of available expertise and attitudes regarding use of an outpatient EMR: A
resources. We believe that surveys such as this longitudinal, multi-practice study. Proceedings/
one can prove to be a useful resource not only to AMIA Annual Symposium, 194-198.
our healthcare organization’s leadership team, but
Gamm, L., Barsukiewicz, C., Dansky, K., &
to those in other healthcare institutions interested
Vasey, J. Pre- and post-control model research on

63
 Assessing Physician and Nurse Satisfaction with an Ambulatory Care EMR

end-users’ satisfaction with an electronic medical satisfaction with two implementations under one
record: Preliminary results. Paper presented at roof. J Am Med Inform Assoc., 11(1), 43-49.
the AMIA.
Penrod, L., & Gadd, C. (2001). Attitudes of
Hier, D., Rothschild, A., LeMaistre, A., & Keeler, academic-based and community-based physicians
J. (2005). Differing faculty and housestaff ac- regarding EMR use during outpatient encounters.
ceptance of an electronic health record one-year Proceedings/AMIA Annual Symposium, 528-532.
after implementation. International Journal of
Pizziferri, L., Kittler, A., Volk, L., Honour,
Medical Informatics, 74, 657-662.
M., Gupta, S., Wang, S., et al. (2005). Primary
Hillestad, R., Bigelow, J., Bower, A., Girosi, F., care physician time utilization before and after
Meili, R., Scoville, R., et al. (2005). Can electronic implementation of an electronic medical record:
medical record systems transform healthcare? A time-motion study. Journal of Biomedical
Potential health benefits, savings and costs. Health Informatics, 38, 176-188.
Affairs, 24(5), 1103-1117.
Scott, J., Rundall, T., Vogt, T., & Hsu, J. (2005).
Joos, D., Chen, Q., Jirjis, J., & Johnson, K. Kaiser Permanente’s experience of implementing
(2006). An electronic medical record in primary an electronic medical record: A qualitative study.
care: Impact on satisfaction, work efficiency and BMJ, 331, 1313-1316.
clinic processes. Paper presented at the AMIA,
Sittig, D., Kuperman, G., & Fiskio, J. (1999).
Washington, DC.
Evaluating physician satisfaction regarding user
Likourezos, A., Chalfin, D., Murphy, D., Sommer, interactions with an electronic medical record
B., Darcy, K., & Davidson, S. (2004). Physician system. Paper presented at the AMIA.
and nurse satisfaction with an electronic medical
Slaughter, L.A., Harper, B.D., and Norman, K.L.
record system. Journal of Emergency Medicine,
(1994). Assessing the equivalence of the paper and
27(4), 419-424.
online formats of the QUIS 5.5. In Proceedings
Linder, J., Schnipper, J., Tsurikova, R., Melnikas, of Mid Atlantic Human Factors Conference, (pp.
A., Volk, L., & Middleton, B. (2006). Barriers to 87-91)Washington, DC.
electronic health record use during patient visits.
Wager, K., Lee, F., & Glaser, J. (2005). Managing
Paper presented at the AMIA, Washington, DC.
healthcare information systems: A practical ap-
Miller, R., & Sim, I. (2004). Physicians’ use of proach for healthcare executives. San Francisco,
electronic medical records: Barriers and solutions. CA: Jossey-Bass.
Health Affairs, 23(2), 116-126.
Wager, K., Lee, F., White, A., Ward, D., & Orn-
O’Connell, R., Cho, C., Shah, N., Brown, K., & stein, S. (2000). Impact of an electronic medical
Shiffman, R. (2004a). Take note: Differential EHR record system on community-based primary care
satisfaction with two implementations under one practices. The Journal of the American Board of
roof. J Am Med Inform Assoc., 11(1), 43-49. Family Practice, 13(5), 338-348.
O’Connell, R., Cho, C., Shah, N., Brown, K., & Windows, S. (2005). Version 14.0.2005. Chicago:
Shiffman, R. (2004b). Take note: Differential EHR SPSS Inc.

This work was previously published in International Journal of Healthcare Information Systems and Informatics, Vol. 3, Is-
sue 1, edited by J. Tan, pp. 63-74, copyright 2008 by IGI Publishing (an imprint of IGI Global).

64
 65

Chapter 5
Information Technology (IT) and
the Healthcare Industry:
A SWOT Analysis

Marilyn M. Helms
Dalton State College, USA

Rita Moore
Dalton State College, USA

Mohammad Ahmadi
University of Tennessee at Chattanooga, USA

aBstract

The healthcare industry is under pressure to improve patient safety, operate more efficiently, reduce
medical errors, and provide secure access to timely information while controlling costs, protecting
patient privacy, and complying with legal guidelines. Analysts, practitioners, patients and others have
concerns for the industry. Using the popular strategic analysis tool of strengths, weaknesses, oppor-
tunities, and threats analysis (SWOT), facing the healthcare industry and its adoption of information
technologies (IT) are presented. Internal strengths supporting further industry investment in IT include
improved patient safety, greater operational efficiency, and current investments in IT infrastructure.
Internal weaknesses, however, include a lack of information system integration, user resistance to new
technologies and processes, and slow adoption of IT. External opportunities including increased use of
the Internet, a favorable national environment, and a growing call for industry standards are pressured
by threats of legal compliance, loss of patient trust, and high cost of IT.

Copyright © 2011, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 Information Technology (IT) and the Healthcare Industry

The healthcare industry faces many well- MetHodology

recognized challenges: high cost of operations,
inefficiency, inadequate safety, insufficient access By categorizing issues into strengths, weaknesses,
to information, and poor financial performance. opportunities, and threats, SWOT analysis is one
For years, many have called for a fundamental of the top tools and techniques used in strategic
change in the way healthcare is delivered. And planning (see Glaister & Falshaw, 1999). SWOT
while there is yet no clear picture of what this assists in the identification of environmental
change will be, many believe a paradigm shift relationships as well as the development of suit-
in healthcare is imminent and that information able paths for countries, organizations, or other
technology (IT) is the catalyst. entities to follow (Proctor, 1992). Valentin (2001)
Increasingly, IT is seen as a way to promote suggests SWOT analysis is the traditional means
the quality, safety, and efficiency of healthcare for searching for insights into ways of crafting
by bringing decision support to the point of care, and maintaining a fit between a business and its
providing vital links and closing open loop sys- environment. Other researchers (see Ansoff, 1965;
tems, and allowing routine quality measurement to Porter, 1991; and Mintzberg, Ahlstrand, & Lam-
become reality. IT can not only reduce operating pel, 1998) agree SWOT provides the foundation
costs, but IT can also ensure a reduction in the to gather and organize information to realize the
number of medical errors. IT in the healthcare desired alignment of variables or issues. By listing
industry provides new opportunities to boost favorable and unfavorable internal and external
patient confidence and reinforce patient trust in issues in the four quadrants of a SWOT analysis,
caregivers and healthcare facilities. With health planners can better understand how strengths
insurers feeling pressure from all directions (new can be leveraged, realize new opportunities, and
regulations, consumers, rising medical costs), understand how weaknesses can slow progress
IT is an even more important asset for carriers or magnify threats. In addition, it is possible to
(Balas, 2000). postulate ways to overcome threats and weak-
When compared to other information intensive nesses (e.g., Hofer & Schendel, 1978; Schnaars,
industries, healthcare organizations currently in- 1998; Thompson & Strickland, 1998; McDonald,
vest far less in IT. For many years, the healthcare 1999; and Kotler, 2000).
industry has experienced only single digit growth SWOT has been used extensively to aid in
in terms of IT investment (Gillette, 2004). As a understanding a variety of decisions and issues
result, current healthcare systems are relatively including: manufacturing location decisions
unsophisticated compared to those in industries (Helms, 1999); penetration strategy design for
such as banking or aviation. With the many is- export promotions and joint ventures (Zhang &
sues and variables surrounding healthcare’s IT Kelvin, 1999); regional economic development
investment, a framework for better understanding (Roberts & Stimson, 1998); entrepreneurship
of the current situation is needed before more (Helms, 2003); performance and behavior of
improvements and enhancements can result. This micro-firms (Smith, 1999), and strategic plan-
article draws upon a comprehensive framework ning (Khan & Al-Buarki, 1992). Hitt, Ireland,
from the strategic planning literature to compile Camp, and Sexton (2001) suggest that identifying
and summarize the major issues facing IT and and exploiting opportunities is part of strategic
the healthcare industry. planning. Thus, SWOT analysis is a useful way
to profile the general environmental position of
a new trend, technology, or a dynamic industry.
By using SWOT analysis, it is possible to apply

66
Information Technology (IT) and the Healthcare Industry

strategic thinking toward the implementation of all levels of the healthcare industry. The strategic
IT in healthcare. By examining the internal and initiative to increase the role of IT in healthcare
external factors interacting both for and against can advance the cause of greater patient safety
IT in healthcare, healthcare providers and supply by enhancing the quality of that care. With com-
chain organizations can formulate a strategic IT prehensive data available in a timely manner,
plan for developing their information resources healthcare providers can make better decisions
over the next several years. By uncovering and about their patients’ care, thereby reducing errors
reviewing the issues, policy makers can enact due to incomplete or insufficient information at
changes to make the process of IT implementation the point of decision (Goldberg, Kuhn, & Thomas,
easier while simultaneously working to change 2002). Lenz (2007) agrees IT has a huge potential
the culture to foster IT benefits for institutions to improve the quality of healthcare and that this
and the patients and other stakeholders they serve. aspect has not been fully explored by current IT
In SWOT analysis, strengths act as leverage solutions. Advanced process management tech-
points for new strategic initiatives, while weak- nology is seen as a way to improve IT support for
nesses are limiting factors. Specifically applied healthcare processes by improving the quality of
to IT in healthcare, strengths should indicate those processes.
areas where either IT or healthcare is particularly The Joint Commission on Accreditation of
strong, (i.e., the technical skills of IT professionals Healthcare Organizations (JCAHO) established
or the quality of existing healthcare information the Healthcare Information Technology Advisory
systems). Weaknesses should display areas where Panel in 2005 to focus attention on the improve-
either IT or healthcare requires improvement, ment of patient safety and clinical processes as new
and may range from personnel issues within IT healthcare information systems are implemented.
to limited healthcare applications beyond routine Members of the panel include researchers, physi-
transaction processing. The threats and oppor- cians, nurses, chief information officers, educators
tunities identified during the external analysis and leaders of healthcare organizations, as well
should be both factual and attitudinal issues that as representatives from the Office of the National
must be addressed in any strategic plan being Coordinator for Health Information Technology,
formulated, and should include both healthcare the American Health Information Management
and IT issues (Martin, Brown, DeHayes, Hoffer, Association, the Agency for Healthcare Research
& Perkins, 2005). The following section presents and Quality, the Veterans Health Administration,
the internal strengths and weaknesses currently and the Healthcare Information and Management
confronting the implementation and proliferation Systems Society. The panel was formed to recom-
of IT in healthcare. mend ways JCAHO’s accreditation process and
the widespread use of technology can be used to
help re-engineer patient care delivery and result
Internal strengtHs in major improvements in safety, quality and ef-
ficiency. The panel was also charged with the task
Improved patient safety of examining such topics as the effect of electronic
health records on performance benchmarking
Patient safety, as expressed in the Hippocratic and public reporting capabilities. Based on the
Oath (Classical Version)—“I will keep them from panel’s recommendations, JCAHO will evaluate
harm and injustice”—is an underlying principle its strategic plan and future direction relative to
of professional healthcare throughout the world. healthcare information technology (Anonymous,
Improving patient safety is a primary objective at 2005b).

67
 Information Technology (IT) and the Healthcare Industry

Two examples of existing information tech- more than three billion prescriptions yearly and
nologies and computer-based information systems pharmacists have to call these physicians 150
contributing to improvements in patient safety million times a year because they cannot read
through better quality of care and reduction of or understand the prescriptions, e-prescribing
errors are smart cards (and/or compact discs) could reduce injuries from medication errors
and computerized physician order entry (CPOE) (Brodkin, 2007).
systems. Smart cards containing a patient’s entire Technology-enabled improvements could also
medical history can be designed to be accessible aid disease prevention and management. Other
only by devices in a hospital, doctor’s office, or benefits could include lowering age-adjusted
other medical facility. They not only eliminate mortality by 18% and reducing annual employee
the problems of lost and comprised hard copies sick days. Lieber (2007) stresses shared experi-
of patient records, but also enable more secure ences regarding pandemic diseases can provide
electronic transfers of patient information to other the best solutions and this is aided by IT solutions
healthcare providers and insurers (Anonymous, and global information exchange.
1997). The technology for Java-based cards can When medical records are available electroni-
securely support applications for multiple health- cally, patients too can have access to their personal
care facilities and be combined with biometric health records. Five large U.S. employers have
measures for identification purposes (Sensmeier, funded an institute where their current and retired
2004). With complete information available, employees and their families can have access to
physicians are able to make better decisions for and maintain their lifelong personal health records
the care of the patient, and order appropriate tests (Five Large Companies, 2007). With access to
and treatments (Goldberg et al., 2002). longitudinal and comprehensive records, patient
Several studies indicate medication errors safety can continue to improve.
are the most likely type treatment error to occur A recent example of IT and improved patient
because drug therapy is one of the most widely safety is the North Mississippi Medical Center
used interventions in healthcare (Kohn, 2001). (NMMC) in Tupelo, Mississippi. Serving 24 rural
Computerized physician order entry (CPOE) sys- counties, NMMC is the largest rural hospital in
tems eliminate transcription errors and can warn the country and the 2006 winner of the Malcolm
of allergies and drug interactions. Such systems Baldrige National Quality Award in the Healthcare
reduce errors by more accurately dispensing the Category. NMMC’s recognition was due largely
correct dosage of the correct medication for the to their success in utilizing IT. Patients’ electronic
correct patient (Bates, Teich, Lee, Seger, Kuper- medical records can be accessed by nurses, by
man et al., 1999;Kuperman, Teich, Gandhi, & partner community hospitals, by physicians in
Bates, 2001; Mekhjian, Kumar, Kuehn, Bentley, their offices, and even by specialists and primary
Teater et al., 2002; Order Entry Rules, 2002; care providers in remote sites, reducing medical
Scalise, 2002; Shane, 2002). errors and duplication of effort. These enhance-
Computerized Physician Order Entry (CPOE) ments have earned NMMC the distinction of one
systems reduce medication errors by 80%, and of the most wired facilities in the country. The
errors with serious potential patient harm by 55% organization has a shared radiography informa-
(Bates et al., 1999). tion system for all its hospitals and clinics which
In pharmaceuticals, e-prescribing, or the elec- reduces report preparation time. For example,
tronic transmission of prescription information patients can have a radiology procedure, see their
from the prescribing physician to a pharmacist can doctor, and obtain their results the same day (Bal-
reduce medical errors. Since Americans receive drige Award Recipient, 2006; Anderson, 2007).

68
Information Technology (IT) and the Healthcare Industry

greater efficiency of operation days denied, and the reassignment of three full-
time equivalent employees to other departments
Information technology, or the digital world of bits within the hospital (Bowen & Bassler, 2006).
and bytes, delivers information faster, smarter, and Automated two-way communications systems for
cheaper (Conger & Chiavetta, 2006). In health- scheduling can greatly improve workflow by man-
care, IT has improved operational efficiency and aging automatic appointment reminders, waiting
increased productivity by reducing paperwork, lists, and cancellation notices. These systems call
automating routine processes, and eliminating patients, and in a pleasant voice, remind them of
waste and duplication. Lieber (2007) reports the a doctor’s appointment, ask them to confirm their
use of electronic health records could save as much intention to keep the appointment, and report the
as $8 billion yearly in California alone through information to the provider’s office (Sternberg,
improvements in delivery efficiency. 2005). RFID technology, particularly in the
Picture archival and communication systems areas of human and material resources, offers
(PACS) not only save providers’ costs for file healthcare facilities a way to measure and control
room, storage space and film supplies, but also their resources as well as the relevant workflow
decrease time spent reporting, filing and retrieving processes (Janz, Pitts, & Otondo, 2005).
records. Web access enables physicians to view Some healthcare providers suggest Emergency
radiological images from their offices, homes, or Department Information Systems (EDIS) can
other remote facilities. IT provides emergency improve operations efficiency in this extremely
rooms with tools for electronic prescriptions, time-critical area by facilitating the flow of patients
order entry, provider documentation, and after- through the emergency department, eliminating
care instructions for patients and their family. redundant patient records, promoting information
Updating electronic instructions is quick and sharing, and providing quicker access to labora-
easy. Purchasing departments are aided by the tory test results and radiology films (Parker,
ability to buy products for specialty areas, such 2004a). Because electronic medical records allow
as anesthesia, infection control, substance-abuse tracking of patients’ conditions and medications,
programs, and home healthcare. Increased pro- emergency room providers and hospitals have
ductivity and positive return on investment are immediate access to detailed information; both
seen in many areas of IT and are continuously patients and providers have a better sense of
improving (Parker, 2004b). what occurred and when. Both groups also report
Three other important technological advances increased satisfaction with the process. Interoper-
for improved productivity are voice-technology ability of systems also makes patient information
systems, two-way communications systems, and available across budgetary and functional units,
radio-frequency identification (RFID). Voice- thereby providing greater continuity of patient
technology systems can significantly reduce the care (Cohen, 2005).
time nurses and admissions personnel spend on Lopes (2007) agrees it is more efficient when
pre-authorizations and pre-certifications required an internal medicine physician can consult with
by third-party payer plans. Within six months of a cardiologist electronically while viewing a
implementing a phone-based voice-technology patient’s medical work-up and history. Such sys-
system, Erlanger Hospital in Chattanooga, Ten- tems create efficiencies, have a positive return on
nessee, reported a greater than 50% drop in phone investment, and there are no misfiled lab results.
transaction times. Moreover, in May, 2006, after IT tools and software are being developed
four years in operation, Erlanger reported a total to meet the growing needs of the healthcare
payback of $920,201, decreased percentages in community. As an example, VHA, Inc. recently

69
 Information Technology (IT) and the Healthcare Industry

introduced an updated version of a Comparative develop, acquire, and integrate applications for
Clinical Measurement tool to give member hos- decision support, benchmarking, facilities man-
pitals flexibility in both collecting and reporting agement, and workflow processes (Cohen, 2005;
clinical improvement data. VHA, Inc. has worked Kay & Clarke, 2005).
with the Joint Commission on Accreditation of If predictions are correct, by 2015 the electronic
Healthcare Organizations to include their mea- medical records market is expected to grow to
sures into the new tool (VHA, 2007). more than $4 billion, up from $1 billion in 2005.
The Kalorama Information Research firm, after
current Investment in It studying the healthcare, diagnostics, pharmaceu-
ticals, and medical devices markets, predicts the
Is there a hospital in the United States that has surge will be led by the increase in IT budgets
not already made an investment in their IT in- of hospitals, physicians’ offices and other U.S.
frastructure? Probably not. In the past ten years, healthcare organizations (Study: U.S., 2007).
advances in health information technologies have Healthcare organizations, having already made
occurred at an unprecedented rate and healthcare investments in their computing and communica-
organizations have responded by increasing their tions hardware and software, application software,
IT investments “threefold” (Burke & Menachemi, and personnel, can leverage their existing IT
2004). Today, albeit at varying levels of sophis- investments as they expand their IT infrastruc-
tication, all hospitals use IT to run their core ad- ture to meet growing demands to achieve more
ministrative and clinical application systems, that efficient operations and more effective levels of
is, patient accounting, insurance billing, human healthcare. Whether or not the IT investments
resources, staff and facilities scheduling, phar- meet the ROI goals of financial departments,
macy, laboratory results reporting, and radiology hospitals are going to implement IT, according
(Cohen, 2005). Most healthcare organizations in to a survey by the Healthcare Information and
the U.S. are spending between 2.1% and 10% of Management Systems Society. Some 88% of
their capital operating budget on IT (Conn, 2007c). hospitals have adopted electronic medical records
Within this spending on IT, healthcare providers and 24% already have them in place. Some 36%
cite electronic health record development as a top are implementing them and 28% have plans to.
priority followed by development and implemen- Only 12% lack IT plans (Greene, 2007).
tation of clinical IT systems to improve patient
care capability (Conn, 2007b).
According to the U.S. Department of Health Internal weaKnesses
and Human Services (HHS), approximately 13%
of the nation’s 4,000+ hospitals use electronic lack of system Integration
medical records and 14% to 28% of the 853,000
U.S. physicians are wired (Swartz, 2005). A re- Integrated systems offer seamless data and pro-
cent study reported, on average, hospitals have cess integration over diverse information systems
acquired 10.6 clinical application systems, 13.5 (Landry, Mahesh, & Hartman, 2005). Since a
administrative application systems, and 50.0 stra- patient’s treatment involves receiving services
tegic application systems (Burke & Menachemi, from multiple budgetary units in a hospital, infor-
2004). Some healthcare organizations including mation system integration should exist between
the Cincinnati Children’s Hospital, Baylor Health- the computer-based applications within a single
care System in Dallas and The Heart Center of hospital. When healthcare organizations coordi-
Indiana are going beyond their core systems to nate and integrate their internal data, they can

70
Information Technology (IT) and the Healthcare Industry

improve operations and decision making; however, is explained by two factors: the system’s per-
most healthcare organizations are not yet at this ceived usefulness and its perceived ease of use.
level of system integration. Clinical, administra- Perceived usefulness is defined as the degree to
tive, and financial systems are not linked, and as which a person believes that using a particular
a result, many healthcare institutions are not yet system would enhance job performance, while
maximizing their IT potential (Cohen, 2005). perceived ease of use is defined as the degree
Moreover, system integration need not be to which a person believes that using a particu-
confined to applications within a single facility. lar system would be free of effort. Subsequent
There are many types of healthcare providers research across a variety of research settings
and healthcare-related agencies in the complex confirms perceived usefulness as the strongest
healthcare network. Since a patient’s treat- predictor of user acceptance (Adams, Nelson,
ment usually involves receiving services from & Todd, 1992; Taylor & Todd, 1995; Venkatesh
multiple providers and interacting with various & Davis, 1996; Mahmood, Hall, & Swanberg,
other healthcare-related entities, information 2001). Some believe that IT implementations in
system integration should also exist between the the healthcare environment, however, encounter
computer-based applications of those separate more resistance than in any other environment
agencies. Immediate benefits of information (Adams, Berner, & Wyatt, 2004).
sharing between different agencies include the The healthcare-related literature suggests phy-
elimination of duplicate work in gathering and sician resistance is a key weakness existing in the
inputting the data, the immediate availability doctor’s office and at the hospital. Consistent with
of the information, a lower probability of error, Davis’ (1989) principle of perceived usefulness,
and greater convenience for the patient. Called a physician acceptance of new IT systems at the
“vision of unsurpassed information technology hospital is linked to the system’s impact on patient
integration,” The Heart Center of Indiana, a joint safety (Rhoads, 2004), while physician acceptance
venture between St. Vincent Health, The Care of new IT at their office largely depends on cost
Group, and CorVasc, reports improved quality (Chin, 2005). Healthcare literature suggests nurse
of care at lower costs through its IT partnership acceptance of new IT has steadily improved as
(Kay & Clarke, 2005). applications demonstrate increased support of the
System integration between agencies could practice of nursing and improvement of patient
also take increased efficiency to the industry or safety resulting from the reduction of human error
national level. A report issued by the Foundation (Sensmeier, 2005; Simpson, 2005).
of Research and Education of the American Health A study of 12 critical access hospitals found
Information Management Association supports a barriers to health information technology included
fully integrated fraud management system which funding, staff resistance to change, staff adapta-
it believes could help address the growing problem tion to IT and workflow changes. Other user
of healthcare fraud (Swartz, 2006a). resistance was noted by the time constraints on
small staff, facility and building barriers, and lack
user resistance of appropriate IT support. While all agree that IT
will improve safety and reduce errors, barriers to
User resistance, more commonly termed user implementation are numerous and must be ad-
acceptance in the information systems literature, dressed (Hartzema, Winterstein, Johns, de Leon,
is nothing new to IT. The original Technology Bailey, McDonald, & Pannell, 2007).
Acceptance Model (TAM) put forth by Davis
(1989) states a user’s level of system acceptance

71
 Information Technology (IT) and the Healthcare Industry

slow It adoption external opportunItIes

Traditionally, healthcare has been slow to adopt the Internet

IT and has lagged significantly behind other in-
dustries in the use of IT (Ortiz & Clancy, 2003; Across the industry, healthcare facilities and pro-
Adams et al., 2004). A 2005 report from the Na- viders are in various stages of incorporating the
tional Academy of Engineering and the Institute Internet into their operations to allow new ways
of Medicine agrees healthcare’s failure to adopt to communicate with the general public, specific
new strategies and technologies has contributed patients, patient groups, physicians, other provid-
to the list of problems now associated with the ers, and employees. Notable Web-based services
industry: thousands of preventable deaths a year, include public Web sites, various telemedicine
outdated procedures, billions of dollars wasted applications for targeted patient audiences, physi-
annually through inefficiency, and costs rising cian portals, physician education sites, and facility
at roughly three times the rate of inflation. Lack intranets which serve an organization’s internal
of competition, resistance to change, and capital audiences. Generally, there is an increased focus
costs are among the major causes for healthcare’s throughout the healthcare industry to improve all
slowness to adopt IT (Hough, Chen, & Lin, 2005). Web-based applications (Sternberg, 2004).
There are signs of progress, however, which Through their public Web sites, hospitals and
offer promise of accelerated change. Many hos- other healthcare agents provide medical infor-
pitals and physician groups are now digitizing mation to the general public (Natesan, 2005).
their medical records and clinical data (Hough et E-Health Web portals offer healthcare services
al., 2005). As noted earlier, some hospitals like and education to people with chronic conditions
Cincinnati Children’s Hospital, Baylor Healthcare and to their caregivers (Moody, 2005). Through
System in Dallas, and The Heart Center of Indi- various telemedicine initiatives, the healthcare
ana have adopted IT at advanced levels (Cohen, industry has reached significant numbers of
2005; Kay & Clarke, 2005). These hospitals are people living in rural areas, providing access
models for the industry, forging a path for other to expert advice and reducing their health risks
healthcare organizations to follow, and emerging (Harris, Donaldson, & Campbell, 2001). Web-
as healthcare leaders in IT whose techniques can based patient support systems educate patients
be benchmarked, emulated and implemented. and allow them better participation in their own
As the healthcare technologies are developed to care. Patients can research detailed information
greater sophistication and functionality, it will for their particular conditions, medications, and
be possible for other healthcare organizations to treatments to understand what is happening and
“leapfrog” over the slow, expensive evolution- to reduce their anxiety. Online surgery videos
ary learning process experienced by the leaders and graphics can be presented in user-friendly
(Conger & Chiavetta, 2006). formats to assist patients in procuring informa-
The following section outlines external op- tion. The Internet has also had a major impact
portunities and threats facing IT and healthcare. in the delivery of information and education to
Specific opportunities are the Internet, the national healthcare professionals (Kiser, 2001). Numerous
environment, and industry standards. Key threats organizations have Web sites for disseminating
include legal compliance, loss of patient trust, and new medical information to physicians. Various
the costs of IT systems, training, implementation, Web-based physician education services have been
and support. established. Some hospitals offer physician portals
allowing physicians to access patients’ medical

72
Information Technology (IT) and the Healthcare Industry

records, lab results, and radiological images and utilizing IT (Sharma, 2004; Grain, 2005; Marino
reports from their offices, homes, or other remote & Tamburis, 2005; Bergmo & Johannessen, 2006;
locations (Cohen, 2005). Fitch & Adams, 2006).
The Internet is also redefining communica- In the U.S., more funds are being made avail-
tion channels between doctors and patients, as able in the form of grants and demonstration
well as between healthcare providers and other projects by the Federal government to encourage
healthcare-related agencies. DeShazo, Fessenden, greater adoption of IT in healthcare. In 2004, of-
and Schock (2005) suggest the top two emerging ficials of the U.S. Department of Health and Human
trends in healthcare are (1) online patient/physician Services (HHS) disclosed a ten-year healthcare
communication and (2) secure connectivity and information infrastructure plan, the “Decade of
messaging among hospitals, labs, pharmacies, and Health Information Technology,” to transform the
physicians. Advances in home technology coupled industry from a paper-based system to an elec-
with the aging of the baby-boom generation have tronic one. More than 100 hospitals, healthcare
created the demand for better communications providers, and communities in 38 states were
with patients about their on-going care and moni- awarded $96 million over three years to develop
toring. Improving the communication between the and use IT for healthcare. Awards were focused
patient’s at-home technology and the provider’s on communities and small and rural hospitals.
technology is also a growth opportunity. Based Five states, Colorado, Indiana, Rhode Island, Ten-
on the adequacy of information transmitted to the nessee, and Utah, were awarded $25 million over
healthcare provider, the physician saves appoint- five years to develop secure statewide networks
ment times and patients are freed from excessive for accessing patient medical information. The
office visits, thereby lowering transaction costs National Opinion Research Center at the Uni-
(Flower, 2005). versity of Chicago was awarded $18.5 million to
The Internet and other advances in IT have create a National Health Information Technology
enabled new models for electronic delivery of a Resource Center to provide technical assistance,
variety of healthcare services. Kalyanpur, Latif, tools, and a best-practices repository as well as
Saini, and Sarnikar (2007) describe the market provide a focus for collaboration to grantees
forces and technological factors that have led to and other federal partners. In all, HHS awarded
the development of Internet-based radiological nearly $140 million in grants to promote the use
services and agree the Internet has provided the of IT, develop state and regional networks, and
platform for cost-effective and flexible radiologi- encourage collaboration in advancing the adop-
cal services. Wells (2007) agrees the practice of tion of electronic health records (Swartz, 2005).
evidence-based medicine requires access to the In 2005, the Agency for Healthcare Research
Internet, mobile devices, and clinical decision- and Quality, part of HHS, awarded over $22 mil-
support tools to assist practitioners in improving lion in grants to 16 institutions in 15 states to aid
preventable medical errors. in implementing healthcare IT projects empha-
sizing patient safety and healthcare quality. The
Favorable external environment grants were designed to encourage the sharing of
information among providers, labs, pharmacies,
There is growing support worldwide for the utiliza- and patients, with the specific goal of decreasing
tion of more IT in healthcare (Caro, 2005). Reports medication errors and duplicate testing. Eleven
from Australia, Great Britain, India, Italy, and of the 16 grants were awarded to small and rural
Norway, for example, document local, regional communities (Anonymous, 2005c).
and national healthcare projects and initiatives

73
 Information Technology (IT) and the Healthcare Industry

In his 2004 State of the Union address, A recent study of several disability compensa-
President Bush called for the transformation of tion programs within the U.S. found each program
electronic health records within the next ten years uses its own terminology and disability definitions
in the United States and urged more healthcare causing non-standard interpretation of terms, mis-
organizations to consider implementing such interpretation of data, and delay in the disability
health information technologies as electronic evaluation process. The study suggests defining
healthcare records (EHR), computerized order- and adopting a standard for disability evaluation
ing of prescriptions and medical tests, clinical could not only eliminate process inefficiencies in
decision support tools, digital radiology images, determining disabilities but could also facilitate
and secure exchange of authorized information, innovative disability technology practices (Tulu,
emphasizing that all of these technologies have Hilton, & Horan, 2006).
been shown to improve patient care quality and Some standardization of data has been intro-
reduce medical errors (Abrahamsen, 2005). In duced with the Health Insurance Portability and
President Bush’s 2008 budget proposal, there Accountability Act (HIPAA) legislation, but for the
is funding for a healthcare system and IT is the most part, standardization projects are voluntary
starting point for the system. Carolyn Clancy, and lack assurances the standards will be adopted
Director of the Agency for Healthcare Research by all parties. The National Quality Forum (NQF)
and Quality, agrees the data generated from the endorsed a voluntary consensus standard for pa-
healthcare system could answer various medical tient safety events and has been adopted by more
inquiries and could draw on the data of EHRs of than 260 healthcare providers, consumer groups,
millions of individuals to advance the evidence professional associations, federal agencies, and
base for clinical care. She further suggests the data research and quality improvement organizations.
could reveal why costs are increasing and what The taxonomy is the first standardized, integra-
risks and benefits are associated with particular tive classification system adopted by a group of
prescription drugs (Lubell, 2007). medical agencies, organizations, providers, and
U.S. states. The standard establishes a common
Industry standards taxonomy for healthcare errors and other patient
safety problems. It can be used to classify data
The development of industry standards for both collected in different reporting systems, allowing
data communications and data taxonomies may the data about patient safety events to be combined
be the most profound of all the opportunities and analyzed (Anonymous, 2005d).
currently facing healthcare. As a crucial first- Standardization can result in greater levels
step in modernizing the U.S. healthcare system, of system integration, increased sharing of data
all industry participants—providers, payers, and between healthcare partners, greater information
regulators—are being urged to adopt interoperable continuity throughout the healthcare industry, and
systems and common data standards for existing more powerful data mining. Enterprise Resource
federal, state, and health networks along with Planning (ERP) systems can offer more online
standard practices to promote data sharing and processing to all users and function, automate
protection of patient privacy (Swartz, 2006b). routine job processes, and redefine existing work
Standard data communications technology and processes (Landry et al., 2005). With integration
standard data definitions are essential for such of systems and standardization of data taxonomies,
health information technologies as electronic the healthcare industry will experience changes
health records and e-prescribing (Brailer, 2004). similar to those in other industries where “enter-
prise” systems have been adopted. Others agree a

74
Information Technology (IT) and the Healthcare Industry

seamless support of information flow for health- uses of their PHI, keep a record of all disclosures
care processes that are increasingly distributed of PHI, and document and disclose their privacy
requires the ability to integrate heterogeneous policies and procedures. Covered entities must
IT systems into a comprehensive system (Lenz, have designated agents for receiving complaints
Beyer, & Kuhn, 2007). and they must train all members of their workforce
System standards resulting in a greater level in proper procedures.
of systems integration is a pressing need. Conn The Security Rule complements the Privacy
(2007a) reports the compromise reached by two Rule and presents three types of security safe-
rival standards groups for data communications guards designated as administrative, physical,
standards can help to bridge the gap between and technical. For each type, the Rule identifies
physicians’ offices and hospitals in the electronic various security standards and names (1) required
health record systems they use. The Continuity implementation specifications which must be
of Care Document standard combines the inde- adopted and implemented as specified in the Act
pendent works by two standards development and (2) addressable implementation specifications
organizations on creating electronic summaries which are more flexible and can be implemented
of care for discharged patients. by the covered entities as deemed appropriate.
Covered entities face potentially severe pen-
alties for failure to comply with the complex
external tHreats legalities of HIPAA and this has caused much
concern throughout the industry. Physicians,
legal compliance medical centers, and other healthcare providers
have experienced increased paperwork and cost
The Health Insurance Portability and Account- to incorporate the requirements of this legislation
ability Act (HIPAA), enacted by Congress in 1996, into their current methods of operation. Future
is the most significant Federal legislation affecting adoptions of new information technologies will
the U.S. healthcare industry since the Medicare be subject to its specifications as well (American
and Medicaid legislation of 1965. Title I of HIPAA College of Physicians, 2006).
legislates improved portability and continuity of
health insurance coverage for American workers. loss of patient trust
Title II addresses “administrative simplifica-
tion” requiring the development of standards The Institute of Medicine (IOM) of the National
for the electronic exchange of personal health Academy of Sciences released a report in 1999
information (PHI). Administrative simplification that caused much attention to be focused on the
requires rules to protect the privacy of personal U.S. healthcare industry. The report stated medical
health information, the establishment of security errors caused between 44,000 and 98,000 prevent-
requirements to protect that information, and the able deaths annually, and medication errors alone
development of standard national identifiers for caused 7,000 preventable deaths (Kohn, Corrigan,
providers, health insurance plans, and employers. & Donaldson, 1999). Within two weeks of the
Two significant sections of HIPAA are (1) the report’s release, Congress began hearings and the
Privacy Rule and (2) the Security Rule. President ordered a government-wide feasibility
The Privacy Rule legislates in detail the col- study for implementing the report’s recommen-
lection, use, and disclosure of personal health dations for (1) the establishment of a Center for
information. To be in compliance with the Privacy Patient Safety, (2) expanded reporting of adverse
Rule, covered entities must notify individuals of events, and (3) development of safety programs

75
 Information Technology (IT) and the Healthcare Industry

in healthcare organizations. According to a study 2005). And although Computerized Physician

by Healthgrades, a leading healthcare ratings Order Entry (CPOE) systems reduce medication
organization, during the period 2000–2002 the errors by 80% (Bates et al., 1999), a 2004 survey
estimated number of accidental deaths per year in by Leapfrog found only 16% of hospitals, clinics,
U.S. hospitals had risen from the 98,000 reported and medical practices expected to be utilizing
by the IOM in 1999 to 195,000 (Shapiro, 2006). CPOE by 2006.
On July 29, 2005, President Bush signed into
law the Patient Safety and Quality Improvement cost
Act, establishing a federal reporting database.
This was the first piece of patient safety legisla- One of the most immediate barriers to widespread
tion since the 1999 IOM report. Under this act, adoption of technology is the high cost of imple-
hospitals voluntarily report “adverse patient mentation. A report by the Annals of Internal
events” to be included in the database, and “pa- Medicine estimated that a National Health Infor-
tient safety organizations” under contract with mation Network (NHIN) would cost $156 billion
the Federal government, analyze the events and in capital investment over five years and $48
recommend improvements. The reports submitted billion in annual operating costs. Approximately
by the hospitals remain confidential and cannot two-thirds of the capital costs would be needed
be used in liability cases. The most recent Health- to acquire the functionalities and one-third for
grades report (April 7, 2007) covering the period interoperability. The present level of spending is
2003–2005, indicates that patient safety incidents only about one-fourth of the amount estimated
have increased over the previous period to 1.16 for the model NHIN. While an NHIN would be
million among the 40 million hospitalizations expensive, $156 billion is equivalent to 2% of
covered under the Medicare program. annual healthcare spending for 5 years (Kaushal
Healthcare must utilize all available means to et al., 2005). Industry reports from Datamonitor,
maximize patient safety and retain patient trust. Gartner, and Dorenfest & Associates predict in-
Healthcare is an information intensive industry creased spending on IT by healthcare providers at
and the delivery of high quality healthcare depends an annual rate of between 10% and 15% (Broder,
in part on accurate data, available at the point of 2004). A study conducted by Partners Healthcare
decision. Handwritten reports, notes and orders, System, Boston, concluded that a national health-
non-standard abbreviations, and poor legibility all care information system would cost $276 billion,
contribute to substantial errors and injuries (Kohn take 10 years to build, and require another $16.5
et al., 1999). A doctor needs to know a patient’s billion annually to operate. However, the study
medical history, ancillary providers need to be also concluded that such a system would save U.S.
able to read the doctor’s orders and patients need hospitals $77.8 billion annually because of more
to be able to understand what the doctor expects efficient communication (Anonymous, 2005a).
of them. IT solutions are available that address According to a study by RAND Health (Health
many of the data accuracy and availability prob- Information Technology, 2005), the U.S. health-
lems in healthcare records (Poston, Reynolds, & care system could save more than $81 billion
Gillenson, 2007); however, the level of adoption annually, reduce adverse healthcare events, and
for these technologies is not impressive. For ex- improve quality of care if it were to widely adopt
ample, a 2005 report predicts that by the end of health information technology. Patients would
2007, only 59% of all medical groups will have benefit from better health and payers would benefit
implemented an Electronic Health Record (EHR) from lower costs; however, some hospitals fear loss
system (Gans, Kralewski, Hammons, & Dowd, of revenue due to reduced patient length of stay. A

76
Information Technology (IT) and the Healthcare Industry

Table 1. SWOT Analysis

Strengths Weaknesses
• Improved Patient Safety • Lack of System Integration
• Greater Efficiency of Operation • User Resistance
• Current Investment in IT • Slow IT Adoption

Opportunities Threats
• The Internet • Legal Compliance
• Favorable External Environment • Loss of Patient Trust
• Industry Standards • Costs

recent study in Florida suggests that this fear may efficiency. Implementation of IT networks to
be unfounded. The results of the study suggest achieve the required level of information and data
that there is a significant and positive relationship communications is complicated by the variety of
between increased levels of IT use and various systems already used by provider organizations
measures of financial performance. The results as well as the lack of system integration within
indicated that IT adoption is consistently related provider organizations.
to improved financial outcomes both overall and Various benchmarking studies are helping to
operationally (Menachemi, Burkhardt, Shewchuk, educate healthcare providers about IT expendi-
Burke, & Brooks, 2006). tures and offer comparison reports on expendi-
Lopes (2007) agrees that fully integrated tures. The availability of systems far exceeds the
electronic medical records systems can replace budget of most organizations to adopt them. How-
paper records and allow hospitals and physicians ever, the improved revenue cycles and cost-benefit
to share medical information electronically to im- offered by IT investments are becoming easier to
prove response and lower costs from duplication. quantify in faster turnaround and processing of
However, the adoption of such systems is slowed patient-related transactions, shared data, reduced
by the high cost of new technology, the complexity duplication of efforts, and increased provider and
of the systems, training, and an unwillingness to customer satisfaction.
adapt work processes to include new information Information technology can help take the paper
technologies. chart out of healthcare, and eliminate error, vari-
ance and waste in the care process. IT can help
connect the appropriate persons, knowledge, and
dIscussIon and conclusIon resources at the appropriate time and location
to achieve the optimal health outcome, increase
Table 1 summarizes the current SWOT analysis of customer service and patient care with industry
IT implementation in the healthcare industry in the leading medication fill rates and timely deliveries,
U.S. The healthcare industry faces multi-faceted cut operation costs through advanced warehouse
challenges to improve patient safety and assure management, reduce internal labor costs, and
information security while containing costs and improve enterprise efficiencies of healthcare
increasing productivity. The key area for address- organizations, all through tightly integrated ap-
ing these concerns is more investment in IT to plications.
facilitate the flow of information and offer access IT can ultimately transform the healthcare
to providers and partners along the healthcare industry. Along with improved safety and greater
supply chain, reduce medical errors, and increase patient trust, adopting IT in healthcare can only

77
 Information Technology (IT) and the Healthcare Industry

improve current conditions and help the United data mining should be studied by those in the
States improve healthcare in general. Concerns management information systems area to deter-
remain about how smaller practices can afford mine cause and effect and recommend changes.
the costs of new systems that require them to As employers seek to contain healthcare costs
move their paper medical records to electronic of their employees, such data can aid in more
media. These costs and start-up expenses mean active involvement in reducing health risks and
an unequal playing field for small practices versus making lifestyle changes (i.e., smoking cessation
larger healthcare systems with more money to programs, dietary counseling, healthy cafeteria
spend on IT integration. food, work-place gyms).
Choosing the best approach to implement
IT systems in healthcare settings is also an area
areas For Future for further study. These systems should meet
researcH healthcare goals in addition to functionality and
integration criteria. Involving physicians and other
IT applications in healthcare are reaching the clinicians in selecting IT systems can increase their
growth phase of the lifecycle. The strengths, support and lessen their resistance to technology.
weaknesses, opportunities, and threats at this stage In fact, the more stakeholders are involved in IT
of the life cycle are clear, but few solutions have selection and implementation planning, the greater
been proposed. Research is needed to forecast their acceptance and rate of adoption will be.
the SWOT issues as IT in healthcare moves from Studies in IT implementation outside the
growth to maturity. Case studies in both large healthcare industry need to be reviewed and ana-
and small physician practices as well as in large lyzed to determine where other industries have
and small healthcare systems are needed to better had success in implementation or have developed
understand the IT implementation timeframe and tools that could aid the healthcare arena. Human
costs. Studies that address ways to overcome hu- resource studies of executive ownership and ac-
man barriers to implementation are also needed. countability can help the healthcare industry better
Using the supply chain model, the healthcare prepare physicians and other practice managers
information system needs to be studied as to ac- to overcome the user resistance to IT.
cess and applicability for other providers including
pharmacists, dietitians, insurance companies,
home health service and equipment providers, reFerences
and other vendors to healthcare. If the healthcare
system it to be truly integrated, these additional Abrahamsen, C. (2005). Washington taps into
players must be included. Protection of patient healthcare technology. Nursing Management,
medical information, access to information, data 36(3), 52.
security, and assurances of privacy should also be
Adams, B., Berner, E. & Wyatt, J. (2004). Applying
studied. International suppliers and other options
strategies to overcome user resistance in a group
for outsourcing should be investigated as a cost
of clinical managers to a business software ap-
containment strategy.
plication: A case study. Journal of Organizational
With the growing body of patient-related
and End User Computing, 16(4), 55-65.
medical and healthcare data, other applications
should be studied. Data is available to postulate Adams, D., Nelson, R. & Todd, P. (1992). Perceived
ways to improve lifelong health and reduce the usefulness, ease of use, and usage of information
incidences of various diseases and ailments. Such technology: A replication. MIS Quarterly, 16(2),
227-247.

78
Information Technology (IT) and the Healthcare Industry

American College of Physicians. (2006). Health Bergmo, T. & Johannessen, L. (2006). The long
Insurance Portability and Accountability Act road from potential to realized gains of infor-
Privacy Rule causes ongoing concerns among mation technology in healthcare – experiences
clinicians and researchers. Annals of Internal from Norway. International Journal of Economic
Medicine, 145(4), 313-316. Development, 8(3), 682-716.
Anderson, K. (2007). The ASRT Salutes Your Bowen, R., & Bassler, P. (2006). Can you hear
Excellence, North Mississippi Medical Center. me?. Healthcare Informatics. Retrieved from
ASRT Scanner, 39(5), 8-11. www.healthcare-informatics.com.
Anonymous. (1997). Just how smart are smart Brailer, D. (2004). Perspective: Translating ideals
cards? Financial Executive, 13(5), 8-10. for health information technology into practice.
Health Affairs, W318-W320.
Anonymous. (2005a). Adoption of health infor-
mation technology standards would save billions. Broder, C. (2004). U.S. healthcare providers plan
(2005). Healthcare Strategic Management, 23(2), IT spending increase. iHealthBeat. Retrieved,
4-6. from http://www.ihealthbeat.org/index.cfm
Anonymous. (2005b). Healthcare information Brodkin, J. (2007). A Growing Divide in Health-
technology panel established. Association of Op- Care IT. Network World, 24(6), 17.
erating Room Nurses, AORN Journal, 82(6), 1018.
Burke, D. & Menachemi, N. (2004). Opening
Anonymous. (2005c). Millions awarded to health- the black box: Measuring hospital information
care IT projects. Quality Progress, 38(12), 12. technology capability. Healthcare Management
Review, 29(3), 207-217.
Anonymous. (2005d). Patient safety event taxon-
omy standard endorsed. Association of Operating Caro, D. (2005). Forging e-health partnerships:
Room Nurses, AORN Journal, 82(6), 1040-1041. Strategic perspectives from international execu-
tives. Healthcare Management Review, 30(2), 174.
Ansoff, H. (1965). Corporate strategy. New York:
McGraw-Hill. Chin, T. (2005). Are physicians at the infotech
tipping point?. American Medical News, 48(10),
Balas E., Weingarten S., Garb C., Blumenthal D.,
20-21.
Boren S., & Brown G. (2000). Improving preven-
tive care by prompting physicians. Arch Intern Cohen, S. (2005). Emerging benefits of integrated
Mcd, 160, 301-308. IT systems. Healthcare Executive, 20(5), 14-18.
Baldrige Award Recipient Profile: Malcolm Conger, T. & Chiavetta, D. (2006). The pulse of
Baldrige National Quality Award 2006 Award industry. Industrial Engineer, 38(1), 30-35.
Recipient, Healthcare: North Mississippi Medi-
Conn, J. (2007a). Agreement could put EHRs on
cal Center. Retrieved from http://www.nist.gov/
fast track. Modern Healthcare, 37(9), 22-23.
public_affairs/releases/mississippi.html
Conn, J. (2007b). Eyes on the HER. Modern
Bates, D., Teich, J., Lee, J., Seger, D., Kuperman,
Healthcare, 37(9), 37-38.
G., et al. (1999). The impact of computerized physi-
cian order entry on medication error prevention. Conn, J. (2007c). Still the matter of money. Modern
Journal of the American Medical Informatics Healthcare, 37(9), 36-37.
Association, 6(4) 313-321.

79
 Information Technology (IT) and the Healthcare Industry

Davis, F. (1989). Perceived usefulness, perceived Hartzema, A., Winterstein, A., Johns, T., de
ease of use, and user acceptance of information Leon, J., Bailey, W., McDonald, K., & Pannell,
technology. MIS Quarterly, 13(3), 319-340. R. (2007). Planning for pharmacy health infor-
mation technology in critical access hospitals.
DeShazo, C., Fessenden, R. & Schock, P. (2005).
American Journal of Health-System Pharmacy,
Realizing IT’s Potential. Physician Executive,
64(3), 315-321.
31(4), 26-29.
Health Information Technology: Can HIT lower
Fitch, C. & Adams, C. (2006). Managing mobile
costs and improve quality?. (2005). Rand Health.
provision for community healthcare support: Is-
Retrieved from http://www.rand.org
sues and challenges. Business Process Manage-
ment Journal, 12(3), 299. Health Grades, Inc. (2007). HealthGrades quality
study: Fourth annual patient safety in American
Five Large Companies Working to Provide Life-
hospitals study. Retrieved from http://www.
long Personal Health Records for Employees.
healthgrades.com/media/dms/pdf/PatientSafety-
(2007). Healthcare Financial Management,
InAmericanHospitalsStudy2007.pdf.
61(2), 11.
Helms, M. (1999). How to be successful in China:
Flower, J. (2005). The 16 challenges. Hospitals
A SWOT analysis. Competitiveness Review, 9(2),
& Health Networks, 4(4), 16-18.
1-10.
Gans, D., Kralewski, J., Hammons, T., & Dowd,
Helms, M. (2003). Japanese managers: Their
B. (2005). Medical groups’ adoption of electronic
candid views of entrepreneurship. Competitive-
health records and information systems. Health
ness Review, 13(1), 24-34.
Affairs (Project Hope), 24(5), 1323-1333.
Hitt, M., Ireland, R., Camp, S., & Sexton, D.
Gillette, B. (2004). IT investment shows marked
(2001). Guest editors’ introduction to the special
increase, but mostly on the clinical side. Managed
issue: Strategic entrepreneurship: Entrepreneurial
Healthcare Executive, 14(9), 58.
strategies for wealth creation. Strategic Manage-
Glaister, K. & Falshaw, J. (1999). Strategic plan- ment Journal, 22(6-7), 479-491.
ning: Still going strong?. Long Range Planning,
Hofer, C. & Schendel, D. (1978). Strategy formu-
32(1), 107-116.
lation: Analytical concepts. St. Paul, MN: West
Goldberg, R., Kuhn, G., Louise, A., & Thomas, Publishing Company.
H. (2002). Coping with medical mistakes and er-
Hough, C., Chen, J. & Lin, B. (2005). Virtual
rors in judgment. Annals of Emergency Medicine,
health/electronic medical record: Current status
40(2), 287-292.
and perspective. International Journal of Health-
Grain, H. (2005). Information systems in the new care Technology & Management, 6(3), 257.
world: An emerging national approach. Australian
Janz, B., Pitts, M. & Otondo, R. (2005). Infor-
Health Review, 29(3), 292-297.
mation systems and healthcare II: Back to the
Greene, J. (2007). The costs & benefits of IT. future with RFID: Lessons learned – some old,
Trustee, 60(2), 8-12. some new. Communications of the Association
for Information Systems, 15, 1.
Harris, K., Donaldson, J. & Campbell, J. (2001).
Introducing computer-based telemedicine in three
rural Missouri counties. Journal of End User
Computing, 13(4), 26-35.

80
Information Technology (IT) and the Healthcare Industry

Kalyanpur, A., Latif, F., Saini, S., & Sarnikar, Lenz, R., Beyer, M. & Kuhn, K. (2007). Inter-
S. (2007). Inter-Organizational e-Commerce national Journal of Medical Informatics, 76(2-3),
in healthcare services: The case of global tele- 201-207.
radiology. Journal of Electronic Commerce in
Lieber, S.(2007). IT goes global. Modern Health-
Organizations, 5(2), 47-56.
care, 37(9), 30.
Kaushal, R., Blumenthal, D., Poon, E., Jha, A.,
Lopes, G. (2007). Paper still rules patient records.
Franz, C., Middleton, B., et al. (2005). The costs
Washington Times (DC), March 5.
of a national health information network. Annals
of Internal Medicine, 143(3), 165-173. Lubell, J. (2007). IT may play Research Role in
’08. Modern Healthcare, 37(5), 4.
Kay, D. & Clarke, J. (2005). Raise care quality,
lower costs successfully with IT partnership. Mahmood, M., Hall, L. & Swanberg, D. (2001).
Journal of Nursing Administration, 12. Factors affecting information technology usage:
A meta-analysis of the empirical literature. Jour-
Khan, G. & Al-Buarki, E. (1992). Strategic plan-
nal of Organizational Computing and Electronic
ning in bahrain. Management Decision, 30(5),
Commerce, 11(2), 107-130.
3-10.
Marino, A. & Tamburis, O. (2005). Technological
Kiser, K. (2001). Doctor, teach thyself. Online
innovation and organisational context informa-
Learning, 23-28.
tion management in the high specialised hospital:
Kohn, L. (2001). The Institute of Medicine report The case of South Italy. International Journal of
on medical error: Overview and implications for Healthcare Technology & Management, 6(3), 276.
pharmacy. American Journal of Hospital Phar-
Martin, E., Brown, C., DeHayes, D., Hoffer, J.,
macy, 58(22), 63-66.
& Perkins, W. (2005). Managing Information
Kohn, L., Corrigan, J., & Donaldson, M. (Eds.). Technology, 5th ed. Upper Saddle River, NJ:
(1999). To err is human: Building a safer health Prentice-Hall.
system. Institute of Medicine, Washington: Na-
McDonald, M. (1999). Marketing plans. Oxford,
tional Academy Press.
UK: Butterworth-Heinemann Press.
Kotler, P. (2000). Marketing management. Upper
Mekhjian, H., Kumar, R., Kuehn, L., Bentley, T.,
Saddle River, NJ: Prentice-Hall.
Teater, P., et al. (2002). Immediate benefits real-
Kuperman, G., Teich, J., Gandhi, T., & Bates. D. ized following implementation of physician order
(2001). Patient safety and computerized medica- entry at an academic medical center. Journal of
tion ordering at Brigham and Women’s Hospital. the American Medical Informatics Association,
Journal of Quality Improvement, 27(10), 509-520. 9(5), 529-539.
Landry, B., Mahesh, S. & Hartman, S. (2005). Menachemi, N., Burkhardt, J., Shewchuk, R.,
The impact of the pervasive information age Burke, D. & Brooks, R. (2006). Hospital infor-
on healthcare organizations. Journal of Health mation technology and positive financial perfor-
and Human Services Administration, 27(3-4), mance: A different approach to finding an ROI.
444-464. Journal of Healthcare Management, 51(1), 40-59.
Lenz, R. (2007). IT Support for healthcare pro- Mintzberg H., Ahlstrand, B. & Lampel, J. (1998).
cesses – Premises, challenges, perspectives. Data Strategy safari: A guide through the wilds of
& Knowledge Engineering, 61(1), 39-58. strategic management. New York: Free Press.

81
 Information Technology (IT) and the Healthcare Industry

Moody, L. (2005). E-Health Web portals: Deliver- Scalise, D. (2002). CPOE: Is it worth it?. Hospitals
ing holistic healthcare and making home the point & Health Networks, 76(1).
of care. Holistic Nursing Practice, 19(4), 156.
Schnaars, S. (1998). Marketing strategy. New
Natesan, C. (2005). The impact of the Internet on York: Free Press.
healthcare services. Journal of American Academy
Sensmeier, J. (2004). Transform workflow through
of Business, 7(2), 247-252.
selective implementation. Nursing Management,
Order entry rules: Healthcare enterprise achieves 35(12), 46-52.
physician acceptance, reduced medication errors
Sensmeier, J. (2005). Here and now: Healthcare’s
and improved patient outcomes through CIS and
top 10 IT breakthroughs. Journal of Nursing
CPOE technology. (2002). Health Management
Administration, October, 8.
Technology, 23(7), 34, 38.
Shane, R. (2002). Computerized physician order
Ortiz, E. & Clancy, C. (2003). Use of information
entry: Challenges and opportunities. American
technology to improve the quality of healthcare
Journal of Health-System Pharmacy, 59(3),
in the United States. Health Services Research,
286-288.
38(2), 10-13.
Shapiro, S. (2006). HealthGrades patient safety
Parker, P. (2004a). Move care to a higher level
study shows increase in U.S. hospital incidents.
with emergency systems. Nursing Management,
Medical News Today, Retrieved from http://www.
35(9), 82-83.
medicalnewstoday.com/printerfriendlynews.
Parker, P. (2004b). Quantify technology’s benefits. php?newsid=40954.
Nursing Management, 35(2), 41.
Sharma, D. (2004). Technologies for the people:
Porter, M. (1991). Towards a dynamic theory A future in the making. Futures, 36(6-7), 733.
of strategy. Strategic Management Journal, 12,
Simpson, R. (2005). In direct proportion: Ratios,
95-117.
IT, and trust. Nursing Management, 36(2), 14.
Poston, R., Reynolds, R., & Gillenson, M. (2007).
Smith, J. (1999). The behavior and performance
Technology solutions for data accuracy and avail-
of young micro firms: Evidence from business
ability problems in healthcare records. Informa-
in Scotland. Small Business Economics, 13(3),
tion Systems Management, 24(1), 59-71.
185-200.
Proctor, R. (1992). Structured and creative ap-
Sternberg, D. (2004). Trends worth watching.
proaches to strategy formulation. Management
Marketing Health Services, 24(2), 44.
Research News, 15(1), 13-19.
Sternberg, D. (2005). Strategic innovation. Mar-
Rhoads, C. (2004). Rx for success. Computer-
keting Health Services, 24(2), 44.
world, 38(34), 27-28.
Study: U.S. electronic medical-records market
Roberts, B. & Stimson, R. (1998). Multi-sectoral
count grow four-fold by 2015. (2006). Business
qualitative analysis: A tool for assessing the com-
Journal (Central New York), 21(7), 8.
petitiveness of regions and formulating strategies
for economic development. Annals of Regional Swartz, N. (2005). U.S. hands out technology
Science, 32(4), 469-494. grants for healthcare. Information Management
Journal, 39(1), 14.

82
Information Technology (IT) and the Healthcare Industry

Swartz, N. (2006a). E-records may end fraud. Valentin, E. (2001). SWOT analysis from a
Information Management Journal, 40(1), 16. resource-based view. Journal of Marketing Theory
and Practice, 9(2), 54-69.
Swartz, N. (2006b). Tech leaders release e-health
roadmap. Information Management Journal, Venkatesh, V. & Davis, F. (1996). A model of the
40(1), 18. antecedents of perceived ease of use: Development
and test. Decision Sciences, 27(3), 451-481.
Taylor, S. & Todd, P. (1995). Understanding
information technology usage: A test of compet- VHA Inc. introduces new version of its compara-
ing models. Information Systems Research, 6(2), tive clinical measurement tool. (2007). Healthcare
144-176. Purchasing News, 31(3), 8-14.
Thompson, A. & Strickland, A. (1998). Crafting Wells, L. (2007). Role of information technol-
and Implementing Strategy, 10th ed. New York: ogy in evidence-based medicine: Advantages
McGraw Hill. and limitations. Internet Journal of Healthcare
Administration, 4(2), 5-10.
Tulu, B., Hilton, B. & Horan. T. (2006. Improv-
ing disability evaluation productivity: Linking Zhang, J. & Kelvin, A. (1999). Strategic entry for
innovative business models with information products and technology into China. Journal of
technology. International Journal of Healthcare International Marketing and Marketing Research,
Technology & Management, 7(1-2), 168. 24(2), 14-29, 85-98.

This work was previously published in International Journal of Healthcare Information Systems and Informatics, Vol. 3, Is-
sue 1, edited by J. Tan, pp. 75-92, copyright 2008 by IGI Publishing (an imprint of IGI Global).

83
84

Chapter 6
Using a Neural Network
to Predict Participation
in a Maternity Care
Coordination Program
George E. Heilman
Winston-Salem State University, USA

Monica Cain
Winston-Salem State University, USA

Russell S. Morton
Winston-Salem State University, USA

aBstract
Researchers increasingly use Artificial Neural Networks (ANNs) to predict outcomes across a broad
range of applications. They frequently find the predictive power of ANNs to be as good as or better than
conventional discrete choice models. This paper demonstrates the use of an ANN to model a consumer’s
choice to participate in North Carolina’s Maternity Care Coordination (MCC) program, a state spon-
sored voluntary public health service initiative. Maternal and infant Medicaid claims data and birth
certificate data were collected for 59,999 births in North Carolina during the years 2000-2002. Part of
this sample was used to train and test an ANN that predicts voluntary enrollment in MCC. When tested
against a hold-out production sample, the ANN model correctly predicted 99.69% of those choosing to
participant and 100% of those choosing not to participant in the MCC program.

IntroductIon ing capabilities needed to support the business

management, customer relations management,
Information technology (IT) plays a pervasive role human resource management and office automa-
throughout the healthcare industry. In addition tions requirements of healthcare organizations,
to providing the data storage and data process- IT also is used increasingly to support decision
making functions.
DOI: 10.4018/978-1-61692-002-9.ch006

Copyright © 2011, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Using a Neural Network to Predict Participation in a Maternity Care Coordination

Most decision support methodologies rely medically related classifications including drug
on the mathematical modeling of historical data. and non-drug chemical compounds (Pehlivanli
Many of these systems, such as the widely accepted et al., 2008), genes (Wang et al., 2008), heart
Acute Physiology and Chronic Health Evaluation sounds (Ari & Saha, 2008), liver abnormalities
(APACHE) system, are based on binary LOGIT (Poonguzhali & Ravindran, 2007) and types of
regression estimations or other statistical analysis epileptic seizures (Najumnissa & ShenbagaDevi,
techniques. This type of modeling requires the 2008). ANNs also have been successfully applied
specification of a priori functional relationships to the diagnoses of cancer (Lisboa & Taktak,
between dependent and independent variables 2006), cardiac state (Samanta & Nataraj, 2008),
based on assumptions such as correct model speci- diabetes (Kuar & Wasan, 2006), gastrointestinal
fication, error-free measurement of independent hemorrhage (Das et al., 2003) and myocardial
variables, and normally distributed, heterosce- infarction (Baxt, 1991; Baxt et al., 2002).
dastistic, independent, zero-mean residuals. It is While medical diagnosis is probably the most
more likely, however, that healthcare decisions common healthcare application for Artificial
will depend on a variety of factors involving Neural Networks, ANNs also have been used
complex, hidden interrelationships of both socio- successfully in other healthcare areas. Examples
demographic and health related characteristics. include assessing community-level vulnerability
To address issues of non-linearity and com- to methamphetamine manufacture (Dalmadge &
plex relationships in study data, many modelers Cain, 2008), evaluating if patient debt is likely to
have turned to other methods of analysis that fall be repaid (Zurada & Lonial, 2005), evaluating the
under the broader categorization of “artificial severity of risks on healthcare non-clinical busi-
intelligence” (AI). AI, which attempts to give ness operations (Okoroh et al., 2007), identifying
computers human-like reasoning capabilities, individuals at risk for high medical costs (Crawford
includes techniques such at expert systems, fuzzy et al., 2005), identifying sources of future high
systems, genetic algorithms, case based reasoning resource demand (Kudyba et al., 2006), and pre-
and a variety of classifier systems like the Artificial dicting nursing staff levels (Seomun et al., 2006).
Neural Network (ANN) used in this study. While ANNs are generally well accepted and
Because of advantages like ease of optimi- frequently used in the healthcare industry, one
zation, prediction accuracy, easy knowledge sector that does not seem to have taken advantage
dissemination, workload reduction and decision of this technology is public healthcare services.
support, Artificial Neural Networks have been Although some high-level examples of ANN use
widely accepted and used for more than a decade for public health issues exist, such as assessing
in the healthcare arena (Lisboa & Taktak, 2006). community-level vulnerability to methamphet-
When used in medical applications, ANNs are amine manufacture (Dalmadge & Cain, 2008) and
known to provide decision support assistance forecasting demand for national immunization
that can produce highly accurate results (Kaur & vaccines (Choy & Kuo, 2006), little has been
Wasan, 2006) and better predictive performance developed at the public program level.
than other modeling alternatives (Alkan et al., While facing the expected pressures for profes-
2005; Alpsan et al., 1995; Goss & Vozikis, 2002). sionalism and quality service, public healthcare
Medical applications of ANN include diverse programs also faces the additional burdens of
examples ranging from the analysis trauma data budgetary restrictions and legislative oversight. As
(Chesney et al., 2006; Eftekhar et al., 2005) to a result, public healthcare programs, like other sec-
predicting the contact map structures of proteins tors of healthcare, are intensifying their focus on
(Chen et al., 2008). ANNs have proven useful in the enhancement of operating efficiency through

85
 Using a Neural Network to Predict Participation in a Maternity Care Coordination

effective resource allocation. One way to enhance The North Carolina Maternity Care Coor-
efficiency is to more accurately identify resource dination program (MCC) is an example of an
demands. Since Artificial Neural Networks excel early prenatal care coordination program, and is
at identifying relationships in historical data for the focus of this study. MCC has the objective
purposes of classification and prediction, it fol- of reducing barriers to Medicaid clients’ use
lows that using an ANN to predict participation of health and social services. The program is
in a public healthcare program should improve geared toward helping eligible women receive
predictive capability, reduce inefficient resource nutritional care, psychosocial counseling, and
allocation, and decrease variability in treatment other resource assistance. For example, women
processes (Kudyba et al., 2006). North Carolina’s in MCC are encouraged to seek eligible services
Maternity Care Coordination (MCC) program such as transportation, housing assistance, and
represents one such public healthcare initiative. job training. Counseling may include social and
The Maternity Care Coordination program emotional support, stress reduction methods,
attempts to coordinate prenatal care for eligible and coaching in healthy behaviors. Referral for
Medicaid clients. MCC provides counseling, refer- WIC enrollment is emphasized, and most women
rals, and resource assistance for women considered enrolled in MCC receive nutritional counseling
to be at high risk for poor birth outcomes. Par- through WIC (Buescher & Horton, 2002).
ticipation in the program, however, is voluntary. Medicaid women who are perceived to be at
The goal of this study is to develop an artificial very high risk for a poor birth outcome are most
neural network that can predict Medicaid women’s likely to be referred to the MCC program by their
voluntary enrollment in the MCC program. physician. Participation in MCC is, however, vol-
untary and thus at the discretion of the Medicaid-
enrolled woman. If she chooses to participate in
MaternIty care coordInatIon MCC, she will have access to expanded prenatal
enHancIng prenatal care care; if she chooses not to participate, she will
receive the standard package of care available
States have attempted to address the issue of poor through the state’s Medicaid program.
neonatal outcomes by incorporating comprehen- Various observable maternal medical and
sive, coordinated prenatal care programs into their socioeconomic risks may influence a pregnant
Medicaid plans. These prenatal care programs woman’s decision to participate in an enhanced
are enhanced beyond the scope of the traditional prenatal care program and may even be the same
medical model to include services such as health risk factors that prompted the healthcare provider
education, psychosocial risk assessment, enroll- to recommend the program in the first place. In
ment in WIC (the Special Supplemental Food actuality, however, it may be the unobserved
Program for Women, Infants, and Children), and relationships among these factors that influence
other types of health promotion interventions her final decision to participate in an enhanced
(Gehshan et al., 2009). Unfortunately, attempts prenatal care program.
to evaluate the effectiveness of comprehensive A variety of complex, hidden personal factors
prenatal care coordination have had mixed results may not be reflected in the medical or intake re-
(Buescher et al., 1991; Cain, 2006; Cain, 2007; cords. These may include health or domestic issues
Clarke et al., 1993; Herman & Berendes, 1996; that the woman is not willing to reveal or able to
Korenbrot & Patterson, 1995; Schulman et al., discern. For example, a woman at higher risk for
1997; Nason & Alexander, 2002). a poor birth outcome may be facing stressors in
the local environment such as violence, use of

86
Using a Neural Network to Predict Participation in a Maternity Care Coordination

alcohol or illicit drug activity in the community. 355,666 observations from the years 2000-2002
These factors may influence her choice to par- resulted in a sampling frame of 137,249 Med-
ticipate, even while remaining unobserved by the icaid births, including 57,635 births to women
healthcare provider. In this study we test whether participating in the MCC program. We randomly
neural network learning is able to model complex selected 60,000 of these observations for use in
nonlinear patterns between a diverse set of predic- the study’s data sample.
tor variables and the choice to participate in MCC.

predIctor VarIaBles
study data
The state of North Carolina collects up to 300
This study uses data developed by the North pieces of maternal and infant information for each
Carolina State Center for Health Statistics (SCHS) birth included in the Composite Linked Birth File.
called the Composite Linked Birth File. The We selected nineteen variables that previously
Composite Linked Birth File is comprised of have been identified as important in predicting if
the linkage of a unique birth certificate record to a woman is at risk for a high risk pregnancy and,
any Medicaid-paid infant claims records, MCC therefore, more likely to participate in the MCC
and WIC enrollment records, and/or infant death program (Paneth, 1995; Schwethelm et al., 1989).
certificate record. The data include a census of all These predictor variables fall into three general
births in the state of North Carolina in years 2000- categories: 1) location, 2) overall health, and 3)
2002. The total number of North Carolina resident pregnancy-related health.
live births was 120,247 in 2000, 118,112 in 2001, Specifically, the predictor variables include
and 117,307 in 2002. Among these, Medicaid- mother’s zip code (MOMZIP), county of residence
paid births based on the infant’s Medicaid status (COUNTY), mother’s age (MOMAGE), mother’s
numbered 49,188 in 2000, 51,720 in 2001, and race (MOMRACE), last year of mother’s school-
48,883 in 2002. Births to women enrolled in the ing (MOMEDUC), marital status (MSTATUS),
Maternity Care Coordination program were 24,694 self-reported tobacco use (TOBACCO), self-
in 2000, 24,328 in 2001, and 19,637 in 2002. reported alcohol use (ALCOHOL), the month
Thus, approximately 50 percent of North Carolina prenatal care began (MMPNCBEG), presence of
Medicaid women who delivered during the study medical risk factor(s) (MEDRISK), and a previous
period were enrolled in the MCC program. death of a live newborn (PREVDETH). The input
The sample includes only Medicaid births variables include several other dummy variables
to white and African-American women aged (0 or 1) indicating the presence of specific medical
15-45 years. Three additional selection criteria conditions such as anemia (ANEMIA), cardiac
were established to address issues of bias in the problems (CARDIAC), diabetes (DIABETES),
sample selection. First, women who enrolled in chronic hypertension (HYPERCH), pregnancy-
the MCC program after 32 weeks gestation were related hypertension (HYPERPR), eclampsia
not included, thereby excluding late joiners. Sec- (ECLAMPSIA), pre-term birth or small for gesta-
ond, only live singleton births were included. And tion age in a previous pregnancy (SGA), and renal
third, births to women who received no prenatal disease (RENAL). MCC is the outcome variable,
care were excluded. In addition to the selection where 1 indicates that the woman chooses to
criteria described above, records with missing participate in the MCC program and 0 indicates
data for any study variables were excluded. The the choice not to participate.
application of the selection criteria to the original

87
 Using a Neural Network to Predict Participation in a Maternity Care Coordination

artIFIcIal neural weighted connections in the input and hidden

networK Model layers. In the beginning the weight values are
randomly set, the input values of the first case in
Back propagation network learning is a commonly the training set are presented to the network’s
used algorithm that has been proven to be a reli- input layer, and the calculated output of each
able tool for general classification applications neuron is fed forward through the network until
(Alspan et al., 1995; Hornik et al., 1989; Medsker an output value is derived. Next, the computed
& Liebowitz, 1994; White, 1990). For this study, output value is compared to the actual value of
the NeuroShell2 software package was used to the output variable in the training case. The error
define and test a three layer, feed forward, back value (the squared difference between actual and
propagation Artificial Neural Network model. calculate output) is then propagated backward
Figure 1 shows a simple propagation network through the network and small adjustments are
of the type used in this study. In the diagram, the made to the weights so that, in the future, the
circles are neurons (mathematical processing network will come closer to calculating the correct
units) and the lines connecting the neurons are nu- output value. This process is repeated for each
meric weights. The models are typically composed case in the training set, and training set is processed
of: 1) a layer of input neurons, wherein each input over and over while the weights are continually
neuron represents one predictor variable, 2) one tweaked.
or more hidden layers of weighting neurons, and Periodically, a holdout (test) set of cases is run
3) an output layer containing a neuron for each against the network to determine if the predictive
dependent variable to be predicted. Our neural power of the network is still improving. The net-
network model includes an input layer with 19 work will never learn an exact predictive function,
neurons, one hidden layer containing 100 neu- but it will slowly approach one. The purpose of
rons, and an output layer containing one neuron the test set is to make sure that the network is not
to predict MCC participation. over trained in recognizing relationships within its
During the training phase of network develop- training set at the expense of loosing its predictive
ment, the ANN must be trained to predict actual power with new data.
output values based a correctly described set of The original sample used in this study con-
sisted of 60,000 observations. This number is a
limitation of the NeuroShell2 software. After the
Figure 1. Diagram of simple propagation artificial sample was selected one observation was found to
neural network have a 4-digit zip code and was dropped, leaving
a total of 59,999 observations. Table 1 presents
the descriptive statistics for the variables used in
the analysis.
Seventy percent (70%) of the observations
(42,000 patterns) were used to train the network
and twenty percent (20%) of the observations
(12,000 patterns) were used as a test set to evalu-
ate model training. Ten percent (10%) of the
observations (5,999 patterns) were used as a
production set to validate the trained model. The
neural network software randomly selected the
observations for each of these three data sets.

88
Using a Neural Network to Predict Participation in a Maternity Care Coordination

Table 1. Descriptive statistics for variables

Variable Min. Value Max. Value Mean Std.Dev.

Zip Code 22717 65109
County 0 100
Mother’s age 15 45 23.57821 5.350374
Mother’s Race 1 2 1.361606 0.480470
1= White; 2= Afr.Am.
Mother’s yrs. schooling 0 16 11.42652 5.421363
Marital Status 1 2 1.600160 0.489869
1=married; 2= single
TOBACCO 1 2 1.793197 0.527171
1=yes; 2 = no
ALCOHOL 1 2 1.867 0.347672
1=yes; 2 = no
MMPNCBEG 0 9 2.808147 1.577052
Dummy Variables
0=no; 1=yes
MEDRISK 0 1 0.123819 0.329377
PREVDETH 0 1 0.018384 0.134335
ANEMIA 0 1 0.027617 0.163875
CARDIAC 0 1 0.003533 0.059338
DIABETES 0 1 0.024734 0.155314
HYPERCH 0 1 0.008467 0.091626
HYPERPR 0 1 0.050718 0.219422
ECLAMPSI 0 1 0.004917 0.069948
SGA 0 1 0.012284 0.110149
RENAL 0 1 0.003217 0.056625
MCC 0 1 0.429857 0.495060

Training continued through 35 epochs of the Following the table is a narrative description of
training set. The model’s weights were evaluated the statistics.
against the test set every 200 events. After 62,000 The coefficient of multiple determination, R2,
events had been processed with no changes to the compares the accuracy of the model to a bench-
weight structure, training was stopped. Finally, the mark model consisting of the mean of all the
production set was processed through the trained samples. An R2 value of 1 indicates a perfect fit,
model to produce the network’s predictions for near 1 indicates a very good fit, and near or below
each pattern in the production file. zero indicates a very poor fit.
The square of the correlation coefficient, r2,
provides a measure of the strength of the relation-
results ship between the actual and predicted outputs. An
r2 closer to 1 indicates a strong linear relation-
NeuroShell2 provides a number of statistical tools ship, while an r2 closer to 0 indicates no linear
to assess the predictive ability of the network relationship.
model. Table 2 presents a summary of their values.

89
 Using a Neural Network to Predict Participation in a Maternity Care Coordination

Table 2. Production set assessment statistics

predicting the number of eligible recipients of
Output 5999 patterns public healthcare services becomes a significant
R 2
0.9917
challenge for program managers, especially when
correlation coefficient r 0.9960
participation is voluntary. To meet this challenge,
r2 0.9920
the use of predictive models has become much
more important in an ever-widening variety of
Mean Squared Error 0.0020
public health programs. Conventional predic-
Mean Absolute Error 0.0120
tion methods, which excel at identifying causal
Min. Absolute Error 0
effects and linear relationships among variables,
Max. Absolute Error 0.8460
require a priori model specification and may not
Percent within 5% 41.357
be able to detect underlying complex relationships
Percent within 5% to 10% 0.1000
among predictor variables. When prediction ac-
Percent within 10% to 20% 0.0170
curacy is more important than assessing causal
Percent within 20% to 30% 0.1830
effects, Artificial Neural Networks offer the user
Percent over 30% 0.7830 a powerful prediction tool without the need to
understand all the subtle relationships that may
exist within the target population. Additionally,
Min Absolute Error is the minimum absolute significant advances in computer hardware and
value of (actual - predicted) for all patterns in software technology have made Artificial Neural
the production set. Max Absolute Error is the Networks an increasingly powerful, available,
maximum absolute value of (actual - predicted) affordable, useful and user-friendly tool. Public
for all patterns. Mean Absolute Error is the mean healthcare resource allocation and planning can
of the absolute value of (actual - predicted) for be greatly enhanced by the predictive power of
all patterns. Mean Squared Error is the mean of ANNs, particularly in terms of voluntary program
(actual – predicted)2 for all patterns. participation.
NeuroShell2 also lists the percent of network This study develops an Artificial Neural Net-
predictions that differ from the actual answers work to predict voluntary enrollment by Medicaid
within a range of specified percentages. When an women in North Carolina’s Maternity Care Coor-
actual answer is 0, a percentage difference can- dination (MCC) program for enhanced prenatal
not be calculated. As a result, the total computed care. Observations from 59,999 randomly selected
percentages may not total 100%. Medicaid births from the years 2000-2002 were
The 5999 patterns in the production set included used in the development and testing of the ANN.
observations for 3453 women who rejected par- The model contained 19 predictor variables related
ticipation in MCC and 2545 women who chose to to location, overall health, and pregnancy-related
participate in MCC. The model correctly predicted health to predict voluntary participation in the
all 3453 non-participants (100%) and 2537 of enhanced prenatal care program. The model cor-
2545 participants (99.69%). rectly predicted the mother’s choice to participate
in MCC over 99% of the time.
Using the superior predictive power of Neural
conclusIon Networks, managers can make better funding,
staffing, and other resource allocation decisions
States face increasing pressure to limit growth based on timely, correct predictions of program
in their Medicaid budgets while still meeting the participation. Neural Network models can be
healthcare needs of vulnerable populations. Thus, updated as new information is obtained and re-

90
Using a Neural Network to Predict Participation in a Maternity Care Coordination

trained to recognize new relationships among Buescher, P. A., & Horton, S. J. (2002). Prenatal
predictor variables that may result from evolv- WIC participation in relation to low birth weight
ing changes in participant characteristics and/or and Medicaid infant costs in North Carolina-An
program modifications. update. Center for Health Informatics and Sta-
This research path may be extended in many tistics Studies, 122, 1–9.
directions. Since so little research has been pub-
Buescher, P. A., Roth, M. S., & Williams, D.
lished on the use of Artificial Neural Networks in
(1991). An evaluation of the impact of maternity
the prediction of voluntary participation in public
care coordination on Medicaid birth outcomes
healthcare programs, one obvious extension is to
in North Carolina. American Journal of Pub-
expand the study of ANNs in the prediction of
lic Health, 81(12), 1625–1629. doi:10.2105/
voluntary participation in other types of healthcare
AJPH.81.12.1625
programs. Other extensions of the study could also
be made to predict voluntary program participa- Cain, M. (2006). A Two-step estimation appli-
tion in other areas outside of health services, such caton in health care management evaluating the
as education, child care, utility subsidies, and effects of choice of public health plan on health
transportation. outcomes. Journal of Academy of Business and
Economics, 6(2), 65–72.
Cain, M. (2007). Self-selection and evaluation
reFerences
of health care programs. Journal of Academy of
Alkan, A., Koklukaya, E., & Subasi, A. (2005). Business and Economics, 7(3), 82–89.
Automatic seizure detection in EEG using logistic Chen, P., Huang, D., Zhao, X., & Li, X. (2008).
regression and artificial neural network. Journal Predicting contact map using Radial Basis Func-
of Neuroscience Methods, 148(2), 167–176. tion Neural Network and Conformational Energy
doi:10.1016/j.jneumeth.2005.04.009 Function. International Journal of Bioinformat-
Alpsan, D., Towsey, M., Ozdamar, O., Tsoi, A. C., ics Research and Applications, 4(2), 123–136.
& Ghista, D. N. (1995). Efficacy of modified back- doi:10.1504/IJBRA.2008.018340
propogation and optimization methods on a real- Chesney, T., Penny, K., Oakley, P., Davis, S.,
world medical application. Neural Networks, 8(6), Maffulli, N., & Templeton, J. (2006). Data Min-
945–962. doi:10.1016/0893-6080(95)00034-W ing Medical Information: Should Artificial Neural
Ari, S., & Saha, G. (2008). Classification of heart Networks Be Used To Analyze Trauma Audit Data.
sounds using empirical mode decomposition International Journal of Healthcare Information
based features. International Journal of Medi- Systems and Informatics, 1(2), 51–64.
cal Engineering and Informatics, 1(1), 91–108. Chiu, R., & Kuo, C. (2006). A decision support
doi:10.1504/IJMEI.2008.019473 model of demand forecast for national immuni-
Baxt, W. G. (1991). Use of an artificial neural zation vaccines. International Journal of Elec-
network for the diagnosis of myocardial infarction. tronic Healthcare, 2(1), 92–104. doi:10.1504/
Annals of Internal Medicine, 115(11), 843–848. IJEH.2006.008699

Baxt, W. G., Shofer, F. S., Sites, F. D., & Hol- Clarke, L. L., Miller, M. K., & Mahan, C. S.
lander, J. E. (2002). A neural computational aid (1993). The Effectiveness of Florida’s Improved
to the diagnosis of acute myocardial infarction. Pregnancy Outcome Program. Journal of Health
Annals of Emergency Medicine, 39(4), 366–373. Care for the Poor and Underserved, 4(2), 117–132.
doi:10.1067/mem.2002.122705

91
 Using a Neural Network to Predict Participation in a Maternity Care Coordination

Crawford, A. G., Fuhr, J. P., Clarke, J., & Hubbs, Hornik, K., Stinchcombe, M., & White, H. (1989).
B. (2005). Comparative Effectiveness of Total Multilayer feedforward networks are universal
Population versus Disease-Specific Neural Net- approximations. Neural Networks, 2(5), 359–366.
work Models in Prediciting Medical Costs. Dis- doi:10.1016/0893-6080(89)90020-8
ease Management, 8(5), 277–287. doi:10.1089/
Kaur, H., & Wasan, S. K. (2006). Empirical study
dis.2005.8.277
on Applications in Data Mining Techniques in
Dalmadge, C. L., & Cain, M. (2008). Utilizing Healthcare. Journal of Computer Science, 2(2),
Neural Networks o Assess Community-Level 194–200. doi:10.3844/jcssp.2006.194.200
Vulnerability to the Methamphetamine Epidemic.
Korenbrot, C. C., Gill, A., & Patterson, E. (1995).
Issues in Information Systems, 9(2), 258–266.
Evaluation of California’s statewide implementa-
Das, A., Ben-Menachem, T., Cooper, G. S., Chak, tion of enhanced perinatal services as Medicaid
A., Sivak, M. V. Jr, Gonet, J. A., & Wong, R. C. benefits. Public Health Reports (Washington,
(2003). Predicition of outcome in acute lower-gas- D.C.), 110(2), 125–133.
trointestinal hemorrahage based on and artificial
Kudyba, S., Hamar, G. H., & Gandy, W. M.
neural network: Internal and external validation
(2006). Utilising neural network applications to
of a predictive model. Lancet, 362(9392), 1261–
enhance efficiency in the healthcare industry:
1266. doi:10.1016/S0140-6736(03)14568-0
prediciting populations of future chronic illness.
Eftekhar, B., Mohammad, K., Ardebili, H. E., International Journal of Business Intelligence
Ghodsi, M., & Keabchi, E. (2005). Comparison and Data Mining, 1(4), 371–383. doi:10.1504/
of artificial neural network and logistic regression IJBIDM.2006.010780
models for prediction of mortality in head trauma
Lisboa, P. J., & Taktak, A. F. G. (2006). The use
based on clinical data. BMC Medical Informatics
of artificial neural networks in decision support
and Decision Making, 5(3), 1–8.
in cancer: A systematic review. Neural Networks,
Gehshan, S., Belnap, D., & Snyder, A. (2009). 19, 408–415. doi:10.1016/j.neunet.2005.10.007
Medicaid Outreach and Enrollment for Pregnant
Medsker, L., & Liebowitz, J. (1994). Design
Women: What Is the State of the Art?Washington,
and Development of Expert Systems and Neural
DC: Urban Institute.
Networks. New York: McMillan.
Goss, E., & Vozikis, G. S. (2002). Improv-
Najumnissa, D. & ShenbagaDevi, S. (2008).
ing Health Care Organizational Management
Intelligent identification and classification of
Through Neural Network Learning. Health
epileptic seizures using wavelet transform. In-
Care Management Science, 5(3), 221–227.
ternational Journal of Biomedical Engineering
doi:10.1023/A:1019760901191
and Technology, 1(3), 293–314. doi:10.1504/
Herman, A. A., & Berendes, H. W. (1996). The IJBET.2008.016963
evaluation of the effectiveness of a community-
Nason, C. S., Alexander, G. R., Pass, M. A., &
based enriched model of prenatal intervention
Bolland, J. M. (2001). An evaluation of a Med-
project in the District of Columbia. Health Services
icaid managed maternity program: The impact
Research, 33(5), 609–621.
of comprehensive care coordination on utiliza-
tion and pregnancy outcome. Journal of Health
and Human Resources Administration, 24(3/4),
493–521.

92
Using a Neural Network to Predict Participation in a Maternity Care Coordination

Okoroh, M. I., Ilozor, B. D., & Gombera, P. (2007). Schwethelm, B., Margolis, L. H., Miller, C., &
An artificial Neural Networks approach in man- Smith, S. (1989). Risk Status and Pregnancy
aging healthcare. International Journal of Elec- Outcome among Medicaid Recipients. American
tronic Healthcare, 3(3), 279–292. doi:10.1504/ Journal of Preventive Medicine, 5(3), 157–163.
IJEH.2007.014548
Seomun, G., Chang, S. O., Lee, S. J., Kim, I. A.,
Paneth, N. S. (1995). The Problem of Low Birth & Park, S. (2006). A Prediction Model for Patient
Weight. The Future of Children: Low Birth Weight, Classification according to Nursing Need: Using
5(1), 21–34. Data Mining Techniques . In Park, H., Murray,
P., & Delaney, C. (Eds.), Consumer-Centered
Pehlivanli, A. C., Ibrikci, T., & Ersoy, O. K. (2008).
Computer-Supported Care for Healthy People.
Consensual classification of drug and nondrug
Fairfax, VA: IOS Press.
compounds. International Journal of Computa-
tional Biology and Drug Design, 1(3), 224–234. Subasi, A., & Ercelebi, E. (2005). Classification
doi:10.1504/IJCBDD.2008.021416 of EEG signals using neural network and logistic
regression. Computer Methods and Programs
Poonguzhali, S., & Ravindran, G. (2007). Evalu-
in Biomedicine, 78(2), 87–99. doi:10.1016/j.
ation of feature extraction methods for classifica-
cmpb.2004.10.009
tion of liver abnormalities in ultrasound images.
International Journal of Biomedical Engineering Wang, H., Zheng, H., & Hu, J. (2008). Poisson
and Technology, 1(2), 134–143. doi:10.1504/ approach to clustering analysis of regulatory
IJBET.2007.015856 sequence. International Journal of Computa-
tional Biology and Drug Design, 1(2), 141–157.
Samanta, B., & Nataraj, C. (2008). Automated
doi:10.1504/IJCBDD.2008.020206
diagnosis of cardiac state in healthcare systems
using computational intelligence. International White, H. (1990). Connectionist nonparametric
Journal of Services Operations and Informatics, regression: multilayer feedforward networks can
3(2), 162–177. doi:10.1504/IJSOI.2008.019331 learn arbitrary mappings. Neural Networks, 5(6),
535–549. doi:10.1016/0893-6080(90)90004-5
Schulman, E. F., Sheriff, D. J., & Momany, E.
(1997). Primary Case Management and Birth Zurada, J., & Lonial, S. (2005). Comparison Of
Outcomes in the Iowa Medicaid Program. The Performance Of Several Data Mining Methods
American Journal of Public Health, 87(1), 80–84. For Bad Debt Recovery In The Healthcare Indus-
doi:10.2105/AJPH.87.1.80 try. The Journal of Applied Business Research,
21(2), 37–53.

93
94

Chapter 7
Can IT Act as a Catalyst
for Change in Hospitals?
Some New Evidence
Teemu Paavola
LifeIT Plc, Finland

aBstract
This chapter presents a succesful reorganization of a patient care process that was carried out in a
middle sized Finnish hospital. The reorganization of the patient care process for joint replacement sur-
gery succeeded in achieving a 50 per cent increase in operations. This study proposes that IT may have
an indirect influence on the achievement of goals, such as productivity, as soon as the IT investment has
been decided upon; in other words, IT benefits start accruing before the IT component is even in place.
This is a new feature to add to the previous definitions, because this particular benefit cannot be logi-
cally derived from any of the features of the actual IT system. Paying enough attention to this phenomen
at the planning stage can be vital to the success of new IT system investment.

IntroductIon with business goals. There are, however, many

phenomena at play shaping the practices of the
In the last two decades, the assessment of the health care sector and identifying and allowing
benefits of IT has given rise to an interactive for these phenomena may be the key to successful
dialogue between management sciences and IT system projects, indeed even more important
information systems science in particular, but in than the technology itself.
health care the subject has received little atten- The literature on both IT management and
tion. In the field of health care, IT investments process development is quite unanimous in its
are still seen as primarily an acquisition to replace belief that both are necessary for achieving more
earlier technology and expand current use, not efficient operation and a productivity increase.
as an investment project to be managed in line Therefore an IT system project is often a change
project by nature, which can make it challenging
DOI: 10.4018/978-1-61692-002-9.ch007 particularly in the field of health care (Berg, 2001;

Copyright © 2011, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Can IT Act as a Catalyst for Change in Hospitals?

Littlejohns, Wyatt, & Garvican, 2003), where process ManageMent

resistance to change is virtually a characteristic tHeorIes
of the profession (Weick & Sutcliffe, 2003).
The Act on Specialized Medical Care concern- The term ‘process thinking’ refers to a number
ing the maximum times to arrange treatment, of management theories that have been used by
which came into force in Finland in March 2005, industry in its quest for better operating processes
has made many healthcare units look at the arrange- over the last few decades. In many of these, the
ment of the services they produce in a new light. use of IT also has a significant role. Indeed, IT
Particular attention is fixed on the legal obligation has become more important in a number of areas,
concerning the waiting times between treatment including health care; yet process thinking has not
decisions and treatment measures, which is to be always been employed.
no more than six months. The need to increase The populations of Europe and the Americas
the number of operations has become a matter of are ageing quickly. The healthcare system is strug-
current debate particularly in orthopaedics, where gling with the combination of rising demand and
the length of queues has become unlawfully long escalating costs in specialist medical care, while
at several hospitals in Finland. Improvements at the same time, there is strong support for re-
in controlling the queues have previously been duced public-sector healthcare spending but firm
achieved by the more efficient handling of refer- rejection of any cuts in service levels. If the two
rals (Harno et al., 2000), but with orthopaedics targets are to become reality simultaneously, the
this was felt to be ineffective (Harno et al., 2001). methods enabling them to be achieved should be
In special operative areas, making use of all the chosen on the basis of how deep the cuts should be.
development potential available within the tradi- Cosmetic improvements would be fairly pain-
tional treatment chains should be explored as a less: for example, Total Quality Management
permanent remedy, after first-aid obtained in the (Crosby, 1979; Deming, 1991) would result in
form of outsourced services. long-term improvements in operating processes
This chapter illustrates a case where process as a more efficient use of resources would bring
management and process development tools gradual savings. Some scholars have, however,
where exploited to support new ways of work and likened some quality management theories to a
improve productivity in healthcare. In Finland, rain dance (Schaffer & Thomson, 1992). In their
Seinäjoki Central Hospital implemented a project view they look good, sound good and allow those
to revise processes in order to reduce queues in involved to feel good, while at the same time they
surgery, particularly artificial-joint surgery. The may have no influence on the rain itself. There
project in question was originally classified as an are also other management theories in the field
IT project which also incorporated process devel- of process thinking.
opment. Over the course of the project, however, According to the time-based management ap-
the balance between the two components shifted, proach, all development should focus on process
and there was no time to incorporate the new IT lead-time (Stalk & Hout, 1990). In such an ap-
system before the changes were implemented. This proach all other positive aspects, improved quality,
chapter describes the results of the experiment for cost savings and customer satisfaction will follow
the benefit of, for example, other operation units automatically. However, development measures
that are taking a close look at their operations and do not need to mean squeezing more out of the
of developers and management in health care as stages intended to boost the value of the treatment
support for decision-making. Parts of the study process. In fact, industrial companies have been
have been published in Finnish language (Jokipii able to find larger savings in the way they use the
et al., 2006). time that brings no added value, which, after all,

95
 Can IT Act as a Catalyst for Change in Hospitals?

accounts for more than 95% of the total (Stalk & for replacing systems already in use, yet they still
Hout, 1990). resulted in significant increases in productivity.
In contrast to Total Quality Management, which Anderson et al. proposed this as a new, inverse
emphasizes continuous development, Business phenomenon to be explained, calling it the New
Process Reengineering (BPR) proposes a radical Productivity Paradox. According to their theory,
revision of the business process. The aim is to start the weak IT impact on productivity especially im-
from scratch without the burden of old operating mediately after the turn of the millennium (2000-
approaches (Oliver, 1993; Hammer & Champy, 2002) can be attributed to companies simply not
1993). The reengineering starts with a definition investing enough in IT, in direct contradiction to
of the desired end result. This will form the basis the IT Productivity Paradox.
for the planning of the new process functions and Lee and Menon (2000) argue that hospitals that
sequences. The aim is to maximize value-adding are characterized by hight technical efficiency are
functions and to get rid of all operations not add- no more productive than hospitals characterized
ing to the value. Extensive use of the information counterwise. In fact, in the hospitals studied IT
technology is often used as the means for achieving capital seemed to have a negative correlation to
the desired results. productivity. They explain this by the fact that
Effecting the operational changes required by although hospital’s processes may have been ef-
BPR has been somewhat problematic. Resistance ficient, resource allocation and budgeting between
to change, which is inherent in human nature, and various categories of capital and labor had not been
the fact that reengineering is often a zero-sum efficient. Devaraj and Kohli (2000) believe that
game, make the implementation of the change the effect of IT on performance can been seen only
process more difficult (Buchanan, 1997). The after a time lag and cannot necessarily be observed
theory, though somewhat worn-out, is still useful, in cross-sectional or snapshot data analyses. With
as it underlines the importance of the information data collected from eight hospitals, their study
technology in performance improvement. Great indicates support for the impact of technology
potential for applying the theory and informa- contingent on BPR practiced by hospitals.
tion technology can be found in sectors that for
ages have relied on well-entrenched operating
models, such as health care (Evans, Hwang, & startIng poInts For
Nagarajan, 1997). cHange and experIMent
The importance of IT for productivity in differ-
ent organizations has been discussed for decades. Seinäjoki Central Hospital wanted to reorganize
The discussion led by Straussmann (1990) and the operations for artificial-joint patients so that
Brynjolfsson (1993) has particularly focused on three operations could be performed in the same
explaining what is known as the IT Productivity operating room in the course of a normal day’s
Paradox. Although Brynjolfsson et al. (1993 and work instead of two. The introduction of a new
1996) later declared that the problem had disap- IT system was also planned as part of the project.
peared by 1991, not all researchers have agreed, The experimental period lasted from November
and interest in explaining the phenomenon remains 2004 to November 2005.
high — so much so that Texas-based reseacher The revision of the treatment process utilized
Mark Anderson et al. (2003) proposed a new IT process thinking and process development tools.
Productivity Paradox to replace the original. They Of these, the Theory of Constraints (Goldratt,
observed a growth in the market worth of busi- 1990) was thought the best applicable for examin-
nesses after Y2K investments leading up to the turn ing the process for treating artificial-joint patients.
of the millennium. These investments were made The point in this approach is to identify those stages

96
Can IT Act as a Catalyst for Change in Hospitals?

in the process that dictate the maximum current mum number of 200 operations recommended by
throughput. By allocating additional resources the Ministry of Social Affairs and Health is only
and development action to these bottlenecks, the exceeded in 25 units. Every year the Seinäjoki
throughput can be improved without needing to Central Hospital performs between 550 and 600
interfere in the other stages of the process. The artificial-joint operations.
main change for increasing the usage of the op- In the study, quantitative material was collected
erating room capacity was transferring the anaes- from the operating days in the experimental period
thetic stage from the operating room to separate on which three artificial-joint operations were
induction facilities. Experiments on this had been carried out (147 patients); because of the small
reported earlier in medical journals (Hanss et al., number of orthopaedists, there were 2-3 of these
2005; Sandberg et al., 2005; Torkki et al., 2005). days in a week. Comparative material consisted of
In the new arrangement, the anaesthetic stage the days on which two artificial-joint operations
was transferred outside the operating room. At the were carried out between January 1 and June 30,
same time, one anaesthesia nurse was added to 2004 (54 patients).
the operating team, working both in the operating The time when patients were in the operating
room and in anaesthetic. Another anaesthetic nurse room and changeover times were recorded in the
took the next patient in good time to the recovery operation database. The time-monitoring material
room or to the operating room’s induction facili- consisted of the times when the operating room
ties to be anaesthetized. As soon as the operating was in use. The median time that patients were in
room was cleaned after the previous operation, the the operating room and the median changeover
next patient could be prepared for surgery. The time, when there is no patient in the operating
next patient was brought to the operating room room, were used for comparison purposes.
already anaesthetized and in the correct position Qualitative material was collected through
for the operation. interviews during the experimental period and
The duties of the orthopaedist that were not by means of a work-satisfaction questionnaire
part of the operations or preparation for them were carried out among doctors and nurses a year after
scheduled outside the operation days. Thus, the sur- the experiment started.
geon whose turn it was to operate was able to focus The new operating model made it possible to
exclusively on the work in the operating room. At carry out three operations during a normal working
the beginning of the experimental period, the same day (see Figure 1). The orthopaedists examined
orthopaedist operated for one week at a time, but the patients during a pre-operative visit or on the
this practice had to be changed so that the operation day preceding the operation. The first patient of
days were rotated among different practitioners. At the morning was in the operating room in time,
the beginning of 2005, there were five orthopaedists and the operation started on time at 8.30 a.m. The
working at Seinäjoki Central Hospital. anaesthetization stages for the second and third
patients, which were carried out staggered with
the operation, took slightly longer than if carried
MaterIal, MetHods out in the operating room.
and results As it was possible to separate some of the steps
previously carried out in the operating room and
In Finland, every year about 6,800 artificial joint have them done outside, the hospital succeeded
operations are carried out on the hip and some in increasing the throughput of the process by
7,200 on the knee, and there are more than 1,700 50%, even though the usage capacity of the op-
instances of further surgery. These operations are erating room remained almost the same.
performed in almost 70 hospitals, but the mini-

97
 Can IT Act as a Catalyst for Change in Hospitals?

Figure 1. The old arrangement and the new operating model for artificial joint surgery

Adding fourth nurse to the operating team (now practice where outsourced services or increasing
2 in anaesthetization and 2 in the operation) made the number of a hospital’s own operating rooms
it possible to shorten the changeover times con- are seen as the only options for increasing output.
siderably: the average time was reduced from 54 From our experiences the throughput of the process
minutes to 13 minutes. This was because the team for artificial-joint operations can be increased
was able to take coffee and meal breaks in turn. while the usage capacity of the operating room
One of the operation nurses was able to help the remains the same or even decreases. Focusing
orthopaedist as necessary. In the three-operation the operations on one operating room proved to
model, anaesthetizing the second or third patient of be effective.
the day in separate facilities reduced the time the An increase in the throughput of the opera-
patient was in the operating room by 20 minutes tion process was sought without increasing the
(149 minutes vs. 129 minutes). workload of the staff. The hospital succeeded in
According to the questionnaire, fifty per cent doing this by firstly dealing with idle waiting.
or more of the doctors who took part in the experi- Targeting greater efficiency here and a simulta-
ment felt that the meaningfulness of their work neous improvement in the throughput required
and work motivation had increased and thought development in several areas, e.g. adding one
that the three-operation experiment should become nurse to the operating team, a bigger work con-
a permanent fixture. The nursing staff felt that tribution from the hospital attendant in preparing
minimizing the idle waiting improves the atmo- patients, preparing the anaesthetic in a new way
sphere and increases work motivation to some and changing the orthopaedist’s work schedule.
extent. The doctors felt the new operating model The justification for adding one nurse was that in
improves the meaningfulness of the work and the revised staggered operation stage, there was
work motivation more than the nursing staff did. also one patient more.
It was not possible to anticipate all the ef-
fects of the change. In order to ensure that things
dIscussIon went smoothly, specialist experienced doctors
acted as anaesthetists and orthopaedists during
The usage capacity of the operating room is gener- the experimental period, but at the same time
ally considered to be the bottleneck in the opera- the arrangement narrowed the opportunities for
tion process. This generalization leads easily to a training specializing doctors. Furthermore, not

98
Can IT Act as a Catalyst for Change in Hospitals?

enough preparation was made for the increase and to make it easier to organize follow-up visits
in the number of operations at all stages of the in a timely fashion and based on actual needs. The
treatment process. At times, the growth in the new IT system would further reduce the workload
throughput caused congestion on the ward and at the hospital by making it possible to carry out
especially in further treatment at health centres. post-operative patient monitoring at local health
In financial terms, the transfer to the practice centres rather than at the hospital outpatient clinic.
of three operations was worthwhile. The resources
for arranging three operations were obtained prin-
cipally by utilizing the fixed costs of the hospital conclusIon
more efficiently. In alternative cost accounting
comparing the additional cost caused by a hospi- This chapter presents a succesful reorganization
tal’s own activities with the cost of an artificial- of the care process for an artificial joint patient.
joint operation acquired from the private sector or The project by a middle sized Finnish hospital
another provider (minus the costs of the prosthesis, offers an encouraging example of a way to exploit
materials and cost of the treatment days) shows a process management tools in health care. Seinäjoki
difference of some USD 4,000 between the hospi- Central Hospital succeeded in obtaining a 50%
tal’s own work and outsourcing with regard to the increase in flow-through in the process for treat-
added third primary operation per day. Because ing artificial-joint patients with the transfer of the
of the limited number of orthopaedists, however, anaesthesia stage outside the operating room in the
it was not possible in the experimental period to reorganization. For every two joint replacement
run ‘flat out’ five days a week. operations previously conducted, there were now
15 complete operations a week would be three operations performed in the same theatre
enough to meet the need for artificial-joint sur- and in a normal working day. In the longer term,
gery in the entire hospital district, and the revised the arrangement would mean that in Finland the
treatment process would generate annual savings entire country’s need for artificial-joint surgery,
of between USD 700,000 and USD 800,000 for about 15,600 operations per annum, could be dealt
the Hospital District of South Ostrobothnia, even with in 30 operating rooms. This could consider-
taking into account the additional recruitment ably streamline the publicly-financed health care
required. Savings come from cutting back on system in Finland, as these operations are currently
services purchased from private hospitals, as the performed in almost 70 hospitals.
hospital itself can now perform a larger percentage The introduction of the new patient care process
of the joint replacement surgery required. demonstrated that the operating theatre capacity
Although planned as part of the project, the was not causing a bottleneck, but that it was the
introduction of the new IT system did not take place orthopaedic surgeons brought in at the various
during the present change project. The reason for intervals who formed the key resource. The reor-
this was that the software was not complete when ganized care process for patients requiring joint
the project was launched. Nevertheless, the IT replacement surgery should produce annual cost
system appears to have played a noteworthy role savings of USD 700,000 to USD 800,000 for the
in the launching of the project because key players Seinäjoki Central Hospital. Following the experi-
whose commitment was essential to the successful ence gained in the project, the Seinäjoki Central
implementation of the change were motivated by Hospital has decided to adopt the project model
the eventual benefits of the IT system to support on a permanent basis. A similar reorganization
the project. The IT system was intended to work is also possible in other hospital districts. This
as a tool for monitoring patients in rehabilitation

99
 Can IT Act as a Catalyst for Change in Hospitals?

observation, however, should only be applied to Brynjolfsson, E., & Hitt, L. (1996). Paradox lost?
orthopaedic joint replacement surgery. Firm level evidence on the returns to information
An interesting detail in this change project was systems spending. Management Science, 42(4),
that the role of the IT system itself was completely 541–558. doi:10.1287/mnsc.42.4.541
marginal. All kinds of justifications for investing in
Buchanan, D. (1997). The limitations and oppor-
IT systems can be given, from economic calcula-
tunities of business process re-engineering in a po-
tions to managerial intuition (Paavola, 2007), but
liticized organizational climate. Human Relations,
in the change project described in this chapter, IT
50, 51–72. doi:10.1177/001872679705000103
seemed to act only as a catalyst (Paavola, 2008).
Perhaps the project could not have been success- Crosby, P. B. (1979). Quality is free. New York:
fully implemented to the extent it was had not an McGraw-Hill.
IT component, which “always requires changes
Deming, W. E. (1991). Out of crisis. Cambridge,
in the operational process”, been included. This
MA: Cambribge University Press.
observation of IT acting as a catalyst for perfor-
mance enhancement and hence improved produc- Devaraj, S., & Kohli, R. (2000). Information
tivity deserves further study. It would be worth technology payoff in the health-care industry:
exploring to what extent indirect impact should a longitudinal study. Journal of Management
be considered in making IT investments alongside Information Systems, 16(4), 41–67.
direct effects, and how this phenomenon could be
Evans, J., Hwang, Y., & Nagarajan, N. (1997). Cost
studied. Are there cases where IT investment was
reduction and process reengineering in hospitals.
motivated as leverage for achieving a particular
Journal of Cost Management, 11(3), 20–27.
change in the operating environment, even though
the same impact on productivity could have been Goldratt, E. M. (1990). Theory of constrains.
achieved without IT investment? Croton-on-Hudson. NY: North River Press.
Hammer, M., & Champy, J. (1993). Reengineer-
ing the corporation - A manifesto for business
reFerences
revolution. New York: Harper Business.
Anderson, M., Banker, R., & Ravindran, Hanss, R., Buttgereit, B., Tonner, P. H., Bein, B.,
S. (2003). The new productivity paradox. Schleppers, A., & Steinfath, M. (2005). Overlap-
Communications of the ACM, 46(3), 91–94. ping induction of anesthesia: An analysis of ben-
doi:10.1145/636772.636776 efits and costs. Anesthesiology, 103(2), 391–400.
Berg, M. (2001). Implementing information doi:10.1097/00000542-200508000-00023
systems in health care organizations: myths and Harno, K., Arajärvi, E., Paavola, T., Carlson,
challenges. International Journal of Medical C., & Arnala, I. (2001). Clinical effective-
Informatics, 64(2/3), 143–156. doi:10.1016/ ness and cost analysis of patient referral by
S1386-5056(01)00200-3 videoconferencing in orthopaedics. Journal
Brynjolfsson, E. (1993). The productivity paradox of Telemedicine and Telecare, 7(4), 219–225.
of information technology. Communications of the doi:10.1258/1357633011936435
ACM, 36(12), 67–77. doi:10.1145/163298.163309

100
Can IT Act as a Catalyst for Change in Hospitals?

Harno, K., Paavola, T., Carlson, C., & Viikinkoski, Sandberg, W. S., Daily, B., Egan, M., Stahl, J. E.,
P. (2000). Patient referral by telemedicine: effec- Goldman, J. M., Wiklund, R. A., & Rattner, D.
tiveness and cost analysis of an intranet system. (2005). Deliberate perioperative systems design
Journal of Telemedicine and Telecare, 6(6), improves operating room throughput. Anesthesi-
320–329. doi:10.1258/1357633001935996 ology, 103(2), 406–418. doi:10.1097/00000542-
200508000-00025
Jokipii, P., Kalliovalkama, K., & Paavola, T.
(2006). Tekonivelpotilaan hoitoprosessin uud- Schaffer, R., & Thomson, H. (1992). Successful
istaminen. [Finnish Medical Journal]. Suomen change programs begin with results. Harvard
Lääkärilehti, 61(14), 1626–1627. Business Review, 22(1), 80–89.
Lee, B., & Menon, N. (2000). Information tech- Stalk, G., & Hout, T. (1990). Competing against
nology value through different normative lenses. time - how time-based competition is reshaping
Journal of Management Information Systems, global markets. New York: The Free Press.
16(4), 99–119.
Strassmann, P. (1990). The business value of
Littlejohns, P., Wyatt, J., & Garvican, L. (2003). computers. New Canaan, CT: The Information
Evaluating computerised health information Economic Press.
systems: hard lessons still to be learnt. British
Torkki, P. M., Marjamaa, R. A., Torkki, M. I.,
Medical Journal, 326, 860–863. doi:10.1136/
Kallio, P. E., & Kirvela, O. A. (2005). Use of
bmj.326.7394.860
anesthesia induction rooms can increase the
Oliver, J. (1993, August). Shocking to the core. number of urgent orthopedic cases completed
Management Today, 18–22. within 7 hours. Anesthesiology, 103(2), 401–405.
doi:10.1097/00000542-200508000-00024
Paavola, T. (2007). Justifying IT system invest-
ments. International Journal of Healthcare Weick, K., & Sutcliffe, K. (2003). Hospitals as
Technology and Management, 8(1/2), 97–106. cultures of entrapment. California Management
doi:10.1504/IJHTM.2007.012104 Review, 45(2), 73–84.
Paavola, T. (2008). Exploring IT system benefits
in health care. Tampere University of Technol-
ogy Publication 756. Tampere: Tampereen Yli-
opistopaino.

101
102

Chapter 8
Informatics Application
Challenges for Managed Care
Organizations:
The Three Faces of Population
Segmentation and a Proposed
Classification System

Stephan Kudyba
New Jersey Institute of Technology, USA

Theodore L. Perry
Health Research Corporation, USA

Jeffrey J. Rice
Independent Scholar, USA

aBstract

Organizations across industry sectors continue to develop data resources and utilize analytic techniques
to enhance efficiencies in their operations. One example of this is evident as Managed Care Organiza-
tions (MCOs) enhance their care and disease management initiatives through the utilization of popula-
tion segmentation techniques. This article proposes a classification system for population segmentation
techniques for care and disease management and provides an evaluation process for each. The three
proposed operational areas for Managed Care Organizations are: 1) Risk Status: early identification
of high-risk patients, 2) Treatment Status: compliance with treatment protocols, and 3) Health Status:
severity of illness or episodes of care groupings, all of which require particular analytic methodologies
to leverage data resources. By applying this classification system an MCO can improve its ability to
clarify internal goals for population segmentation, more accurately apply existing analytic methodolo-
gies, and produce more appropriate solutions.

Copyright © 2011, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Informatics Application Challenges for Managed Care Organizations

IntroductIon Some of these include an increase in the aging

population, costs for defensive medicines, opti-
Population segmentation is the term broadly ap- mizing existing health care facility usage (e.g.,
plied to technologies used to correctly identify staffing doctors and nurses along with designated
and target the right patients for care and disease bed utilization rates) and the introduction of new
management program interventions. Many MCOs organizations such as HMOs and PPOs (Smith-
are using these technologies, including those Daniels, Schweikhart, & Smith-Daniels,1988).
containing predictive modeling techniques and One way Managed Care Organizations are at-
quantitative applications, to enhance patient tempting to improve their efficiencies in treat-
care and optimize available resources. With the ing illnesses is through the development and
rapidly growing use of population segmentation management of robust data resources and the
and predictive modeling today, it is essential to utilization of analytic techniques to identify pat-
understand the relative strengths and weaknesses terns and trends in patient populations. With this
of different types of population segmentation information, efficiency can be enhanced by more
techniques, including those employing predictive accurately identifying the sources of resource
methods. The purpose of this article is to propose demand of specific customer segments and initi-
a classification system for the different types of ating strategic health care management policies
population segmentation techniques and their and better allocating available resources to meet
usefulness in addressing the independent and those demands (Heskett, 1983; McLoughlin, Yan,
interactive roles of risk status, treatment status, & Van Deirdonck, 1995).
and health status in patient evaluation initiatives.
Organizations across industry sectors have analytic Methods for Health care
intensified their initiatives to increase operational Management
efficiency through effective resource allocation,
and the health care sector is no exception. Given The utilization of analytic techniques in strategic
the increased level of competition in today’s’ management is increasing (Shook, 2000). More
digital, information economy, organizations are formal analytic techniques such as stochastic trees
faced with the task of increasing productivity by have been utilized to help increase operational
more efficiently allocating available resources in efficiencies by enhancing the decision making
producing goods and services to meet the demands process in medical treatment procedures (Ha-
of their customers. One of the greatest issues facing zen, 1992, 2000). Other analytic methodologies
the health care industry today is managing patients involving data mining techniques enable decision
who suffer from chronic illnesses. Currently, ap- makers to identify patterns in clinical-, claims-
proximately 100 million Americans have at least and activity-based historical data, to better un-
one chronic condition and this is expected to rise derstand explanatory relationships in data and
to over 150 million Americans in the next 20 to create models to more accurately predict future
30 years (Faughty, 1999; Institute of Medicine, resource demand (Xiaohua, 2005). Artificial
2001). Furthermore, chronic conditions are the neural networks are computer algorithms that
leading cause of death, disability, and illness in the identify relationships in historical data that can
United States, accounting for over 75% of direct be used for classification and prediction (Bishop,
medical expenditures (Landro, 2002). 1995; Swingler, 1996). Reducing the uncertain-
The health care industry has been faced with ties in process resource requirements through
a number of additional factors that have increased enhanced predictive capabilities is seen to increase
the complexity of managing available resources. efficiency across industry sectors (Kudyba &
Hoptroff, 2001).

103
 Informatics Application Challenges for Managed Care Organizations

More effective disease management through ing to enhance care and disease management
data analysis can assist our current health care programs by correctly focusing resources and
system to fill the ever-increasing gaps in health interventions on the segment of a population who
care service, management, and treatment. Disease would benefit the most from interventions (Ash,
management programs utilize both cost-based and Zhao, Ellis, & Sclein, 2001; Cousins, Shickle, &
health-based rationales for prevention and health Bander, 2002; Ridinger & Rice, 2000). While
promotion, emphasizing medical best practices most analytic approaches share a common goal, to
and evidence-based medicine. Usually, outcomes enhance a care or disease management program’s
are evaluated on the basis of clinical/therapeutic ability to improve quality of care while lowering
improvement or compliance, financial/cost re- overall cost, they are optimized to address this
duction, and behavioral/emotional enrichment goal in different ways. We propose a classifica-
(Hunter & Fairfield, 1997; Leider & Krizan, 2001). tion system to clarify the objective of the most
Ultimately, the success of disease management commonly used population segmentation and
programs often depends upon the ability to target predictive modeling approaches.
an appropriate intervention to the appropriate The proposed classification system divides the
patient, where population segmentation enhances methodologies based upon their useful functional-
MCOs’ ability to accomplish this. ity in addressing three primary operational areas
The following section of this article provides for MCOs. The three proposed classifications are
more detailed background on the application of Risk Status, Treatment Status, and Health Status
population segmentation techniques to enhance (see Figure 1). The benefits of this classification
health care efficiency and introduces three primary system are to enable MCOs to clarify their pri-
areas where they can be utilized by MCOs.A de- mary objective, thereby enabling them to select a
tailed analysis of the required analytic objective methodology that best fits their need and ultimately
corresponding to three primary health care assess- achieve better disease management solutions.
ment areas (e.g. Risk Status, Health Status and Disease management programs make use of
Treatment Status) is then provided. The criteria population segmentation to improve the efficiency
for categorizing population segmentation methods and effectiveness of their interventions. Population
according to the three primary assessment areas segmentation approaches often serve different
are then described. objectives such as early identification of high-risk
patients, identifying compliance with treatment
protocols, or categorizing patients by severity
approacHes to populatIon of illness for payment adjustments or outcomes
segMentatIon and predIctIVe reporting. It is important to use the appropriate
ModelIng and a proposed methodology to best meet the information re-
classIFIcatIon systeM quirements of a specific intervention. Depending
upon the MCOs’ focus and intervention(s), some
Research addressing the use of quantitative-based disease management programs might use one
decision support systems to enhance efficiency approach while other programs might use two
in the health care sector is on the rise given the or more approaches.
development of data resources and availability For disease management purposes, predictive
of sophisticated analytic methods (Walchzak, modeling is defined as utilizing currently avail-
Brimhall, & Lefkowitz, 2006; Raghupathi, 2006). able data to prospectively identify an individual’s
There is great interest today in the application of risk of a specific outcome. Thus, one may look at
population segmentation and predictive model- a patient’s claims or survey data to determine if

104
Informatics Application Challenges for Managed Care Organizations

Figure 1.
RISK STATUS

Future Risk of
High Cost and Utilization
(e.g., Predictive Modeling)

Risk Status

Total
Patient
Health
Patient
Total Health
Health Treatment
Status Status
Current Status of Current Type and
Physical and Mental Health Intensity of Health Care
(e.g., Current Health Assessment) (e.g., Medical Best Practices)

HEALTH TREATMENT
STATUS STATUS
A single patient may be looked at from each of the three perspectives.

the patient has an increased risk of future hospi- outcomes (e.g., emergency department or hospital
talization, high cost of care, or some morbidity visits). It can also refer to the probability of future
event (Kudyba, Hamer, & Gandy, 2005). As risk for developing certain medical conditions,
will be described next, each of the population such as obesity or cardiac disease (Perry, 2007).
segmentation applications (i.e., health status, Risk status focuses on the probability of future
treatment status, and risk status) has a predic- events, making this classification of population
tive component, but not all are optimized for segmentation optimally predictive in nature.
predicting future risk. The following section will Predictive modeling techniques for this area
describe the objectives required for each of the range from simple linear equations to complex
three areas of application (assessment) and clarify neural network forecasting techniques (Grana,
the appropriateness of corresponding population Preston, McDermott, & Hanchak, 1997; Kiernan,
segmentation approaches. Kraemer, Winkleby, King, & Tylor, 2001). These
may include seasonal adjustment methods and
trend lines (Cote & Tucker, 2001), and heuristics
tHree prIMary areas oF including uncomplicated rules-based algorithms
assessMent For Mcos (rIsK to multifaceted decision support techniques (Fer-
status, treatMent status and reira et al., 2001). Obviously, the mathematical
HealtH status) or logical technique used to assess risk status
is highly dependent upon the type of risk being
risk status evaluated. Regardless of the methodology, the
objective of evaluating a patient’s risk status is
Risk status refers to a patient’s likelihood for the same, prediction of an unknown future event.
specific clinical outcomes (e.g., myocardial A recent study conducted by Kiernan et al.
infarction), financial outcomes (e.g., significant (2001) compared two prediction techniques, lo-
increases in future health care costs), or utilization gistic regression and signal detection, to assess

105
 Informatics Application Challenges for Managed Care Organizations

individuals who are at risk for being overweight. Risk Status is best suited for correctly identifying
This study was based on survey data from 1,635 and targeting the highest risk patients and thus is
White and Hispanic men and women. Body mass the most predictive of the three classifications. The
index (BMI) was used to define “overweight” next operational area which can be addressed by
for this population, and predictor variables used population segmentation methodologies involves
included gender, ethnicity, age, and educational Treatment Status for patients.
level. Results from the study demonstrate that
both methodologies had similar predictive accu- treatment status
racy and identified a similar set of risk predictor
variables. Nevertheless, these methods did not Treatment status focuses on the actual care that
classify the same individuals into population a patient is receiving, or in other words, the type
subgroups. Notably, a very high risk group (less and intensity of health care delivered. Treatment
educated, young Hispanic adults) was hidden by status encompasses an assessment of medical
the logistic regression analysis but was revealed best practices or evidence-based medicine pro-
by the signal detection analysis. These results tocols. It is often evaluated as part of a physician
demonstrate the importance of selecting the best profiling system. Obviously, not all patients are
technique of population segmentation to identify alike; there is variability in both mental and
a high-risk population. physiological responses to treatment regiments.
Logistic regression was utilized to predict the Likewise, not all physicians are alike; there are
probability of asthma-related hospital admissions differences in education and training as well as in
for asthma members in a large HMO (Grana et personal attitudes. Nevertheless, evidence-based
al., 1997). A predictive model was built from clinical practice guidelines can help to manage
administrative data (e.g., medical, pharmacy, this inevitable variation. Within a disease man-
laboratory, and enrolment files) associated with agement framework, treatment status becomes a
over 54,000 asthma patients. Asthma-specific very important population segmentation applica-
utilization, pharmacy data, and length of enrol- tion, especially when attempting to manage and
ment were found to be the best predictors of coordinate the health care delivery for thousands
future asthma-related admissions. Results were of chronically ill patients. Often, there is a gap
evaluated by the predicted number of admissions between recommended standards of care and
compared with the actual number of admissions, the treatment received by chronically ill patients
broken down into 10 deciles. Analysis of the top (Muney, 2002). Only through careful evaluation
three deciles resulted in a sensitivity and speci- of a patient’s treatment status can these treatment
ficity score of 0.70 and 0.71, respectively. These gaps be identified and addressed.
results demonstrate how claims data alone can be Research on treatment status includes stud-
used to accurately predict future hospitalizations ies such as that conducted by O’Connor, Sperl-
for patients within a population. Hillen, Pronk, and Murray (2001), to investigate
Risk status technologies are optimized for clinical-based practice characteristics related to
predicting future risk. In addition to applications best practices within chronic disease care. The
in disease and care management programs, this objective of the study was to identify those features
category of population segmentation, which em- shared by successful primary care clinics. Seven
ploys predictive modeling techniques, is also the primary care practices managing patients with
type of segmentation technology that is currently diabetes, hypertension, lipid disorders, or heart
emerging in underwriting, rating, and payment ad- disease were examined. Data from each of these
justment. Population segmentation that focuses on clinics were compiled and treated as individual

106
Informatics Application Challenges for Managed Care Organizations

case studies. Results from this study illustrate current health relative to other patients. Health
that focusing on treatment status of chronically Status technologies generally serve one of two
ill patients can facilitate significant improvements broad purposes: 1) classifying patients by health
in health outcomes within 1 to 2 years in adults status for outcomes measurement and comparison
with diabetes, hypertension, or lipid disorders. purposes, or 2) categorizing patients into similar
For example, a 20% reduction in risk for a major severity levels for assessing treatment intensity
cardiovascular event (on a population basis) was or for payment adjustment design. Other health
reported. This study, as well as others (Heller & status methods are utilized to group patients
Arozullah, 2001; Solberg, Reger, & Pearson et according to their likely treatment needs and
al., 1997; Wagner, Austin, & Korff, 1996), illus- resource consumption during a current episode
trates the benefits of improving treatment status of care (Baker, 2002). Currently, many health
in chronically ill patients, which results in better care organizations are using proprietary grouper
clinical and financial outcomes. software for patient/provider profiling, utilization/
While treatment status has some predictive clinical benchmarking, disease/case management
value and is an important component of care activities, and quality improvement initiatives.
and disease management, it is not optimized for There is a relationship between current health
true predictive modeling. For example, we can status and health outcomes. Goetzel, Anderson,
use a population segmentation methodology to Whitmer et al. (1998) conducted a study of the
address treatment status to identify two patients Health Enhancement Research Organization
that have elevated cholesterol and are not being (HERO) database in order to estimate the impact
appropriately treated with antilipidemic medica- of modifiable health status factors on health care
tion. Both patients are at increased risk for poor expenditures. Overall, individuals who had poor
clinical outcomes, so there is some predictive health status had significantly higher expenditures
component to this methodology. However, we than individuals who had a better health status in 7
cannot immediately infer if both patients are at of the 10 health status categories (depression, high
co-equal risk. To add to the example, if one of the stress, high blood glucose levels, high/low body
patients is 35 years of age with a family history of weight, former tobacco users, current tobacco us-
hyperlipidemia and cardiac disease and the other ers, and high blood pressure). Moreover, individu-
patient is 85 years of age without any current als with poor health status in multiple categories
cardiac disease, then clearly the two patients are had much higher medical expenditures than those
not at equal risk for cardiovascular complications without. For example, those with heart disease,
in the future. Treatment Status is optimized to psychosocial, and stroke profiles had 228%,
identify gaps in treatment, and is therefore not 147%, and 85% higher expenditures, respectively.
optimized to identify patients’ Risk Status or for This study indicates that patients’ health status
predictive modeling. One final area of focus for is associated with at least near-term increases in
MCOs entails analytic requirements for Health the likelihood of incurring future high medical
Status applications. costs (Anderson, Whitmer, & Goetzel, 2000;
Leutzinger, Ozminkowski, Dunn et al., 2000).
Health status Despite the correlation between health status
and future health outcomes, general population
Health status refers to the current standing of a segmentation technologies that are optimized for
patient’s clinical, physical, and mental health. Health Status are not optimized for predictive
Population segmentation techniques to address modeling. For example, two patients who are in
Health Status are utilized to describe a patient’s critical condition may score similarly on a GHA.

107
 Informatics Application Challenges for Managed Care Organizations

Both patients are likely to incur substantial costs This is important when balancing trade-offs in
in the near term and are obviously at short-term performance. Once this is achieved, the best class
risk for poor clinical outcomes. However despite of population segmentation technology must be
the same GHA score, we cannot immediately selected to meet this goal. For Risk Status applica-
infer if both patients are at the same risk for long- tions, the goal is to apply and evaluate the predic-
term health outcomes or long-term expenditures. tive modeling accuracy. The focus is on accuracy
To continue the example, if one of the patients metrics such as positive predictive value, true/false
is 70 years of age with terminal cancer and the positive rates, true/false negative rates, sensitivity
other patient is 25 years of age and was recently and specificity at various screening thresholds, and
in a traumatic accident, then the outcomes may ROC or R-squared values. For each metric, it is
be dramatically different over a 12-month time important to ascertain how the metrics have been
frame. Health Status is optimized to correctly validated and if they will generalize to a particular
classify patients vis-à-vis their current health, MCO setting. For Treatment Status applications,
and is therefore not optimized to identify patients’ the goal is to evaluate the comprehensiveness
Risk Status or for predictive modeling. and accuracy of the assessments. The focus is on
number of diseases covered, number of evidence-
based standards covered, the quality of the algo-
eValuatIon crIterIa oF rithms used (i.e., are they disclosed or hidden in
populatIon segMentatIon “black box” technology?), and the source of the
MetHodologIes For Mco standards (evidence-based, proprietary, etc.). The
areas oF assesMent Treatment Status system’s accuracy must also be
considered. This typically involves an assessment
An important first step in evaluating population of the sensitivity and specificity of classification
segmentation and predictive modeling techniques with respect to treatment status. Applications
is to define the specific purpose and goal of based upon both survey and claims data invariably
the modeling initiative. To date, the evaluation produce some errors. Consequently, it is important
process has often been complicated because of that the assessments produce reasonable informa-
a lack of clarity in the specific goal for potential tion that will be clinically acceptable, providing a
applications. In addition, there is often a lack of sound basis for intervention decisions. For Health
clarity within the MCO for what would be done Status applications, the evaluation focuses on the
with the information once obtained. For example, system’s ability to correctly classify patients into
the care and disease management departments comparable groups or cohorts. For example: (1)
might envision early identification of at risk pa- What is the accuracy of the system’s prediction in
tients, the provider network department might categorizing patients into a certain level of health
envision provider profiling on quality metrics, or episode of care? (2) Can the system be gener-
and the finance department might plan to use the alized and can the results be used to compare to
information to enhance underwriting or rating. other groups or over time? and (3) Is the system
While all of these are reasonable goals, they may well accepted by the provider community if it is
not all be obtainable with a single technology, and to be included in payment schemes? If a single
it is highly unlikely that a single technology will system purports to provide current health status
perform optimally well for all purposes. and also predicted future health outcomes, then
The evaluation process should begin with a it is important to evaluate the accuracy metrics
definition of the purpose and goal of the applica- as described above for both the current health
tion, where multiple goals, should be prioritized. status prediction and the future health outcomes
prediction.

108
Informatics Application Challenges for Managed Care Organizations

Existing population management approaches the segment of the population with the highest
fail to fully utilize information derived from probability of future hospitalizations. However,
these three statuses. The difficulty in integrating the analysis of each assessment application is
multiple data sources and data types has resulted optimized differently: Risk Status is optimized
in the optimization of operational systems based for the prediction of future events, Treatment
on single sources. Obviously, positive outcomes Status is optimized for the identification of gaps
have certainly been achieved using this type of in medical care, and Health Status is optimized
single source methodology, which focuses inter- for the classification of patients’ current state of
ventions vis-à-vis highly specific, singular data physical and mental health. Clearly, in order to
sources or assessments (Huffman, 2005; Ropka, capitalize on the strengths of population seg-
2002). However, as described in this article, when mentation methodologies, health plans need to
faced with the penultimate operational challenge understand the primary functionality of a given
of population management (i.e., delivering limited population segmentation methodology in conjunc-
resources to the right patients at the right time) it tion with the required objective of the particular
becomes increasingly important to use informa- assessment application and, subsequently, select
tion derived from multiple sources in order to the most appropriate modeling approach that best
decrease the probability of incorrect allocation fits their purposes.
of these resources. As MCOs increasingly rely upon care and
The informatics approach described in this disease management programs to fill gaps in health
article is not without limitations. This approach care service and delivery, correctly identifying and
requires extensive IT resources and IT system targeting patients for program interventions has
architectural changes. It is also very data intensive risen in importance. Though risk status is clearly
and requires the coordination and management the application that is optimally predictive in na-
of large amounts of data. ture, it cannot alone be used to evaluate, assess,
and allocate interventions to a given population.
Population segmentation methodologies address-
conclusIon ing current treatment status and health status,
as well as anticipated risk status, are the most
In this article we have proposed a classification and effective and logical way to move forward with
evaluation method for population segmentation an overall intervention strategy. Meanwhile, it
approaches used in care and disease management. is important to differentiate between these three
Risk Status, Treatment Status, and Health Status approaches, because they have been optimized for
assessments each provide useful information different goals. Clarity in purpose and focused
to forecast the best allocation of resources and evaluation against that purpose will provide
interventions within a population of interest. For MCOs with the best techniques. Consequently,
example, a disease management company might MCO disease management programs that truly
be interested in managing those patients who integrate/utilize all three approaches will best
are at high risk for future hospitalizations. In succeed in accomplishing the main goal of focus-
this example the disease management company ing the right resources on the right patients while
could use information from population segmen- increasing quality of care and decreasing overall
tation approaches in order to focus resources on medical expenditures.

109
 Informatics Application Challenges for Managed Care Organizations

reFerences Grana, J., Preston, S., McDermott, P.D., &

Hanchak, N.A. (1997). The use of administrative
Anderson, D.R., Whitmer, R.W., & Goetzel, data to risk-stratify asthmatic patients. American
R.Z., et al. (2000). Health enhancement research Journal of Medical Quality, 12, 113-119.
organization (HERO) research committee. The
Hazen, G.B. (1992). Stochastic trees: A new tech-
relationship between modifiable health risks and
nique for temporal medical decision modeling.
group-level health care expenditures. American
Medical Decision Making, 13, 227-236.
Journal of Health Promotion, 15, 45-52.
Hazen, G. (2000). Preference factoring for stochas-
Ash, A.S., Zhao, Y., Ellis, R.P., & Schlein, K.M.
tic trees. Management Science, 46(3), 389-403.
(2001). Finding future high-cost cases: Compar-
ing prior cost versus diagnosis-based methods. Heller, C., & Arozullah, A. (2001). Implementing
Health Services Research, 36, 194-206. change: It’s as hard as it looks. Disease Manage-
ment and Healthcare Outcomes, 9, 551-563.
Baker, G.B. (2002). Integrating technology and
disease management—the challenges. Healthplan Heskett, L. (1983). Shouldice Hospital Ltd, Rev.
Magazine, 43, 60-66. 1/88. Boston, MA: Harvard Business School.

Bishop, C. (1995). Neural networks for pattern Huffman, M. (2005). Implementing outcome-
recognition. Oxford University Press. based homecare: A workbook of OBQI, care
pathways and disease management. Jones and
Cousins, M.S., Shickle, L.M., & Bander, J.A.
Bartlett.
(2002). An introduction to predictive modeling for
disease management risk stratification. Disease Hunter, D.J., & Fairfield, G. (1997). Managed care:
Management, 5, 157-167. Disease management. British Medical Journal,
315, 50-53.
Cote, M.J., & Tucker, S.L. (2001). Four methodolo-
gies to improve health care demand forecasting. Institute of Medicine. (2001). Crossing the quality
Healthcare Financial Management, 5, 54-58. chasm: A new health system for the 21st century.
Washington, DC: National Academy Press, Com-
Ferreira, F.L., Peres, Bota, D., Bross, A., Melot,
mittee on Quality of Health Care in America.
C., & Vincent, J.L. (2001). Serial evaluation of
the SOFA score to predict outcome in critically ill Kiernan, M., Kraemer, H.C., Winkleby, M.A.,
patients. Journal of American Medical Associa- King, A.C., & Taylor, C.B. (2001). Do logistic
tion, 286, 1754-1758. regression and signal detection identify different
subgroups at risk? Implications for the design of
Foughty, T. (1999). Accomplishing health im-
tailored interventions. Psychological Methods,
provement through diabetes disease management.
6, 35-48.
Drug Benefit Trends, 12, 43-45.
Kudyba. S., Hamer, B., & Gandy, W. (2005, De-
Goetzel, R.Z., Anderson, D.R., Whitmer, R.W.,
cember). Enhancing efficiency in the health care
et al. (1998). The relationship between modifiable
industry. Communications of the ACM, 28(12),
health risks and health care expenditures: An
107-110.
analysis of the multi-employer HERO health risk
and cost database. Journal of Occupational and Kudyba, S., & Hoptroff, R. (2001). Data mining
Environmental Medicine, 40, 843-854. and business intelligence: A guide to productivity.
Hershey, PA: Idea Group.

110
Informatics Application Challenges for Managed Care Organizations

Landro, L. (2002, January 4). Technology pro- Ridinger, M.H., & Rice, J.J. (2000). Predictive
grams make disease management easier. The Wall modeling points way to future risk status. Health
Street Journal, Health Journal. Management Technology, 21, 10-12.
Leider, H.L., & Krizan, K. (2001). Disease man- Ropka, M.E. (2002). Symptom status and func-
agement—a great concept, but can you implement tional status outcomes: Humanistic outcomes in
it? Disease Management, 4, 111-119. obesity disease management. Obesity Research,
10, 42S-49S.
Leutzinger, J.A., Ozminkowski, R.J., Dunn,
R.L., et al. (2000). Projecting future medical care Smith-Daniels, V., Schweikhart, S., & Smith-
costs using four scenarios of lifestyle risk rates. Daniels, D. (1988). Capacity management in
American Journal of Health Promotion, 15, 35-44. health care services: Review and future research
directions. Decision Sciences, 19, 889-919.
McLaughlin, C., Yang, S., &Van Dierdonck, R.
(1995). Professional service organizations and Solberg, L.I., Reger, L.A., & Pearson, T.L., et al.
focus. Management Science, 14(7), 1185-1193. (1997). Using continuous quality improvement
to improve diabetes care in populations: The
Muney, A.M. (2002). Evidence-based medicine
IDEAL model: Improving care for diabetics
needs to be promoted more vigorously. Managed
through empowerment, active collaborations and
Care, 11, 10-12.
leadership. Joint Commission Journal on Quality
O’Connor, P.J., Sperl-Hillen, J.M., Pronk, N.P., Improvement, 23, 581-592.
& Murray, T. (2001). Primary care clinic-based
Swingler, K. (1996). Applying neural networks:
chronic disease care. Features of successful pro-
A practical guide. New York: Academic Press.
grams. Disease Management and HealthCare
Outcomes, 9, 691-698. Wagner, E.H., Austin, B.T., & Korff, M.V. (1996).
Improving outcomes in chronic illness. Managed
Perry, T. (2007), December). Identification and
Care, 4, 12-25.
prediction of chronic conditions for health plan
members using data mining techniques: In data Walczak, S., Brimhall, B., & Lefkowitz, J. (2006).
mining methods and applications. Taylor Francis. Nonparametric decision support systems in medi-
cal diagnosis: Modeling pulmonary embolism.
Raghupathi, W. (2006). Designing clinical deci-
International Journal of Healthcare Information
sion support systems in health care: A systemic
Systems and Informatics, (1), 2.
approach. International Journal of Healthcare
Information Systems and Informatics, 2(1). Xiaohaua, H. (2005). Temporal rule induction for
clinical outcome analysis. International Journal
of Business Intelligence and Data Mining, 1(1),
122-136.

This work was previously published in International Journal of Healthcare Information Systems and Informatics, Vol. 3, Is-
sue 2, edited by J. Tan, pp. 21-31, copyright 2008 by IGI Publishing (an imprint of IGI Global).

111
112

Chapter 9
Scrutinizing the Rule:
Privacy Realization in HIPAA
S. Al-Fedaghi
Kuwait University, Kuwait

aBstract

Privacy policies, laws, and guidelines have been cultivated based on overly verbose specifications.
This article claims that privacy regulations lend themselves to a firmer language based on a model of
flow of personal identifiable information. The model specifies a limited number of situations and acts
on personal identifiable information. As an application of the model, the model is applied to portions of
the Privacy Rule of Health Insurance Portability and Accountability Act (HIPAA).

IntroductIon lower level “machinery” questions which should

be left to domestic implementation. (I. GENERAL
The notion of privacy is becoming an important BACKGROUND, 19, e)
feature of modern society. In this context, deciding
how to regulate the processing of personal identifi- According to Bennett (2001), privacy prin-
able information (PII) is a vital issue. Responding ciples enfold different interpretations,
to the public’s awareness of the importance of
protecting privacy of personal identifiable infor- There are disputes for example: about … the dis-
mation, guidelines have evolved and converged tinction between collection, use and disclosure of
around a set of basic privacy principles (e.g., information, and whether indeed these distinctions
OECD, 1980). How successful are these privacy make sense and should not be subsumed under
principles? According to the OECD (1980) report, the overarching concept of “processing” … How
these and other statutory issues are dealt with
The choice of core principles and their appropriate will, of course, have profound implications for the
level of detail presents difficulties… In particular, implementation of privacy protection standards
it is difficult to draw a clear dividing line between within any one jurisdiction.(p. 12)
the level of basic principles or objectives and

Copyright © 2011, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Scrutinizing the Rule

Mechanisms to protect the privacy of personal systems, tries to go beyond the HIPAA Privacy
identifiable information include legal measures, Rule (Westin & Gelder, 2005, p. 6). Nevertheless,
policies and privacy-enhancing technologies. Leg- it is also important to develop clearly defined
islations such as the Health Insurance Portability privacy terms. Recent developments in the areas
and Accountability Act (HIPAA) and systems of privacy and information assurance have placed
such as P3P are not sufficient to safeguard privacy the elements of information privacy on firmer
because “they do not address how personal data theoretical ground. This includes a model for
is actually handled after it is collected” (He & personal information flow (Al-Fedaghi, 2006)
Antón, 2003, p. 1). Also, according to Fischer- that systematically categorizes subprocesses in-
Hübner and Ott (1998, p. 1) “privacy cannot volved in personal information processing. This
be efficiently implemented solely by legislative article illustrates a sample of the benefits of such
means. Data protection commissioners are there- an approach through inspecting portions of the
fore demanding that legal privacy requirements Privacy Rule of HIPAA. Our purpose is not to
should be technically enforced and should be take a position on any of the privacy principles or
design criteria for information systems.” subprinciples, and neither is it to highlight con-
Among types of personal identifiable informa- flicts and contradictions, but rather it is to provide
tion, health information is ranked as being the most precise specifications of privacy notions in order
sensitive, at the same level as financial information to apply them to different privacy standards, leg-
(GPC Alberta, 2003). In the health information islations, or codes of practice. Understanding the
area, “Survey research found that the public … constitutive elements of privacy principles will
was deeply worried about how their personal help to determine any disparities and consisten-
medical information was being accessed and used cies/inconsistencies when incorporating them in
in other sectors, for secondary purposes such as different laws and in their interaction with other
insurance, employment, licensing, research, law requirements (e.g., enforcement mechanisms).
enforcement, public health, and media activities”
(Westin & Gelder, 2005, p. 4). personal Identifiable Information
The Health Insurance Portability and Account-
ability Act (HIPAA) of 1996 is the most significant This section focuses on personal identifiable
health care legislation in U.S. history. The U.S. information (PII) as our main object of study,
Department of Health and Human Services (HHS) and its flow model (Al-Fedaghi, 2005a, 2006).
issued the Privacy Rule (45 CFR Parts 160 and It is typically claimed that what makes the data
164) to implement the requirement of HIPAA. “private” or “personal” is either specific legisla-
According to U.S. Department of Health & Hu- tion, for example, a company must not disclose
man Services (2003), “The Standards for Privacy information about their employees; or individual
of Individually Identifiable Health Information agreements, for example, a customer has agreed
(“Privacy Rule”) establishes, for the first time, to an electronic retailer’s privacy policy. However,
a set of national standards for the protection of this line of thought blurs the difference between
certain health information” (HSS, 2003). personal identifiable information and other “pri-
While privacy laws have an important role, vate” or “personal” information. Personal identifi-
other approaches are valuable. The “Privacy able information has an “objective” definition in
by Design” approach that incorporates the Fair the sense that it is independent of such authorities
Information Practices standards into information as legislation or agreement.

113
 Scrutinizing the Rule

Definition of Personal Identifiable anonymous form in the processing stage.

Information Double arrows in Figure 1 denote two acts on
PII (will be defined) that for simplicity’s sake can
In the information sphere, information is clas- be considered as one type of act. For example,
sified as personal identifiable information (PII) the double arrow to Storage indicates storing or
and nonidentifiable information (NII). Personal retrieving data. The double arrow between Col-
identifiable information is information about lecting and Disclosing reflects the fact that every
singly-identifiable persons, called its proprietors. disclosure act of information is accompanied
PII is information that has referents that are natural by a collecting of information by another agent.
persons. There are two types of PII: For example, if someone discloses his/her PII to
an insurance company, then simultaneously the
1. Atomic personal information where the company is collecting this PII. The double arrow
information refers to a single proprietor, for between Processing and Mining is necessary to
example, John is 25 years old. “Referent,” indicate that mining, as a type of processing,
here, implies an identifiable (natural) person. produces two types of information: (1) processed/
2. Compound personal information where the mined information that is not new (e.g., implied
information refers to more than one propri- information), and (2) processed/mined informa-
etor, for example, Mary donated her kidney tion that is new (e.g., classifying a customer as a
to Alice. risk, based on statistical analysis of all customers).
Any compound PII is privacy-reducible to a set Example: Consider the situation where we have
of atomic PII (Al-Fedaghi, 2005a). one proprietor and two PII agents (e.g., companies,
agencies, other individuals). Suppose that the
PII Flow Model roles of these three actors are defined as follows:

The PII Flow Model divides functionality into Proprietor: Creates, stores, and discloses PII to
stages that include informational entities and agent 1.
processes, as shown in Figure 1. Each stage may Agent 1: Collects, stores, utilizes, and discloses
have several specific units that denote autonomous PII to agent 2.
functions such as storage/retrieval and utilization Agent 2: Collects and processes PII through a
of collected information. New PII is created by mining technique that creates new PII that
proprietors, nonproprietors (e.g., medical diag- is stored and utilized in some applications
nostics by physicians), or is deduced by someone (e.g., decision making).
(e.g., data mining that generates new information
from existing information). The created informa- The PII flow model for this simple environ-
tion is used either in some type of utilization (e.g., ment can be drawn based on Figure 1, as shown
decision making), stored, or it is immediately in Figure 2. For each actor in this scenario we
disclosed. The processing stage involves acting make a copy of the PII flow model. However,
(e.g., storing, anonymization, data mining, sum- because the proprietor does not collect or process
marizing, translating) on PII for whatever purpose PII, these stages are not shown in his/her region.
it is collected. The disclosure stage involves re- The creation and processing stages are not shown
leasing PII to insiders or outsiders. The “disposal” for agent 1 because it does not create or process
or disappearance of PII can happen anywhere in PII. Similarly, the disclosure stage is not shown
the model, such as the transformation of PII to an in the region of agent 2. Let t be a piece of PII of
the proprietor. It originates in the Creating box in

114
Scrutinizing the Rule

Figure 1. PII flow model.

Creation Proprietor Non-proprietor

stage

Creating PII

Storage Utilize

Collection
Collecting PII
stage 6 5

3
Storage Utilize

Processing
stage Processing PII
Mining

Storage Utilize

Disclosure
stage Disclosing PII

Figure 2. PII flow model for a proprietor and two agents.

Proprietor’s Region 1st Agent’s Region 2nd Agent’s Region

Creating Store Utilize

Utilize

Creating
Storage
Storage

Collecting Collecting

Mining

Disclosing Disclosing Processing

115
 Scrutinizing the Rule

the proprietor’s region. It is stored in Store, and Examples:

moves through the Disclosing box to the Collect-
ing box of the agent 1’s region. It is stored and Act A: A person discloses PII to a hospital (i.e.,
utilized there, and moves through the Disclosing a colleting agent).
box to the Collection phase of agent 2. There, it Act B: A hospital discloses a student’s PII to an
moves to the Processing and to the Mining boxes insurance company.
where it generates new PII in the Creation phase Act C: PII is produced by a mining program (e.g.,
for Storage and later Utilization. We can add John is a high-risk customer) of a bank is
details to any stage as the situation requires. For disclosed to crediting company.
example, agent 2 may add Storage to keep a copy Act D: Processed PII (changing the original data
of the original PII or a Disclosure stage can be to another form through such operations as
added if agent 2 discloses the resultant new PII modification, translation, summarization,
to a third agent. generalization) is released from a hospital
to an insurance company.
Act E: A hospital stores PII in its automated or
tHe proprIetor-otHers manual system, accessing stored data, era-
arcHItecture sure of stored data.
Act F: A hospital utilizes PII to communicate
Using the PII flow model, we can built a system with its patients (e.g., e-mails).
that involves a proprietor on one side and others Act G: A hospital stores processed PII (e.g., mined
(other persons, agencies, companies, etc.) who PI) in its system, accessing stored data.
perform different types of activities in the PII Act H: An agent utilizes processed data for
transformations between different stages. We will research.
refer to any of these as PII agents. PII agents may Act I: A collecting agent sends collected PII to
include anyone who participates in activities over a processing agent in the same enterprise.
PII. The proprietor is not accepted as agent with Act J: An agent mines PII to extract implied PII
respect to his/her own PII. (e.g., the information Z is the grandfather
The EU Privacy Directive (1995) and HIPAA of Y is taken from Z is the father of W and
manage the type of system that organizes (1) The W is the father of Y).
relationship between the proprietor and agents that Act K: An agent Stores PII produced by a min-
utilize his/her PII, and (2) The relationship among ing program.
the agents. It is a “binary” system that involves the Act L: An agent makes decisions based on PII
proprietor on one side and all agents on the other that is produced from a mining program.
side. As a result, we need two types of PII flow Act M: An agent produces new PII using a min-
regions: one for proprietors and one for agents. ing program (e.g., an analysis of customers
The proprietor’s region includes activities on his/ produces the information that John is a
her PII and the others’ region includes activities high-risk person).
on the proprietor’s PII, as shown in Figure 3. We Act N: A newspaper writer publishes PII about
assume that there is no interest in how proprietors the proprietor.
collect and process their own PII. Act O: An agent collects PI from another agent.
We distinguish here among types of acts on Act P: An agent creates PI (gossip) and processes it.
PII (labeled A through Q) as shown in Figure 3 Act Q: Informing a proprietor of results of medi-
and described in Table 1. cal tests.

116
Scrutinizing the Rule

Figure 3. Architecture of proprietor/agent PII flow.

Proprietor’s Region Agent’s Region

N
Non-proprietor Creating

Utilize Store K L
F Utilize
A
Disclosing Collecting M
P
O E
I J Mining
Store
Processing
B Store
Utilize
G H
D
C
Q
Collecting Disclosing

The advantage of using this type of categorization while it is difficult to apply this law to PII created
is that we can specify different types of codes, by people about other people.
requirements, or restrictions for each type of act Example: Consider the act of disclosing PII.
on PII. It includes four types of disclosures. This type
Example: Consider the act of creating PII by includes A (disclosing PII by a proprietor), B
others. This type includes M (creating PII by a (disclosing PII by a collecting agent), C (disclos-
mining agent), and N (creating PII by nonpropri- ing PII by a creating agent), and D (disclosing
etor). These two creating acts are different from PII by a processing agent). It is clear that the
the act of a proprietor creating PII (about him/ requirements and restrictions of disclosing PII
herself) and disclosing it (act A). For example, the will differ with respect to who is disclosing the
EU Directive requirement that “the data subject information: a proprietor disclosing his/her PII
must be in a position to learn of the existence or others disclosing their PII to one another. For
of a processing operation” (EU Directive, 1995, example, The OECD’s restriction directed at such
Recitals 38), is applied to M and N, but not to practices as “deceiving data subjects to make them
A. Additionally, acts M and N are two different supply information” (OECD, 1980, I. GENERAL
types of creating PII. Act M creates information BACKGROUND: The Problems (52)) makes
by automatic means as in the case of Jane is a high sense only in act A. Furthermore, the difference
risk person based on analysis of PII of all custom- extends to the type of agent. We apply different
ers of a company. Act N involves nonautomatic considerations to an agent who is creating new or
means such as an informer claims that Jane is implied PII using a mining technique compared
a terrorist based on his/her surveillance. These to a simple collecting agent. The point is that we
two types of creating PII may require different have a well-defined set of acts on PII that can
treatment in different contexts. For example, a be used in different applications such as writing
law may forbid creating PII by automated means codes and guidelines.

117
 Scrutinizing the Rule

Table 1.Types of acts on personal identifiable information

Description of the Act Comments

A Disclosing PII by a proprietor Act A also represents collecting PII (by a collecting agent).
B Disclosing PII by a collecting agent B implies O (collecting PI by another collecting agent)
Disclosing PII by a creating agent In Figure 3, B implies O, that is, the disclosed PII flows from a
C
collecting agent to another collecting agent.
In Figure 3, D implies B, that is, the disclosed PII flows from a
D Disclosing PII by a processing agent
processing agent to a collecting agent.
E Storing PII by a collecting agent E (double arrow in figure 3) includes retrieval of PII.
F Utilizing PII by a collecting agent “Utilize” indicates noninformational operations.
We can separate the storing and retrieval acts as two independent
G Storing PII by a processing agent
acts in the model.
Utilizing processed PII may be different from utilization of other
H Utilizing PII by a processing agent
types of PII.
I Processing by an agent PII flows from the collecting stage to a processing stage.
Mining is a type of processing. This type of mining produces
J Mining PII by a mining agent
implied PII but not new PII.
K Storing PII by a creating agent
L Utilizing PII by a creating agent
M Creating PII by a mining agent Automatic creation of PII. Mining is a type of processing.
N Creating PII by a nonproprietor Non-automatic creation of PII (e.g., gossip).
O occurs simultaneously with B: If an agent discloses PII then
O Collecting PII from nonproprietor
there is an agent that collects that PII.
Nonproprietor may create PII and process it immediately without
P Processing of created PI
storing or acting on it.
Q Disclosure to proprietor E.g., Informing a person of results of medical tests.

Each of the PII acts can be supplemented national standards for handling health care in-
with various purposes (Utilizations) as shown. formation. The 169-page law is supplemented
For example, purposes for Disclosing PII by a with about 1000 pages of related regulations and
proprietor can be Internet purchase, government directions. The U.S. Department of Health and
form, and security. While the number of acts is Human Services (HHS, 2003) issued the Privacy
limited, the number of purposes (Utilizations) is Rule (45 CFR Parts 160 and 164) to implement
unlimited. Also, purposes that are applied to a the requirement of HIPAA. The rule assures a
certain act may not be suitable for another act. right of patient access to/track of his/her health
For example, a nonproprietor who creates PII for information and mandates notification of how it
an Internet purchase is an illegal act except with will be used and disclosed. It applies only to a
the appropriate authorization. subset of health providers.
We will apply the PII FLOW MODEL archi-
tecture to some of the notions of the Privacy Rule
HIpaa in order to illustrate how our model can enhance
the conceptual framework under which these
The Health Insurance Portability and Account- notions can be interpreted.
ability Act (HIPAA) of 1996 establishes uniform

118
Scrutinizing the Rule

definitions in the rule is not health information because it is not created

or collected by the named agents. This approach
The Privacy Rule develops a set of definitions confuses the ordinary meaning of health infor-
starting with health information, moving to mation with the meaning of health information
Individually identifiable health information and in HIPAA. Common language is important when
lastly what is called Protected Health information. we think of the wide applications of HIPAA in
According to HIPAA (1996): the life of every citizen.
The definition also complicates health infor-
Health information means any information, wheth- mation by bringing in a certain application area
er oral or recorded in any form or medium, that: payment. Do we change the definition of health
information if we have completely free health
1. Is created or received by a health care pro- services? A better approach is to define health
vider, health plan, public health authority, information, and then include or exclude such
employer, life insurer, school or university, applications.
or health care clearinghouse; and So what is health information? In the PII
2. Relates to the past, present, or future physi- flow model, it is simply defined in the context
cal or mental health or condition of an in- of personal identifiable information, as will be
dividual; the provision of health care to an described next.
individual; or the past, present, or future According to the Health Insurance Portability
payment for the provision of health care to an and Accountability Act (HIPAA, 1996, PART
individual (PART C--ADMINISTRATIVE C -ADMINISTRATIVE SIMPLIFICATION -
SIMPLIFICATION, section 1171(4) of the DEFINITIONS, SEC. 1171):
Act).
Individually identifiable health information is
The first question is, why are health care provid- information that is a subset of health information,
er, health plan, public health authority, employer, including demographic information collected from
life insurer, school or university, or health care an individual, and:
clearinghouse mentioned in the definition of health
information? If we want to restrict the handling of 1. Is created or received by a health care
a particular type of information to certain players provider, health plan, employer, or health
then such a restriction is not part of the defini- care clearinghouse; and
tion of that information. Imagine that we define 2. Relates to the past, present, or future physi-
scientific information as information created or cal or mental health or condition of an in-
received by schools and universities. Building a dividual; the provision of health care to an
foundation for the Privacy Rule requires giving, individual; or the past, present, or future
first, a definition of health information, and then payment for the provision of health care to
later declaring that only information utilized by an individual; and
some users is considered in the context of HIPAA. i. That identifies the individual; or
This is a typical approach, for one reason, that ii. With respect to which there is a reason-
if the set of users changed in the future then we able basis to believe the information can
do not have to change the definition of health be used to identify the individual.
information.
The definition is also constructed such that if The definition involves, in addition to previous
you measure your own temperature then the result criticisms, ambiguity. What about John has been

119
 Scrutinizing the Rule

infected by Mary’s disease, then is it John’s iden- • Operational HPI: John is injected with
tifiable health information or Mary’s identifiable Penicillin.
health information? Consider the information This
animal has Parkinson’s disease, then is it Parkin- Notice that there is no mention about whether the
son identifiable health information? It certainly HPI is for payment or freely provided, or who
identifies the known individual: Parkinson. “creates or receives” it.
In the PII flow model, John has been infected The definitions of PII and HPI can be comple-
by Mary’s disease is compound PII that can be mented by restricting HPI to HPI handled by
reduced to atomic PII. Also, the definition of a health care provider, health plan, employer,
personal identifiable information is based on or health care clearinghouse. For example, the
referents–a well-known logical term—of type so-called protected health (personal) informa-
individual (i.e., a natural person). Thus, This tion (PHI) is defined in HIPAA (Definitions,
animal has Parkinson’s disease is not PII because 160.103) as:
it does not refer to an individual.
Based on the definition of personal identifiable • Protected health information means indi-
information, we define health personal informa- vidually identifiable health information:
tion (HPI) as a special kind of personal identifiable 1. Except …
information related to the health of a referent of 2. Protected health information excludes
type person. There are many kinds of HPI: individually identifiable health informa-
tion in …
• Strict HPI: (physical and emotional descrip-
tions): John’s blood type is A. In our framework, it is merely a restricted
• Financial HPI: John’s kidney operation is subset of HPI. A hierarchy of definitions (Figure
paid for by insurance XYRE123. 4) simplifies the conceptual application of the
• Location HPI: John’s kidney operation is Privacy Rule:
performed in General Hospital.
• Compound HPI: John’s physician is Ed- This conceptual framework simplifies understand-
ward. ing of the application of the rule. For example,

Figure 4. Relationships among types of information

World Personal Identifiable Information (PII)

Reference (uniquely About health

identification) to a
person
Health PII

R estri cti on s

Protected Health PII

120
Scrutinizing the Rule

consider a researcher who obtains the informa- every risk of an incidental use or disclosure of
tion from student records maintained by a school. protected health information be eliminated.” We
From Figure 4, it is clear that such an act is not can conclude from these samples that the Privacy
protected health information because it is simply Rule means by “use and/or disclosure” all types
restricted to specified agents. Notice that in the of operations or processing on PII. The rule’s “use
PII flow model an agent may have two roles: as and/or disclosure” refers in the PII flow model to
a PII system and as a HPI system. For example, collecting, storing, utilizing, processing, mining,
a hospital may have a PII database system as an creating, and disclosing HPI regardless of their
employer and a HPI database system as a HIPAA types.
covered entity. Each system has its own PII flow Examining “use,” by itself, we find that HIPAA
model with its own privacy rules such that the defines it as follows:
other system is treated as another collecting agent.
Use means, with respect to individually identifi-
use and disclosure able health information, the sharing, employment,
application, utilization, examination, or analysis
Going beyond the basic definitions of health of such information within an entity that main-
related information, we examine now a central tains such information (Emphasis added). (Sec.
notion in the Privacy Rule: use and/or disclosure. 164.501)
According to HHS (2003),
The terms, “sharing, employment, application,
The Privacy Rule standards address the use utilization, examination, or analysis” are over-
and disclosure of individuals’ health informa- loaded terms. The PII flow model distinguishes
tion—called “protected health information” by between two types of using PII:
organizations subject to the Privacy Rule — called
“covered entities,” as well as standards for indi- 1. The informational acts on PII that are
viduals’ privacy rights to understand and control hard-wired in the PII flow model such as
how their health information is used. (p. 1) collecting, storing, processing, and so forth.
In these acts information is what is acted
Apparently, use and/or disclosure is a very on.
important term in the Privacy Rule. In the SUM- 2. In the PII flow model, Utilize denotes non-
MARY OF THE HIPAA PRIVACY RULE, U.S. informational acts where what is acted on
Department of Health & Human Services (HSS, is not PII (e.g., a doctor treats a patient).
2003), “use or disclosure” occurs 32 times while It is the “use” to which health PII is put
“use(s) and disclosure(s)” occurs 35 times. (treatment, payment).
The following are samples of how the rule
utilizes use and/or disclosure. HIPAA’s “use” does not distinguish types of
When “defining” Business Associate, the rule acts. For example, sharing is an act on PI. It im-
describes a business associate as “a person or plies disclosure/collecting of PII among agents. If
organization … that performs certain functions a doctor requested the medical file then this situ-
or activities … or provides certain services to, a ation represents, the (hospital) system discloses
covered entity that involve the use or disclosure the file to an internal agent (doctor). If the first
of individually identifiable health information” doctor who receives the file shares the information
(HHS, 2003, p. 3). Also, HHS (2003, p. 6) states with a new doctor then this is also disclosure (and
that, “The Privacy Rule does not require that collection) of PII between two internal agents.

121
 Scrutinizing the Rule

Sharing in these cases involves disclosing and consent, deliberately omitting to tell the patient
collecting information. However, “Employment” the consequences of their consent or the use to be
(see HIPAA definition of “use” mentioned previ- made of their information would constitute an act
ously) is not necessarily this type of act. of malfeasance.” But, “the consequences of their
Employment (partially) means utilizing consent” and “the use to be made of their informa-
information in the noninformational act such tion” are two completely different actions. Consent
as treatment of patient. It is the “use” to which to sharing health PII does not imply consent for
health PII is put. Treatment is represented in the a course of medical treatment. Informed consent
PII flow model in the utilize box, where a doctor in the sharing of health PII involves further acts
as a collecting agent utilizes the medical record on PII such as disclosure and processing (e.g., for
disclosed to him/her by the system to treat (non- research), while Informed consent in the context
informational act) a patient. of a course of medical treatment involves medical
Thus, the rule mixes acts on information and complications, side effects, and so forth.
the “use” to which health PII is put. Acts on in- In the Privacy Rule, consent is not required for
formation are limited as described in the PII flow purposes of the imprecisely defined operations of
model; however, the “use” to which information “treatment, payment and health operations.” The
is put is infinite. Regulations for acts on PII ought rule relates consent (acknowledgement) to use
to be very specific for acts on information, while information in these operations not to the actual
regulations for the “use” to which information is treatment, payment or health care operations. It
put are usually open-ended. states that: “a covered entity is permitted, but
Other terms in the definition of “use” (appli- not required, to use and disclose protected health
cation, utilization, examination, or analysis) can information, without an individual’s authoriza-
also be clarified in similar fashion. The distinc- tion, for … Treatment, Payment, and Health Care
tion is important for defining several notions. Operations.” Treatment is defined in the Privacy
For example, “consent” for informational acts is Rule at 45 CFR 164.501 as:
different from consent for noninformational acts.
Consent to disclose information does not imply “Treatment” generally means the provision,
consent for the use to which health PII is put as coordination, or management of health care and
described next. related services among health care providers or
by a health care provider with a third party, con-
sultation between health care providers regarding
consent a patient, or the referral of a patient from one
health care provider to another.
According to Galpottage and Norris (2005, p.
6), after reviewing several definitions, “The Use and/or disclosure for treatment is an infor-
term ‘patient consent’ implies that the patient mational act (act on information). However, we
is willing to share personal health information observe that the information can be disclosed
and, where appropriate, to submit to a course of internally (covered entity’s own treatment) or
medical treatment.” Notice how this definition disclosed externally to altogether different cov-
assumes a single consent for two different types ered entity. The ideal situation for disclosing PII
of acts: informational acts (sharing health PII) to another covered entity is shown in Figure 5.
and noninformational acts (a course of medical The consulted covered entity collects (act A)
treatment). Also, according to Galpottage and the disclosed PII and uses it as a consulted entity
Norris (2005, p. 6) “In the context of informed (F) to create medical opinion (N) that is disclosed

122
Scrutinizing the Rule

back to the original entity (C followed by Q). The Even if it is decided that an individual’s
PII is not acted on in any other way. There are authorization is not required in either case, an
several subsets of acts that may be specified for acknowledgement of the two cases ought to be
entities receiving the information. If the new entity specified explicitly for several purposes includ-
is a specialist who actually treats the individual ing for the sake of individual’s awareness and the
then the phrase without an individual’s authoriza- rule’s implementer’s alertness.
tion seems meaningless because the patient has Health Care Operations can be analyzed in
already given consent for treatment (hence PII). a similar way to the discussion of “treatment”
The PII flow model presents a precise description and “payment.”
of “minimally necessary” acts for disclosure for
this type of treatment. This analysis can be used in disclosure
rewriting the Privacy Rule or as a supplementary
interpretation to the rule. In contrast, the rule The rule states examples of “disclosures that would
gives a blank check to disclose protected health require an individual’s authorization include dis-
information, without an individual’s authoriza- closures to a life insurer for coverage purposes,
tion, for external covered entities based on the disclosures to an employer of the results of a pre-
ambiguous term treatment. employment physical or lab test, or disclosures
The second use that a covered entity is per- to a pharmaceutical firm for their own marketing
mitted to disclose protected health information purposes” (HHS, 2003, Authorized Uses and
without an individual’s authorization is for pay- Disclosures section). However, in light of our
ment. Payment is defined in the Privacy Rule at categorizations of disclosures, “disclosure” may
45 CFR 164.501: mean the disclosure of the original PII collected
from the proprietor, the disclosure of collected
“Payment” encompasses the various activities PII from other collecting agent, the disclosure of
of health care providers to obtain payment or processed PII, or ever the disclosure of PII created
be reimbursed for their services and of a health by mining program or input by another person
plan to obtain premiums, to fulfill their coverage (e.g., physician, health worker).
responsibilities and provide benefits under the Example: According to the rule, “A covered
plan, and to obtain or provide reimbursement for entity may use or disclose, without an individual’s
the provision of health care. authorization, the psychotherapy notes, for ...”
(HHS, 2003, Authorized Uses and Disclosures
Again we see here a blank check for any section). Psychotherapy notes are defined:
internal or external activities. There is no dif-
ference between disclosing PII to an employee Psychotherapy notes” means notes recorded
of a covered entity and to another covered entity (in any medium) by a health care provider who
across the nation. Consider the case of misusing is a mental health professional documenting or
a social security number by an employee of the analyzing … that are separated from the rest of
covered entity in contrast to misusing it while in the of the individual’s medical record. Psycho-
the possession of consumer reporting agencies. therapy notes excludes medication prescription
Notice that other portions of the Rule may help and monitoring, counseling session start and stop
in interpreting a certain part of it, however, the times, the modalities and frequencies of treatment
claim we make is the wording of each part can furnished, results of clinical tests, and any sum-
be made clearer using our approach. mary of the following items: diagnosis, functional

123
 Scrutinizing the Rule

status, the treatment plan, symptoms, prognosis, conclusIon

and progress to date.
This article has introduced a systematic descrip-
Figure 6 shows simplified versions of different tion of the “processing” of personal identifiable
regions in this situation. The psychotherapist col- information. The description is based on a flow
lects PII from four sources shown in the figure. model that limits handling personal identifiable
information to creating, collecting, processing and
1. The original medical record disclosing information. It also specifies a limited
2. Created PII by the psychotherapist during number of acts on personal information. This
the session can provide a more methodological scheme for
3. New PII revealed by the patient during the privacy policies, statues, and guidelines specifica-
session tions. To substantiate this claim we applied the
4. Created PII related to the session (e.g., start model to some portions of the Privacy Rule of the
and stop times, e.g., John’s session is at 4 Health Insurance Portability and Accountability
pm, John [the patient] arrived late for the Act (HIPAA). Our analysis maps the notion of
session). “use,” “disclosure,” and “consent” in the rule to
its corresponding acts in the PII flow model. This
These are clear categorizations of information analysis can be used in rewriting the rule or as a
that can be supplemented by the type of descrip- supplementary interpretation to the rule.
tions given by the rule (e.g., summary of this type Another aim of this article is to show that
or that type). Suppose the patient made threats privacy regulations lend themselves to a firmer
during the session to kill a person named Alice. language based on a PII flow model. While this ar-
How does the rule handle the issue of disclosure ticle uses the model to scrutinize the Privacy Rule
of this PII? In the PII flow model, (3) above can of HIPAA, it is also applicable to the statutory-
be divided into two types of PII: (a) atomic PII, based privacy Directive of the European Union
and (b) compound PII. Because the threat involves and to codes and guidelines (Al-Fedaghi, 2007).
Alice, then she has the right as a proprietor to
know this information (Al-Fedaghi, 2005b).
We conclude that the PII flow model provides a reFerences
framework for deciding on disclosure issues such
as the psychotherapist notes based on categoriza- Al-Fedaghi, S. (2005a). How to calculate the
tion of types of PII instead of the rule’s general information privacy. In Proceedings of the Third
description that lack preciseness. The PII flow Annual Conference on Privacy, Security and
model furnishes a foundation for categorizing Trust: How to Calculate Information Privacy, St.
PII and acts on PII. This involves distinguishing Andrews, New Brunswick, Canada. Retrieved
different types of PII, different types of regions January 29, 2008, from http://www.lib.unb.ca/
of agents, and different types of acting on PII. Texts/PST/2005/pdf/fedaghi.pdf
This discretion of types of PII, of acts on PII, of
regions of agents clearly enhances the important Al-Fedaghi, S. (2005b). Privacy as a base for
requirement of releasing “the minimum amount confidentiality. In Proceedings of the Fourth
reasonably necessary to achieve the purpose of Workshop on the Economics of Information
the disclosure” (HSS, 2003, Limiting Uses and Security, Harvard University, Cambridge, MA.
Disclosures to the Minimum Necessary section). Retrieved January 29, 2008, from http://infosecon.
net/workshop/schedule.php

124
Scrutinizing the Rule

Al-Fedaghi, S. (2006, June 20-23). Aspects of GPC Alberta. (2003). Oipc stakeholder survey
personal information theory. In Proceedings of 2003—highlights report. Office of the Informa-
the Seventh Annual IEEE Information Assurance tion and Privacy Commissioner of Alberta (Tech.
Workshop (IEEE-IAW): Aspects of Personal In- Rep.).
formation Theory. West Point, NY: United States
He, Q., & Antón, A. I. (2003, June 16-17). A
Military Academy.
framework for modeling privacy requirements
Al-Fedaghi, S. (2007, July 12-14). When reinvent- in role engineering. In Proceedings of the Inter-
ing principles is necessary. In Proceedings of the national Workshop on Requirements Engineering
Seventh International Computer Ethics Confer- for Software Quality (REFSQ 2003), Klagenfurt
ence, University of San Diego, CA. / Velden, Austria. Retrieved January 29, 2008,
from http://crinfo.univ-paris1.fr/REFSQ/03/pa-
Bennett, C. J. (2001, June 19). What government
pers/P14-He.pdf
should know about privacy: A foundation paper.
In Paper prepared for the Information Technology HHS. (2003). Summary of the HIPAA privacy rule.
Executive Leadership Council’s Privacy Confer- U.S. Department of Health & Human Services.
ence (Revised August 1, 2001). Retrieve January Retrieved January 29, 2008, from http://www.
29, 2008, from http://www.accessandprivacy.gov. hhs.gov/ocr/privacysummary.pdf
on.ca/english/pub/wgskap.doc
HIPAA. (1996, August 21). Health Insurance
EU Directive. (1995, October 24). Directive Portability and Accountability Act of 1996, Public
95/46/EC of the European Parliament and of Law 104-191, 104th Congress. Retrieved January
the council. Retrieved January 29, 2008, from 29, 2008, from http://aspe.hhs.gov/admnsimp/
http://eur-lex.europa.eu/LexUriServ/LexUriServ. pl104191.htm
do?uri=CELEX:31995L0046:EN:HTML
OECD. (1980). Guidelines on the protection
Fischer-Hübner, S, & Ott, A. (1998, October 5-8). of privacy and transborder f lows of per-
From a formal privacy model to its implementa- sonal data. Retrieved January 29, 2008, from
tion. In Proceedings of the 21st National Infor- http://www.oecd.org/document/18/0,2340,
mation Systems Security Conference, Arlington, en_2649_34255_1815186_1_1_1_1,00.html
VA. Retrieved January 29, 2008, from http://www.
Westin, A. F., & Gelder, V. (2005, August). Build-
cs.kau.se/~simone/niss98.pdf
ing privacy by design in health data systems.
Galpottage, A., & Norris, A. (2005). Patient Center for Social and Legal Research: A report
consent principles and guidelines for e-consent: by The Program on Information Technology,
A New Zealand perspective. Health Informatics Health Records and Privacy of the Center for
Journal, 11(1), 5-18. Retrieved January 29, 2008, Social and Legal Research. Retrieved January
from http://jhi.sagepub.com/cgi/reprint/11/1/5 29, 2008, from http://www.amia.org/inside/initia-
tives/healthdata/2006/ehrrept9-6-05_westin.pdf

This work was previously published in International Journal of Healthcare Information Systems and Informatics, Vol. 3, Is-
sue 2, edited by J. Tan, pp. 32-47, copyright 2008 by IGI Publishing (an imprint of IGI Global).

125
126

Chapter 10
In What Ways Does Web
Technology Support the
Individual in Choice Reforms
in Health Care?
A Comparison among Norway,
Denmark, and Sweden

Agneta Ranerup
Göteborg University, Sweden

aBstract

The aim of this article is to evaluate the provision of Web support in choice reforms in health care in
Norway, Denmark, and Sweden. Two main issues are investigated: (1) What institutional frameworks
for choice in health care exist, and how is the exercise of choice supported by Web technology in these
countries? (2) As a consequence of this, what roles of the individual are mediated by this technology? The
present study provides a critical analysis of current technologies for providing information about health
care. It is concluded that in Norway the individual is equipped to be a reasonably informed consumer,
customer, and citizen. A similar situation exists in Denmark, but here the consumer role is even more
prominent. In Sweden, there has been little technological support for these roles, but recently national
actors have initiated a project aimed at creating a national portal for public health care.

During the last 10 years, patients have become all core issues in several fields of research such
more informed than they previously were about as e-health and Consumer Health Informatics
various aspects of health care, often by using (Eysenbach, 2001; Mureo & Rice, 2006; Nelson
information technology (IT) (Josefsson, 2005; & Ball, 2004; Tan, 2005). There are differences
Tovey, 2006). IT, patients, and health care are between these fields related to their history and

Copyright © 2011, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
In What Ways Does Web Technology Support the Individual in Choice Reforms in Health Care?

their focus of interest. However, it can be stated the OECD countries. Part of this renewal was
without controversy that a central rationale in an increase in the individual’s rights of choice
both of these fields is to support the individual in health care, mostly among public providers,
from the point of view of the medical rational- but sometimes also among private providers (Le
ity associated with the role of being a patient in Grand & Bartlett, 1993). When discussing the
need of care. This support might be provided introduction of choice reforms in health care, it
in the form of supporting the provision of care is obvious that the institutional frameworks of
and medical knowledge associated with the choice (Table 1) varies among countries. This
individual’s illnesses and treatments, including is also why comparative studies of institutional
the larger administrative process in a hospital frameworks of choice in different countries are
(Forducey, Kaur, Scheiderman-Miller, and Tan, of interest (Silber, 2004; Vrangbæk, Østergren,
2005). The patient role and its associated medical Winblad-Spångberg, & Okkels, 2006).
rationality are seen in contrast to other roles, for In a challenging way, a theoretical perspective
example, that of citizen. A simple but yet telling outlined by Anttiroiko (2004) connects the issue of
example of the former role is a piece of research patients, IT, and institutional frameworks of choice
evaluating government health portals (Glenton, reform in health care (Table 1). According to this
Paulsen, & Oxman, 2005). This article contains perspective, IT and the institutional frameworks
a broad comparative study of health portals in in conjunction mediate the relationship between
different countries. The focus is on assessing to the individual and public-sector services, for ex-
what extent government health portals provide ample, in the form of health care. The relationship
access to relevant, valid, and understandable between the individual and the public sector is,
information about the effects of various specified Anttiroiko argues, influenced by the coevolution
interventions (“treatments”). of technology and institutional frameworks. In
Against the background of the increasing this way, the individual will be in a position to
importance of IT for patients, it is interesting to pursue different roles (Table 1) in his interactions
note an emerging phenomenon in health care: with the public sector. This relationship is today
the introduction of choice reforms (Table 1). mediated by technologies such as e-mail, workflow
This development has its origins in the 1970s systems, and customer relationship management
and the renewal of the public sector, often re- (CRM) systems, as well as the Web, which is
ferred to as “New Public Management” (NPM), used in the production and communication of
which dominated the reform agenda in many of

Table 1. Overview of fundamental theoretical concepts

Fundamental concepts Brief explanations

A type of public-sector reform emanating from the 1970s, which
appeared in many OECD countries. Among other things, it intended
to offer the individual a larger role either in making a direct choice
Choice reform
among different providers of publicly financed services, or in exert-
ing an indirect influence through the representative of a purchasing
agency.
Laws or other forms of regulative agreements that delimit the right
Institutional frameworks of choice
of choice of hospital for individuals in these countries.
Different roles as well as accompanying rationalities and behavior
Patient, citizen, consumer, customer
that the individual may exercise toward health care.

127
 In What Ways Does Web Technology Support the Individual in Choice Reforms in Health Care?

health care services. In the relationship between sented, with a focus on the framework for choice
the public sector and the individual, there exists in health care in the three countries studied and a
also institutional frameworks of choice such as thorough description of the technical facilities in
laws and rights (Table 1), in this case in particular, each case. This is followed by a critical analysis
those that allow the individual certain rights to of these observations in the light of the potential
choose a preferred hospital. roles of the individual that might be mediated by
This perspective on health care may be primar- Web technologies in the three countries. Finally,
ily of interest to those countries in which health some conclusions and issues for further research
care is predominantly publicly financed (the end the article.
United Kingdom, the Nordic countries, Spain,
Italy, Australia, and Canada). However, it is of a
more general interest to countries with predomi- InstItutIonal FraMeworKs
nately private health systems (the United States For cHoIce wHIcH aFFect
and others) (Palier, 2005) because it contributes tHe role oF tHe IndIVIdual In
to a richer understanding of how IT affects the HealtH care
individual’s potential to examine available ser-
vices in today’s health care environment. For One basis for this study is the viewpoint that the
example, quality rankings that are offered through national institutional frameworks that regulate
the Web are relevant for individuals that are try- choice are of importance when discussing the
ing to inform themselves about both public and rights of choice in health care (Anttiroiko, 2004;
private health care. Vrangbæk et al., 2006). However, questions about
With this as a background, the aim of this institutional frameworks and individual rights
article is to evaluate the role of IT in general, are seldom addressed in e-health and Consumer
and the Web in particular, in choice reforms in Health Informatics (Eysenbach, 2001; Mureo &
health care in three Nordic countries (Norway, Rice, 2006; Nelson & Ball, 2004). One reason
Denmark, and Sweden) from the point of view of for this might be that, at face value, incorporat-
the individual seeker of care. Two main issues to ing institutional issues into the discussion would
be considered are the following: (1) What institu- diminish the validity of studies for health care and
tional frameworks for choice in health care exist, patients in different countries. Even in research
and how is the exercise of choice supported by in which the focus is on the individual’s informa-
Web technology in these countries? and (2) As a tion needs in consultation situations (Coulter,
consequence of this, what roles of the individual Entwistle, & Gilbert, 1999; Tovey, 2006), this
are mediated by this technology? In summary, this aspect is missing. Much research in these fields
research contributes to e-health and Consumer is based on experiences from the United States,
Health Informatics by taking the issue of institu- where health care is generally not based on uni-
tional frameworks seriously when discussing the versal rights of large groups in society to receive
roles of the individual as mediated by technology. subsidized health care (Palier, 2005; Tan, 2005).
The article continues as follows: the first This might explain why it is less common to take
section briefly elaborates on the positioning of the institutional framework into consideration in
the study in relation to previous research and is this context; however, there are a few exceptions.
followed by an overview of theoretical concepts One example is found in a book about Consumer
for discussing the potential roles of the individual Health Informatics which discusses collaborative
toward health care. Then the research method is healthware with a special focus on a group of
described. After that, the empirical study is pre- patients receiving Medicaid support (Goldsmith

128
In What Ways Does Web Technology Support the Individual in Choice Reforms in Health Care?

& Safran, 2004). In a discussion of evidence- tic views of being a citizen. The individualistic
based patient choice, Elwyn and Edwards (2001) perspective is the moral standpoint, meaning
include aspects of the institutional framework in that the relationship with the collective becomes
the medical choice situation, using experiences important and individual rights are emphasized.
from the United Kingdom. In contrast, Peckham The collective tradition puts the individual’s
(2002) discusses the harmfulness of omitting the interests aside to value collective ideas. In this
institutional framework in research about patients, case, it is the citizen’s obligation to the collective
especially when dealing with issues of choice in that is emphasized (Østergren, 2004). In a similar
health care. The focus of the present study is on vein, Willems writes about doctors as gatekeep-
the institutional and technological components ers guaranteeing fairness in health care. Thus,
of choice reform in health care. In line with the he argues, this arrangement “[…] constructs, in
theoretical framework proposed by Anttiroiko other words, patients as citizens” (Willems, 2001,
(2004), studying the institutional as well as the p. 27). A further dimension stems from Baggot
technological components of choice reform will (2005), who describes citizens’ specificity in their
be taken as the basis for discussing the individual potential to exercise “voice” rather than “choice.”
roles that are mediated in each country. Potential “voice” activities are, for example, taking
part in societal discourse on health care or even
pre-election debates about health care issues. This
concepts For dIscussIng line of thinking can be contrasted with other ways
tHe role oF tHe IndIVIdual In of having an impact on health care by choosing
HealtH care certain services and providers rather than others
(“choice”), and being in this way an active part
It is interesting to note that the concepts of patient, of the larger development and renewal of health
citizen, and consumer describe roles related to the care (Clarke, 2006). Thus, the concept of citi-
individual’s relationship with health care that are zen implies both individual capacities, such as
central in current health care discussions (Bag- knowledge about rights, and more collectivistic
got, 2005; Harrison, Dowswell, & Milewa, 2002). attributes such as having a sense of fairness and
When discussing the general public in e-health and being equipped to take part in discourse about
Consumer Health Informatics, the most prevalent the conditions of health care.
words used are “patient” and “consumer,” whereas The role of consumer in health care can be
the concept of “citizen” is much less common (Oh, defined by incorporating aspects of the ideal
Rizo, Enkin, & Jadad, 2005). Some authors argue of the “calculating consumer” emanating from
that these concepts are marked by lack of clarity economics (Elwyn & Edwards, 2001). In line
in terminology and usage (Baggot, 2005; Elwyn with this thinking, to behave as a consumer in the
& Edwards, 2001). The intention here is not to choice of a doctor or hospital, the individual must
provide an extensive discussion of the definitions be able to obtain an overview of alternatives and
of these concepts. However, an overview of how to compare and rank these alternatives (Greener,
roles and rationalities are defined in research is a 2003). No less important, when discussing the is-
necessary basis for the rest of this study. sue of offering individuals choice in health care
The first role to discuss in the individual’s reform, Østergren (2004) argues that the capacity
relationship with health care is that of citizen. to exit is the essential attribute of consumers. In
In her study of choice in health care, Østergren other words, they can vote with their feet (Clarke,
(2004) describes individualistic and collectivis- 2006). In contrast to these definitions and roles,
when Howgill (1998) speaks about health care

129
 In What Ways Does Web Technology Support the Individual in Choice Reforms in Health Care?

consumers, he is talking about the private market Because of the exploratory nature of the issues
for care, in which the individual depends on private investigated here, the chosen research approach is
economic resources to pay for care received. Thus, the case-study method (Yin, 1999). This method
the consumer role implies capacities to consider allows for the study of a single phenomenon within
different alternatives in a more elaborate way and its real-life context, while at the same time toler-
to make an informed choice among them. ating the condition that the boundaries between
Alternatively, one may consider the related role a phenomenon and its context are unclear. Data
of health care customers. According to Greener collection should ideally cover several sources.
(2003), the preferred as well as de facto attainable In addition, an operational framework should
role of individuals in their relationship toward be in place, even though the current intention is
health care is to be a health customer. This means exploratory. Having responded to requests for
being, not a full-fledged calculating consumer, a consciously planned research strategy rather
but an individual who is able to judge services than an informal procedure, this type of study
received based on preferences and expectations. can generate reliable results. All these aspects
This judgment might be in form of a performance have inspired the research approach of this study.
measurement regime in which actual patients The institutional frameworks for choice in the
would be asked for their opinions on the level of three countries are investigated by using other re-
service they have received, as well as the progress searchers’ writings on the issue of choice in health
they have made as a result of treatment. Last but care. Furthermore, official documents describing
not least, there is of course also the role of being the institutional frameworks in the three countries
a patient, or in other words, pursuing activities have been consulted. The intention has been to
related to medical rationality (Glenton et al., 2005), give an adequate, although not detailed, picture
as presented in the introduction. of the various frameworks. However, the main
focus of attention is the analysis of technologies
in each country from the point of view of choice
researcH context and reform in health care. Only public national portals
MetHod are included in this study, because they provide a
balanced view of how significant societal agencies
The context of this study is the institutional, but present and support choice reforms in health care
even more so, the technological component of in these countries. More specifically, the portals
choice reforms in health care in three of the Nordic are owned and implemented by national public
countries: Norway, Denmark, and Sweden. The agencies and associations in the three countries,
institutional framework of all these countries of- for example, the Ministry of Health in Norway, the
fers the option for a patient to choose a hospital Association of County Councils, the Department
other than the one that is closest to his or her place of the Interior, and the Department of Health in
of residence (Vrangbæk et al., 2006). These three Denmark, and the Association of County Councils
countries are marked by significant similarities in Sweden.
from an economic and cultural point of view. In The focus of attention is the content and
addition, it is no news that the access to Internet functionality of the three countries’ technologies.
in these countries is high. Thus, it is argued here This means that in the Norwegian case, the Web
that they are an interesting test bed for examining site www.sykehusvalg.no, introduced in 2003 as
the “state of the art” of technologies that support a prominent component of choice reform, is in-
the choice of hospitals. cluded as a whole. In Denmark, a national portal,
www.sundhed.dk, for health care was introduced

130
In What Ways Does Web Technology Support the Individual in Choice Reforms in Health Care?

in 2003. In this portal, all texts explaining choice application of the concepts for comparisons (Yin,
reforms and other associated facilities are included 1999) of the mediated roles.
in this study. This also means that links to other
national public agencies’ facilities related to choice
reform in health care are included in the analysis: results
a Web site describing waiting times in health
care for different treatments (www.venteinfo. overview of regulations
dk), a Web site describing available public and
private hospitals included in the choice reform Norway. In the early 1990s, free choice of hospitals
(www.sygehusvalg.dk), and an extended facility was established in one region as a pilot project. In
for the evaluation of quality of care (www.sund- Norway, a new act expanding choice to the whole
hedskvalitet.dk). In Sweden, there is no national country was passed in 1999 and became effective
public portal explicitly designed for supporting in January 2001 (Sosial- og helsedepartementet,
choice in health care. However, there are facilities 1998). It allowed choice among all public hospi-
offering an overview of waiting times for differ- tals. This reform was a part of a patients’ rights
ent treatments (www.vantetider.se). For the sake act which also included rights to assessment and
of completeness, a brief view is also given of the second opinions, treatment within national and not
Web site Sjukvårdsrådgivningen (Health advice only regional capacity limits, as well as the right
online), www.sjukvardsradgivningen.se, which of involvement and the right to information. One
is owned by the Association of County Councils provision was that the patients themselves should
in Sweden. pay little or no travel costs. In September 2004,
This author provides a straightforward, non- hospital choice was extended to private hospitals
theoretical walkthrough of the various facilities and particular hospital units within multihospital
that are available through the portals in the three trusts, as well as to child and psychiatric care. A
countries. An account of this investigation is also waiting-time guarantee, procompetition legisla-
offered in the results section. This is used as the tion, and expanded capacity were also put into
basis for analyzing the roles of the individual in place (Vrangbæk et al., 2006). This meant that
health care, which are mediated by their contribu- once a time limit has been exceeded in a region,
tion in view of the different role concepts presented the national office for social insurance must
in the literature (see “Concepts for discussing the help the patient find either a private hospital or
role of the individual in health care”). More spe- a hospital abroad (Sosial- og helsedirektoratet,
cifically, the features of the different role concepts 2004; Vrangbæk et al., 2006). This reform has
are compared to the results of this technological been characterized as late, but, at that time, it
walkthrough to see whether they “match.” For was more radical than anything in Denmark or
example, with regard to the role of a customer Sweden (Vrangbæk & Østergren, 2004).
in health-care facilities, seeing evaluations of Denmark. In the early 1990s, free choice of
services from the ordinary service users’ point of hospitals became an issue in public discourse.
view is fundamental for individual seekers of care The county councils initially opposed the idea,
(Greener, 2003). Furthermore, it is argued here but entered a voluntary agreement before a formal
that the mediated roles must be discussed with the parliamentary decision was made. When a formal
help of an approximate, informal qualitative scale decision was made in 1992, the issue of free choice
rather than an absolute and quantitative measure. was formulated as “extended choice” (Vrangbæk
The reason is the exploratory and qualitative & Østergren, 2004). This meant that hospitals
nature of study in this field and the associated were allowed to refuse access to extended-choice

131
 In What Ways Does Web Technology Support the Individual in Choice Reforms in Health Care?

patients in times of heavy workload. The patients referral from their general practitioner (GP) or
themselves had to pay their travel costs to receive another doctor. Some authors argue that patients
nonacute treatment in other counties. Moreover, lack information about rights, and because of this,
choice was limited to the same level of specializa- they are unable to benefit from them (Vrangbæk
tion and did not include private hospitals. This can et al., 2006; Winblad-Spångblad, 2003). In addi-
be characterized as a somewhat restricted model tion, the GP plays a significant role as a source of
(Vrangbæk & Østergren, 2004). knowledge and also in the role of issuing referrals
However, since May 2002, patients have been (Vrangbæck et al., 2006). This would imply that
free to choose any public Danish hospital and a the individual patient is in a difficult situation in
few hospitals owned by patient associations for regard to his or her potential to make use of these
elective treatments, in case their GP finds that this rights of choice.
is relevant. If a patient has waited more than 2
months in his or her home county, this right also overview of technologies for
includes private hospitals and hospitals abroad supporting choice in Health care
(Pedersen, Christiansen, & Bech, 2005; Vrangbæk
et al., 2006). Norway. In Norway, there has been some ex-
Sweden. In contrast to Norway and Den- perimentation with Web-based league tables
mark, free choice of hospitals in Sweden is not containing quality indicators and free telephone
mandated in formal law, but was adopted as a lines in 1998-2001 (Vrangbæk et al., 2006), or in
recommendation by the Federation of County other words, the period immediately preceding
Councils (Landstingsförbundet) in 1989. This the introduction of choice reform in health care.
meant that since the beginning of 1991, patients, However, when choice in health care was intro-
with some exceptions, have had the right to seek duced de facto in 2001, it was soon supported
care throughout the entire country at primary- by a Web site called Free Hospital Choice (Fritt
care centers and certain hospitals and private sykehusvalg), www.sykehussvalg.no, which was
clinics. In 2000, these recommendations were publicly launched in May 2003. A text on this
clarified and simplified (Landstingsförbundet, Web site, accessed on November 24, 2006* (all
2000a), but they were still perceived as contro- information marked by * comes from the same
versial by some county councils. In 2003, this source), explains its aim:
new recommendation was accepted by all county
councils (Vrangbæk et al., 2006). Through this [It] supports the government’s goal of facilitating
recommendation, patients are given the right to patients’ right to choose where to receive treat-
choose a hospital or a specialist anywhere in the ment. The service offers patients, next of kin, and
country, except in the case of highly specialized clinical personnel up-to-date quality information
care. In some counties (8 out of 21), the patient concerning patient’s rights, waiting times, and
must obtain a referral from his or her GP to seek quality information about the different hospitals,
hospital care outside his or her own county. In as well as other relevant information.
the other county councils, this is not necessary
(Vrangbæk et al., 2006). Moreover, according to this text, this purpose
Summary. The above discussion has revealed of this Web site is to empower Norwegian citizens
that all three countries have introduced institu- and to enable them to make better-informed deci-
tional frameworks that give the individual the right sions about which hospital to choose for different
of choice of hospital in the health care system. types of treatment. The stated background to the
To choose a different hospital, patients need a Web site is the Patients’ Rights Act. This law gives

132
In What Ways Does Web Technology Support the Individual in Choice Reforms in Health Care?

a patient the right to choose in which hospital he hospital, and overall experience of care received;
or she is to be treated. and (3) the number of patients residing in cor-
The Web site contains several sections with ridors as opposed to ordinary patient rooms at
different features. The first section is called Wait- the hospital.
ing times (Ventetider).* A comprehensive list of A third section is called Rights (Rettigheter).*
17 kinds of treatments and illnesses (pediatric This contains a general and easily understandable
treatments, heart ailments, psychiatric disorders, document about the right to choose a hospital for
etc.) is found on the front page of this section. treatment. There are documents stating the types
By clicking on one of these areas, the user can of hospitals that are eligible for choice and an over-
access a list of treatments. By selecting one of view of how health care will help the individual
these treatment choices, he or she obtains a list defray the costs of going to another hospital. There
of waiting times in number of weeks for medi- are also links to the official documents describing
cal examination, walk-in clinic treatment, and the framework for regulating choice.
inpatient treatment for all hospitals in the country A fourth section is called, For health care
that have adequate competence in each particular personnel (For helsepersonal).* These facilities
case. One can also go to an alphabetical list of can also be accessed by patients and are similar
treatments or click on a picture of the human to the ones provided to patients. However, in this
body. Here one can select a type of treatment and facility, the waiting-time data and quality indica-
in this way obtain the relevant waiting time. The tors contain a longitudinal dimension. With regard
database containing the waiting times for treat- to the quality indicators, through the facilities for
ments is searchable, but the search is limited to health personnel, it is possible to compare a larger
one or a few of the five hospital regions in Norway. number of indicators for each hospital than by
In the list of waiting times for treatments, a user using the facilities designed for citizens.
can also find information about quality indicators Denmark. In Denmark, a portal called Health
and type of institution (public or private) that are Care (Sundhed), www.sundhed.dk, was launched
relevant for each hospital. in 2003. On its home page, accessed November 24,
A second section is called Attention to qual- 2006** (all information marked by ** comes from
ity (Pekepinn på kvalitet).* Here a text explains the same source), is the statement: “Sundhed.dk
the concept of quality indicators, saying that the is your main entrance to health care in Denmark.
general public can use the indicators to get in- Here you get information that is useful for patients
formation about the quality of different services, and health personnel.” There are also facilities that
predominantly at the hospital level. It also states enable the individual patient to follow his or her
that the significant factor is the experiences of recorded treatments and diagnoses through the
patients, whereas the indicators do not necessarily facilities provided. The portal contains several
say much about the result of the medical treatment sections: Health, Treatments, Medicine, What
as such. There are links leading to descriptions about the law?, Facts and figures, Organization
of various quality indicators, such as: (1) the of health care, and News. There are also links to
number of planned operations that subsequently all the regions in Denmark and their hospitals.
are postponed; (2) indicators expressing different Furthermore, there is a special section for patients
views of quality of care from the perspective of containing links to the complaints center, facilities
both walk-in clinic patients and inpatients. For for self-service, donation of organs, and patient
each one of these, the figures are compared with records. A button labeled Treatments provides
mean value statistics for all hospitals in Norway: access to a special database with waiting times in
standard of premises, communication, organiza- weeks for treatment, Waiting times (Ventetider).**
tion, information, ease of getting around in the

133
 In What Ways Does Web Technology Support the Individual in Choice Reforms in Health Care?

This facility is created and administered by the treatments; (2) expected waiting time: for medi-
National Board of Health. A text says: “Here cal examination, hospital inpatient treatment, and
you can find waiting time for elective treatments walk-in clinic treatment; (3) complications. In the
and operations in public and private hospitals. case of heart failure, quality indicators include:
This information is only approximate, which (1) types of medical examination performed:
is why you should seek further advice through electrocardiogram (EKG), X-ray; (2) treatments:
your GP, patient advisor, or hospital personnel.” physical exercise, patient education, beta-blocker
This facility offers the option to select one of 21 use; (3) mortality. In contrast, for some types of
kinds of treatment, as well as associated treat- treatments, for example infertility treatments,
ment subtypes. Then the relevant hospitals for the there are no indicators related to the medical
selected treatment appear, as well as the waiting examinations or results of treatment, but only
time. It is possible to include private hospitals in the expected waiting time for treatment. There
this search. There is also an option to restrict the are 21 kinds of treatment to choose from, fol-
search to various regions in Denmark. lowed by associated treatments for each primary
Moreover, by clicking on the Treatments, treatment chosen. Search can also be limited on a
button, one can also access the new facility that geographical basis (by region). When using these
was introduced in November 2006: Quality in facilities, the quality indicators for the hospitals
Health Care (Sundhedskvalitet).** A press release implementing the selected treatments are shown.
introducing this facility says that, “This Web site The Treatments section in the Health care
is the first version of how a complete picture of (Sundhed) portal also contains a subsection called
quality and service can be made available through Hospital statistics. Here it is possible to search
the Internet offering information about individual statistics for both public and private hospitals to
hospitals.”** This intention is further described on determine the total number of patients and the
its home page as a means to support the choice of most prevalent types of treatment for each hospital.
hospital. It is also stated that by looking at statistics There are also statistics for the duration of treat-
about different aspects of hospital operation, it is ments in general (for inpatients) and for the most
possible to obtain information about the quality prevalent types of treatment in each hospital. In
of both public and private hospitals. The user is the Treatments section, there is a final subsection
advised to make a choice of quality indicators from related to choice in health care: Quality of care.
those that are available: (1) standard of premises: Here one can find information and databases
number of beds per room, plus toilets; 2) sanitary related to different quality studies in association
conditions: hand washing, postoperative infec- with different types of treatments. There is also
tions, kitchen facilities, and sanitary services in a database containing patient views of care, de-
general; (3) rights: contact person, waiting-time rived from a patient survey conducted in 2004. It
guarantee, and extended choice of hospital; (4) contains patient views about, for example, their
patient safety: mistakes in medication, injuries relationship with hospital personnel, their sense
during surgery; and (5) patient satisfaction: of inclusion in care, information provided when
general, sense of inclusion in care, safety when leaving the hospital, hospital premises, and wait-
leaving the hospital. ing times. Another section accessible from the
There are also quality indicators for the dif- home page, What about the law?, is also relevant
ferent types of treatments or kinds of illness. to the issue of choice in health care. This section
For example, for cataract operations, there are has several subsections, of which one is called
quality indicators for: (1) activities: number of Free choice of hospitals. Here the frameworks for
patients, average number of days for inpatient choice are summarized, with links to documents

134
In What Ways Does Web Technology Support the Individual in Choice Reforms in Health Care?

containing the actual frameworks for choice. strengthen the position of patients by increasing
Equally important, on the Health care portal, there the transparency of health care and making the
is also a link to a Web site called Free Hospital various waiting times public. In 2005, waiting
Choice (Sygehusvalg.dk)** implemented by the times for both specialized care and primary care
Association of County Councils. The home page were included. This Web site could be accessed
of this Web site states that: directly as well as through the Web sites of the
county councils in Sweden.
Patients that do not receive care by a public hos- The home page of Waiting Times in Care, ac-
pital after two months of waiting now have the cessed on November, 26, 2006*** (all information
option to seek care at private clinics in Denmark marked by *** comes from the same source),
as well as hospitals abroad. […] To use the right indicates that the site contains waiting times for
of free choice of hospital, there is a precondition both specialized and primary care. There are also
that the chosen hospital has an agreement with the links to a news section as well as to reports. Both
county councils and hospitals about the relevant of these are primarily directed toward specialists
treatment.** because their main focus is on the finer technicali-
ties of waiting times and how they are reported.
At this level, there is a link called Patients. By The database offers the possibility of searching
clicking on this link, one finds general information for a certain hospital, treatment, kind of illness,
about patient rights as well as the Web site itself. or region. 26 kinds of treatment are available for
There are also links to waiting times (the database choice, six types of specialized medical examina-
described above) and a leaflet about the free choice tion, and approximately 40 treatments. For each
of hospitals. There is an option to search among all treatment, the relevant hospitals, waiting time in
hospitals included in the free choice of hospitals weeks, information about free capacity, and con-
after the 2-month waiting period. In this database, tact information of hospitals with free capacity are
there are options to search in an alphabetical list presented. The information about free or excess
containing all private hospitals and clinics, to capacity is defined as the potential of each clinic
search within selected geographical boundaries, to receive patients coming from county councils
and to search among 18 areas of treatment (breast, other than its own. The search can be limited to
heart, orthopedic, and others). For each one of data about first visit, medical examination, and
these alternatives, there is a selection of relevant treatment.
treatments. Clicking on any one of these brings Concerning the issue of choice in health care,
up a list with all the private hospitals that are in a document*** available on the Web site expands
a position to offer the selected type of treatment. on the relationship between the waiting-time guar-
Sweden. In Sweden, a Web site called Waiting antee and the free choice of hospitals and explains
Times in Care (Väntetider i vården), www.van- that the ways of exercising choice are decided by
tetider.se, was introduced in April 2000, contain- the individual county council. It is also pointed out
ing a database of waiting times for various types that a prerequisite for exercising the right of choice
of treatments. It was created by the Association is that a doctor in your home county council has
of County Councils (Landstingsförbundet). At issued a referral. This treatment can then be car-
that time, the database contained waiting times ried out in another county council’s jurisdiction.
for approximately 30 different treatments and Sometimes, when the treatment is costly, some
included about 30 specialist hospitals (Landst- county councils may request special permission
ingsförbundet, 2000b). The database is part of before the right of choice can be requested. The
efforts by the county councils and regions to rights to information about options are described

135
 In What Ways Does Web Technology Support the Individual in Choice Reforms in Health Care?

as follows: “Normally, it is the responsibility guarantee, but no information about the rights of
of the patient to find an alternative provider of choice in health care.
care.” This formulation can be contrasted with In 2006, the Association of County Councils in
the recommendation about choice in health care Sweden initiated a project to build a national portal
that since 2003 has been adopted by all county for health care: Health care on the Web (Vården
councils in Sweden: “It is an important task of the på webben). The portal Health advice online and
county councils to inform their inhabitants on a its developer provided the basic infrastructure for
continuous basis about the options for choice in this work, but since 2007, all 21 county councils
health care. Such information must be directed have joined in. A prominent goal behind the new
towards the population as a whole as well as to national portal is that an individual should have
those that are seeking care” (Landstingsförbundet, access to information about all kinds of providers
2000a, p. 2). of care, including those that are not part of the
Furthermore, in 2001, the National Board of publicly financed health care system. A further
Health and Welfare in Sweden published a report goal is that the national portal should enhance
on quality indicators in health care written as a gov- the individual’s ability to compare conditions and
ernment project. In this report, the use of quality regulations, waiting times, and quality of care
indicators from internal and external viewpoints (MD, National project, personal communication,
was very briefly discussed as a basis for choice of March 29, 2007). This ambition has been backed
health care provider. The Web site, Waiting Times up by government policies (Socialdepartementet
in Care, was characterized as an important basis et. al., 2007). The launch of the national portal is
for the choice of health care provider as pursued planned for 2009. The design of individual facili-
by patients and doctors (Socialstyrelsen, 2001). A ties was discussed intensively in the autumn of
new report including a broad spectrum of quality 2007, and requirements documentation is planned
indicators for different treatments was published to be available for the purchasing process to start
in 2006 (SKL & Socialstyrelsen 2006). However, in December 2007 (Designer, National project,
this report did not form part of an online service. personal communication, September 18, 2007).
The indicators included showed data that compare
the different county councils, not the individual
hospitals. Furthermore, the indicators were explic- dIscussIon
itly introduced as not to be used as a basis for the
choice of hospital (SKL & Socialstyrelsen 2006, technology in choice reforms in
p. 12). In the autumn of 2007, a new report was norway, denmark, and sweden
produced in which the individual hospital’s values
for some of the indicators were published as a test There is a national Web site aimed at supporting
case (SKL & Socialstyrelsen, 2007). choice of hospitals in Norway, whereas in Den-
In Sweden, there is also a portal, Health advice mark there is a national portal for health care as
online (Sjukvårdsrådgivningen), www.sjukvard- a whole that contains many facilities to support
sradgivningen.se, owned by the county councils in choice. Sweden, by contrast, does not have facili-
Sweden. The focus of this Web site is on illnesses, ties at the national level for supporting the free
injuries, anatomy, health, drugs, and treatments. choice of hospitals other than those dedicated to
There is also a section on rights, focusing mostly the waiting-time guarantee and some very brief
on disabilities, means of assistance, and dental information about rights of choice in connection
care. There is information about the waiting-time with this. These three cases illustrate the different
ways that technology is implemented by important

136
In What Ways Does Web Technology Support the Individual in Choice Reforms in Health Care?

national authorities and associations in health making information available at a distance and
care. More specifically, such an implementation around the clock.
might be in the form of a standalone facility for Against this background, it is possible to
supporting choice of hospital, as in Norway, or of conclude that the Web technology features imple-
a facility that is part of a larger portal incorporat- mented in these three countries are fairly simple:
ing facilities from different public agencies, as adding capabilities to restrict search in various
in Denmark. Thus, in the Danish case, there is a ways, connecting various types of information
common interface to health care as a whole. This which support choice, and providing a choice of
is good from the point of view of the individual preferred interfaces. However, there also exists the
for whom the concept of a one-stop portal is con- potential to provide more advanced, composite in-
venient, offering a single window to all available teractive decision support to the individual seeker
services (Wimmer, 2002). of care. In this case, simple and straightforward
From a technological viewpoint, the facilities Web technology might be used for supporting
for providing these kinds of information in the choice by arranging a logical chain of operations
three countries are quite simple. However, when it that the individual can follow. One example is
comes to the facilities showing waiting times, Nor- from the field of educational and career guidance,
way provides several types of interfaces: graphical for which computer-based decision support has
(a picture of the human body) as well as textual been provided through the Web for some years
(lists with sublevels for kinds of illnesses and (Tait, 1999). Moreover, in the Swedish labor mar-
treatments, or an alphabetical list of treatments). ket, the National Labor Market Board has, since
The two other countries provide textual interfaces 2004, provided computerized decision support
only. The Norwegian technology in this respect through the Web through which an individual
gives the individual the option of using various states his or her preferences regarding choice of
interfaces according to his or her preferences and education and vocation. This part of the decision
levels of knowledge. For finding relevant hospitals chain is followed by further facilities to delimit
for treatments, it is possible to restrict the search search, examine available options, and rank these
in several ways, by geographical area as well as options (Norén & Ranerup, 2005). Recently, a
by private or public provider. Also in Norway, the similar model has been introduced to provide
facilities that show waiting times provide links computerized decision support to the individual
to quality information for the chosen hospitals, citizen in the new premium pension system in
thus making it easier for the individual to combine Sweden (Ranerup, 2006).
these two aspects of choice. Last but not least, the
Danish facilities for quality indicators provide Mediated roles of the Individual with
various options for restricting search (illnesses, respect to Health care
treatments, geography, or type of quality indica-
tor). This is good if, as in Denmark, a broader Norway. In Norway, the facilities provided in
spectrum of quality indicators is offered and care www.sykehusvalg.no have the potential to support
is taken to avoid information overload when the the individual in becoming a reasonably informed
results are shown. Moreover, preferences regard- citizen with regard to the rights of choice (Øster-
ing indicators might vary among individuals and gren, 2004). An individual user can find more
situations. These are the most obvious examples of general information about rights and benefits as
the fact that technological facilities can provide an well as links to the relevant official documents
extra performance-enhancing capability beyond regulating choice. The sense of fairness (Willems,
the most obvious and simple aspect of the Web: 2001) in the distribution of and access to care is to

137
 In What Ways Does Web Technology Support the Individual in Choice Reforms in Health Care?

some extent supported by the waiting-time infor- and treatment for a specific disease as well as
mation for each care provider. More specifically, with the results of treatment are absent. This
fairness is supported by transparency regarding means that the individual in Norway is equipped
such strategic information, which is fundamental to compare different hospitals, although not in a
from a democratic point of view: in this way, the deep and elaborate way. It is argued here that the
differences in waiting times for various hospitals quality indicators provided in Norway are quite
and types of illness become visible. As in the limited in the sense that the individual is equipped
provision of all kinds of public services, there is to function as a customer rather than as a fully-
the potential for citizens to exercise voice as op- fledged calculating consumer who is capable of
posed to choice (Baggot, 2005), in this particular comparing and ranking a considerable number
context by taking part in the societal discourse of available options. According to Greener: “[H]
on health care. This is supported by information ealth customers would judge the level of service
about rights, waiting times, available options in they believe they have received, and assess health
the form of public and private hospitals, and, services according to their own expectations of
last but not least, quality indicators. A limitation the services as well as to national standards”
here is the focus on choice in health care in the (Greener, 2003, p. 86). The most important thing
technological facilities provided, which means here is, according to Greener, that it is the patients’
that the individual is not offered an overview of opinion of the level of service they have received
health care in general through a one-stop por- that is the focus of attention.
tal (Wimmer, 2002). Such an approach would Moreover, the role focusing on medical or
make the individual more knowledgeable about patient rationality is supported by the facilities
health care in general, but would also imply that provided, because these facilities offer informa-
individuals are using the available facilities on a tion that is of relevance to those seeking care.
regular basis. A general overview of available hospitals and
In Norway, the individual is also reasonably their competence in various of treatments and
well equipped through the Web to behave as a illnesses is part of this information. However, in
consumer (Greener, 2003; Harris, 2003), by be- the Norwegian case, as already noted, there is
ing offered an overview of available hospitals the limitation that the focus of the information
for requested treatments and information about on facilities is solely on the choice of hospital.
quality of care. There are various types of quality Denmark. In Denmark, individuals are reason-
indicators in health care (Rygh & Mörland, 2006; ably well equipped to function as citizens through
Socialstyrelsen, 2001): indicators for structural the available technological facilities which provide
issues (capacity for care, education of personnel), an overview of rights of choice and fairness in a
for process issues (to what extent clinical praxis manner equivalent to Norway. However, in their
is in accordance with the optimal process for role as consumers, the Danes are a little better
medical examination and treatment), for the result off than the Norwegians. This is mainly because
of treatments, for example mortality, and for the of the recently implemented facilities offering
patient’s view and experiences (Socialstyrelsen, quality indicators for health care, Quality of
2001). In the Norwegian quality indicators, there Care (Sundhedskvalitet), which provide a basis
is a focus on structural issues in general, which for making comparisons. In fact, the introduction
appears to some extent also in the form of pa- and continued development of quality indicators
tients’ views on structural conditions. However, was explicitly a part of supporting choice reform
indicators for comparing the treatment received in health care (Socialstyrelsen, 2001). In terms of
with the optimal course of medical examination the categories of quality indicators classified as

138
In What Ways Does Web Technology Support the Individual in Choice Reforms in Health Care?

structural-, process- and results-oriented (Rygh online (Sjukvårdsrådgivningen), the individual

& Mörland, 2006), the Danish facilities contain is, to some extent, equipped to act as a patient,
instances of all of these. Interestingly, this in- because many forms of support in line with
cludes the result of treatments both at a general the patient or medical rationality are offered.
level (complications, mortality) and connected However, these are not connected to individual
with specific treatments. As for the issue of sup- hospitals or primary-care providers, but are kept
porting individuals as customers, there are also on a general level. In Sweden, the regulatory
indicators expressing the more personal views of framework says that it is an important task for
patients (Socialstyrelsen, 2001) with regard to care the county councils to keep the general public,
received. These are in the form of five indicators as well as those that are seeking health care, con-
covering different aspects of patients’ safety and tinuously informed about their rights of choice
satisfaction from the patients’ point of view. In fact, (Landstingsförbundet, 2000a), but not much is
in Norway and Denmark, the individual patient is said about how or in what forms this should be
the target of these attempts that, as a further step done. However, in the information about choice
in these processes, are made available through that does exist at a national and regional level
the Web and therefore more accessible as well (see above) the individual is defined as being
as searchable. However, this is not to say that all responsible for finding information about the
patients have the indicators they need to make a hospitals that he or she wants to use as a part
choice or that they have a sufficient understand- of the right of choice. This exposes a mixed
ing of available indicators to use them in their attitude toward the individual in regard to his
decisions (Rygh & Mörland, 2006). or her rights to information in choice reform
Last but not least, the Danes are also better in general and information about available op-
off with regard to facilities offered to support the tions for choice in particular. Of course, there
individual as a patient. Not only are facilities that is a significant difference between, on the one
are relevant for choice offered (available hospitals hand, Norway and Denmark and, on the other
and treatments), but also many types of general hand, Sweden. In Norway and Denmark there are
and specific information about health care, cover- laws guaranteeing the right of choice, whereas in
ing quality of care as well as health in general. Sweden there is an agreement among the county
Recently (autumn 2006), the connections between councils. In Sweden, regional authorities have a
the portal www.sundhed.dk and the individual comparatively large influence when it comes to
patient have been made stronger by the introduc- de facto regulation of citizens’ rights. In addi-
tion of technological facilities which enable the tion, as argued by Karlsson (2003), in Sweden,
patient to review his or her own treatments. In the individual generally has no rights in his or
this manner, the facilities provided can be seen her role as a citizen that can be requested from
as a means of connecting the individual to many an authority or institution in health care or
aspects of health care, including those that have elsewhere, in the way that such a request could
to do with institutional and medical issues. be made, for example, in the United Kingdom
Sweden. The Web facilities supporting choice (UK). It is seen as a paradox that less generous
in Sweden must be characterized as modest. The welfare states, for example the UK, offer citi-
brief information about rights and the waiting- zens more individual rights than does Sweden
time information to some extent support the (Karlsson, 2003).
individual in the role of citizen. As for support Furthermore, the quality-indicator projects in
for the role of consumer, there is not much to Sweden are marked by the absence of indicators
be found. In the updated portal, Health advice supporting choice, because there are no indica-

139
 In What Ways Does Web Technology Support the Individual in Choice Reforms in Health Care?

tors for individual hospitals. However, attempts conclusIon

are being made to introduce such indicators as
of the autumn of 2007 (SKL & Socialstyrelsen, This study contributes to research in e-health
2007). The indicators cover various aspects and Consumer Health Informatics by investigat-
such as waiting times, treatments used, results ing the roles of the individual that are mediated
of treatments, and patients’ views. The rationale by technology in choice reforms in health care,
behind the work with indicators in Sweden can thus emphasizing the joint role of institutional
be expressed in the following way: “Citizens frameworks of choice and technology. From a
and patients have the right to know what health general point of view, the technological support
care accomplishes and to compare their own for this objective is considerable in Norway and
county council with others. An equally important Denmark, but much less developed in Sweden.
aspect is that comparing results enhances learn- However, it is fair to say that the Web technology
ing and the further development of health care” features that are implemented are fairly simple,
(Socialstyrelsen & SKL, 2006, p. 10). Thus, it offering capacities for delimiting search, connect-
appears that a discursive rationale about enhanc- ing various types of information, and choosing
ing transparency for citizens through quality the preferred type of interface. However, in the
indicators is considered important in Sweden, facilities offered by the three countries, there is
although implemented rather half-heartedly. The no more advanced computerized decision support
reason for this judgment is that the actual reports to enhance the choice of hospital. In spite of this,
describing the quality indicators are not more Web technology offers information about hospi-
specifically targeted to be understood by citizens. tals, rights, and quality of care right at hand to
During the autumn of 2006, there was a the individual, thus bypassing, for example, doc-
change of government in Sweden. The intention tors, who have often had the role of information
to increase choice in health care is supported in gatekeepers (Winblad-Spångberg, 2003).
policy writings of the new coalition (Allians för As for the roles of the individual as mediated
Sverige, 2006). It is argued that hospitals should by technology, in Norway, the technology sup-
be responsible for informing the individual about ports the individual in becoming a reasonably
where he or she can obtain care with minimal informed citizen regarding the rights of choice, as
waiting time. In this way, current political events well as a consumer with some capacity to make
might change the technological aspects of choice comparisons and choose a hospital. In Denmark,
reform in health care in Sweden. It is also interest- the individual is equipped to behave as a reason-
ing to note that recently (January 2007), all the ably informed citizen, but also to a significant
county councils aligned themselves with a new extent as a consumer because of having access to
project which aims to launch a national health more fully developed information about services
care portal in 2009. The purchasing process is (hospitals, treatments, quality indicators) and
planned to start in 2008, followed by a complex facilities to compare these. In Sweden, there is
implementation phase involving the 21 county considerably less potential for the individual to
councils. This initiative can be interpreted as become an informed citizen or consumer by means
a late attempt to combine the resources of the of available technology. A difference between
comparatively independent county councils, the three countries is that the rights of choice in
with, to date, an unclear prognosis when it health care are implemented by means of laws
comes to the chances of a successful result in a in Norway and Denmark, but in Sweden, rights
reasonable time. of choice are a recommendation that has been
accepted by all the County Councils.

140
In What Ways Does Web Technology Support the Individual in Choice Reforms in Health Care?

The focus of this study has been on tech- the Web. Another interesting issue for research
nological support for choice and not on other is what indicators citizens actually prefer as a
potential sources of support, for example patient basis for choice.
advisors or paper documents that are not a part Yet another issue for further research is the
of a technological infrastructure such as the Web. potential to introduce computerized decision sup-
The approach chosen has been considered to be port that combines different stages and aspects
relevant against the background of the increas- of a decision process for hospital choice, in line
ing importance of IT for patients and health care with similar attempts in the fields of educational
alike. It is based on a theoretical framework that and career guidance and premium pension issues,
sees the relationship between the individual and as discussed earlier. These forms of support are
the state as more than that previously mediated based on a model for relevant decision-making
by institutional arrangements and technology in the area of concern. In the area of educational
(Anttiroiko, 2004) and thus contributes to the and career guidance, there are certain theories
discourse in this field. that support this (Law, 1999). In regard to the
The research presented here also contributes choice of premium pension funds, the decision
to the contemporary praxis of health care. At a support provided is designed on the basis of
general level, this study has shown how different financial theory and psychology (SOU, 2005).
types of technology supporting choice in health An interesting issue, then, is what theoretical
care can be designed in today’s health care envi- frameworks might form a part of similar design
ronment. This is of interest to countries that are attempts in health care.
expanding choice reform in health care and wish Last but not least, the newly initiated project
to use IT as a part of this effort. A recent example of implementing a national health care portal is
in this respect is the UK, with its intention to both relevant for the issues discussed here and
offer technological facilities to support choice a challenging issue for further research. This is
(Department of Health, 2003), which have been even truer because one of the major objectives
more recently (2005-2007) rolled out. of the portal is to provide information about all
As for further research into technology to kinds of health care providers as well as to sup-
support choice in health care, the finer aspects of port comparisons of rights, quality, and waiting
providing quality indicators to patients through times. However, implementing an infrastructure
the Web are of interest. The problem is one of for a portal at a national level is one thing, but
providing understandable and relevant informa- ensuring that the 21 county councils in Sweden ac-
tion about important aspects of care and how this tively implement the requested regional structure
might be communicated to the individual (Rygh with information and interactive services is quite
& Mörland, 2006). It has been argued that the another. As summarized by a high-ranking civil
use of quality indicators is relevant from many servant, “The local leadership of this process is
viewpoints. However, the fact that there are meth- the most important thing” (MD, The Association
odological problems with, for example, reporting of County Councils Purchasing Agency, personal
the result of health care makes the use of these communication, December 10, 2007).
indicators for comparing hospitals problematic
when they are used as a basis for choice (Schen,
2005). The question is, then, whether this is a acKnowledgeMent
surmountable problem, or whether new types of
quality indicators for this particular area of use Thanks are due to the Bank of Sweden Tercente-
will have to be developed and provided through nary Foundation for funding this research.

141
 In What Ways Does Web Technology Support the Individual in Choice Reforms in Health Care?

reFerences Forducey, P., Kaur, K., Scheiderman-Miller, C., &

Tan, J. (2005). Surveying the e-health landscape:
Allians för Sverige. (2006, August 23). Election Cases and applications. In J. Tan (Ed.), E-health
program [Valmanifest]. Retrieved January 29, care information systems. An introduction for
2008, from www.maktskifte06.se/fileadmin/ students and professionals (pp. 203-224). San
Upload/pdf/Valmanifest_2006.pdf Fransisco, CA: Jossey Bass.

Anttiroiko, A. (2004). Introduction to democratic Glenton, C., Paulsen, E., & Oxman, A. (2005).
e-governance. In M. Mälkiä, A.-V. Anttiroiko, & Portals to wonderland: Health portals lead to con-
R. Savolainen (Eds.), E-transformation in gover- fusing information about the effects of health care.
nance: New directions in government and politics Medical Informatics and Decision Making, 5(7).
(pp. 22-49). Hershey, PA: Idea Group.
Goldsmith, D., & Safran, C. (2004). Collaborative
Baggot, R. (2005). A funny thing happened on healthware. In R. Nelson & M. Ball (Eds.), Con-
the way to the forum? Reforming patient and sumer informatics. Applications and strategies
public involvement in the NHS in England. Public in cyber health care (pp. 9-19). Berlin: Springer.
Administration, 83(3), 533-551.
Greener, I. (2003). Patient choice in the NHS: The
Clarke, J. (2006). Consumers, clients or citizens? view from economic sociology. Social Theory &
Politics, policy and practice in reform in social Health, 1, 72-89.
care. European Societies, 8(3), 423-442.
Harris, K. (2003). How do patients choose phy-
Coulter, A., Entwistle, V., & Gilbert, D. (1999). sicians? Enrolees in employment-related health
Sharing decisions with patients: Is the information plans. Health Services Research, 38(2), 711-732.
good enough? BMJ, 318, 318-322.
Harrison, S., Dowswell, G., & Milewa, T. (2002).
Degoulet, P., Fieschi, M., Jaulent, M-C., & Ménard, Guest editoral: Public and user “involvement” in
J. (2004). Patient empowerment, cybermedicine, the UK National Health Service. Health and Social
and citizen education. In R. Nelson & M. Ball Care in the Community, 10(2), 63-66.
(Eds.), Consumer informatics. Applications and
Howgill, M. (1998). Health care consumerism, the
strategies in cyber health care (pp. 120-128).
information explosion, and branding: Why ’tis
Berlin: Springer.
better to be the cowboy than the cow. Managed
Department of Health. (2003). Choice of hospitals: Care Quarterly, 6(4), 33-43.
Guidance for PCTs, NHS Trusts and SHAs on of-
Josefsson, U. (2005). Coping with illness online:
fering patient choice of where they are treated.
The case of patients’ online communities. The
London: Department of Health.
Information Society, 21(2), 143-153.
Elwyn, G., & Edwards, A. (2001). Evidence-based
Karlsson, L. (2003). Conflict or harmony? Indi-
patient choice? In A. Edwards & G. Elwyn, G.
vidual rights in Swedish and British health care
(Eds.), Evidence-based patient choice. Inevitable
[Konflikt eller harmoni? Individuella rättigheter
or impossible? (pp. 3-18). Oxford University Press.
och ansvarsutkrävande i svensk och brittisk sjuk-
Eysenbach, G. (2001). What is e-health? Journal vård]. Doctoral dissertation, Göteborg: CEFOS,
of Medical Internet Research, 3(2), Article e20. Göteborgs universitet.

142
In What Ways Does Web Technology Support the Individual in Choice Reforms in Health Care?

Landstingsförbundet. (2000a). Recommendation Peckham, S. (2002). Book review. Health and

about choice in health care. [Rekommendation Social Care in the Community, 10(6), 520-522.
om valmöjligheter inom hälso- och sjukvården].
Pedersen, K., Christiansen, T., & Bech, M. (2005).
Dnr Lf924/00. Stockholm: Landstingsförbundet.
The Danish health care system: Evolution–not
Landstingsförbundet. (2000b). Waiting times in revolution–in a decentralized system. Health
care on the Internet [Väntetider i vården på Inter- Economics, 14, 41-57.
net. Pressmeddelande] 2000-04-10. Stockholm:
Ranerup, A. (2006). Electronic markets connect-
Landstingsförbundet.
ing citizens to pension reform. Electronic Markets,
Law, B. (1999). Career-learning space: newDOTS 16(4), 282-291.
thinking for careers education. British Journal for
Rygh & Mörland. (2006). Hunting for good quality
Guidance & Counselling, 27(1), 35-55.
indicators [Jakten på de gode kvalitetsindikato-
Le Grand, J., & Bartlett, W. (1993). Introduction. rerne]. Tidsskrift for Den Norsk Lægeforening,
In J. Le Grand & Bartlett W. (Eds.), Quasi-markets 21, 2822-2825.
and social policy (pp. 1-12). London: Macmillan.
Silber, D. (2004). The case for e-health. In T.
Mureo, M., & Rice, R. (2006). The Internet and Iakovidis, P. Wilson, & J. Healy (Eds.), E-health
health care. Theory, research, and practice. NJ: (pp. 3-27). The Netherlands: IOS Press.
Lawrence Erlbaum.
SKL & Socialstyrelsen. (2006). Open compari-
Nelson, R., & Ball, M. (2004). Consumer informat- sons of quality and efficiency in health care. A
ics. Applications and strategies in cyber health comparison between county councils in 2007
care. Berlin: Springer. [Öppna jämförelser av hälso- och sjukvårdens
kvalitet och efektivitet. Jämförelser melan lands-
Norén, L., & Ranerup, A. (2005). The Internet
ting 2007]. Stockholm: SKL, Socialstyrelsen.
Web portal as an enrolment device, In B. Czar-
niawska & T. Hernes (Eds.), Actor-network theory SKL & Socialstyrelsen. (2007). Open compari-
and organizing (pp. 188-207). Malmö: Liber and sons of quality and efficiency in health care. A
Copenhagen Business School Press. comparison between county councils in 2007
[Öppna jämförelser av hälso- och sjukvårdens
Oh, H., Rizo, C., Enkin, M., & Jadad, A. (2005).
kvalitet och efektivitet. Jämförelser melan lands-
What is e-health (3): A systematic review of
ting 2007]. Stockholm: SKL, Socialstyrelsen.
published definitions. Journal of Medical Internet
Research, 7(1), article e1. Socialdepartementet, SKL, Socialstyrelsen.
(2007). National IT strategy in health care. Where
Østergren, K. (2004). The institutional construc-
do we stand? Report 2007 [Nationell IT-strategi för
tion of consumerism. A study of implementing
vård och omsorg. Var står vi idag? Lägesrapport
quality indicators (Working paper 15). Bergen:
2007]. Stockholm: Socialdepartementet, SKL,
Stein Rokkan Centre for Social Studies.
Socialstyrelsen.
Palier, B. (2005). How to understand the dif-
Socialstyrelsen. (2001). A comprehensive view
ferences between various health care systems
of quality indicators in health care. A proposal
[Att förstå skillnadena prestation mellan olika
by the National Swedish Board of Health and
hälso- och sjukvårdssystem]. Stockholm: Sveriges
Welfare [Övergripande kvalitetsindikatorer inom
kommuner och Landsting.
hälso- och sjukvården] Stockholm: Socialsty-
relsens förslag.

143
 In What Ways Does Web Technology Support the Individual in Choice Reforms in Health Care?

Sosial og helsedepartementet. (1998). Law about Vrangbæk, K., Østergren, K., Winblad-Spång-
patient rights [Lov om patientretttigheter]. Ot berg, U., & Okkels, B. (2006). Patients’ reactions
prpnr 12 (1998-99). Oslo: Sosial- og helsedepar- to free choice of hospitals in Norway, Denmark
tementet. and Sweden. Bergen: Norwegian School of Eco-
nomics and Business Administration.
Sosial-og helsedirektoratet. (2004). Law on
patient rights. Instructions [Lov om pasientret- Willems, D. (2001). Balancing rationalities:
tigheter. Rundskriv]. IS12/2004. Oslo: Sosial- og Gatekeeping in health care. Journal of Medical
helsedirektoratet. Ethics, 27, 25-29.
SOU. (2005). Difficult waters? Premium pension Wimmer, M. (2002). Integrated service model-
savings on course [Svårnavigerat? Premiepen- ling for online onestop government. Electronic
sionssparande på rätt kurs?], SOU 2005:8, Swedish Markets, 12(3),149-156.
Government Official Reports. Stockholm: Fritzes.
Winblad-Spångberg, U. (2003). From decision to
Tait, A. (1999). Face-to-face and at a distance: The practice: The role of the physician in implement-
mediation of guidance and counselling through ing patient choice in Swedish health care [Från
new technologies. British Journal of Guidance beslut till verklighet: Läkarnas roll vid imple-
& Counselling, 27(1), 113-122. menteringen av valfrihetsreformer i hälso- och
sjukvården]. Uppsala: Institutionen för folkhälso-
Tan, J. (2005). E-health care information systems.
och vårdvetenskap, Uppsala Universitet.
An introduction for students and professionals.
San Fransisco, CA: Jossey Bass. Yin, R.K. (1999). Enhancing the quality of case
studies in health service research. Health Service
Tovey, D. (2006). The informed patient. Singapore
Research, 34(5), Part II, 1209-1224.
Medical Journal, 47(9), 741-745.
Vrangbæk, K., & Østergren, K. (2004). The
introduction of choice in Scandinavian hospital
systems. Arguments and policy processes in the
Danish and the Norwegian case (Working pa-
per 5). Bergen: Stein Rokkan Centre for Social
Studies.

This work was previously published in International Journal of Healthcare Information Systems and Informatics, Vol. 3, Is-
sue 2, edited by J. Tan, pp. 32-47, copyright 2008 by IGI Publishing (an imprint of IGI Global).

144
 145

Chapter 11
Characteristics of Good
Clinical Educators from Medical
Students’ Perspectives:
A Qualitative Inquiry Using a Web-Based
Survey System
Gary Sutkin
University of Pittsburgh School of Medicine, USA

Hansel Burley
Texas Tech University, USA

Ke Zhang
Wayne State University, USA

Neetu Arora
Texas Tech University, USA

aBstract

Medical educators have a unique role in teaching students how to save lives and give comfort during
illness. This article reports a qualitative inquiry into medical students’ perspectives on the key qualities
which differentiate excellent and poor clinical teachers, using a Web-based questionnaire with a purpose-
ful sample of third- and fourth-year medical students. Thirty-seven medical students responded with 465
characteristics and supportive anecdotes. All participants’ responses were analyzed through reviewing,
coding, member checking, recoding and content analysis, which yielded 12 codes. Responses from 5
randomly chosen participants were recoded by two authors with an inter-rater reliability coefficient of
0.72, implying agreement. Finally, 3 larger categories emerged from the data: Content Competence,
Teaching Mechanics, and Teaching Dynamics. We incorporate these codes into a diagrammatic model
of a good clinical teacher, discuss the relationships and interactions between the codes and categories,
and suggest further areas of research.

Copyright © 2011, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 Characteristics of Good Clinical Educators from Medical Students’ Perspectives

IntroductIon dynamic, automatic features such as e-mail invita-

tions for participation, participation tracking and
Medical teachers must be creative and effective records, e-mail reminders and other necessary
teachers in addition to being sound clinicians follow-ups about the survey research. Thirdly, a
and successful researchers (Bowen, 2006). Cur- Web-based questionnaire may easily incorporate
riculum reform increases the need for skilled interactive components into the survey such as
teachers in our medical schools. Although most a pop-up reminder, as necessary, to encourage
faculty do not undergo formal teaching training participants to provide more verbiage in their
during their medical training, many United States responses to open-ended questions. Additionally,
medical schools have created faculty development a Web-based survey makes it easy to collect data
workshops to help faculty, among other things, fast and in a predetermined analyzable format.
become better teachers (Searle, Hatem, Perkowski, Finally, the Web-based approach is usually less
& Wilkerson, 2006). But what makes a good medi- expensive than the other options (Schleyer, 2000).
cal school teacher? Some educators have attempted
to answer this question by surveying their own
medical students: some through the coding and reVIew oF tHe lIterature
categorization of answers from surveys (Boender-
maker, Conradi, Schuling, Meyboom-de-Jong, & Medical educators enjoy high status among pro-
Swierstra, & Metz, 2003; Cote & Leclere, 2000; fessional educators, with good reason. Complex
Ker, 2003; Pinsky, Monson, & Irby, 1998) and medical innovations, falling mortality rates, and
others from the analysis of answers to Likert-type the growth of the field of medical education have
scales (Morrison, Hitchcock, Harthill, Boker, & made the need for good medical teaching all the
Masunaga, 2005; Cox & Swanson, 2002; Elzubeir more critical. What makes an effective medical
& Rizk, 2002). educator, in particular, and an effective teacher, in
In our medical school, student assessments of general? Interestingly, with a metaphorical flour-
clinical faculty who teach third and fourth-year ish, Oser and Baeriswyl (2001) compared a teacher
medical students are heterogeneous, performed on to an expert in an emergency room, someone who
a departmental basis, inconsistently administered, reacts constantly to the immediate events, despite
and not standardized or validated. Multiple sources having a plan. They describe a lesson period as
are recommended to improve the content of edu- a chain of operations guided by rules, with the
cational assessment (Epstein, 2007); students are teacher employing innumerable helping, piloting,
one source of assessment of faculty teaching. To and controlling activities to meet lesson goals. In
identify quality teaching in clinical settings, it is the last 3 decades, in research on public school
critical to understand what contributes to good teaching, the emphasis has been on linking such
clinical teaching from students’ perspectives. teaching behaviors to student performance (Bo-
The advancement of Internet technologies also rich, 1996). The real answer to the above question
provides new opportunities for dynamic, instant is that teaching is a very complex activity that is
Web-based data collection efforts. In addition to influenced by myriad factors, both personal and
the fast speed and higher accuracy rate of data environmental.
collection, a Web-based questionnaire also makes In an empirical study, Good and Brophy (2000)
it possible to reach and engage prospective partici- identified effective elementary and secondary
pants despite the physical distances and diverse teacher behaviors that were associated with
geographical locations. More importantly, a well increased student performance. These factors
designed Web-based survey system may provide included teacher efficacy, student opportunity to

146
Characteristics of Good Clinical Educators from Medical Students’ Perspectives

learn, classroom management and organization, satisfaction. For example, Prichard and Sawyer
curriculum pacing, active teaching, teaching to (1994) identified the instructor’s personality as an
mastery, and a supportive learning environment. important factor in how postsecondary teachers
The Handbook of Research on Teaching (4th ed), teach, along with philosophy, teaching methods,
atomized the topic, deeply surveying the litera- outside influences and past role models. The
ture by subject field, including writing, reading, central theme is that personality and method
mathematics, science, health education, physical must be the right fit for instructor, subject matter,
education, visual arts, history, social studies and and students. In another study, Lempp and Seale
more (Richardson, 2001), and emphasized observ- (2004) interviewed medical students on the qual-
able teaching behaviors. The act of teaching itself ity of their teaching in British medical schools.
seems to be as complex as the content in any field While students reported positive role models, they
being taught. also found a hidden curriculum, one focused on
However, educational research in postsecond- the importance of hierarchy in the teaching of
ary institutions is less focused on teacher behavior medicine. Students reported many incidents of
and more focused on faculty development, even humiliation. They concluded that clinical practice
though faculty development can be a fuzzy con- and research have dominated clinical educator’s
cept. Menges and Ausin (2001) surveyed the lit- time. Teaching in a clinical setting is considered
erature on postsecondary teaching and suggested to be a lesser activity, unrecognized in incentive
that discipline and institutional contexts shape plans, and in fact, few physicians had received any
instructional practices. In particular, they con- training in teaching, learning theory, or learning
cluded that instructional practices diverge based assessment.
on differences between the paradigmatic hardness
of the field, with hard fields (e.g., chemistry and
physics) having rather fixed practices and softer MetHods
fields (e.g., sociology and English) being more
receptive to instructional innovation. In the post- Characteristics of effective clinical teachers have
secondary literature, instructional development not been studied extensively in the literature.
programs are often buried in faculty development Qualitative inquiry methods are well-suited for
efforts, with “faculty development” often used relatively underexplored areas of research and for
synonymously with effective teaching develop- theory-building. The primary goal in this study
ment. There also appears to be considerable criti- was to identify key qualities that differentiate
cism of faculty development efforts. Murry (2003) good and poor clinical teachers using a qualitative
suggests that these programs lack cohesiveness research design. Thirty seven medical students
and should be focused on teaching. He called for from a large Southwestern medical school vol-
formal programs, strong support from leadership untarily responded to an open ended, Web-based
and ties of teaching effectiveness to the reward questionnaire regarding characteristics of good
structure. Gottlieb, Rogers, and Rainey (2002) and poor clinical teachers.
suggested that self-assessment of instruction is Internet as a qualitative research tool is gaining
important to teaching effectiveness, assessing popularity (Mann & Stewart, 2000) and has many
skill, educational needs, progress, and strengths advantages such as low cost (Schaefer & Dillman,
and weaknesses of performance. 1998), convenience, and context of noncoercive
Finally, some research on postsecondary and antihierarchical dialogue (Boshier, 1990),
teaching identifies instructor personality as an which lends itself to collaborative research. The
important precursor to student performance and Internet has become an important mode of com-

147
 Characteristics of Good Clinical Educators from Medical Students’ Perspectives

munication and safe expression. Its reach within Thirty seven students participated in this
the academic community is even higher. When research (21 males; 16 females). The majority
research participants are difficult to reach and the of them were White-American (70%) and in the
data of interest is sensitive in nature, as was the age-group range of 25-29 years (59%). Nearly
case in this research, the Web offers the unique half of the participants (49%) were in their third
advantage of collecting and generating data in a year of medical school. Their preferred medical
centralized and noninvasive manner. It provides specialties varied widely.
an opportunity for researchers to capture larger Due to their textual nature, Web-based ques-
audiences, thereby enhancing cost effectiveness. tionnaire responses lend themselves to content-
Participants of this study were asked in an analysis, which was our choice of methodology.
electronic questionnaire to list three characteris- According to Weber (1990), content analysis is
tics of a good clinical teacher and a 3-5 sentence “a systematic research method for analyzing
description of an instance when a faculty member textual information in a standardized way that
demonstrated that quality in his or her teaching ac- allows evaluators to make inferences about that
tivities. Additionally, they were asked to describe a information.” In terms of the sampling units, we
time when a faculty member was a nonexample of received 465 total response posts, in 12-point
that characteristic. The questions were pilot tested Font, which totaled approximately 30 pages of
by three recent graduates to enhance the clarity, single-spaced text data. Three authors (GS, HB,
relevance, user-friendliness, and applicability of NA) independently reviewed and coded the tex-
the questionnaire. Their recommendations were tual data to find patterns among the responses.
incorporated into the final version (see appendix These codes represented overarching themes of
for an abbreviated Web form). characteristics of good clinical teachers as identi-
Upon receiving approval from the Institutional fied by the students.
Review Board, an invitation for participation was Use of research team members to interpret
sent to a pool of potential participants. Because we and double-check the coding schemes is a way
were interested in understanding the perceptions to enhance reliability in qualitative research. A
of medical students regarding clinical teaching, commonly used means of establishing reliabil-
the sample selected for this research was pur- ity is the use of multiple coders and the closely
poseful. In general, qualitative studies rely on related technique of peer examiners (LeCompte
the power of carefully selecting information-rich & Preissle, 1993), which reduces potential bias in
participants for whom the topic is meaningful the analysis and reporting phase by using multiple
and relevant (Sandelowski, 1995). The sample perspectives to validate results (Kvale, 1996).
of participants for this research consisted of (a) The authors of this study met weekly for 2
medical students in the last month of their third- months, exploring the patterns they observed in
or fourth-year, and (b) graduates in their first the data. One author (KZ) attended these meet-
year of residency. Their voluntary participation ings as an external auditor in order to maximize
was solicited through personal contact, e-mails, reliability in the multiple rater coding process. The
and flyers distributed at shelf examinations. The preliminary labels for the codes were agreed upon,
flyers and e-mails contained information about and they were routinely evaluated for agreement.
informed consent and listed a Web-link to the Discrepancies were resolved by discussion. This
questionnaire. In order to ensure anonymity of process of refining led to establishing the coding
responses, no identifying information was col- schemes. Codes were retained if at least two of
lected from participants. We invited dialog as the three authors agreed. Codes were expressed
participants’ schedules permitted. in the positive sense. For example, quotes about

148
Characteristics of Good Clinical Educators from Medical Students’ Perspectives

an attending not having time to spend with the 1. We defined knowledgeable according to
student and about an attending taking extra time the ACGME core competency definition:
with the student were both classified under the demonstrates knowledge of the biomedi-
code “Available.” cal sciences and its application to patient
During the final stages, the authors defined the care (ACGME, 2007). Examples of quotes
codes and classified them into larger categories. include:
The data analyses resulted in 12 codes which were
classified into three final categories. The resul- Dr. X is a man of extreme knowledge; he has
tant codes and categories were then used by two many years of experience on top of many
authors (GS, NA) to re-code responses from five years of education.
randomly chosen students. This recoding yielded
an adequate inter-rater reliability coefficient of I think knowledge is essential to a teacher
0.72 for qualitative research. because how can you be a great teacher
without having information to teach?

FIndIngs 2. We defined forming positive relationships

with patients according to the ACGME core
The findings presented below, describe the codes competency definition: provides patient care
and the categories that emerged from content that is compassionate, appropriate and effec-
analysis. The analysis yielded 12 codes, which tive (ACGME, 2007). Students frequently
we classified into three larger categories: (1) used descriptors like “respect,” “caring,”
Competence, (2) Teaching Mechanics, and (3) “approachable,” and “positive.” Examples
Teaching Dynamics. See Tables 1, 2, and 3 for of quotes include:
full descriptions, definitions, and positive and
negative examples. In our examples, we have Dr. X gives each patient’s concerns the time
replaced faculty names with “Dr. X” and gender and attention that he or she desires. As a
pronouns with alternating use of “he” and “she.” result, Dr. X’s approach to interacting with
everyone is something that students can feel
comfortable emulating.
coMpetence
More examples and full definitions of these
We defined Competence as the set of knowledge, codes can be found in Table 1.
skills, and abilities that physicians must be able
to perform and demonstrate. Many students
described effective clinical teachers as excellent teacHIng MecHanIcs
physicians worthy of modeling. These comments
could be separated into two main codes: knowl- We defined Teaching Mechanics as the instruc-
edgeable and relationships with patients. The tional toolset that helps students analyze and syn-
students were impressed not only by their teachers thesize biomedical information. The good teachers
who were intelligent, well-read, and technically used these techniques in their every-day clinical
skilled, but also by those adept at interacting with teaching. These tools ranged from good question-
their patients. ing techniques to unique learning activities. We
categorized these instructional practices into five
codes: guided practice, presentation techniques,

149
 Characteristics of Good Clinical Educators from Medical Students’ Perspectives

Table 1. Competence–the knowledge, skills, and abilities essential to being a physician

Characteristics of Good Medical School Teachers

(Codes, Definitions and Examples)

Codes Definitions Quotes: Positive Examples Quotes: Nonexamples

♦ “Dr. X can quote you all the ♦ “Some teachers cannot
landmark journal articles to sup- stand to be wrong or
Demonstrates knowl- port all sides of a given clinical admit they do not know
edge of the biomedi- situation and acknowledges that something”
Knowledgeable cal sciences and its there are multiple views/ways of ♦ “Occasionally you have
application to patient doing things” questions that do not have
care ♦ “a database of information…she an answer. Bad teachers
not only explains the details, but will try to sidetrack and
the process behind the details confuse you.”
♦ “I admire the way that Dr. ♦ “talking about patients …
X treats all patients … with with unkind remarks”
Provides patient care respect. He initiates these inter- ♦ “Many would try to make
Forms positive
that is compassion- actions by being approachable up for lack of … patient
relationships
ate, appropriate and and making people feel comfort- interaction (on rounds)
with patients
effective able.” with the ‘what would you
♦ “taught the patient about his guys like me to talk about
disease” today’ speech”

patient-related teaching, constructivist teaching, Dr. X’s method of diagramming concepts

and enrichment. is not unique but it does help make the vast
amount of information manageable. By or-
1. The technique of guided practice allows the ganizing thoughts in a concept diagram, a
teacher to take a student step by step through foundation can be formed on which details
the learning process. It can be accomplished can be added.
verbally with questions and prompts, or
visually, such as the supervised teaching 3. Many students appreciated patient-related
of a procedure. One student noted: teaching. Two examples are:

Dr. X would often put the instrument in your When he visits a patient on rounds, each
hand and walk you through a procedure. patient becomes an opportunity to learn
2. Superior presentation techniques were also from him. He knows his patients very well
often mentioned. They sometimes took the and can tell each one’s story in remarkable
form of an organized lecture or a novel detail with little to no paperwork to prompt
way of expounding on clinical topics. Two him. He explains patients’ conditions and
examples are: disease processes in ways that everyone in
the room can understand.
Dr. X had very clear learning objectives
and made sure that we all participated and She makes us present in front of the patient
completed learning goals. She knew what and discusses the case openly in front of the
your limitations were and had reasonable patient so as to include them in their treat-
expectations of students at our level. ment plan and disease process.

150
Characteristics of Good Clinical Educators from Medical Students’ Perspectives

4. Comments were also made about faculty nonthreatening, available, patient, enthusiastic,
skilled in constructivist teaching, or interac- and respectful.
tive teaching that helps students construct
the meaning for themselves from a given 1. Nonthreatening:
context. Constructivist teaching takes ad-
vantage of what a student already knows, Dr. X is non-intimidating and doesn’t make
and then fits new information to construct you feel like you’re constantly being graded.
a deeper level of understanding (Caine &
Caine, 1991). For example: Dr. X never talked down to us or belittled
us.
Dr. X would bring picture atlas to rounds.
He would bring a list of board questions 2. Available:
and even made an amazing handout that
seemed to fit perfect for boards. I felt free to ask Dr. X any question that I
wanted, regardless of how simple.
5. Enrichment is used by teachers to allow stu-
dents to come to a much deeper understand- 3. Patient:
ing of a topic. Students often commented
on feeling “challenged” and “intellectually Dr. X gives the students and residents time
stimulated.” Two examples are: to present the H&P without interrupting.
Listens to everything we say.
Dr X had the perfect combination of asking
questions to pull knowledge and make you Dr. X allowed me to close the fascia and
think, but balanced it with a quick didactic was very patient. When Dr. X noticed how
giving her thoughts. nervous I was, Dr. X used humor to make
me feel better and reassured me that every
Dr. X would have you read all the time over physician has been there.
your patients and expected that you would
know EVERYTHING about them. He always 4. Enthusiastic:
pushed you to know the next step.
Dr. X has an energetic teaching style... (and)
More examples and full definitions of these definitely keeps our attention.
codes can be found in Table 2.
Dr. X loved to teach and would tell you that
daily.
teacHIng dynaMIcs
5. Respectful:
We defined Teaching Dynamics as the personal-
ity characteristics that improve the conditions of Dr X loves to hear students’ ideas about
learning for the students. These characteristics management and treatment. We talk it out
generally represented the more “human” side amongst (us) so we are almost all equals.
of the doctor and included attitudes, emotions,
values, and reflective ability. Effective teachers More examples and full definitions of these
were often described as “caring” about their stu- codes can be found in Table 3.
dents. We divided the comments into five codes:

151
 Characteristics of Good Clinical Educators from Medical Students’ Perspectives

Table 2. Teaching Mechanics—instructional toolset that helps students analyze and synthesize biomedi-
cal information
Characteristics of Good Medical School Teachers
(Codes, Definitions and Examples)

Codes Definitions Quotes: Positive Examples Quotes: Nonexamples

♦ “Dr. X spends many hours ♦ “Faculty members that
making sure that students un- lose patience in an OR
derstand her lectures. She asks setting and grab the
Questions, prompts, to see if every student knows needle driver out of your
Guided prac- and models in order to how to perform a good physical hand”
tice clarify what learners examination on a newborn. ♦ “Dr. X. got mad at
are to do. Having [been] checked by Dr. me for not driving the
X, students feel confident in (laparoscopic) camera
their physical [examination] right but wouldn’t show
skill.” me how to do it”
♦ “Presents knowledge in an ♦ “Unorganized … it was
understandable easily learnable very confusing”
Delivers educational
format” ♦ “when they don’t know
content with superior
Presentation ♦ I’ve never heard REI/amen- how to convey that
organization, scope,
techniques orrhea explained in a more knowledge into some-
sequencing, and pace.
memorable/simple fashion than thing interesting to listen
he did” to”
♦ “how the material
♦ “Includes patient in conversa-
Uses the patient as a related to patient care
tions”
teaching tool, integrat- and how it related to
♦ “Taught clinical skills by
Patient-related ing knowledge and other topics remained a
demonstrating them and then
teaching skill while maintaining mystery”
having us briefly practice the
patient dignity. ♦ “Fast rounds in the hall,
skills on patients”
no teaching.”
♦ “Dr. X demonstrates an abil-
ity to be creative in both his
lectures as well as his clinical
teaching points. He gave a
lecture just the other day in ♦ “Things don’t seem to
which he had us recreate the stick as well when there
Teaches interactively
pelvic floor with play-do. This is very little interaction”
to help learners “con-
Constructivist will forever be in my mind ♦ “One-way conversation
struct” meaning for
teaching because it was interactive and is particularly poor in
themselves.
imaginative. Not just another small group settings
regurgitation of material. This such as rounds”
type of anatomy lecture was
made much more memorable
by imagination and the interac-
tive nature of the lecture.”
♦ “what you need to be
♦ “and I gain so much more
learning is the core
understanding of the topic be-
medical knowledge, not
Provides in-depth cause Dr. X explains the “why”
random ‘resident level’
content or challenging and “how” of pathophysiology”
facts”
Enrichment examples in order to ♦ “Later in the week Dr. X re-
♦ “Many staff would sim-
deepen the students’ reviewed the discussion with us
ply follow bullets from
understanding briefly. That helped solidify the
power points without
points he made“
making any clinical cor-
♦ “gave us extra assignments”
relations”

152
Characteristics of Good Clinical Educators from Medical Students’ Perspectives

Table 3. Teaching Dynamics–personality characteristics that improve learning conditions for students

Characteristics of Good Medical School Teachers

(Codes, Definitions and Examples)

Codes Definitions Quotes: Positive Examples Quotes: Non-examples

♦ “I never heard Dr. X raise ♦ “One attending is so freakin’
her voice … she instructed intense and intimidating that you
Non- Uses nonmenac-
without humiliating” stress out the entire day, night”
threaten- ing language and
♦ “Dr. X didn’t criticize me ♦ “You are afraid to ask questions
ing actions
when I didn’t know the because he will make you feel like
answer” it is a stupid question”
♦ “Many physicians are pressured
♦ “Says ‘not now, we’ll discuss
by the bottom line, seeing patients
it later’ and in fact, does
and billing. Teaching should take a
discuss later”
priority …”
♦ “We had a number of conver-
♦ “There were a couple of instances
Spends extra sations and she really listened
in which the faculty made me feel
Available time teaching the – this from someone who was
as though I was an impediment
students working late into the night to
to his/her work and that I was igno-
take care of her patients”
rant and unwanted.”
♦ “Took the time to … dem-
♦ “Some attendings would not initi-
onstrate how he did that
ate interactions with students. Any
surgery”
interaction is limited to residents”
♦ “Remains calm/patient in all ♦ “Not staying calm”
Appears calm, settings” ♦ “when we’re in their clinic, they
Patient composed, steadily ♦ “Ensured everyone under- emphasize that we need to go fast
persevering stood” and if you’re not done, they’ll walk
♦ “Calm and unwavering” in on you”
♦ “She took sincere joy in
showing us proper exam
technique” ♦ “reviewed topics in a very de-
Passionate about ♦ “Because he enjoys his work, tached fashion”
Enthusi-
being a doctor and he places his specialty in a ♦ “These faculty members were
astic
eager to teach positive spotlight & encour- merely teaching as a fulfillment of
ages students to follow in his a requirement of their contract”
footsteps”
♦ “She loves the field”
♦ I admire the way that Dr. X
♦ “Another attending … was even
treats all patients, students,
more rude, inflammatory and
Approaches others and staff with respect.
derogatory to me”
with a sense of ♦ Dr. X at all times main-
Respectful ♦ “Other teachers were obviously not
deference for their tains respect for everyone’s
interested in … showing students
worth and esteem personal predilections and
any respect”
teaches with an abject neutral-
ity

“pIMpIng” Two examples of positive “pimping” included:

Although we did not put “pimping” as a separate I think being pimped is an ok way to teach, as long
code or category, we noted the frequency of pas- as it is not the kind of pimping where you are SO
sionate comments made it. Pimping was noted as freaked out about their next question and getting
both malignant and positive. it wrong that... you can no longer concentrate. Dr.
X was not like that at all. Dr. X’s questions were

153
 Characteristics of Good Clinical Educators from Medical Students’ Perspectives

relevant to each patient and... it created such a student into the teaching moment. These qualities
better learning environment. are not exhaustive, but are certainly illustrative.
Even though the data easily separated into
One day in the OR, Dr. X asked basic anatomy these 12 codes and three categories, interaction
questions concerning branches of the abdominal and overlap among them was evident, and thus
aorta. I did not recall at that time from anatomy 2 we created a pictorial representation of their
1/2 years prior all the answers Dr. X was looking interaction (Figure 1). We call this troika of do-
for. That experience pushed me to learn more. mains the Student Inspired Model for Effective
Clinical Teaching (SIMECT). According to our
Pimping could also be threatening and non- model, each category is affected by its associ-
educational: ated teaching characteristics (i.e., codes), which
can have either a positive or negative effect on
Dr. X made us feel like we were being separated its effectiveness. The personality attributes were
and stoned. especially easy to view in multiple intersecting
continuums (for example, ranging from threaten-
Some attendings like to ask very specific and ing to nonthreatening, disrespectful to respect-
pointed questions with the intent of being able to ful, and the like). We incorporated the negative
ask you questions that you are unable to answer. comments into our model and drew the arrows
This method of learning has the advantage of to show how these continuums affected clinical
being easily remembered (because it is always teaching. We believe this model to be predictive
easier to remember uncomfortable situations in nature, so that the right mix of educator com-
that made you feel unintelligent); however, they petence, teaching dynamics, and use of teaching
do not incorporate this knowledge into a usable mechanics can increase or decrease and speed
framework. Instead you just fill your mind with or slow student learning of critical knowledge,
factoids that stand alone. skills, and dispositions.
We assigned each example of “pimping” to The codes were stable when analyzed by teach-
relevant codes. For example, “if you didn’t know ers and were not altered by changes in context. A
the answer, Dr. X would make you feel stupid” was clinical educator who would take time to teach
coded as a negative example of nonthreatening, during a stressful surgery, for example, was also
and “Dr. X expects you to know everything about described as a nonthreatening questioner, enthu-
your patients” was coded as a positive example siastic about teaching, and patient with students
of enrichment. and those under his or her care. The domains are
connected in such a way that improving practice
in one area may help behavior in another area.
dIscussIon This model can be a very useful tool in develop-
ing teacher education programs. For instance,
Through coding and categorizing the descriptive a program designed to improve the enrichment
responses obtained using a Web-based question- (teaching mechanics) of a clinical lesson would
naire, we were able to construct a paradigm of also improve the students’ perception of the
the excellent clinical teacher from the students’ teacher’s enthusiasm (teaching dynamics) and
perspective. The excellent clinical educator is both knowledge of the content (competence).
knowledgeable and good with patients, utilizes a Although all three domains influenced effec-
toolset of effective teaching techniques, and pos- tive clinical teaching, we learned through our
sesses personality characteristics that draw the coding process that Teaching Dynamics was the

154
Characteristics of Good Clinical Educators from Medical Students’ Perspectives

Figure 1. Student inspired model for excellent clinical teching (SIMECT)

student Inspired Model for excellent clinical teaching
(sIMect)

Knowledgeable

Competence

Interaction with
patients

Non-
Available
Enthusiastic

Threatening

Respectful

Impatient Outcome:
Teaching Excellent
Dynamics Clinical
Patient
Teaching
Disrespectful

Non-
Threatening

Busy Enthusiastic

Guided
Practice

Enrichment

Presentation Teaching
Techniques Mechanics

Constructivist
Teaching
Patient
Related
Teaching

most frequently mentioned. In fact, it seemed to be nonthreateningly appeared to have presentation

the centerpiece feature of students’ observations techniques more effective than one who berated
of the effective clinical educator, in that it served while presenting. This finding was completely
as a catalyst for the overall teaching-learning unexpected, and it dominated respondent dis-
process. Teaching dynamics often made evident cussions and stories. Without dissent, students
the instructor’s competence, while influencing concluded that the positive teaching personalities
(positively or negatively) any particular strategy mediated both the display of competence and the
used to deliver content to the learner. For ex- use of teaching techniques.
ample, a knowledgeable instructor who taught

155
 Characteristics of Good Clinical Educators from Medical Students’ Perspectives

The Teaching Dynamics finding is reflective of like using pimping to humiliate, in the hands of
personality research found in personality devel- a cold and distant clinical educator could shut
opment psychology, particularly the personality down learning for students.
orientation known as the Big Five Model. The Big Although our qualitative data were taken from
Five personality factors are agreeableness, con- one medical school, we suspect that these ideal
scientiousness, extraversion, emotional stability, clinical teaching qualities have some universality
and openness (Goldberg, 1993). These dimensions in the relationship among competence, teaching
of personality were derived from factor analyses dynamics, and teaching methods, with teaching
of language people used to describe themselves, dynamics being the feature of teacher that most
and they are useful because they can represent affects adaptations in the face of student, con-
diverse systems of personality description (Oliver textual, and environmental differences. Elzubeir
& Srivastava, 1999). According to Oliver and and Rizk (2001) asked medical students what they
Srivastava (1999), these traits are stable and are look for in a faculty role model and then coded
predictive of behaviors, including workplace be- and categorized the results into three categories,
haviors. Our findings suggest that “expressions” similar to ours. Their students especially appreci-
of these personality traits (e.g., nonthreatening, ated faculty who demonstrated respect, honesty,
available, and patient) are also predictive of teach- politeness, and enthusiasm.
ing effectiveness that can be targeted for personal Paukert and Richards (2000) reviewed written
development and change. For example, a clinical descriptions of faculty who had “significantly and
educator can consciously use language that is less positively influenced their clinical education.”
threatening or can learn more effective ways of They coded and categorized the responses into
providing feedback on student performance. five categories: person, physician, teacher, super-
The use of pimping as an instructional tech- visor, and unspecified (global). While the first
nique is one example of how an expression of three were similar to our categories, “supervisor”
personality type can affect teacher-student re- included attendings that provided their students
lationships. We were not surprised to receive so with opportunities to participate in patient care
many comments surrounding the art of pimping. and practice skills.
“Pimping” is a loosely defined term that connotes Irby, Ramsey, Gillmore, and Schaad (1991)
a person in power publicly probing a junior col- used observations of faculty teaching to create a
league’s knowledge base. It usually manifests as 44-item, 7-point scale questionnaire, which they
the questioning of a medical student in front of administered to students to identify the charac-
their peers. Wear, Kokinova, Keck-McNulty and teristics of effective ambulatory clinical teachers.
Aultman (2005) recently surveyed fourth-year They found that good teachers were described as
medical students and discovered that although being actively involved with learners, promoting
pimping can be intended for humiliation or can learner autonomy, and demonstrating patient
be inappropriate for the student’s level, it is of- care skills.
ten seen by students as a positive pedagogical The dynamic relationships between the three
tool. Our students also noticed this dual nature major components of good clinical teaching also
of pimping, and many appreciated the stimulus have practical implications on the design and
for additional learning. Therefore, a highly mo- implementation of quality e-learning in health
tivated and extraverted clinical educator could education. Teaching dynamics, which were so
pimp enthusiastically and respectfully, increas- important to our students, can be incorporated
ing students’ interest in learning a content or a and translated into e-learning or blended learning
skill. Conversely, a negative teaching dynamic, teaching activities.

156
Characteristics of Good Clinical Educators from Medical Students’ Perspectives

The Web-based questionnaire, designed and entered by the respondents, rarely needing to be
developed particularly for this study, worked printed and never needing to be retyped. We es-
very well as a data collection instrument with a pecially wanted to avoid the delay and disconnect
few customized features. We were able to track associated with large amounts of raw data in the
and record participant’s login attempts, identify hands of a transcriptionist. Additionally, we hoped
incomplete submissions and provide reminders that the speed and reliability of the Web-survey
for participants to complete the survey in one itself would be replicated with the data analysis
or multiple attempts. The Web-based system performed by the qualitative software package.
also provided friendly prompts as necessary to After reviewing several qualitative software
encourage elaborations on open-ended ques- packages, we chose Qualrus, a relatively new
tions. These technology-enhanced features have product that promised artificial intelligence (AI)
certainly helped increase response rate and algorithms that would help speed data analysis
obtain complete and richer data, in a more time- (Brent, Slusarz, & Thompson, 2002). This soft-
efficient, user-friendly, and cost-efficient fashion. ware application has graphical coding tools that
We highly recommend future developments of can guide and simplify the coding process. Our
similar Web-based assessment tools in health choice of this particular product was based on
education research. the earlier experience of one of the researchers
As with most studies of this type, our study with the product. We also liked the idea of the
limitations include a small sample that may not AI features. Having the software learn from user
be representative of the population under study. codes and make recommendations of possible
We anticipate that our backgrounds and biases codes could be very helpful.
might have influenced our coding and subsequent However, ours was a simple project, made even
grouping into characteristics. Another group of simpler by the Web survey; so we did not fully
coders might classify the data in a different way. use the software. Our project had one timepoint
We attempted to mitigate our biases through for respondents to answer the survey. Addition-
collaboration between medical (GS) and general ally, our data was already highly structured, and
(HB, KZ, NA) educators. we were interested in speed. The data was easily
The Web-based survey greatly expedited the organized by simply cutting and pasting with a
pilot testing of our survey. The qualities submitted word processor.
by our pilot-testers became our first coding cat- The software’s process mirrors classical
egories, helping us to develop an initial framework qualitative analysis procedures, including coding,
for categorization. Adjustments to the Web survey grouping the codes, establishing relationships
after pilot-testing were finished immediately. From among the groups of codes, and managing reli-
then on, respondents could access the survey from ability issues. We decided to meet face-to-face and
the Internet at their convenience, and the data was used this exact logic to resolve coding, recoding,
easily retrievable by the researchers. grouping, and reliability issues. Though the meet-
Because of the physical and academic distances ings were time-consuming, no software package
from each other and because this project was could replace the dynamism and efficient problem-
funded by a time-sensitive grant, the researchers solving of those meetings. Like Seidel (1991), we
initially planned to analyze the collected data desired to be closer to our data and did not want
using a qualitative software program, which we our software “bridge” to become an impediment.
believed would help to bridge those differences Qualrus software is better suited for complex
among us. The idea was to keep the process as longitudinal projects that amass large amounts
seamless and paperless as possible, with data of data from multiple sources using various me-

157
 Characteristics of Good Clinical Educators from Medical Students’ Perspectives

dia. If we had massive amounts of data with no Boendermaker, P.M., Conradi, M.H., Schuling, J.,
ostensible organization, it is easy to see how our Meyboom-de-Jong, B., Swierstra, R.P., & Metz,
implicit inductive and iterative processes could J.C.M. (2003). Advances in Health Science Edu-
have been made more explicit, and hence more cation Theory Practice, 8, 111-116.
manageable, using qualitative software.
Boendermaker, P.M., Schuling, J., Meyboom-de-
Future areas of research might include content
Jong, B., Zwierstra, R.P., & Metz, J.C.M. (2000).
analysis of in-person interviews with medical
What are the characteristics of the competent
students and faculty as well as direct observation
general practitioner trainer? Family Practice 17,
of faculty teaching medical students. Although
547-553.
these methodologies have been previously used
(Boendermaker, Schuling, Meyboom-de-Jong, Borich, G. (1996). Effective teaching methods (3rd
& Swierstra, & Metz, 2000; Irby, 1994; Irby ed.). Englewood Cliffs, NJ: Merrill/Prentice Hall.
et al., 1991), we would like to conduct them in
Boshier, R. (1990). Socio-psychological factors
our medical school and compare the findings to
in electronic networking. International Journal
the SIMECT model. Our findings can have an
of Lifelong Education, 9(1), 49-64.
important impact on medical school assessment
of clinical teaching. Our medical school faculty Bowen, J.L. (2006). Medical education: Educa-
are not regularly evaluated on the 12 teaching tional strategies to promote clinical diagnostic
characteristics identified as most important by reasoning. New England Journal of Medicine,
our medical students. Future research might 355, 2217-2225.
aim at testing some of these characteristics in
Brent, E., Slusarz, P., & Thompson, A. (2002).
order to create new assessment toolsets. Finally,
Qualrus: The intelligent qualitative analysis
our study does not address the effects of these
program (manual). Columbia, MO: IdeaWorks.
shared characteristics on student learning. Further
research on instructor intentions and attitudes Caine, R.N., & Caine, G. (1991). Making connec-
may help illuminate the effect of these teacher tions: Teaching and the human brain. Alexandria,
characteristics on student learning. Eventually, VA: Association for Supervision and Curriculum
interventions based on these domains could be Development.
developed to improve clinical instruction.
Côté, L., & Leclère, H. (2000). How clinical teach-
In conclusion, we present a paradigm of ef-
ers perceive the doctor-patient relationship and
fective clinical teaching as viewed by our medi-
themselves as role models. Academic Medicine,
cal students: a good clinical teacher possesses a
75, 1117-1124.
comfortable competence, specific instructional
strategies, and a dynamic set of positive person- Cox, S.S., & Swanson, M.S. (2002). Identifica-
ality traits. tion of teaching excellence in operating room
and clinic settings. American Journal of Surgery,
183, 251-255.
reFerences
Elzubeir, M.A., & Rizk, D.E.E. (2001). Identifying
ACGME Outcome Project. (2007). Accredita- characteristics that students, interns and residents
tion Council for Graduate Medical Education. look for in their role models. Medical Education,
Retrieved January 29, 2008, from http://www. 35, 272-277.
acgme.org/outcome/comp/compFull.asp. Epstein, R.M. (2007). Assessment in medical
education. New England Journal of Medicine,
356, 387-396.

158
Characteristics of Good Clinical Educators from Medical Students’ Perspectives

Goldberg, L. (1993). The structure of phenotypic Morrison, E.H., Hitchcock, M.A., Harthill, M.,
personality traits. American Psychologists, 48, Boker, J.R., & Masunaga, H. (2005). The online
26-34. clinical teaching perception inventory: A “snap-
shot” of medical teachers. Family Medicine, 37,
Good, T., & Brophy, J. (2000). Looking in class-
48-53.
rooms (8th ed.). New York: Longman.
Murray, J. P. (1999). Faculty development in a na-
Gottlieb, R., Rogers, J., & Rainey, K. (2002). Us-
tional sample of community colleges. Community
ing curriculum evaluation as a basis for clinical
College Review, 27(3), 47. Retrieved January 29,
faculty development. College Student Journal,
2008, from Questia database: http://www.questia.
36(2), 279+. Retrieved January 29, 2008, from
com/PM.qst?a=o&d=5001886124
Questia database, http://www.questia.com/
PM.qst?a=o&d=5000799148 Oliver, J., & Srivastava, S. (1999). The Big Five
taxonomy: History, measurement, and theoreti-
Irby, D.M. (1994). What clinical teachers in
cal perspectives. In L. Pervin & O.P. John (Eds.),
medicine need to know. Academic Medicine, 69,
Handbook of personality: Theory and research
333-342.
(2nd ed.). New York: The Guilford Press.
Irby, D.M., Ramsey, P.G., Gillmore, G.M., &
Oser, F., & Baeriswyl, F. (2001). Choreographies
Schaad, D. (1991). Characteristics of effective
of teaching: Bridging instruction to learning. In
clinical teachers of ambulatory care medicine.
V. Richardson (Ed.), Handbook of research on
Academic Medicine, 66, 54-55.
teaching (4th ed). Washington, DC: American
Ker, J.S. (2003). Attributes of trainers for post- Educational Research Association.
graduate training in general surgery–a national
Paukert, J.L., & Richards, B.F. (2000). How medi-
consensus. Surgeon, 1, 215-20.
cal students and residents describe the roles and
Kvale, S. (1996). Interviews: An introduction characteristics of their influential clinical teachers.
to qualitative research interviewing. Thousand Academic Medicine, 75, 843-845.
Oaks, CA: Sage.
Pinsky, L.E., Monson, D., & Irby, D.M. (1998).
LeCompte, M.D., & Preissle, J. (1993). Ethnogra- How excellent teachers are made: Reflecting on
phy and qualitative design in educational research success to improve teaching. Advances in Health
(2nd ed.). San Diego, CA: Academic Press. Science Education, 3, 207-215.
Lempp, H., & Seale, C. (2004). The hidden cur- Prichard, K. W., & Sawyer, R.M. (Eds.). (1994).
riculum in undergraduate medical education: Handbook of college teaching: Theory and ap-
Qualitative study of medical students’ perceptions plications. Westport, CT: Greenwood Press. Re-
of teaching. British Medical Journal, 320, 770-773. trieved January 29, 2008, from Questia database:
http://www.questia.com/PM.qst?a=o&d=9413558
Mann, C., & Steward, F. (2000). Internet commu-
nication and qualitative research: A handbook for Richardson, V. (2001). (Ed.). Handbook of re-
researching online. Thousand Oaks, CA: Sage. search on teaching (4th ed). Washington, DC:
American Educational Research Association.
Menges, R., & Austin, A. (2001). Teaching in
higher education. In V. Richardson(Ed.), Hand- Sandelowski, M. (1995). Focus on qualitative
book of research on teaching (4th ed). Washington, methods: Sample size in qualitative research.
DC: American Educational Research Association. Research in Nursing and Health, 18, 179-183.

159
 Characteristics of Good Clinical Educators from Medical Students’ Perspectives

Schafer, D.R., & Dillman, D.A. (1998). Develop- Seidel, J. (1991). Method and madness in the ap-
ment of a standard e-mail methodology: Results plication of computer technology to qualitative
of an experiment. Public Opinion Quarterly, 62, data analysis. In N. Fielding & R.M. Lee (Eds.).
378-397. Using computers in qualitative research (pp.
107-116). London: Sage.
Searle, N.S., Hatem, C.J., Perkowski, L., & Wilk-
erson, L. (2006). Why invest in an educational Wear, D., Kokinova, M., Keck-McNulty, C., &
fellowship program? Academic Medicine, 81, Aultman, J. (2005). Pimping: Perspectives of 4th
936-40. year medical students. Teaching and Learning in
Medicine, 17, 184-191.
Weber, R.P. (1990). Basic content analysis (2nd
ed.). Newbury Park, CA.

160
Characteristics of Good Clinical Educators from Medical Students’ Perspectives

appendIx: clInIcal teacHIng weB-Based QuestIonnaIre

(aBBreVIated)

clinical teaching

General Instructions

Please read the following instructions before completing the questionnaire. The purpose of this survey
is to identify characteristics you believe are associated with effective clinical teaching. Participation is
voluntary. Your responses will be kept anonymous. Filling out this survey implies consent to participate
in this project. Use examples from the third and fourth years of medical school.

Clinical Teaching

Instructions: The following open-ended questions include several subquestions. Please answer each of
the subquestions to the best of your knowledge.
In your opinion, what makes a good clinical teacher? Please list the 3 characteristics or qualities of
good clinical teaching in the appropriate text boxes below. Please list only ONE characteristic in each
of 3 cells. Below each characteristic, please also: A) nominate a faculty member from the third or fourth
year of medical school who best exemplifies a positive example of that characteristic, B) write a short
story that illustrates how that faculty member demonstrated that quality in his/her teaching activities,
and C) write a short story that describes a situation in which any faculty member was a nonexample of
that characteristic. Please do NOT list the name of the faculty member in part C. For these short stories,
please write a minimum of 3-5 sentences, but feel free to write as much as the box will allow. Please
report examples from the third and fourth year of medical school.

This work was previously published in International Journal of Healthcare Information Systems and Informatics, Vol. 3, Is-
sue 2, edited by J. Tan, pp. 69-86, copyright 2008 by IGI Publishing (an imprint of IGI Global).

161
162

Chapter 12
Open Source Software:
A Key Component of E-Health
in Developing Nations
David Parry
Auckland University of Technology, New Zealand

Emma Parry
National Women’s Health, Auckland District Health Board, New Zealand

Phurb Dorji
Jigme Dorji Wanchuck National Referral Hospital, Bhutan

Peter Stone
University of Auckland, New Zealand

aBstract
The global burden of disease falls most heavily on people in developing countries. Few resources for
healthcare, geographical and infrastructure issues, lack of trained staff, language and cultural diver-
sity and political instability all affect the ability of health providers to support effective and efficient
healthcare. Health information systems are a key aspect of improving healthcare, but existing systems
are often expensive and unsuitable. Open source software appears to be a promising avenue for quickly
and cheaply introducing health information systems that are appropriate for developing nations. This
paper describes some aspects of open-source e-health software that are particularly relevant to devel-
oping nations, issues and problems that may arise and suggests some future areas for research and
action. Suggestions for critical success factors are included. Much of the discussion will be related to a
case study of a training and E-health project, currently running in the Himalayan kingdom of Bhutan.

organIZatIon oF tHIs paper source software that make it attractive for software
development in the health domain for low income
This paper is organised around a number of sec- countries. The methodology section then intro-
tions. The introduction outlines the rationale of duces the framework of assessment that is being
the paper and deals with some aspects of Open used. The majority of this paper describes a case
study of a project run by the authors in Bhutan in
DOI: 10.4018/978-1-61692-002-9.ch012

Copyright © 2011, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Open Source Software

Figure 1. Research domains

the obstetric domain. Critical success factors for
such a project are then analysed and some conclu-
sions are drawn. The discussion covers some of
the issues that have arisen from this experience,
and articulates some lessons learned. This.

IntroductIon

This project deals with the intersection of a number

of domains, as shown in Figure 1

e-Health
Health Information systems
E-health has become a popular term for the
transformation of healthcare that has occurred Health information systems (HIS) often have
through the use of electronic communications, three main objectives, to improve patient care,
in a conscious imitation of “ebusiness”. E-health improve management and form part of a qual-
encompasses more than the traditional electronic ity improvement programme. However, these
health record. It involves the use of information objectives – as described by (Littlejohns, Wyatt,
and communications technologies in the widest & Garvican, 2003) are not always achieved. As
sense, including telemedicine, web-based health part of a HIS implementation there are often
and mobile devices for healthcare. A definition major changes to workflow and practice, large
has been proposed, after comprehensive analysis, expenditures on hardware including computing
in (Pagliari et al., 2005) and communications, and system integration,
as well as software development, training and
“e-health is an emerging field of medical infor- implementation. (Littlejohns et al., 2003) Points
matics, referring to the organization and delivery out that failures occur in HIS development – often
of health services and information using the due to a lack of understanding of the complexity
Internet and related technologies. In a broader of the project. Interestingly OSS appears to answer
sense, the term characterizes not only a techni- some of these issues by providing more stable – if
cal development, but also a new way of working, less feature-rich – software and providing a gener-
an attitude, and a commitment for networked, ally larger pool of developers and users than for
global thinking, to improve health care locally, proprietary software.
regionally, and worldwide by using information
and communication technology” open source software

This definition is actually adapted from a pre- Open source software (OSS) has gained very
vious one in an editorial(Eysenbach, 2001). The wide acceptance particularly in the web server
globalised and networked aspects are particularly community. Projects such as Apache (Mockus,
important in our case study – the emphasis is on Fielding, & Herbsleb, 2000) have involved
communication and collaboration rather than large scale participation, and dominant market
distance share.. In the healthcare domain, Sourceforge.

163
 Open Source Software

net lists 58 applications for download. Many of ease on individual households can be very
the applications are extremely specialized, but large. For example up to 100% of house-
the other hand, some like WIRM (Jakobovits, hold income being spent on end-stage
Rosse, & Brinkley, 2002) are effectively complete care for AIDS patients in some nations in
development environments. This paper will argue Africa(Russell, 2004).
that successful development and use of OSS in • Developing nations often have a diverse
healthcare requires a number of critical success mixture of groups within them, and it
factors, and that these reflect both the nature of OSS cannot be assumed that all citizens have
projects in the wider world and particular aspects a common language. Even when a com-
relevant to healthcare. OSS can be seen as part mon language exists, it may be spoken by
of a wider movement that has been characterized a relatively small percentage of the world’s
as innovation from the user community (Hippel, population, and commercial development
2001). This emphasizes the point that that OSS is of software using that language may not be
not just “free” but also is able to be modified by feasible.
the community that uses it. There is a very large • Infrastructure and resource constraints in
and continuing project (OpenEHR) Kalra, 2005 particular for network connectivity may
#3034, that is attempting a to produce reliable, reduce the utility of high-performance sys-
semantically correct representations of health tems routinely used in the west. For exam-
information. This project is not really focussed on ple PACS systems involving transmission
developing specific implementations, but rather of large images via network connections
a common means of representation. may not be practical, but memory-stick
based approaches may be feasible (Parry,
developing nations and Sood, & Parry, 2006).
the case of Bhutan • Open source approaches allow the devel-
opment of expertise in multiple sites away
Health information systems are important for de- from large commercial organizations.
veloping nations as well as industrialized ones. A Therefore they can encourage the upskill-
large review of the use of information technology ing of software developers in smaller cen-
in primary care in developing countries (Tomasi, tres. This expertise can be applied to the
2004) identified five main areas of application – localization of standard packages and the
data processing in the health care system, decision development of a solid base for software
support, electronic data transmission, electronic support. In this aspect both the develop-
patient records and telemedicine. Many develop- ment and the use of open source software
ing countries have low levels of trained clinical can be beneficial in the education sector.
staff, and this can increase the load on second- Developing nations often have large and
ary and tertiary providers. In order to audit their increasing numbers of young, educated
performance, and increase efficiency, electronic people available for project development.
records and workflow systems can reduce the OSS tools are attractive for teaching infor-
workload on the staff available. mation systems development because of
Both of these aspects are particularly important cost, wide availability of documentation
for developing nations for a number of reasons. and localized versions being available. For
example, linux is available in x language
• Developing nations have extremely lim- versions including dzonga – the national
ited health budgets, but the burden of dis- language of Bhutan.

164
Open Source Software

• Commercial software suppliers may be some recent papers on the use of OSS in health
reluctant to sell advanced software pack- information systems, focused mainly on developed
ages in developing nations because of the country applications,(Kantor, Wilson, & Midgley,
difficulty of arranging support and the per- 2003; McDonald et al., 2003). Interestingly these
ceived threat of piracy papers point out that the use of OSS in healthcare
• Developing nation’s health systems often is not new and that although perhaps small-scale
have a complex collection of groups work- this work has been consistently ongoing. However,
ing within them including governments, these papers do emphasize the potential gains to
commercial organizations, local and inter- be had by the use of OSS in healthcare both in
national charities and international official terms of health providers and also developers.
organizations. The requirement for report- Because of the widely varying state of commu-
ing and data analysis may well be more nications infrastructure in developing countries,
complex than in industrialised nations. western models of development which emphasize
• Infrastructure developed to support call- in-hospital systems linked by fixed line high ca-
centre development or tourism, including pacity networks may not be appropriate. In the
internet and telecommunications technolo- context of less developed countries, there have
gy, is easily adapted to allow links between been a number of telemedicine projects, often
nations. Because OSS tools are supported concerned with communication from centres of
via the web, this approach avoid the reli- excellence in western nations eg (Swinfen, Swin-
ance on expensive and out-of-date paper fen, Youngberry, & Wootton, 2005), or within
manuals and development kits. OSS’s li- developing countries (Deodhar, 2002), but a shared
censing structure allows cross-national approach to development is vital (Wooton, 2001).
projects to be completed much more easily.
• Mobile devices have particular promise
for e-health in developing nations(Iluyemi, MetHodology
Fitch, & Parry, 2007). Mobile OSS de-
velopment is a particularly active area In order to analyses a case study, some sort of
of research(Raento, Oulasvirta, Petit, & framework of analysis should be adopted. The
Toivonen, 2005). development project was actually quite complex
• An exhaustive survey of African nations with elements of telemedicine, knowledge man-
e-health status(Kirigia, Seddoh, Gatwiri, agement and information processing included in
Muthuri, & Seddoh, 2005) has shown that the overall design (see Figure 2).
many nations currently have very low pre- A survey of Telemedicine projects in India
paredness, given the relative paucity of (Pal, Mbarika, Cobb-Payton, Datta, & McCoy,
internet and telephone connections, how- 2005) identified six critical success factors for
ever recent developments in increasing Telemedicine success, these were used as practi-
bandwidth capacity see for example http:// cal and simple measures that could be applied in
www.fibreforafrica.net/ may be improving this complex if small-scale project. The success
this state of affairs. factors identified were::

The case study in this paper deals with the inter- 1. Set clear program objectives
section of a number of research domains, (Figure 2. Garner Government Support
1), which means that the choice of methodology 3. Adapt User-Friendly Interfaces
for analysis may be challenging. There have been

165
 Open Source Software

Figure 2. Overall system plan

4. Determine Accessibility Via walk from the nearest road head. Bhutan has had
Telecommunications and Internet Access major successes in increasing life expectancy and
5. Implement Standards and protocols improving health care but avoidable neonatal and
6. Measure Cost-effectiveness and User maternal mortality and morbidity remains an is-
Satisfaction sue. Current figures for Bhutan suggest an infant
mortality rate of 67/1000 and a maternal mortality
The case study will deal with these areas, rate of 4.2/1000 – compare with New Zealand’s
although the project is wider than a simple tele- rate of 4/1000 and 0.07/1000 respectively (World
medicine project as it includes database develop- Health Organization, 2006a).
ment and integration with the audit system, along Large numbers of preventable neonatal deaths
with web-based protocols. continue to occur in the less developed counties.
However, recent work has suggested (Darmstadt
et al., 2005) that evidence based interventions
case report e-HealtH in antenatal and intrapartum care could reduce
support For oBstetrIc these rates by between 37% and 67%. These
serVIces In BHutan interventions are not complex and are relatively
inexpensive. The overarching imperative is to
Bhutan is a small Buddhist Kingdom located in the ensure appropriate care for pregnant women that
Himalayas, with a population of under 700,000. involves patient education and cooperation with
Land transport is extremely slow because of the antenatal and intrapartum services. Although there
geography - for example it takes 3 days to travel have been many studies on the use of e-health in
across the country, a distance of around 300km. obstetrics and perinatology and in less developed
There is only one airport and no facility for heli- nations (Deodhar, 2002)there remains relatively
copter transport. Seventy percent of the population little work on the evaluation of these systems,
live in rural areas with 30% more than 1 hours especially in terms of outcome and integration

166
Open Source Software

of these systems into existing structures, and the to hospital-based consultants even in cases where
changes that occur because of their introduction, primary care would be more appropriate. This and
although the 1:45 cost benefit ratio quoted in this the paucity of qualified obstetricians, results in
study is impressive. a large workload and the Obstetricians are busy
The World Health Organisation has been and often difficult to contact for advice. A current
running a “Making Pregnancy Safer” Initiative project is running to introduce the emergency
(World Health Organization) in order to reduce the obstetric care (EMOC) system to Bhutan, and
level of neonatal and maternal mortality. Previous the protocols are being integrated with these to
work in Bhutan had developed a protocol book for provide seamless care.
emergency obstetric care (EMOC). Other coun-
tries using EMOC have recorded improvement in project description
outcomes, for example Bangladesh (Islam MT,
2006) and Peru (Kayango et al., 2006).One of The aim of the project was to collaboratively
the major lessons learned in these trials was that develop a number of treatment and/or diagnosis
a record of outcomes via a perinatal database, and protocols to allow the clinical staff to apply appro-
the wide dissemination of protocols, for example priate evidence-based care for the major problems
those that identify potentially high-risk patients, that would be dealt with by a perinatal service. The
are vital for success. Surprisingly perhaps, the role of the Perinatal service is described in (Mas-
identification of appropriate procedures for deal- carenhas, Eliot, & MacKenzie, 1992), essentially
ing with high-risk patients has been shown to be it provides care for mothers and baby between
effective in reducing the demand for interventions conception and birth, and aims to reduce the risks
(Islam MT, 2006). to mother and baby in this process, by appropriate
intervention and monitoring In addition to the staff
the Bhutanese Health system applying the protocols in practice, the aim is to
raise awareness of the issues that affect perinatal
Healthcare is free in Bhutan, and is delivered via outcomes amongst others, for example referring
a tertiary structure. The primary healthcare unit clinicians. The development of a collegial edit-
is the “Basic Health Unit” (BHU), of which there ing and review process involving Clinical staff in
are around 170 around the country. These units Bhutan and New Zealand was also seen as a vital
do not always have medically qualified staff, but part of the project. The project also included the
they run outreach and clinic services and usually development of a perinatal web-based database
a number of beds are available. Delivery services to allow for more effective management of the
are sometimes available run by nurses. District service on a day-to-day basis and also allow for
health units (around 30) will have at least one analysis of clinical performance.
generalist medical officer, some of these units OSS occurs in a number of places in the system.
have the capacity to perform caesarean sections The perinatal database is written in PHP with a
and ultrasound scanning. There are 3 Referral mySQL database engine. Web page development
hospitals in the country which have at least one was done using open source tools, as was the
obstetrician and theatre services. The Jigme Dorji xml protocol development. However Microsoft
Wanchuck National Referral Hospital (JDWNRH) products were used for the operating systems and
in the captial has four obstetricians and is the ter- web server software in Bhutan, along with the
tiary referral unit for the whole country. Because Linux/Apache setup for the webserver in New
there is no general practitioner service, patients Zealand.. In addition, standard MSOffice products
have the opportunity to refer themselves directly were used to develop the protocols.

167
 Open Source Software

reVIew oF success Factors adapt user-Friendly Interfaces

clear objectives The user interface adopted was a standard web-

browser, whatever the source of the data – even
The objectives of the project were identified in locally stored protocols would display in a browser.
initial discussions and codified in the agreement The native protocols were stored as XML docu-
signed between the stakeholders. The objectives ments, which were then displayed in a human-
included the development of a perinatal medicine readable format via a web browser. XML was
service, continuing support for this service and chosen for ease of updating – in that the editing
standardisation of treatment based on the best process could alter content without a great deal
possible evidence. An additional objective was of formatting issues and with the awareness that
sustainability of the service, OSS supported this other display methods such as voice responses
by allowing low or no cost technical documenta- or mobile devices may be used in the future. As
tion and development tools to be made available an open standard, XML is very well suited to
to local staff. this approach. The XML design is intended to be
expandable and able to represent both diagnostic
government support and therapeutic protocols.
A fragment of the XML representation is
The Royal Government of Bhutan (RGOB) is shown in Table 1. The initial outline was based
the sole supplier of healthcare in the kingdom. on the PubMED schema, but simplified to remove
The RGOB runs a series of five-year plans excessive bibliographic elements. The XML
which identify objectives and priorities as well as documents identify the responses to particular
sources of funding. Plans developed by overseas diagnoses or symptoms which would be expected
providers are examined and extensive negotiation to be encountered commonly. The aim is to allow
takes place to ensure that the country receives ap- clinical staff - who may be at a remote site – to
propriate and sustainable help that is consistent identify what emergency care is needed, whether
with the RGOB objectives. This process began the patient needs to be referred or transferred
in the case of this project, two years before the and the degree of urgency of that referral. Also
initiation, when representatives of the funders – the protocol can identify what additional tests or
The Magee Family – met with other stakeholders procedures need to be performed.
including government representatives, clinical
staff from New Zealand and Bhutan associated determine accessibility
with the project, and UNICEF. This resulted in a
project agreement that was signed off in a formal Although land transport is difficult, Bhutan is
ceremony. The project composed a number of in the process of increasing the availability of
other elements including funding for hardware internet access. Apart from dial-up connections
and training of clinical workers in the perinatal there are microwave links and recently the In-
medicine area. Continuing involvement of the ternational Telecommunications Union (ITU)
stakeholders has been a great asset to the project. e-post initiative(International Telecommunica-
RGOB department of IT has been running a long- tions Union, 2006), has recently been launched in
term project to support OSS and is getting closer Bhutan using very small aperture (VSAT) satellite
to the development of a policy on its use (Bhutan ground stations for rural access to electronic com-
Department of Information Technology, 2007) munications, and this may be useful for rollout to

168
Open Source Software

Table 1. XML Protocol Fragment

<Root_Element>
<Name>Cord prolapse</Name>
<Definition>
The cord that normally presents itself is within intact membranes. When the mem-
branes rupture, the cord prolapses. This is an emergency as cord compression and/
or occlusion can cause fetal asphyxia.
</Definition>
<Keywords>
<Keyword>Rupture of Membrane</Keyword>
<Keyword>Prolapse</Keyword>
</Keywords>
<Diagnosis>
<Diagnostic_step>Palpable cord on vaginal exam</Diagnostic_step>
<Diagnostic_step>Observed cord protruding onto vulva</Diagnostic_step>
</Diagnosis>
</Root_Element>

remote areas. OSS tools are often very efficient in standards in health IT and the importance of
terms of file size, and machine footprint (use of standards in the open source environment.
processor time and memory) so they can be used
in a wider range of scenarios than might be pos-
sible for the latest proprietary operating systems
or applications.
Table 2. Major Elements of the Protocol Document
Implement standards Element Comment
Name Name of protocol
The protocols themselves were developed in a Definition
standard format (Table 2), and as seen above, Keywords Used for searching
implemented using XML. In addition to this an Diagnosis For diagnostic protocols
attempt was made to standardize the production Diagnostic Step
of the protocols, so that candidate protocols from
Procedure For procedural protocols
other sources would go through an editorial process
Procedure Step
and be routinely revised. This process has been
Audience Intended user, includes country
followed by paper-based systems previously, but and location
electronic approaches allow instantaneous updat- Evidence A small selection of the sup-
ing of the live protocol without fear of version porting evidence
control issues and also allow a trail to be kept of Author Multiple Authors Possible
previous versions that can be linked to any events Last Update
linked historically to the implementation. Review Date
Recent work (Yellowlees, Marks, Hogarth, &
Turner, 2008) has reiterated the importance of

169
 Open Source Software

Measure cost-effectiveness which may only be available online, may well

and user satisfaction increase the quality and speed of development..

This aspect is perhaps the most difficult part of

the project. As part of the process protocols will dIscussIon and Future worK
be regularly reviewed by stakeholders. In addi-
tion, a Perinatal database is being implemented in There are some general issues that affect eHealth
order to record outcomes, and assess performance initiatives, and the use of OSS in the developing
against that expected in the protocol – in particular world, in particular connectivity, computing re-
areas where the protocols are not being followed, sources and skills.
and whether the protocols or behavior or both
should be modified. It is hoped that improvements connectivity
in the mortality and morbidity figures will also
be noticeable. Finally a rise in awareness of the Less developed nations have generally much lower
general maternity service and increased access to availability of fixed telephone lines. In addition,
it by women – only half of whom currently have geographic, economic and governmental issues
an attended birth – will accompany improved often conspire to make conventional dial-up access
outcomes among those who have contact with less common than in western countries. However,
the Perinatal service. Current work is focused wireless and satellite solutions such as VSATS
on examining the effect of improved ultrasound including the International Telecommunications
training and outreach clinics. Union (ITU) e-post initiative (International Tele-
communications Union, 2006) are overcoming
these issues. It is important to recognize that not
lessons learned every nation’s infrastructure is developing in the
same way, and many nations may leapfrog to wire-
Integration of protocols from diverse sources less solutions without the use of landline-based
was one of the major challenges facing the team. solutions. However high bandwidth solutions may
Protocols were sourced from National Women’s not be appropriate for developing countries. One
Hospital Auckland, the World Health Organisation of the most successful e-health projects has been
and EmOC protocols in Bhutan. the Swinfen Project – currently expanding in Iraq
Collaborative review of protocols was ex- (Swinfen et al., 2005). This project uses e-mail
tremely important, as buy-in from clinical staff in a store-and forward model, between clinicians
is vital. However the process of maintaining a in various countries. The prospects of advanced
common electronic repository was technically telepresence approaches being effective in routine
difficult as each of the reviewers tended to work care seem slight because of issues concerning
asynchronously using paper copies. The final ap- quality of service, bandwidth and reliability of
proach used was to produce paper prototypes and connection. Even though the “trauma pod” and
distribute them, collect back annotated versions other projects financed by the US Department of
and then combine them in a final word document. Defence are beginning to show results (Romano,
This was then converted to XML. Development of Lam, Moses, Gilbert, & Marchessault, 2006), costs
the Perinatal database was restricted by the very are likely to render this approach problematic in
small numbers of users available to test and com- other contexts.
ment on the system, and a wide user community,

170
Open Source Software

computing often have skills that are no longer available in

more developed nations. Collaboration in train-
Devices such as the Simputer (The Simputer ing of medical professionals, where trainees from
Trust, 2000) and the sub $100 laptop (OLPC, different nations are exchanged, can improve the
2006), promise much cheaper access to computing training in both systems. This can be supported by
power. It should be emphasized that for e-health the use of e-health tools such as websites, e-mail
applications, the computing device can be fairly and instant messaging.
simple, indeed mobile devices may become the Other skills required include the support of
preferred means of access. Along with cost, the the e-health infrastructure in terms of technical
ability to survive rough treatment, extremes of support for computing devices and connectivity.
temperature and humidity and long battery life – Fortunately the requirement of tourists from west-
or even the use of clockwork power in the case ern countries for internet connectivity wherever
of the sub $100 laptop – are more important in they are, along with the burgeoning industries of
developing countries than in Organization for call centres and ‘off-shoring’ of software devel-
economic cooperation and development (OECD) opment are providing a strong push for training
member countries. Parts supply and transport in these areas.
cost can make the repair of computers extremely OSS use in education and training allows
expensive. However organizations such as Global nations with limited resources to devote more
Assistance for Medical Equipment (GAME) funding to the human side of education, as well
(http://www.global-medical-equipment.org/ as allowing projects that involve software local-
whatwedo.html) have established links between ization to advance quickly. Open-source clinical
professional organizations in the developed and protocols may become important repositories of
less-developed world. These approaches move clinical knowledge allowing rapid development
beyond the shipping of obsolescent equipment, and input from experience, especially is based on
to an integrated and well thought and sustainable standard electronic forms.
out collaboration between donors and recipients. Another important aspect of skill transfer and
collaboration is the use of early warning networks
skills and Information for disease surveillance such as the Global Out-
break Alert and Response Network (GORAN) that
At present consumer e-health is of limited use- played a very large part in the early detection of
fulness in the developing world. Low levels of SARS (Heymann & Rodier, 2004). Such networks
literacy and information literacy cause difficulties. link health workers throughout the world and the
However the fact that the vast majority of web transfer of information is by no means one-way.
resources are written in English, and are US-centric There remains a dearth of well-controlled stud-
in terms of organization of healthcare, availability ies of e-health initiatives in developing nations,
of drugs and medical devices and naming make but the need for effective collaboration remains
even materials designed for health consumers in paramount (Wooton, 2001). However there are a
the OECD countries less useful for those in other number of pointers to success;
nations. However, these issues are much less im-
portant when the provision of e-health services 1. The e-health system must be compatible
for medical professionals is considered. Adapting with existing organizational and cultural
general principles to specific cases is a key skill structures. Some “western” assumptions
of medical professionals. Indeed the traffic is not do not apply in less developed nations and
all one-way, less developed nation professionals vice versa. For example routine ultrasound

171
 Open Source Software

examination in early pregnancy has not been has a web presence, the technical barriers to such
shown to be effective in reducing mortal- collaboration are much lower than they were even
ity in a Cochrane review(Neilson, 1998). ten years ago. It is hoped that further work will
However an environment where mortality refine the system sufficiently to allow the software
due to unsuspected problems is much greater, to be placed in a repository such as Sourceforge.
and the availability of on-demand scans is net. Furthermore, it is hoped that such an ap-
lower, may give different results. proach will encourage increasing collaboration
2. Collaboration and training between the pro- and development in this area.
fessionals involved is vital. This applies to
both clinical and technical staff. This may
in fact be the area of greatest benefit. acKnowledgMent
3. Ingenuity is more important than technology.
Store-and-forward email may be of greater This project work would not have been possible
utility than telepresence. without the generosity of the Magee family and
4. Open-source technology is particularly the hard work of the representatives of UNICEF
suited to this area of work. Lower costs, and the Royal Government of Bhutan. Staff of
availability of technical skills, greater range Jigme Dorji Wanchuck National Referral Hospital,
of customized languages and often lower Thimphu assisted in a very wide range of roles
technology requirements make Open Source and continue to work on this project
approaches and especially web-based Open-
source tools particularly attractive.
reFerences
Future work in this area will include greater
use of multicentre collaboration, both within Bhutan Department of Information Technology.
existing networks such as GORAN and GAME (2007). Bhutan’s Journey Towards Open Source.
and outside them. Lower bandwidth costs, and Paper presented at the DebConf7, Edinburgh,
easier access to high bandwidths will enable Scotland. https://penta.debconf.org/~joerg/
richer media to be used, such as telesonography events/25.en.html
via store and forward (Parry et al., 2006).Common Darmstadt, G., Bhutta, Z., Cousens, S., Adam, T.,
health problems are starting to afflict North and Walker, N., & Berni, L. d. (2005). Evidence-based,
South – ageing populations, the rapid spread of cost-effective interventions: how many newborn
new infectious diseases and chronic conditions. babies can we save? Lancet, 365(9463), 977–988.
Common approaches to these issues – including doi:10.1016/S0140-6736(05)71088-6
the use of low-cost assistive technology, and off-
shoring of medical procedures such as radiology Deodhar, J. (2002). Telemedicine by email-
(Larson & Janower, 2005) may be controversial, -experience in neonatal care at a primary
but at least the discussion has started. There are care facility in rural India. Journal of Tele-
enormous potential benefits in the development medicine and Telecare, 8(Suppl 2), 20–21.
of e-health in collaboration with developing na- doi:10.1258/135763302320301867
tions, and the benefit to the people of world may Eysenbach, G. (2001). What is e-health? Journal of
be immense. An additional benefit of the open Medical Internet Research, 3(2), e20. doi:10.2196/
source approach may be an increased ability for
jmir.3.2.e20
IT specialists in developed nations to assist people
around the world. As virtually every nation now

172
Open Source Software

Heymann, D. L., & Rodier, G. (2004). Global Kirigia, J., Seddoh, A., Gatwiri, D., Muthuri, L.,
Surveillance, National Surveillance, and SARS. & Seddoh, J. (2005). E-health: Determinants, op-
Emerging Infectious Diseases Journal, 10(2). portunities, challenges and the way forward for
Retrieved 1st December 2006, from http://origin. countries in the WHO African Region. BMC Public
cdc.gov/ncidod/EID/vol10no2/03-1038.htm Health, 5(1), 137. doi:10.1186/1471-2458-5-137
Hippel, E. v. (2001). Innovation by User Com- Larson, P. A., & Janower, M. L. (2005). The
munities: Learning from Open-Source Software. Nighthawk: Bird of Paradise or Albatross? Jour-
MIT Sloan Management Review, 42(4), 82. nal of the American College of Radiology, 2(12),
967–970. doi:10.1016/j.jacr.2005.08.002
Iluyemi, A., Fitch, C. J., & Parry, D. T. (2007, April
18-20th 2007). Mobile and Wireless Technologies Littlejohns, P., Wyatt, J. C., & Garvican, L. (2003).
for e-Health Provision in Africa: Opportunities Evaluating computerised health information sys-
for GSM-Based Patient Centric Services. Paper tems: hard lessons still to be learnt. British Medi-
presented at the Med-e-Tel Luxembourg Inter- cal Journal, 326(7394), 860–863. doi:10.1136/
national Telecommunications Union. (2006). bmj.326.7394.860
Bhutan to be Testbed for ITU’s E-Post Venture
Mascarenhas, L., Eliot, B. W., & MacKenzie, I. Z.
with Universal Postal Union. Retrieved 20th
(1992). A comparison of perinatal outcome, ante-
April 2006, from http://www.itu.int/newsarchive/
natal and intrapartum care between England and
press_releases/2002/10.html
Wales, and France. British Journal of Obstetrics
Islam, MT, H. Y., Waxman R, Bhuiyan AB. (2006). and Gynaecology, 99(12), 955–958.
Implementation of emergency obstetric care train-
McDonald, C. J., Schadow, G., Barnes, M., Dexter,
ing in Bangladesh: lessons learned. Reproductive
P., Overhage, J. M., & Mamlin, B. (2003). Open
Health Matters, 14(27), 61–72. doi:10.1016/
Source software in medical informatics--why,
S0968-8080(06)27229-X
how and what. International Journal of Medi-
Jakobovits, R. M., Rosse, C., & Brinkley, J. F. cal Informatics, 69(2-3), 175–184. doi:10.1016/
(2002). WIRM An Open Source Toolkit for Build- S1386-5056(02)00104-1
ing Biomedical Web Applications. Journal of the
Mockus, A., Fielding, R. T., & Herbsleb, J. (2000).
American Medical Informatics Association, 9(6),
A case study of open source software develop-
557–570. doi:10.1197/jamia.M1138
ment: the Apache server. Paper presented at the
Kantor, G. S., Wilson, W. D., & Midgley, A. (2003). Software Engineering, 2000. Proceedings of the
Open-source software and the primary care EMR. 2000 International Conference on.
Journal of the American Medical Informatics As-
Neilson, J. P. (1998). Ultrasound for fetal assess-
sociation, 10(6), 616. doi:10.1197/jamia.M1403
ment in early pregnancy. Cochrane Database of
Kayango, M., Esquiche, E., Luna, M. R., Frias, G., Systematic Reviews, 4.
Vega-Centeno, L., & Bailey, P. (2006). Strength-
OLPC. (2006). One laptop per Child. Retrieved
ening emergency obstetric care in Ayacucho,
1st December 2006, 2006, from http://laptop.org/
Peru. International Journal of Gynaecology and
Obstetrics: the Official Organ of the International Pagliari, C., Sloan, D., Gregor, P., Sullivan,
Federation of Gynaecology and Obstetrics, 92, F., Detmer, D., & Kahan, P. J. (2005). What Is
299–307. doi:10.1016/j.ijgo.2005.12.005 eHealth (4): A Scoping Exercise to Map the Field.
Journal of Medical Internet Research, 7(1), e9.
doi:10.2196/jmir.7.1.e9

173
 Open Source Software

Pal, A., Mbarika, V. W. A., Cobb-Payton, F., Datta, The Simputer Trust. (2000). Simputer(TM) - Wel-
P., & McCoy, S. (2005). Telemedicine diffusion come. Retrieved December 1, 2006, from http://
in a developing Country:The case of India (march www.simputer.org/
2004). Information Technology in Biomedicine.
Tomasi, E. F., Luiz Augusto and Maia, Maria
IEEE Transactions on, 9(1), 59–65.
de Fatima Santos. (2004). Health information
Parry, E. C., Sood, R., & Parry, D. T. (2006). In- technology in primary health care in developing
vestigation of optimization techniques to prepare countries: a literature review. Bulletin of the World
ultrasound images for electronic transfer [Ab- Health Organization, 82(11), 867–874.
stract]. Ultrasound in Obstetrics and Gynecology,
Wooton, R. (2001). Telemedicine and developing
28(4), 487-488(482).
countries–successful implementation will require
Raento, M., Oulasvirta, A., Petit, R., & Toivonen, a shared approach. Telemedicine and Telecare,
H. (2005). ContextPhone: A Prototyping Platform 7(Suppl 1), 1–6. doi:10.1258/1357633011936589
for Context-Aware Mobile Applications, 4, 51-59.
World Health Organization. (2006a). Estimates
Romano, J. A., Lam, D. M., Moses, G. R., Gilbert, of child and adult mortality and life expectancy
G. R., & Marchessault, R. (2006). The Future of at birth by country. Retrieved January 17, 2007,
Military Medicine Has Not Arrived Yet, but We from http://www.who.int/healthinfo/statistics/
Can See It from Here. Telemedicine and e-Health, mortlifeexpectancy/en/index.html
12(4), 417-425.
World Health Organization. (2006b, 6th June
Russell, S. (2004). The economic burden of illness 2006). Introduction to the ‘Making Pregnancy
for households in developing countries: a review Safer’ initiative. Retrieved October 1, 2007, from
of studies focusing on malaria, tuberculosis, and http://w3.whosea.org/pregnancy/introf.htm
human immunodeficiency virus/acquired im-
Yellowlees, P. M., Marks, S. L., Hogarth, M., &
munodeficiency syndrome. Am J Trop Med Hyg,
Turner, S. (2008). Standards-Based, Open-Source
71(2_suppl), 147-155.
Electronic Health Record Systems: A Desirable
Swinfen, P., Swinfen, R., Youngberry, K., & Future for the U.S. Health Industry. Telemedicine
Wootton, R. (2005). Low-cost telemedicine in and e-Health, 14(3), 284-288.
Iraq: an analysis of referrals in the first 15 months.
Telemedicine and Telecare, 11(Suppl 2), 113.

174
 175

Chapter 13
An Empirical Investigation into
the Adoption of Open Source
Software in Hospitals
Gilberto Munoz-Cornejo
University of Maryland Baltimore County, USA

Carolyn B. Seaman
University of Maryland Baltimore County, USA

A. Güneş Koru
University of Maryland Baltimore County, USA

aBstract

Open source software (OSS) has gained considerable attention recently in healthcare. Yet, how and
why OSS is being adopted within hospitals in particular remains a poorly understood issue. This re-
search attempts to further this understanding. A mixed-method research approach was used to explore
the extent of OSS adoption in hospitals as well as the factors facilitating and inhibiting adoption. The
findings suggest a very limited adoption of OSS in hospitals. Hospitals tend to adopt general-purpose
instead of domain-specific OSS. We found that software vendors are the critical factor facilitating the
adoption of OSS in hospitals. Conversely, lack of in-house development as well as a perceived lack of
security, quality, and accountability of OSS products were factors inhibiting adoption. An empirical
model is presented to illustrate the factors facilitating and inhibiting the adoption of OSS in hospitals.

IntroductIon large and fast-growing number of OSS users and

software products in a large variety of domains.
The open source software (OSS) phenomenon OSS is already being adopted and used as a soft-
has become an important area of interest in ware platform in a number of fields other than
information systems research due in part to the healthcare (Dedrick & West, 2003; 2004; Norris,

Copyright © 2011, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 An Empirical Investigation into the Adoption of Open Source Software in Hospitals

2004; Waring & Maddocks, 2005), and it has the managers were chosen to represent the hospitals’
potential to be equally promising for the hospital perspective on this topic. The following three
industry (Fitzgerald & Kenny, 2004). Studying questions guided this investigation:
OSS adoption in any domain can help reveal pat-
terns and phenomena that are applicable to adop- 1. What are the types and names of OSS prod-
tion in general, in addition to revealing insights ucts that hospitals choose to adopt?
into the domain being studied. In particular, the 2. What is the extent of OSS adoption for these
adoption and use of OSS in a hospital context products in hospitals?
remains a poorly understood phenomenon; only a 3. What are the factors facilitating and inhibit-
handful of researchers have addressed the factors ing the adoption of OSS in hospitals?
inhibiting or facilitating such adoption. Such an
understanding is important in helping hospitals To research these questions, a survey and inter-
make better decisions about whether and how views were used to acquire both breadth and depth
adoption of OSS could benefit them. of understanding. The purpose of the survey was
The first step in developing a better understand- to answer the first two questions—to explore and
ing is to explore the current state of OSS adoption, characterize the types of OSS products adopted in
and the factors inhibiting and influencing it in hospitals and to discover the extent to which these
hospitals. Such an exploration is the goal of this products have been adopted. The interviews were
study. Once this current state is well described, used to answer question three to attain a deeper
it will be possible to seek answers to higher-level understanding of the factors that are facilitating
questions about the pros and cons, the costs and and inhibiting the adoption of OSS in hospitals.
benefits, the advantages and disadvantages of OSS In the following sections of this article, we
adoption in this domain, which is the second goal. first present the related work in this area. Then,
Therefore, the present study is of considerable we introduce the methodology for our survey
interest for both practitioners and researchers. It and interview studies. After that, we present our
will provide hospitals and healthcare organizations data analysis and results. Then, we introduce
that are considering the adoption of OSS technolo- our empirical model of the adoption of OSS in
gies with an understanding of how technological, hospitals. Finally, we present our conclusions and
environmental and organizational factors affect the implications of our work.
the adoption process. This way hospital IT prac-
titioners, or others attempting to introduce OSS
technology into hospitals, can prepare against lIterature reVIew
the expected barriers and can utilize the facilita-
tors for successful adoption. This research also open source software adoption in
provides scholars with an empirical model for Healthcare
better understanding facilitating and inhibiting
factors, as well as providing the foundations for Over the past few years, a small number of
further research that may validate and expand on researchers have focused on the study of the
the empirical model in other healthcare organiza- potential advantages and risks of adopting and
tions and other domains. implementing OSS in the healthcare domain.
The main objective of this investigation was to Prior research encouraged the adoption and use of
explore and analyze the extent of OSS adoption OSS in healthcare organizations because of OSS’s
in hospitals, along with the factors influencing potential to both enhance healthcare delivery and
or inhibiting this adoption process. Hospital IT lower software acquisition costs (Carnall, 2000;

176
An Empirical Investigation into the Adoption of Open Source Software in Hospitals

Kantor, Wilson, & Midgley, 2003; McDonald in the healthcare industry. They also proposed
et al., 2003; Valdes, Kibbe, Tolleson, Kunik, & that OSS might allow rapid scientific advance-
Petersen, 2004). ment due to the sharing of information and soft-
OSS could potentially be more reliable and se- ware (Erickson et al., 2005; Scarsbrook, 2007).
cure than proprietary software because its source Other authors such as DeLano (2005) presented
code can be inspected and reviewed (Carnall, some reasons for the potential success of OSS
2000). Past research introduced and extended the predicting that the pharmaceutical research and
idea of OSS as a software development model development process may benefit from the OSS
that could definitively improve clinical and re- development model.
search software in the field of medical informat-
ics (Yackel, 2001). A paper by Kantor, Wilson, open source software adoption in
and Midgley (2003) also presents the potential Hospitals
benefits that OSS could provide in the area of
primary care. Kantor et al., also proposed that A case study of OSS adoption was conducted at
the adoption of OSS would reduce the excessive the Beaumont hospital in Ireland, where the IT
costs, the frequent turnover of vendors, and the department, under limited financial resources,
lack of common data standards that are afflict- made the decision to adopt OSS. Several OSS
ing electronic medical records (EMR) systems products were adopted and implemented suc-
in primary care. cessfully. The authors reported that there were
More recently, McDonald, Schadow, Barnes, important initial start-up and future operational
et al. (2003) also investigated the potential role costs when OSS products were preferred in the
that the OSS model of software development may hospital (Fitzgerald & Kenny, 2004).
have in the medical informatics area. They also Another study by Glynn, Fitzgerald and Exton
described a number of OSS products that have (2005) investigated the commercial adoption of
been used in the medical informatics domain OSS using an innovation adoption theory frame-
over the years, including: OpenEMed, a patient work based on Tornatzky and Fleischer’s (1990)
record system; OSCAR, a family practice office model. They derived a framework that was then
management and medical record system; as well used to investigate the adoption process of OSS
as the internationally well-known VistA system, in the case of the Beaumont hospital (Fitzgerald
a computer-base patient records system (CBPR) & Kenny, 2004).
developed in MUMPS (Massachusetts General The OSS products and processes were also seen
hospital Utility Multi-Programming System) by as promising in terms of enabling rapid evolution
the U.S. Department of Veteran’s Affairs (Brown, and proliferation of applications in the medical
Lincoln, Groen, & Kolodner, 2003; Longman, domain through their use of open standards and
2007). A more recent study by Valdes et al. higher degrees of interoperability (Raghupathi &
(2004) also pointed out that OSS could be an ef- Gao, 2007). The authors argued that the devel-
fective solution for the problems that distress the opment processes in the Eclipse project (http://
healthcare industry such as high costs, business eclipse.org) could improve scalability, prevent
failures and barriers of standardization (Valdes vendor lock-ins, and reduce costs in the medical
et al., 2004). Other papers by Erickson, Langer, information systems including electronic health
and Nagy (2005), Scarsbrook (2007) and Nagy record and clinical decision support systems.
(2007) supported the growth and adoption of OSS There are some recent studies focusing only
in radiology because OSS may significantly lower on the managerial and technical barriers to the
the entry cost for standards-compliant practices adoption of OSS (Holck, Larsen, & Pedersen,

177
 An Empirical Investigation into the Adoption of Open Source Software in Hospitals

2005). Past research on OSS and healthcare also industry. Only a handful of researchers have ad-
proposed that OSS would reduce the number of dressed the factors inhibiting or facilitating OSS
bugs and failures in medical systems, as well as adoption in hospitals (Carnall, 2000; Kantor et
reduce their overall cost (Yackel, 2001). A study al., 2003; Valdes et al., 2004; Glynn, Fitzgerald &
by Hogarth and Turner (2005) focused on creating Exton, 2005). Each of the aforementioned studies
a catalogue of existing OSS clinical projects and in this section found that top management support,
on determining metrics for their viability. The limited financial resources, past experiences using
authors mentioned that many of the factors that OSS-like systems, and the flexibility to modify,
are required to make a “successful and vibrant” combine, and tailor OSS are the most important
OSS community within the mainstream software facilitating factors for the adoption of OSS within
applications systems (e.g., Linux, Apache, etc.) a hospital scenario. The factors inhibiting adop-
may not necessarily be applicable to the clinical tion range from the fear of IT personnel becoming
software applications systems. de-skilled by not using mainstream commercial
Another study by Kantor, Wilson and Midgley applications, the lack of OSS-literate IT person-
(2003) presented a set of potential advantages that nel, the lack of other successful OSS examples in
the adoption of OSS may provide with regards to the industry, to the lack of reliable procurement
lowering the resistance of hospitals to the adop- models for the adoption of OSS. Finally, many of
tion of electronic medical records (EMR). These the papers and studies reported are cases from
included: 1) the potential of OSS to reduce EMR European countries, with healthcare systems that
ownership and software development costs, 2) the are very different from that in the U.S. Table 1
removal of vendor lock-in, and 3) the adherence of presents only a summary of the facilitators and
OSS to standards for the compatibility and data inhibitors shown to influence the adoption of OSS
interchange among systems. as found in the literature.
In another study by Valdes, Kibbe, Tolleson, et
al. (2004) dealing with the barriers to the prolifera-
tion of electronic health records/electronic medical MetHodology
records (EHR/EMR), the authors concluded that
OSS is a viable solution to the barriers of high A mixed methods design was used in this research
cost, business failure and standardization that to explore the extent of OSS adoption in hospi-
the healthcare industry is facing when adopting tals as well as to investigate the influencing and
EHR/EMR. The authors mentioned that, for ex- inhibiting factors. The exploratory approach of
ample, interconnectivity problems are more easily this study is warranted by the fact that, as of yet,
solved when using OSS, since no technical infor- the adoption and use of OSS in U.S. hospitals has
mation can be hidden. They also added that OSS not been accompanied by any theoretical ground-
can help alleviate the high costs associated with ing or by empirical analysis that explains how or
the adoption and implementation of EHR/EMR why OSS products are being adopted and used.
(Valdes et al., 2004). Although this article presents That is, thus far, there are few existing concep-
a good case for the adoption of OSS solving the tual frameworks to guide a research effort in this
barriers that EHR/EMR is facing, the authors do area. Similarly, there are no theoretical guidelines
not support their case with empirical data. that have been empirically evaluated to support
In summary, even though we have witnessed a a rigorous understanding of the complex factors
widespread, significant OSS research and industry that inhibit the adoption and successful imple-
adoption of OSS, there are still few studies on mentation of OSS technologies in hospitals. For
OSS adoption and use, especially in the hospital these reasons, a mixed methods approach using a

178
An Empirical Investigation into the Adoption of Open Source Software in Hospitals

Table 1. Main facilitators/inhibitors of OSS adoption

Author(s) Major Factor Findings

Facilitators Inhibitors
Fitzgerald and • Limited financial resources • Lack of support from vendors
Kenny (2004) • Top management support • Perception that OSS would threaten local proprietary
• Software functionality software companies
• User’s past experience • Fear by users to become de-skilled
Gynn, Fitzger- • Perception that the benefits of OSS outweigh its • Perception of work under-valued if using OSS products
Adoption of Open Source Software in Hospitals

ald and Exton disadvantages • Having to change operating model to OSS

(2005) • OSS-literate IT personnel • Fear by users to be de-skilled
• Top management support • Lack of OSS champion example
• Personal support for OSS ideology • Lack of tolerance to technical problems with OSS
• Network externalities • Favorable arrangements with proprietary vendors
• The OSS champion example
Holck, Larsen • Limited financial resources • Lack of reliable procurement models
and Pedersen • Pressure to upgrade IT systems  Legal (licenses)
(2005) • Top management support  Technical (functionality, security, usability)
• User’s past experience  Corporate and business policy (vendor,
• Government support customer support, and software alliances)
Tomas Yakel • Access to real-world systems • Lack of a mature OSS beyond prototype phase
(2001) • Reduction of bugs in medical systems • High level of technical expertise required for OSS
• Reduction of software ownership and develop- • Proprietary mindset of the medical community
ment cost • Technology complexity of the medical domain
• Lack of OSS-IT personnel support, specifically for
medical software applications
MacDonald et • Public policy encouraging that all software • Medical software currently in use is proprietary
al. (2003) developed by the government must be released software
under an OSS license • Leadership and top management in healthcare is risk
• Information mechanisms to disseminate to the adverse
community about OSS developments and benefits • Elimination of in-house personnel due to outsourcing
Adoption of Open Source Software in Healthcare

• Technology complexity of the medical domain

Hogarth and • Reduction of software ownership and develop- • Lack of OSS-IT personnel support, specifically for
Turner (2005) ment cost medical software applications
• Disappearance of vendor lock-in • Technology complexity in the medical domain
• OSS adherence to standards for compatibility • Success of mainstream applications might not translate
and data interchange to clinical software
Kantor et al. • OSS can reduce EMR ownership and development
(2003) cost
• Disappearance of the vendor lock-in
• OSS adherence to standards for compatibility
and data interchange

179
 An Empirical Investigation into the Adoption of Open Source Software in Hospitals

grounded theory perspective was selected over a turally. The survey sample was selected from the
confirmatory or causal research design approach. Healthcare Information and Management Systems
Grounded theory is a systematic, qualitative re- Society (HIMSS) from their electronic mailing
search procedure used to develop an inductively list database of chief information officers (CIO),
grounded theory that explains a process, an ac- chief technology officers (CTO), vice presidents
tion, or interaction about a phenomenon (Glaser & (VP) (of information technology (IT), informa-
Strauss, 1967; Glaser, 1978; 1999; Creswell, 1994; tion systems (IS) and management information
2005; Strauss & Corbin, 1998; Charmaz, 2006). systems (MIS)), and directors and managers of
other IT departments within hospitals. HIMSS
was selected because it is a leading non-profit
study desIgn organization dedicated to improving healthcare
through the application of information technology
The data collection methods used in this research (HIMSS, 2006). This research takes a key infor-
are a survey and interviews, allowing both breadth mant approach that allowed the responses of the IT
and depth of information concerning the adoption managers to represent those of the hospital being
of OSS in hospitals. We focused on Baltimore, surveyed. The use of managers as key informants
Washington and Northern Virginia (BWNV) has been successfully applied in many IT studies
area hospitals instead of a nationwide area. This that involve organizations (Huff & Munro, 1985;
allowed us to spend more time cultivating each Gatignon & Robertson, 1989; Chau & Tam, 1997;
contact from the target population through initial Eyler et al., 1999; Goode, 2005).
phone calls, and to obtain richer data in the form
of personal face-to-face and telephone exchanges.
First, a survey was used to gather data from a surVey adMInIstratIon
wide variety of hospitals dispersed across a geo-
graphic area. This was done in order to explore Prior to sending the survey invitation e-mail out,
and characterize the extent and the types of OSS an attempt was made to contact each of the IT
products adopted by hospitals. Following the sur- managers in the target population by telephone in
vey, semi-structured interviews were conducted an effort to encourage participation and receive
in-person and by telephone with IT managers in a verbal commitment from them to complete the
order to attain deeper understanding of the factors survey. After the initial telephone contact, an
that facilitate or inhibit OSS adoption in hospi- e-mail invitation letter was sent to the potential
tals. Interviews are the quintessential qualitative respondents. The survey link was appended to the
method for data collection and one of the most bottom of the e-mail cover letter and upon click-
widely used techniques for acquiring qualitative ing the survey link, the participant was directed
data in order to collect impressions and opinions to the online survey (Appendix A).
about the particular research issue (Tashakkori & Descriptive statistics, such as frequency
Teddlie, 1998; Patton, 2002). distributions, percentages, standard deviations,
The target population for the study consists of confidence intervals, Chi-square and Fisher’s tests
hospital executives, directors and managers that were computed in order to analyze the survey
are involved in IT within BWNV area hospitals. results. Moreover, to ensure better reporting and
Although we selected the BWNV area largely complete description of our Web-based survey
because of our own location, it is an appropriate results, we applied a checklist of recommenda-
choice because it is one of the most diverse areas tions from the Checklist for Reporting Results
in the U.S. socioeconomically, politically, and cul-

180
An Empirical Investigation into the Adoption of Open Source Software in Hospitals

of Internet e-Surveys (CHERRIES) (Eysenbach, constant comparative method. Each interview was
2004) (Appendix B). treated as an individual case. NVivo® was used
to assist the qualitative analysis process, to man-
age data, to store the interview transcripts, and to
InterVIews help in coding text (Bazeley & Richards, 2000).

The interview population consisted of the subset

of survey respondents who responded positively results
to a survey item that specifically asked if they
were willing to share their thoughts and expe- survey results
riences in an interview. A total of 11 survey
respondents initially agreed to be interviewed. This research finds that 23% (n=7) of the hospi-
All such respondents were sent an e-mail letter tals within the survey sample have adopted OSS.
introducing the objectives of the interview and Conversely, 76% (n=23) of the hospitals indicated
asking to schedule a meeting. By the end of this that they have not adopted any type of OSS. All
process, only five IT hospital managers ultimately of the hospital adopters of OSS reported having
agreed to be interviewed. The other six manag- general-purpose products. Among them, only 57%
ers, for reasons unknown, chose not to respond (n=4) reported having adopted domain-specific
to the many invitations by e-mail and telephone products. Table 2 presents descriptive statistics
to participate and were unreachable to be inter- profiling the hospitals in our survey sample.
viewed. Each interview lasted 30-60 minutes and Key findings from this research indicate that
was conducted between January and May 2007. hospitals are adopters of both general-purpose and
The interviews focused on the organizational, domain-specific products, but they have adopted
technological, and environmental factors that general-purpose products to a greater extent than
facilitated or inhibited the adoption of OSS at domain-specific products. General-purpose OSS
their hospitals (Appendix C). Before conduct- adoption in hospitals clusters mainly in databases,
ing each interview, the participant was briefed desktop software, programming languages, and
on the nature and purpose of the study. All the operating systems, as well as Web development
participants were asked for their authorization to tools and server products. Well-known OSS
be recorded during the interview and were asked products such as MySQL, Linux, Apache, Firefox,
to sign an informed consent. PHP and Perl were the leading software products
The interviews were coded and analyzed that hospitals selected to adopt. The scale used
employing grounded theory consistent with the in the survey to indicate extent of adoption was
systematic procedures recommended by Strauss adapted following Fichman and Kemerer (1997)
and Corbin (1998), namely open coding, axial and it ranges from unawareness (no knowledge of
coding and selective coding. Coding is the pro- OSS), to awareness, interest (actively learning),
cess that dissects, differentiates, combines, and evaluation/trial (acquisition and initiation of an
discovers concepts and relevant features from evaluation or trial version), commitment (use
the data (Seaman, 1999). We developed concepts for one or more deployment projects), limited
and categories emerging from the data using the deployment (regular, but still limited, deploy-
line-by-line analysis as described by Straus and ment and use), and general deployment stages (a
Corbin (1998) and Glaser and Straus (1967). The stable and regular part of the IT infrastructure).
concepts and categories were generated by our The survey results show that the vast majority
analysis of the data and validated applying the of general-purpose products are positioned from

181
 An Empirical Investigation into the Adoption of Open Source Software in Hospitals

Table 2. Descriptive statistics of surveyed hospitals

Frequency Percent
(n =30) %
Hospital type
Healthcare system hospital 12 40.0
Hospital as a part of a multi-system network 11 36.7
Stand-alone hospital 5 16.7
Ambulatory care facility 1 3.3
Other 1 3.3
Number of beds in the hospital
<50 beds 1 3.3
101-200 Beds 2 6.7
201-300 Beds 6 20.0
301-400 Beds 4 13.3
401-500 Beds 2 6.7
>501 11 36.7
Not classified by beds 4 13.3
Hospital’s annual gross revenue
< $5M 1 3.3
$5M-$25M 3 10.0
$26M-$50M 3 10.0
$51M-$200M 2 6.7
$201M- $350M 8 26.7
$351M-$500M 2 6.7
> $501M 11 36.7
Annual IT operating budget
<2% 5 16.7
2.1-3.0% 15 50.0
3.1-4.0% 1 3.3
4.1-5.0% 5 16.7
5.1-6.0% 1 3.3
>8% 3 10.0
Type of IT personnel
In-house 27 90.0
Outsourced 3 10.0
Number of in-house IT staff employed full time
≤10 5 16.7
10-30 10 33.3
31-60 7 23.3
≥91 8 26.7

continued on the following page

182
An Empirical Investigation into the Adoption of Open Source Software in Hospitals

Table 2. continued

Years of experience of in-house IT staff

≤2 years 1 3.3
2-5 years 6 20.0
5-10 years 15 50.0
≥10 years 8 26.7

the evaluation/trial stages to the limited deploy- adoption of OSS in hospitals. Further, hospitals
ment stages. Well-known OSS products, for rely heavily on software vendors for all of their
example, MySQL, Linux, Apache, and Perl, are IT solutions. The results also show that hospital
in the limited deployment stages, whereas OSS software vendors enlarge their product lines and
desktop software applications, such as Firefox and the services they provide to hospitals to include
Mozilla, are in the evaluation/trial stages. The general-purpose and domain-specific OSS prod-
extent of adoption of domain-specific products ucts. In addition, IT managers have a positive
is lower than that of general-purpose products. satisfaction level, in general, with the software
The predominant adoption stages for all the vendor services and products, and, overall, have
domain-specific OSS products are awareness to a good relationship with them. Table 3 presents a
interest. Domain-specific adoption occurs mainly concise summary of the results of the interviews.
in the telemedicine, electronic medical records, The majority of the hospital IT managers
radiology, laboratory and pharmacy information reported that lack of in-house development,
systems products. and a perceived lack of security, quality, and
Furthermore, the results of the survey provide accountability of OSS products were the most
information about relevant contextual and struc- significant factors that inhibit the adoption of OSS
tural characteristics of the hospitals that tend to in hospitals. IT managers also identified the lack
adopt OSS. These characteristics may have a of medical informaticians, patient-privacy protec-
determinant effect on the adoption of OSS. First, tion and privacy legislation as major inhibitors to
the majority of the adopting hospitals are very adopt OSS, particularly domain-specific products.
large hospitals, with 500 beds or more. Second, Based upon our findings (from both the survey
these hospitals tend to have high annual revenue, and interviews), the following section presents an
more than $500 million. Third, hospital adopters empirical model describing the factors facilitating
of OSS have a propensity to have a large number and inhibiting the adoption of OSS in hospitals
of IT support staff. Finally, hospitals that have and the relationships between them.
adopted OSS also tend to have IT budgets that
are less than 3% of the hospital’s total budget.
adoptIon oF oss In HospItals:
Interview results an eMpIrIcal Model

This study also identifies, through the interview We have used Strauss and Corbin’s (1998) para-
data, key categories that facilitate and inhibit the digm to develop an empirical model describing
adoption of OSS in the hospitals within the sample. the adoption of OSS in hospitals, based on our
The interview data reveal that hospital software data. This empirical model helps us to develop
vendors are the most critical factor influencing the and propose connections between the factors that

183
 An Empirical Investigation into the Adoption of Open Source Software in Hospitals

Table 3. Emerging code categories and subcategories of the adoption of OSS in hospitals

Core Categories Subcategories

1. Hospital IT human resources • In-house software development
• IT personnel
• Medical informaticians
2. Hospital regulatory landscape • Patient-privacy protection and privacy legislation
• Lack of liability/accountability provided by OSS
3. Hospital software vendors • Software vendor providers of OSS
• Satisfaction level with software vendors
• New software business models
4. Hospital organizational factors • Hospital organizational culture
• Hospital organizational structure
5. Hospital technological factors • Perceived lack of quality
• Perceived lack of security
6. International development of OSS • Labor cost and qualified programmers
• Type of healthcare systems

emerged from our findings. Figure 1 presents the construct in any way causes the second, but that
empirical model that lays out the analysis of the the mix of factors and actions described by the
factors that emerged from our results and the first construct influence the mix of factors and
relationships between them. actions described by the second construct in any
The empirical model identifies temporal particular instance. The constructs of the model
and inferential, rather than causal, relationships are described in more detail below.
between the factors relevant to the adoption of
OSS in hospitals. For example, the mix of causal causal conditions
conditions in a particular hospital at a particular
point in time (as defined by the level of in-house Causal conditions, as the term is being used in our
development, the number of IT personnel, etc.) empirical model, are factors that are identified as
sets the stage and shapes what happens when an influencing the core category. There is evidence
event occurs related to the core category (e.g., from our findings that all of these causal conditions
when a software vendor offers an open source have an influence on whether or not a hospital is
solution to the hospital). This core category then open to an offer of OSS by a software vendor.
directly influences the strategic actions (i.e., The subject of technical personnel in hospitals
adoption or non-adoption of OSS) that lead to came up often in our interview data (see “Inter-
the consequences. The contextual factors and view Results” and Table 3). Hospital IT managers
intervening conditions moderate and mediate the report that the lack of in-house development is
strategic actions that are employed to bring about the rule rather than the exception; hospitals do
certain consequences (Strauss & Corbin, 1998; not develop their own software systems, and thus
Creswell, 2005). So, in terms of the symbology they depend on software vendors for all their IT
in Figure 1, an arrow from one construct to an- operations and software needs. Managers also
other cannot be interpreted to mean that the first mentioned that much of their IT staff personnel

184
An Empirical Investigation into the Adoption of Open Source Software in Hospitals

are exclusively devoted to the on-site support of product strongly influences the decision to adopt
IT systems provided by vendors. The degree to products from software vendors. As one IT man-
which a hospital lacks in-house development activ- ager expressed: “the factor that caused us not
ity, and IT personnel with technical development to adopt OSS is the support and accountability
skills influences how dependent they are on their that comes with writing a check to a commercial
software vendors, and thus influences how they software vendor.” The negative perceptions of
would react to the offerings of their vendors. Such quality, security and lack of liability reinforce
a dependence would make a hospital more likely the hospitals’ dependence on software vendors.
to accept a technology solution from a vendor that Finally, our findings report that IT managers
included OSS. A related causal condition is the have a positive satisfaction level, in general, to-
lack of personnel who possess an amalgamation wards the products, support and services that soft-
of medicine and information systems expertise ware vendors provide in their hospitals, as noted
and thus who would be able to develop and main- in the interview results. This further reinforces
tain software systems tailored to hospitals and the hospitals’ dependence on software vendors.
healthcare organizations. In summary, these causal conditions all shape
The perceived lack of general quality and and impact the core category, that is, they influ-
perceived lack of security of OSS products are ence what happens when and if a vendor offers a
persistent themes that emerged from our data hospital a solution that includes OSS.
analysis (these are described under the core cat-
egory “Hospital technological factors” in Table 3). core category
As one manager commented “OSS is not going to
have the same level of quality and not nearly the The mix of causal conditions in a particular hos-
same level of documentation and rigor you can pital setting sets the stage for the “core category,”
get from a corporate environment.” Another IT that is, the hospital software vendors. While our
manager opined that “the majority of the OSS are survey did not address the issue of software ven-
probably of inferior quality because they are just dors, there was unanimous consensus amongst all
gifts that any research lab puts together and hands the hospital IT managers interviewed that hospital
out from a couple graduate students.” Managers software vendors play a pivotal role in the adoption
also perceive OSS as a high-risk product when it process of OSS in hospitals, as discussed in the
concerns security. As one manager commented, “It interview results. IT managers identify hospital
is not the fact that the OSS won’t be able to provide software vendors who supply OSS products and
the functionality that we need in the hospital. The services as the key facilitators for the adoption
major concern is going to be how secure OSS is.” of both general-purpose and domain-specific
Managers perceive OSS to be highly vulnerable OSS products. In terms of the empirical model
to attacks from hackers or other parties, which presented in Figure 1, the actions of the software
may inhibit them from adopting OSS, even from vendors is the trigger, or the gateway, that creates
a vendor. These quality and security factors will the situation where a hospital must decide to adopt
color a hospital’s openness to a vendor’s offer of or not adopt OSS. Such a decision does not even
an open source solution. arise except through the actions of a software
The lack of accountability of OSS providers vendor, according to the findings of this study.
is also a concern for the hospital IT managers As one manager commented, “hospitals are so
we interviewed. Having a vendor that can be dependent on vendors of hospital IT products
held liable or accountable if there is inadequate that we are not in the position to kind of ‘buck the
or insufficient quality or security of the software rules’ and go it alone for the adoption of OSS.”

185
 An Empirical Investigation into the Adoption of Open Source Software in Hospitals

Figure 1. Empirical model for the adoption of OSS in hospitals

Contextual factors

· Hospital type
· Hospital size
· Hospital IT budget
· Hospital organizational
culture
· Hospital organizational
structure
Causal conditions Consequences

· Lack of in-house · Reducing software

Core category Strategic actions
software development development and
· Lack of IT personnel implementation costs
· Lack of medical · Hospital software · Adoption of OSS
· Avoiding vendor
informaticians vendors
lock-in
· Lack of general quality · Promoting common
· Lack of security data standards
· Lack of liability of OSS · Increasing software
· Satisfaction level with quality
software vendors · Increasing security
Intervening conditions

· Patient-privacy
protection and privacy
legislation
· Type of health care
system
· International
development of OSS

However, sometimes this decision is not even & Corbin, 1998, p. 132). Our data, especially the
explicit. As one IT manager adopter of OSS survey data presented in “Survey Results”, reveal
expressed, “we don’t have a conscious decision that several contextual factors are expected to
to adopt OSS because our hospital outsources moderate the adoption of OSS in hospitals. The
a lot of our technical knowledge to vendors, so combined qualitative and quantitative results of
the adoption of OSS is coming throughout the this study provide evidence that the following
vendor’s decisions for the most part.” contextual factors may facilitate or inhibit the
The hospitals’ decision to adopt OSS from adoption of OSS in hospitals: 1) hospital type, 2)
software vendors is linked to their belief that the hospital size, 3) hospital IT budget, 4) hospital
OSS offered this way has “a professional level of organizational culture, and finally 5) hospital
quality control” that is greater than the OSS avail- organizational structure. These factors are dif-
able from other sources, such as the Internet. As ferent from the causal conditions listed earlier,
one IT manager who adopted vendor-supported in that they are more general, static factors that
OSS stated, “I am very happy using OSS because, apply to the hospital as a whole and do not specifi-
for me, the best of two worlds is when vendors cally form the hospital’s attitude towards OSS, or
support an OSS solution. I am willing to pay for towards the software vendor.
OSS, because I feel I have professional quality Depending on the hospital type (such as a stand-
and control over the software.” alone hospital versus a multi-hospital network, or
a university hospital versus a private hospital, and
contextual Factors so on), the importance of IT adoption within the
hospital may differ. Different types of hospitals
Contextual factors are the “specific set of condi- seem to have different requirements to adopt
tions (patterns of conditions) that intersect di- software. For example, a university hospital may
mensionally at this time and place to create a set allow experimentation with new software products
of circumstances or problems to which persons while a private hospital in a multi-hospital network
respond through actions/interactions” (Strauss may not allow any type of experimentation. Such

186
An Empirical Investigation into the Adoption of Open Source Software in Hospitals

factors may have an effect on the adoption of OSS because they perceived OSS as posing a threat to
by hospitals. Hospital size is likely to be related patients’ privacy and confidentiality as well as to
to organizational characteristics such as slack in HIPAA compliance mandates. Consequently, we
resources or a large professional workforce that conclude that the aforementioned three interven-
can also have a positive effect on the adoption of ing conditions also mediate the adoption of OSS
OSS in hospitals. Hospital IT budget is another in hospitals.
contextual factor that emerged in our study as a
meaningful factor since hospitals with smaller strategic actions
relative IT budgets (with 3% or less of the total
hospital budget) have a propensity to adopt OSS. Strategic actions are “purposeful or deliberated
Other contextual factors within the hospital acts that are taken to resolve a specific problem”
such as organizational culture and organizational (Strauss & Corbin, 1998, p. 133). The interaction
structure can also have an effect on the adoption outcome of the core category (hospital software
of OSS. As one manager commented, “the or- vendors) with the contextual factors and the in-
ganizational design of the hospitals has a major tervening conditions may result in a decision by
influence on the adoption of software within hospitals to make full use of a technology—in this
the hospital, I don’t want to use the word power case OSS—as a plausible or implausible alterna-
structure, but it is almost the political landscape tive to proprietary (closed source) or commercial
of the organization that influences the way we software products.
adopt any technology.”
Our findings support the effect that all the consequences
aforementioned factors have on the strategic ac-
tions (i.e., adoption or non-adoption) as depicted Consequences are the outcomes of the interaction
in Figure 1 with regards to OSS adoption within of the core category with the contextual factors,
hospitals. intervening conditions and the strategic actions.
The outcomes of this empirical model are closely
Intervening conditions aligned with the potential benefits of OSS claimed
in the literature reviewed in literature section
Intervening conditions are those conditions that of this article. However, we can only speculate
“mitigate or otherwise impact causal conditions” about the actual consequences, as that part of
(Strauss & Corbin, 1998, p. 131). The intervening the model is beyond the scope and objectives of
conditions identified in this study included: 1) this research. However, investigating the conse-
patient-privacy protection and privacy legisla- quences of OSS adoption in hospitals is a vital
tion, 2) type of healthcare system, and 3) inter- area for future research.
national development of OSS. These intervening
conditions are factors, external to the immediate
hospital setting that may inhibit the adoption of IMplIcatIons
OSS in hospitals. IT managers we interviewed
(see Table 3) report that factors such as patient- Implications for the literature
privacy protection and privacy legislation may act
as deterrents for the adoption of OSS in general, OSS has created a stir of interest in many disci-
especially with regards to the domain-specific OSS plines ranging from computer science to sociology,
products. For example, hospital IT managers were and a growing body of literature has emerged to
reluctant to adopt domain-specific OSS products explain many aspects of OSS. However, no work

187
 An Empirical Investigation into the Adoption of Open Source Software in Hospitals

has investigated the adoption of OSS in hospitals. as a critical factor in how OSS would be used in
The research presented here addresses this gap. the future, even before getting to the technology
A number of respondents from the interviews portion of the adoption of OSS by hospitals.
noted that the lack of IT personnel and the lack of This study also shows that the hospital industry
medical informaticians are inhibiting factors for is a very conservative industry when it concerns
adoption of OSS by hospitals. This is consistent adopting new technologies. Managers repeatedly
with previous authors (Yackel 2001; MacDonald indicated the “conservative aspects and risk ad-
et al., 2003; Fitzgerald & Kenny, 2004; Hogarth verse” behavior of the hospital industry to adopt
& Turner, 2005; Waring & Maddocks, 2005) not only OSS but also any new technology. This
who have noted the importance of IT personnel finding is consistent with MacDonald et Al. (2003)
with high levels of technical expertise required and Glynn (2005) who also pointed out hospitals’
in order to deal with OSS applications and the risk averse behavior when adopting IT.
technological complexity in the medical domain Finally, our core finding about the central
that needs personnel that understand both medi- role of software vendors in the adoption deci-
cine and information systems. sion in hospitals has some relationship to prior
In contrast to other studies claiming that the literature. Some existing studies have indicated
reduction of ownership and development cost that avoiding vendor lock-in is perceived to be an
is one of the main advantages of adopting OSS advantage of adopting open source (Carr, 2003;
in healthcare (Yackel, 2001; Kantor et al., 2003; 2004; Fink, 2003; Kantor et al., 2003; Fitzgerald,
Fitzgerald & Kenny, 2004; Glynn et al., 2005; Hog- 2004; Goldman & Gabriel, 2005; Goode, 2005).
arth & Turner, 2005; Holck, Larsen, & Pedersen, In contrast, in our study, the role of vendors
2005), the findings from this research indicated emerged quite differently. The role of vendors
that cost factors are not a core, important category as OSS adopters, who then transfer their adop-
for hospital IT managers when deciding to adopt tion decisions on to their client hospitals, has not
OSS. The IT managers in our study were found previously been described in the literature. This
to be more concerned about the quality, security finding describes vendors as innovating the way
and liability issues surrounding OSS than about they develop, distribute, support and maintain
the potential cost-benefit factors associated with software systems within hospitals. Prior studies
the adoption and use of OSS. This finding also have not shown software vendors to be such key
compares with a prior study by Goode (2005), enablers of OSS in the hospital industry.
which also noted that managers see software with
high cost as an indicator of quality. Implications for Future research
Prior research (Fitzgerald & Kenny, 2004;
Glynn et al. 2005; Holck et al. 2005; Waring & This research is unique within the field of OSS
Maddocks, 2005) has noted the importance of top and healthcare. That is, there is no study that has
management support for the successful adoption been published to date presenting an empirical
of technology within organizations. Our findings, model for the adoption of OSS in hospitals. This
by contrast, show that not only top management model, grounded in empirical data collected from
support is important to the adoption of OSS by surveys and interviews, identifies the factors and
hospitals, but clinical personnel within hospitals relationships facilitating and inhibiting the adop-
(e.g., physicians, nurses, etc.) also exert a signifi- tion of OSS in hospitals. This model provides the
cant influence on the decision to adopt not only basis for future testing of the interactions among
OSS but any technology. Many IT managers the key concepts proposed in this study. Further-
recognized the political influence of these groups more, there are numerous significant issues for

188
An Empirical Investigation into the Adoption of Open Source Software in Hospitals

researchers, including ourselves. Our findings, 3. Further empirical investigation into the rela-
while not highly generalizable due to the limita- tionship between hospital software vendors
tions of the study, provide sufficient grounding and adoption of OSS.
for future confirmatory studies.
In particular, a number of very interesting Implications for practice
propositions or hypotheses are suggested by our
empirical model, and by the survey and interview The present research provides a better understand-
data. Future research aimed at validating these ing to hospital IT managers and practitioners
hypotheses would be a significant contribution to about the extent of OSS adoption in hospitals
the field. Examples of such propositions include: in conjunction with the factors facilitating or
inhibiting this adoption process. Hospitals and
• Proposition: Adoption of OSS is more likely healthcare organizations that are considering
to be found in hospitals that have in-house the adoption and implementation of OSS tech-
technical staff with experience in software nologies need to understand how technological,
development, OSS, and/or with medical environmental and organizational factors affect
informatics. the adoption process. This way IT hospital practi-
• Proposition: Hospitals with an existing re- tioners can prepare against the expected barriers
lationship with a software vendor who offers and can utilize the facilitators for successful OSS
OSS solutions are more likely to adopt OSS. technology adoption.
The likelihood increases with the degree of The first implication for practitioners is that,
dependence on the vendor and the degree of contrary to theoretical and anecdotal expectations
satisfaction with the vendor. about the cost-benefit advantages of OSS versus
• Proposition: The adoption of OSS in a proprietary or commercial-based software, the
hospital is more likely when there is a cen- findings from this research indicated that financial
tralized IT strategy within the hospital. factors are not deemed to be a core concern for
• Proposition: The likelihood of a hospital’s IT managers when deciding to adopt OSS. The
adoption of OSS is negatively correlated with IT managers in our study were found to be more
the IT manager’s perception of the general concerned about the quality, security and liability
quality and security of OSS products. issues surrounding OSS. This implies that, when
building a business case, or justification, for the
To further validate propositions such as those adoption of OSS, the analysis must take into
above, as well as the whole empirical framework account issues related to quality, security, and
derived from this study, the following future accountability with at least as much prominence
research is planned: as cost-benefit issues.
The second implication for hospital IT prac-
1. Validation of the model by collecting data titioners would be to involve all the stakeholders
from a large sample of hospitals, either in within the hospital in the adoption decision-
the U.S. and/or internationally, would allow making; for this particular point, our finding
for further conclusions about the causal indicated that physicians, nurses, and other clinical
relationships and interactions suggested by personnel are key stakeholders to address in the
our empirical model. adoption process of not only OSS but any type
2. A case study in a hospital setting to ana- of technology introduced to a hospital. Thus the
lyze the consequences of the adoption and receptivity to the idea and philosophy of OSS
implementation of OSS. must be assessed with these stakeholders, and

189
 An Empirical Investigation into the Adoption of Open Source Software in Hospitals

any ideas and concerns that might surface during industry. Therefore, such partnerships could be a
the assessment must be documented and taken potentially transforming development in promot-
into account. ing and adopting OSS in hospitals.
The third recommendation for hospitals that
are considering OSS is that they can start adopt-
ing OSS with a small pilot project in order to lIMItatIons oF tHe study
test and experiment with the quality issues of
interest, as well as the costs and benefits, of OSS Notwithstanding the important contributions of
to the hospital. In addition, it is very important the current study, it has its own shortcomings.
to collect data and metrics from the pilot project For example, our findings may not apply to the
and communicate the results to all the stakehold- full spectrum of U.S. hospitals. This research is
ers, including vendors, within the hospital. It is exploratory in nature, so that the design, data
important to mention that OSS is not “free,” and collection methods, and analysis were broad by
never will be without a cost. design, and not intended for confirmation. This re-
Another implication for practitioners who want search also examined the adoption or non-adoption
to promote the use of OSS within the hospital of OSS in a limited geographical area and over a
and healthcare industry is for them to liaise with particular time period, which makes any attempts
hospital software vendors and the OSS com- to generalize the results across hospitals in the
munity. Coordinating with hospital IT vendors U.S. difficult without further empirical analysis
is important because, as our findings reported, and investigation.
any tendency towards adoption of OSS in hospi- Another limitation was the modest sample
tals is occurring because healthcare IT vendors size of response in the survey (n=30) and in-
are embracing, providing, and maintaining OSS terviews (n=5). Through the evolution of this
products. Under this business model, hospital study, it became clear that IT managers in the
software vendors are not only offering the software hospital industry in the BWNV area were less
to hospitals but also offering services for instal- than enthusiastic about discussing and sharing
lation, customization, and maintenance of OSS information about open source adoption within
applications, either domain-specific or general- their hospitals. Many attempts to influence a
purpose. Furthermore, there are good examples higher rate of response and interview participation
of software partnerships amongst IT businesses, were made, including initial contacts, follow-up
open source communities, and researchers such as contacts, reminders, and even financial incentives.
Eclipse and even Linux (Capek, 2005; Goldman & While the small sample size affects the ability to
Gabriel, 2005; Zeller & Krinke, 2005) that can be generalize results, it does not affect what was the
replicated in the hospital and healthcare industry. intent of the study, to explore and identify relevant
Moreover, the hospital industry is probably the issues and factors for further study. However, it
most influential and powerful industry operating is important to mention that these are important
today in the healthcare area. If this industry sees limitations for any future similar study because
the benefits from OSS, then partnerships between of the unwillingness of the managers and execu-
IT businesses, OSS communities, and universities tives to share their views on issues concerning
could result in research, development and promo- IT adoption.
tion of OSS hospital products and policies that Finally, another limitation of this research is
further the evolution of the OSS movement, as that the data appears not to represent all types of
well as provide substantial benefits to the hospital OSS products. While it was not the intent of the

190
An Empirical Investigation into the Adoption of Open Source Software in Hospitals

study, it is clear from the responses (in particular reFerences

the types of OSS that survey respondents report
adopting) that our respondents were referring pri- Bazeley, P., & Richards, L. (2000). The NVivo
marily to large, enterprise-level OSS applications qualitative project book. London; Thousand Oaks,
(e.g., database servers, Web servers, operating CA: Sage Publications.
systems, etc.). This limits our ability to extend our
Brown, S., Lincoln, M., Groen, P., & Kolodner,
findings to the entire spectrum of OSS products
R. (2003). VistA—U.S. Department of Veterans
available to hospitals. It also limits our ability to
Affairs national-scale HIS. International Journal
compare the results of this study to prior research,
of Medical Informatics, 69(2-3), 135-156.
which mostly addresses the adoption of smaller,
stand-alone, download-and-install types of OSS Capek, P., Frank, S., Gerdt, S., & Shields. D. (2005).
applications. A history of IBM’s open source involvement and
strategy. IBM Systems Journal, 44(2), 249-248.
Carnall, D. (2000). Medical software’s free future.
conclusIon
BMJ, 321(7267), 976.
This research identifies the factors that could lead Carr, N. (2003). IT doesn’t matter. Harvard Busi-
to more effective adoption of OSS by hospitals. In ness Review, 81(5), 41-49.
addition, this research sheds light and broadens the
Carr, N. (2004). Does IT matter?: Information
understanding of OSS adoption within hospitals
technology and the corrosion of competitive ad-
by offering to IT practitioners information on the
vantage. Boston, MA: Harvard Business School
extent of that adoption currently.
Press.
Finally, the insight gained from this research
serves as a guide and foundation for future work Charmaz, K. (2006). Constructing grounded
to investigate more determinants of OSS adop- theory. London; Thousand Oaks, CA: Sage
tion in hospitals and healthcare organizations. It Publications.
is also the researcher’s hope that this study will
Chau, P., & Tam, K. (1997). Factors affecting
be the seminal stone to pave the way for future
the adoption of open systems: An exploratory
studies on OSS adoption and implementation in
study. MIS Quarterly: Management Information
organizations, public and private, national and
Systems, 21(1), 1-20.
international.
Creswell, J. (1994). Research design: Qualitative
& quantitative approaches. Thousand Oaks, CA:
acKnowledgMent Sage Publications.
Creswell, J. (2005). Educational research: Plan-
The authors would like to thank Khaled El Emam,
ning, conducting, and evaluating quantitative
Medha Umarji, Roy Rada, Katherine Stewart
and qualitative research (2nd ed.). Upper Saddle
Dongsong Zhang and Stephen Russell, as well as
River, NJ: Merrill.
the anonymous IJHISI reviewers for their edito-
rial insights. Additionally, we deeply appreciate Dedrick, J., & West, J. (2003). Why firms adopt
the time, interest, and participation of hospital open source platforms: A grounded theory of
IT managers in the Baltimore-Washington area innovation and standards adoption. Paper pre-
surveyed and interviewed in this study. sented at the Proceedings of the Workshop on
Standard Making: A Critical Research Frontier
for Information Systems. Seattle, WA.

191
 An Empirical Investigation into the Adoption of Open Source Software in Hospitals

Dedrick, J., & West, J. (2004). An exploratory Glaser, B. (1999). The future of grounded theory.
study into open source platform adoption. Paper Qualitative Health Research, 9(6), 836-845.
presented at the Proceedings of the 37th Hawaii
Glaser, B., & Strauss, A. (1967). The discovery
International Conference on System Sciences.
of grounded theory; strategies for qualitative re-
DeLano, W. (2005). The case for open source search. Chicago, IL: Aldine Publishing Company.
software in drug discovery. Drug Discovery
Glynn, E., Fitzgerald, B., & Exton, C. (2005).
Today, 10(3).
Commercial adoption of open source software:
Erickson, B., Langer, S., & Nagy, P. (2005). An empirical study. Paper presented at the Pro-
The role of open source software in innovation ceedings of the 2005 International Symposium
and standardization in radiology. Journal of the on Empirical Software Engineering (ISESE’05).
American College of Radiology, 2(11), 927-931. Noosa Heads, QLD, Australia.
Eyler, A., Mayer, J., Rafii, R., Housemann, R., Goldman, R., & Gabriel, R. (2005). Innovation
Brownson, R., & King, A. (1999). Key informant happens elsewhere: Open source as business
surveys as a tool to implement and evaluate physi- strategy. Amsterdam; Boston, MA: Morgan
cal activity interventions in the community. Health Kaufmann.
Educ. Res., 14(2), 289-298.
Goode, S. (2005). Something for nothing: Man-
Eysenbach, G. (2004). Improving the quality of agement rejection of open source software in
Web surveys: The checklist for reporting results of Australia’s top firms. Information and Manage-
Internet e-surveys (CHERRIES). J Med Internet ment, 42(5), 669-681.
Res, 6(3), e34.
HIMSS. (2006). HIMSS bylaws. Retrieved June
Fichman, R., & Kemerer, C. (1997). Object tech- 1, 2006, from http://www.himss.org/content/files/
nology and reuse: Lessons from early adopters. himss_bylaws.pdf.
Computer, 30(10), 47-59.
Hogarth, M., & Turner, S. (2005). A study of clini-
Fink, M. (2003). The business and economics of cal related open source software projects. Paper
Linux and open source. Upper Saddle River, NJ: presented at the American Medical Informatics
Prentice Hall PTR. Association Conference (AMIA). Washington,
D.C.
Fitzgerald. (2004). A critical look at open source.
Computer, 37(7), 92-94. Holck, J., Larsen, M., & Pedersen, M. (2005).
Managerial and technical barriers to the adop-
Fitzgerald, & Kenny. (2004). Developing an infor-
tion of open source software. Berlin-Heidelberg:
mation systems infrastructure with open source
Springer.
software. Software, IEEE, 21(1), 50-55.
Huff, S., & Munro, M. (1985). Information tech-
Gatignon, H., & Robertson, T. (1989). Technology
nology assessment and adoption: A field study.
diffusion: An empirical test of competitive effects.
MIS Quarterly, (December), 327-338.
Journal of Marketing, 53(1), 35-49.
Kantor, G., Wilson, W., & Midgley, A. (2003).
Glaser, B. (1978). Theoretical sensitivity: Ad-
Open source software and the primary care EMR.
vances in the methodology of grounded theory.
J Am Med Inform Assoc, 10(6), 616.
Mill Valley, CA: Sociology Press.

192
An Empirical Investigation into the Adoption of Open Source Software in Hospitals

Longman, P. (2007). Best care anywhere: How Strauss, A., & Corbin, J. (1998). Basics of quali-
America’s most socialized healthcare system be- tative research: Techniques and procedures for
came its finest: What the VA miracle means for you developing grounded theory (2nd ed.). Thousand
and your country. Sausalito, CA: PoliPointPress. Oaks, CA: Sage Publications.
McDonald, C., Schadow, G., Barnes, M., Dexter, Tashakkori, A., & Teddlie, C. (1998). Mixed
P., Overhage, J., Mamlin, B., et al. (2003). Open methodology: Combining qualitative and quan-
source software in medical informatics—why, titative approaches. Thousand Oaks, CA: Sage
how and what. International Journal of Medical Publications.
Informatics, 69(2-3), 175-184.
Valdes, I., Kibbe, D., Tolleson, G., Kunik, M., &
Nagy, P. (2007). Open source in imaging infor- Petersen, L. (2004). Barriers to proliferation of
matics. Journal of Digital Imaging, 20(0), 1-10. electronic medical records. Informatics in Primary
Care, 12(1), 3-9.
Norris, J. (2004). Mission-critical development
with open source software: Lessons learned. Van Latum, F., Van Solingen, R., Oivo, M., Hoisl,
Software, IEEE, 21(1), 42-49. B., Rombach, D., & Ruhe, G. (1998). Adopting
GQM-based measurement in an industrial envi-
Patton, M. (2002). Qualitative research and
ronment. Software, IEEE, 15(1), 78-86.
evaluation methods (3rd ed.). Thousand Oaks,
CA: Sage Publications. Waring, T., & Maddocks, P. (2005). Open source
software implementation in the UK public sec-
Raghupathi, W., & Gao, W. (2007). An eclipse-
tor: Evidence from the field and implications for
based development approach to health information
the future. International Journal of Information
technology. International Journal of Electronic
Management, 25(5), 411-428.
Healthcare, 3(4), 433-452.
Yackel, T. (2001). How the open source devel-
Seaman, C. B. (1999). Qualiative methods in
opment model can improve medical software.
empircal studies of software engineering. Soft-
Medinfo, 10, 68-72.
ware Engineering, IEEE Transactions on, 25(4),
557-572. Zeller, A., & Krinke, J. (2005). Essential open
source toolset: Programming with Eclipse, JUnit,
Scarsbrook, A. (2007). Open source software
CVS, Bugzilla, Ant, Tcl/Tk and more. Chichester,
for radiologists: A primer. Clinical Radiology,
England: John Wiley & Sons.
62(2), 120-130.

This work was previously published in International Journal of Healthcare Information Systems and Informatics, Vol. 3, Is-
sue 3, edited by J. Tan, pp. 16-37, copyright 2008 by IGI Publishing (an imprint of IGI Global).

193
 An Empirical Investigation into the Adoption of Open Source Software in Hospitals

appendIx a. surVey InstruMent

http://userpages.umbc.edu/~gimunoz1/Appendix%20A-%20Survey%20Instrument.pdf

appendIx B. cHerrIes

http://userpages.umbc.edu/~gimunoz1/Appendix%20B-%20CHERRIES.pdf

appendIx c. InterVIew guIde protocol

http://userpages.umbc.edu/~gimunoz1/Appendix%20C-%20Interview%20Guide.pdf

194
 195

Chapter 14
Intelligent Agent Framework for
Secure Patient-Doctor Profiling
and Profile Matching
Masoud Mohammadian
University of Canberra, Australia

Ric Jentzsch
Compucat Research Pty Limited, Australia

aBstract

Radio frequency identification (RFID) is a promising technology for improving services and reduction
of cost in health care. Accurate almost real time data acquisition and analysis of patient data and the
ability to update such a data is a way to improve patient’s care and reduce cost in health care systems.
This article employs wireless radio frequency identification technology to acquire patient data and in-
tegrates wireless technology for fast data acquisition and transmission, while maintaining the security
and privacy issues. An intelligent agent framework is proposed to assist in managing patients’ health
care data in a hospital environment. A data classification method based on fuzzy logic is proposed and
developed to improve the data security and privacy of data collected and propagated.

IntroductIon nard, D. 2007]. This chapter research considers

the use of RFIDs and its potential in hospitals and
Research into the use of developing and evolving similar environments. Furthermore RFIDs are
technologies needs to be expanded in order that used to collect data at its source while developing
society as a whole can benefit. Radio Frequency profiles for patients and their care. There are four
Identifiers (RFID) have been around for many areas where using RFIDs and their data collection
years. Their use and projected use has only begun can have significant positive effects in hospitals.
to be researched in hospitals [Fuhrer, P. and Gui- These four areas are:

Copyright © 2011, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 Intelligent Agent Framework for Secure Patient-Doctor Profiling and Profile Matching

• Care tracking: this is getting the right care RFIDs are used in hospitals for tracking high-
to the right patient at the right time. value assets and setting up automated maintenance
• Quality of care: improving the services routines to improve operational efficiencies.
given to the right patient at the right time However the use of RFIDs in tracking beds and
in a timely manner. tracking mobile equipment is in its infancy. RFIDs
• Cost of care: finding ways to be effective is used to monitor equipment for example how
in the use of available resources such that long a bed was used at a particular location to
the cost per patient per incident does not ad- determine a sterilization schedule as well as bed
versely increase to the cost of the resources. location tracking.
• Service of care: better, more timely informa- However RFID technology is already being
tion for a more informed decision making deployed across the pharmaceutical industry to
process, to provide more knowledgeable combat drug counterfeiting, drugs shelf life track-
individual tailored care. ing [Kowalke, M. 2006]. Managing expensive,
often difficult to replace, and legal drugs can only
RFID tags and readers are most commonly are be improved using RFIDs.
associated with tracking goods in manufacturing The management of patients and their condi-
and warehousing, but hospitals are starting to ap- tion is paramount in a hospital. RFIDs can assist
ply RFID to new purposes [Kowalke, M. 2006]. in asset and personnel tracking, patient care, and
RFID technology does not require contact or billing where unnecessary expenses will be cut,
line of sight for communication, like bar codes. the average length of stay of a patient is reduced,
RFID data can be read through the human body, where more patient lives will be saved due to timely
through clothing, read wirelessly, and through efficient services, and where patient records are
non-metallic materials. actively continuously updated to provide better
Both research and practical application of patient care. [Kowalke, M. 2006].
the use of RFIDs in hospitals continues to be An RFID chip stores the wearer’s data that can
of importance. For hospitals this has meant the be accessed by a hand-held reader. This makes
potential of managing inventories in a more ef- patient identification more reliable, provides
ficient manner. Inventories in hospitals take on a updated patient condition nearly instantly, and
variety of differences than to manufacturing. The improves the cost of health care.
nature of the inventory and assets in a hospital can The health sector is already taking up people-
include various types of equipment (that is often tagging where it allows nurses to radio their loca-
very expensive, comes in many sizes, and uses), tion if they are being assaulted, reduce mother baby
drugs (that come in a variety of sizes, shapes, color, mismatches and baby theft, help severe diabetics
and governing regulations), beds, chairs, patients with getting correct treatment, and monitoring
(the primary reason hospitals exist), and staff. disoriented elderly patients without the need for
The percentage of worldwide radio frequency a dedicated member of staff [Tindal, S. 2008].
identification (RFID) projects concerning people- The need is not to keep track of staff but be
tagging has increased from eight percent to 11 able to locate the staff with the particular skills
percent since 2005 [Tindal, S. 2008]. However, that are needed at the right time and place. Staff
the healthcare sector has yet to quantify or provide wearing badges with RFIDs embedded can be
evidence of the benefit to people-tagging. Human found to help provide that needed and timely
chipping is not new but does bring up a lot of care that a patient may need. However privacy
ethical questions [Angeles, R. 2007]. concerns have been aired over patient tracking
using RFIDs.

196
Intelligent Agent Framework for Secure Patient-Doctor Profiling and Profile Matching

A patient upon arrival to a hospital can be patients’ health care data in a hospital environ-
is issued with an RFID tag which can contain ment. A fuzzy data classification system is also
information concerning the patient such as their developed to improve the application of regulatory
surname, first name, reason admitted to hospital, data requirements for security and privacy of data
date of admission, their doctor’s name, a patient exchange. The chapter is divided into four main
number and a section for monitoring. Monitoring sections. Next section considers issues relate to
could include heart rate, blood pressure, and some data collection and profiling. Section three is based
other vital signs. Monitoring would be settable on the patient to doctor profiling and intelligent
to the need of the patient. For example once an software agents. The fourth section covers RFID
hour might be sufficient for most patients, but background and provides a good description of
for others every 15 minutes might be sufficient. RFIDs and their components. This section dis-
This illustration shows how a particular section cusses several practical cases of RFID technol-
of a hospital might be configured according to the ogy in and around hospitals. It will also list three
needs of that section. Each patient has a patient- possible applicable cases assisting in managing
tag. Each patient’s bed has a bed-tag. Spaced out patients’ medical data. The final section discusses
within rooms, hall ways, and hospital staff stations the important issue of maintaining patients’ data
are receivers. Every 15 minutes the receiver inter- security and integrity and relates that to RFIDs.
rogates the bed and the patient tags. The patient’s
vital signs are sent to the patient database where
the patient’s condition is recorded. The patient data collectIon
care profile is then updated with this informa-
tion. If anything is out of range or an exception is Large amount of health care data such as patients,
identified the nearest nurse station to the patient doctors, nurses, institution itself, drugs and pre-
is then contacted. scriptions, diagnosis, and many other areas is
There is a need for more research into appli- collected and stored in hospitals. It is not feasible
cations and innovative architectures for secure or effective to use RFID to collect and retrieve
access, retrieval and update of data in healthcare such large amount of data. This chapter concen-
systems [Finkenzeller. K. (1999), Glover. B and trates on a subset with the understanding that all
Bhatt H. (2006), Hedgepeth W. O. (2007). La- areas could, directly or indirectly, benefit from
hiri, S. (2005), Schuster E. W., Allen S. J. and the use of RFID and intelligent software agents
Brock D. L. (2007). Shepard S. (2005), Angeles, in a health care environment.
R. (2007), Pramatari, K.C., Doukidis, G.I. and The RFID [Bhuptani, M., & Moradpour, S.
Kourouthanassis, P. (2005), Qiu R, Sangwan R. (2005)] provides the passive vehicle to obtain the
(2005), Mickey, K. (2004), Whiting, R. (2004), data via its monitoring capabilities. The intelligent
Weinstein, R. (2005)]. Although many organiza- software agent provides the active vehicle in the
tions are developing and testing the possible use interpretation profiling of the data and reporting
of RFIDs the real value of RFID is achieved in capacity. By investigating and analyzing collect
conjunction with the use of intelligent software patient data the patient’s condition can be moni-
agents for processing and monitoring data ob- tored and abnormal situations can be reported on
tained via RFIDs. Thus the issue becomes the time. Using this information an evolving profile
integration of these two great technologies for the of each patient can be constructed and analysed.
benefit of assisting health care services. Analyzing the data can assist in deciding what kind
This article considers a framework using RFID of care a patient requires, the effects of ongoing
and Intelligent Software Agents for managing care, and how to best care for this patient using

197
 Intelligent Agent Framework for Secure Patient-Doctor Profiling and Profile Matching

available resources (doctors, nurses, beds, etc…) and patient. A static profile is kept in pre-fixed
for the patient. The intelligent software agent data fields where the period between data field
builds a profile of each patient as they are admit- updates is long such as months or years. A dynamic
ted to the healthcare institution by analyzing the profile is constantly updated as per evaluation of
recorded and stored data about each patient. The the situation in which the situation occurs. The
same way a profile for each doctor is developed updates may be performed manually or automated.
based on stored data about each doctor. Therefore The automated user profile building is especially
patients and doctors profile can be correlated to important in real time decision-making systems.
obtain the specialization and availability of the Real time systems are dynamic. The profiling
doctors to suit the patients. of patient doctor model is based on the patient /
doctor information. These are:
patient profiling
• The categories and subcategories of doctor
Profiling is combined with personalization, and specialization and categorization. These cat-
user modeling [Wooldridge, M. and Jennings, N. egories will assist in information processing
(1995)]. The use of profile in hospitals and health- and patient / doctor matching.
care so far has been limited. Tracking of informa- • Part of the patients profile based on their
tion about consumers’ interests by monitoring symptoms (past history problems, dietary
their movements online is considered profiling restrictions, etc.) can assist in prediction of
or user modeling in e-commerce systems. By the patients needs specifically.
analyzing the content, URL’s, and other informa- • The patients profile can be matched with the
tion about a user’s browsing path/click-stream a available doctor profiles to provide doctors
profile of a user behavior is constructed. However with information about the arrival of patients
patient profiling differ from user profiling in e- as well as presentation of the patients profile
commerce systems. The patient profiling is use- to a suitable, available doctor.
ful in a variety of situations such as providing a
personalized service based on the patient and not A value denoting the degree of association
on symptoms or illness to a particular patient as can be created form the above evaluation of the
well as assisting in identifying the medical facili- doctor to patient’s profile. The intelligent agent
ties in trying to prevent the need for the patient to based on the denoting degrees and appropriate,
return to the hospital any sooner than necessary. available doctors can be identified and be allocated
Patient profiling also assist in matching a doctor’s to the patient.
specialization to the right patient. A patient profile In the patient / doctor profiling the agent will
can also assist in providing information about the make distinctions in attribute values of the pro-
patient on continuous bases for the doctors so that files and match the profiles with highest value. It
a tailored and appropriate care can be provided should be noted that the agent creates the patient
to the patient. and doctor profiles based on data obtained from
the doctors and patient namely:

patIent to doctor proFIlIng • Explicit profiling occurs based on the data

entered by hospital staff about a patient.
A patient or doctor profile is a collection of infor- • Implicit profiling can fill that gap for the
mation that can be used in a decision analysis situ- missing data by acquiring knowledge about
ation between the doctor, domain environment, the patient from its past visit or other relevant

198
Intelligent Agent Framework for Secure Patient-Doctor Profiling and Profile Matching

Figure 1. Agent profiling model using RFID

patient data Patient staff data

database Staff database

RFID
patient data Patient profile Staff profile

Patient Profile Staff Profile

Agent Engine Agent Engine

Rule Base Rule Base

Profile Agent Engine

Rule Base

Matching Profile Agent

databases if any and then combining all these Profile matching [Doan, A-H. Lu, y. Lee, T.
data to fill the missing data. Using legacy Han, J (2003)] performed is based on a vector
data for complementing and updating the of weighted attributes using an intelligent agent
user profile seems to be a better choice than system. To get this vector, the intelligent agent
implicit profiling. This approach capitalizes uses a rule-based system to match the patient’s
on user’s personal history (previous data attributes (stored in patient’s profile) against doc-
from previous visit to doctor or hospital). tor’s attributes (stored in doctor’s profile). If there
is a partial or full match between them then the
The proposed agent architecture allows user doctor will be informed (based on their availability
profiling and matching in such a time intensive from the hospital doctor database). Such a rules
important application. The architecture of the based system is built based on the knowledge
agent profiling systems using RFID is given in of domain experts. This expert system is exten-
Figure 1. sible as new domain knowledge can be added
Profile matching done is based on a vector to its knowledge base as rules. Large amount of
of weighted attributes. To get this vector, a rule research in the area of profiling in e-commerce,
based systems can be used to match the patient’s schema matching, information extraction and
attributes (stored in patient’s profile) against doc- retrieval has been shown promising results [(Do,
tor’s attributes (stored in doctor’s profile). If there H. Rahin, E. (2002), Doan, A-H. Lu, y. Lee, T.
is a partial or full match between them then the Han, J (2003)]. However profiling in healthcare
doctor will be informed (based on their availability is new and innovative.
from the hospital doctor database).

199
 Intelligent Agent Framework for Secure Patient-Doctor Profiling and Profile Matching

Staff and patient / doctor profiling and profile about availability of the doctors are obtained and
matching could be the missing link in providing updated and the profiling agent continuously
more tailored healthcare professionals and facility checks such information from the staff database.
to patients in a hospital environment. The matched doctors then can be ranked as: “high”,
Profile matching may consist of: “medium” and “low”. A fuzzy rule the profile
matching system then may look like:
• determining the matching algorithms re-
quired for matching patient / doctor profile, IF patient_doctor_profile is total match and doctor_availability
• determining the availability of staff and is highly available Then doctor_ranking = high
facilities required for a given patient.
• understanding of government policies related The integration of RFID capabilities and
patient healthcare, intelligent agent techniques provides promising
development in the areas of performance improve-
Another issue related to patient / doctor profil- ments in RFID data collection, inference and
ing is defining the level of matching of the patient knowledge acquisition and profiling operations.
and doctor profile. There is not always possible to Due to the important role of intelligent agents
provide the services of the doctor/s and facilities in this system, it is recognized that there is a need
identified as exact match for a patient healthcare for a framework to coordinate intelligent agents
because the matching doctor may be unavailable so that they can perform their task efficiently.
or unreachable. Some guidelines include issues Intelligent agent coordination [Wooldridge, M.
such as the critical nature of the patient illness, and Jennings, N. (1995). Odell, J. and Bigus, J. P.
its level of sensitivity and regulatory rules. and Bigus, J. (1998). Shaalana, K. El-Badryb,M.
As such the rules that govern patient / doc- and Rafeac, A. (2004)] has shown to be promis-
tor profile matching can be expressed in human ing. The Agent Language Mediated Activity
linguistic terms which can be vague and difficult Model (ALMA) agent architecture currently
to represent formally. Fuzzy Logic (Zadeh, L. A., under research is based on the mediated activity
1965) has been found to be useful in its ability to framework. We believe that such a framework is
handle vagueness. As such the profiling patient able to provide RFID with the necessary frame-
/ doctor matching is based on fuzzy logic. The work to profile a range of internal and external
profiling matching system consists of a fuzzy medical/patient profiling communication activi-
rule based system uses and inference engine to ties performed by wireless multi-agents.
a weighted value between a patient profile and
doctor/s profile.
The matching between a patient profile and rFId descrIptIon
doctor/s is then divided into the following classes:
“total match”, “medium match”, “low match” and RFID or Radio Frequency Identification is a pro-
“no match”. Based on these class categories a gressive technology that has been said to be easy
weighted match of patient / doctor profile can be to use and well suited for collaboration with intel-
identified. The doctors then can be categorized ligent software agents. Basically an RFID can:
and ranked based on the matching profile value.
The doctors can be classified into classes based • be read-only;
on their matching profile as well as their avail- • volatile read/write; or
ability such as “highly available”, “more and less • write once / read many times
available” and “not available”. Of course the data • RFID are:

200
Intelligent Agent Framework for Secure Patient-Doctor Profiling and Profile Matching

• non-contact; and L. (2007). Shepard S. (2005)]. A basic RFID system

• non-line-of-sight operations. consists of four components namely, the RFID
tag (sometimes referred to as the transponder),
Being non-contact and non-line-of-sight will a coiled antenna, a radio frequency transceiver
make RFIDs able to function under a variety of and some type of reader for the data collection.
environmental conditions and while still provid-
ing a high level of data integrity [Finkenzeller. K. transponders
(1999), Glover. B and Bhatt H. (2006), Hedgepeth
W. O. (2007). Lahiri, S. (2005), Schuster E. W., The reader emits radio waves in ranges of any-
Allen S. J. and Brock D. L. (2007). Shepard S. where from 2.54 centimeters to 33 meters. De-
(2005)]. Next section will discuss the environment pending upon the reader’s power output and the
that RFIDs operate in and their relationship to radio frequency used and if a booster is added
other available wireless technologies such as the that distance can be increased. When RFID tags
IEEE 802.11b, IEEE 802.11g, IEEE 802.11n etc… (transponders) pass through a specifically created
in order to fulfill their requirements effectively electromagnetic zone, they detect the reader’s
and efficiently. activation signal. Transponders can be on-line
RFID or Radio Frequency Identification is a or off-line and electronically programmed with
progressive technology that has been said to be unique information for a specific application or
easy to use and well suited for collaboration with purpose. A reader decodes the data encoded on
intelligent software agents. Basically an RFID the tag’s integrated circuit and passes the data to
can be read-only, volatile read/write; or write once a server for data storage or further processing.
/ read many times. RFID are non-contact; and There are four major frequency ranges that
non-line-of-sight operations. Being non-contact RFID systems operate at. As a rule of thumb,
and non-line-of-sight will make RFIDs able to low-frequency systems are distinguished by short
function under a variety of environmental condi- reading ranges, slow read speeds, and lower cost.
tions and while still providing a high level of data Higher-frequency RFID systems are used where
integrity [Finkenzeller. K. (1999), Glover. B and longer read ranges and fast reading speeds are
Bhatt H. (2006), Hedgepeth W. O. (2007). Lahiri, required, such as for vehicle tracking, automated
S. (2005), Schuster E. W., Allen S. J. and Brock D. toll collection, asset management, and tracking
of mobile equipment.

Table 3. Frequency ranges for RFID systems

Frequency Range Applications

Low-frequency 3 feet Pet and ranch animal identification; car key locks
125 - 148 KHz
High-frequency 3 feet library book identification; clothing identification; smart cards
13.56 MHz
Ultra-high freq 25 feet Supply chain tracking: Box, pallet, container, trailer tracking
915 MHz
Microwave: 100 feet Highway toll collection; vehicle fleet identification
2.45GHz

201
 Intelligent Agent Framework for Secure Patient-Doctor Profiling and Profile Matching

coiled antenna read ranges, than active tags, and require a higher-
powered reader.
The coiled antenna is used to emit radio signals
to activate the tag and read or write data to it. Hospital environment
Antennas are the conduits between the tag and
the transceiver that controls the system’s data In a hospital environment, in order to manage
acquisition and communication. RFID antennas patient medical data we need both types; fixed
are available in many shapes and sizes. They can and handheld transceivers. Also, transceivers can
be built into a doorframe, book binding, DVD be assembled in ceilings, walls, or doorframes
case, mounted on a tollbooth, embedded into a to collect and disseminate data. Hospitals have
manufactured item such as a shaver or software become large complex environments. In a hospital
case (just about anything) so that the receiver nurses and physicians can retrieve the patient’s
tags the data from things passing through its zone medical data stored in transponders (RFID tags)
[Finkenzeller. K. (1999), Glover. B and Bhatt H. before they stand beside a patient’s bed or as they
(2006), Hedgepeth W. O. (2007). Lahiri, S. (2005), are entering a ward.
Schuster E. W., Allen S. J. and Brock D. L. (2007). Given the descriptions of the two types and
Shepard S. (2005).]. Often the antenna is pack- their potential use in hospital patient data man-
aged with the transceiver and decoder to become agement we suggest that:
a reader. The decoder device can be configured
either as a handheld or a fixed-mounted device. • It would be most useful to embed a passive
RFID transponder into a patient’s hospital
types of rFId transponders wrist band;
• It would be most useful to embed a passive
RFID tags can be categorized as active, semi- RFID transponder into a patient’s medical
active, or passive. Each has and is being used file;
in a variety of inventory management and data
collection applications today. The condition of Doctors should have PDAs equipped with
the application, place and use determines the RFID or some type of personal area network de-
required tag type. vice. Either would enable them to retrieve some
Active RFID tags are powered by an internal patient’s information whenever they are near the
battery and are typically read / write. Tag data patient, instead of waiting until the medical data
can be rewritten and / or modified as the need is pushed to them through the hospital server.
dictates [Finkenzeller. K. (1999), Glover. B and After examining both ranges for Active and
Bhatt H. (2006)]. The semi-active tag comes Passive RFID tags, we can suggest the following:
with a battery. The battery is used to power the
tags circuitry and not to communicate with the • Low frequency range tags are suitable for
reader [Shepard S. (2005)]. Passive RFID tags the patients’ band wrist RFID tags. Since
operate without a separate external power source we expect that the patients’ bed will not be
and obtain operating power generated from the too far from a RFID reader. The reader
reader. Passive tags, since they have no power might be fixed over the patient’s bed, in
source embedded in themselves, are consequently the bed itself, or over the door-frame. The
much lighter than active tags, less expensive, and doctor using his/her PDA would be aiming
offer a virtually unlimited operational lifetime. to read the patient’s data directly and within
However, the trade off is that they have shorter a relatively short distance.

202
Intelligent Agent Framework for Secure Patient-Doctor Profiling and Profile Matching

• High frequency range tags are suitable for there it is a short distance to be transmitted to
the physician’s tag implanted in their PDAs. the doctor’s PDA, a laptop, or desktop through a
As physicians move from one location to WLAN or wired LAN.
another in the hospital, data on their patients The “Hand Down Polling” techniques as pre-
could be continuously being updated. viously described, provides the ability to detect
all detectable RFID tags at once (i.e. in parallel).
transceivers Preventing any unwanted delay in transmitting
medical data corresponding to each RF tagged
The transceivers / interrogators can differ quite patient. Transponder programmers are the means,
considerably in complexity, depending upon the by which data is delivered to write once, read
type of tags being supported and the application. T\ many (WORM) and read/write tags. Program-
he overall function of the application is to provide ming can be carried out off-line or on-line. For
the means of communicating with the tags and some systems re-programming may be carried
facilitating data transfer. Functions performed by out on-line, particularly if it is being used as an
the reader may include quite sophisticated signal interactive portable data file within a production
conditioning, parity error checking and correc- environment, for example. Data may need to be
tion. Once the signal from a transponder has recorded during each process. Removing the
been correctly received and decoded, algorithms transponder at the end of each process to read the
may be applied to decide whether the signal is a previous process data, and to program the new
repeat transmission, and may then instruct the data, would naturally increase process time and
transponder to cease transmitting or temporarily would detract substantially from the intended
cease asking for data from the transponder. This flexibility of the application. By combining the
is known as the “Command Response Protocol” functions of a transceiver and a programmer, data
and is used to circumvent the problem of read- may be appended or altered in the transponder
ing multiple tags over a short time frame. Using as required, without compromising the produc-
interrogators in this way is sometimes referred to tion line.
as “Hands Down Polling”. An alternative, more It can be concluded from this section that RFID
secure, but slower tag polling technique is called systems differ in type, shape, and range; depending
“Hands Up Polling.” This involves the transceiver on the type of application, the RFID components
looking for tags with specific identities, and inter- shall be chosen. Low frequency range tags are suit-
rogating them in turn. able for the patients’ band wrist RFID tags. Since
Hospital patient data management deals we expect that the patients’ bed not to be too far
with sensitive and critical information (patient’s from the RFID reader, which might be fixed on
medical data). Hands Down polling techniques the room ceiling or door-frame. High frequency
in conjunction with multiple transceivers that range tags are suitable for the physician’s PDA tag.
are multiplexed with each other, form a wireless As physicians move from on location to another in
network. The reason behind this choice is that, the hospital, long read ranges are required. On the
we need high speed for transferring medical data other hand, transceivers which deal with sensitive
from medical equipment to or from the RFID and critical information (patient’s medical data)
wristband tag to the nearest RFID reader then need the Hands Down polling techniques. These
through a wireless network or a network of RFID multiple transceivers should be multiplexed with
transceivers or LANs to the hospital server. From each other forming a wireless network.

203
 Intelligent Agent Framework for Secure Patient-Doctor Profiling and Profile Matching

applications of the rFId technology ii. To an on-line patient monitoring unit or

in a Hospital a nurse’s workstation within the hospital.
iii. Or the acquired patients’ medical data
The following section describes steps involved in can be fed into an expert (intelligent
the process of using RFID in hospital environment agent) software system running on the
for patient information management: hospital server and to be then compared
with other previously stored abnormal
1. A biomedical device equipped with an em- patterns of medical data, and to raise an
bedded RFID transceiver and programmer alarm if any abnormality is discovered.
will detect and measure the biological state 4. Another option could be using the in-built-
of a patient. This medical data can be an embedded RFID transceiver in the biomedi-
ECG, EEG, BP, sugar level, temperature cal device to send the acquired medical data
or any other biomedical reading. After the wirelessly to the nearest RFID transceiver in
acquisition of the required medical data, the the room. Then the data will travel simulta-
biomedical device will write this data to the neously in a network of RFID transceivers
RFID transceiver’s EEPROM using the built until reaching the hospital server.
in RFID programmer. Then the RFID trans- 5. If a specific surgeon or physician is needed
ceiver with its antenna will be used to transmit in a specific hospital department, the medi-
the stored medical data in the EEPROM to cal staff in the monitoring unit (e.g. nurses)
the EEPROM in the patient’s transponder can query the hospital server for the nearest
(tag) which is around his/her wrist. The data available doctor to the patient’s location.
received will be updated periodically once In our framework an intelligent agent can
new fresh readings are available by the bio- perform this task. The hospital server traces
medical device. Hence, the newly sent data all doctors’ locations in the hospital through
by the RFID transceiver will be updated detecting the presences of their wireless
(and may be accumulated as needed) to the mobile device; e.g. PDA, tablet PC or laptop
old data in the tag. The purpose of the data in the WLAN range. Physicians may also
stored in the patient’s tag is to make it easy use RFID transceivers built-in the doctor’s
for the doctor to obtain medical information wireless mobile device.
regarding the patient directly via the doctor’s 6. Once the required physician is located, an
PDA, tablet PC or laptop. alert message will be sent to his\her PDA,
2. Similarly, the biomedical device will also tablet PC or laptop indicating the location
transfer the measured medical data wire- to be reached immediately including a brief
lessly to the nearest WLAN access point. description of the patient’s case.
Since high data rate transfer rate is crucial 7. The doctor enters into the patient’s room
in transferring medical data, IEEE 802.11b or ward according to the alert he/she has
or g is recommended for the transmission received. The doctor wants to check the
purpose. medical status of a certain patient and inter-
3. Then the wirelessly sent data will be routed rogates the patient’s RFID wrist tag with his
to the hospitals main server; to be then sent RFID transceiver equipped in his\her PDA,
(pushed) to: tablet PC or laptop, etc.
i. Other doctors available throughout the
hospital so they can be notified of any
newly received medical data.

204
Intelligent Agent Framework for Secure Patient-Doctor Profiling and Profile Matching

classIFIcatIon oF data For that encrypted data for security professionals or

encryptIon other designated individuals within an organiza-
tion. Since managing the keys for encrypted data
Data security, availability, privacy and integrity can become difficult many large organizations
[McGraw, G. (2006)] are of paramount important choose to encrypt their regulated data. Once
in healthcare and hospital environment. Data data is transmitted wirelessly, security becomes
security and privacy policies in healthcare are a more crucial issue. Unlike wired transmission,
governed by hospital, medical requirements and wirelessly transmitted data can be easily sniffed
government regulations. These requirements out leaving the transmitted data vulnerable to
demand not only for data security but also for many types of attacks. For example, wireless
data accessibility and integrity. data could be easily eavesdropped on using any
Implementing data security using data en- mobile device equipped with a wireless card. In
cryption solutions remain at the forefront for worst cases wirelessly transmitted data could be
data security. Data encryption algorithms are intercepted and then possibly tampered with, or
implemented to protect the actual data. However in best cases, the patient’s security and privacy
data is stored and transferred through several would be compromised. Hence emerges the need
devices and simply protection of data by data for data to be initially encrypted from the source.
encryption fails to secure the resource on which Two main layers of encryption can be used
it is stored or transferred. On the other hand when using RFID’s in hospital environment, they
the issue of keys and overall process of data are, Physical (hardware) layer encryption and Ap-
encryption process remains complex. The best plication (software) layer encryption. This means
data encryption solutions are those that balance encrypting all collected medical data at the source
information protection with on-demand access to or hardware level before transmitting it. Thus,

Figure 2. Sample patient-doctor relational database with and associate fuzzy set

205
 Intelligent Agent Framework for Secure Patient-Doctor Profiling and Profile Matching

we insure that the patient’s medical data would jor difficulty for many organizations as it is an
not be compromised once exposed to the outer expensive and time consuming task.
world on its way to its destination. So even if a Classification of and data process may consist
person with a malicious intent and also possessing of:
a wireless mobile device steps into the coverage
range of the hospitals’ WLAN, this intruder will • Determining the business and corporate ob-
gain actually nothing since all medical data is jectives and the level of protection required
encrypted, making all intercepted data worth- for data in that organization (e.g. security
less. In application (software) layer encryption measures required, intellectual property
all collected medical data at the destination or protection, strategic use of data, privacy
application level is encrypted once it is received. policies),
Application level encryption runs on the doc- • Understanding of government policies for
tor’s wireless mobile device (e.g. PDA, tablet PC data protection and accessibility,
or laptop) and on the hospital server. Once the • Determining the corporate vales of the data
medical data is received, it will be protected by and its sensitivity (private business critical
a secret pass-phrase (encryption\decryption key) data, internal usage of data, public release
created by the doctor who possesses this device. of data),
This type of encryption would prevent any per- • Determining who needs access to the data
son from accessing patient’s medical data if the (e.g. user security level),
doctor’s wireless mobile device gets lost, or even • Determining the processes that manipulate
if a hacker hacks into the hospital server via the the data (internal processes, external applica-
Internet, intranet or some other mean. tions, mix internal-external applications),
Sensitive and mission critical data are stored in • Determine the life time of the data and issues
databases, in server applications, and middleware. related to storage, backups and removal of
However many solutions to data encryption at this data (tape, sever, outsourcing), and
level are expensive, disruptive, and demanding • Determining the level of protection required
intensive resource. Using a data classification for the audience that may view the organiza-
process organizations can identify and encrypt tion’s data.
only the relevant data thereby saving time and
processing power. Without data classification Another issue related to data classification is
organizations using encryption process would defining the level of classification for data. There
simply encrypt everything and consequently is no exact and firm rule on the level of classifica-
impacting users more than necessary (Cline, J., tions. Some guidelines include issues such as the
2007; Butterfield, R., 2007). data type, its level of sensitivity and corporate
Understanding the value of the data is signifi- objective and regulatory rules.
cant information for an organization in determin- Such data classification will be different for
ing and deploying the proper data classification, different organizations based on policies of each
security and risk assessment. Data classification organization and government regulatory polices.
is essential and can assist organizations with As such policies are expressed in human under-
their data security, privacy and accessibility standable language and are vague and difficult to
needs. Such a classification process needs to be represent formally. The excessive gap between
able to determine the value, sensitivity, privacy, precision of classic logic and imprecision and
government regulations and corporate strategic vagueness in definition of polices creates difficulty
objectives. However data classification is a ma- in representing this policies in formal logic. Fuzzy

206
Intelligent Agent Framework for Secure Patient-Doctor Profiling and Profile Matching

Logic [Zadeh, L. A. (1965).] has been found to as described below. For example consider the
be useful in its ability to handle vagueness. In following entities of a relational database system:
this article a data classification method based on
fuzzy logic [Zadeh, L. A. (1965)] is presented to Patient(PatientID, Name, Address, TelNo, InsuranceID)
determine data classification levels for data in an Insurance(InsuranceID, Type, InsuranceProviderID)
organization. The level of sensitivity and corporate InsuranceProvider(InsuranceProviderID, Name, Address,
objective and regulatory rules are determined TelNo, FaxNo)
using this classification method. Doctor(DoctorID, Name, OfficeNo, TelNo, PagerNo)
Classification levels could divide data into PatientDoctor(PatientDoctorID, PatientID, DoctorID, Visit-
classes such as “top secret”, “secret”, “confiden- Date, Notes)
tial”, “mission critical”, “not critical”, “private but
not top secret”, and “public”. Based on these class Adding meta-data values can then be used for
categories the business processes and individuals adaptation and implementation of classification
that access and use the data and the level of encryp- of data in databases for an organization. The
tion can be identified. The users can be categorized meta-data values can be can be obtained from
to determine access to any of these data classes. the knowledge workers of the organization based
The users can be classified into classes based on on organization policies, procedure and business
need-to-know such as “very high”, “high”, “me- rules as well as government requirements for
dium” and “low” users. The need for encryption data privacy and security. For example Table 3
level of the data can also be determined to be shows the metadata value related to security of
high, medium, zero (not necessary). attributes of table Patient based on organization’s
To classify data with minimal resources impact security policy and government security and pri-
and without needing to re-design databases one vacy policy. The values are in the range of 0 to
option is to add extra information to each data 70, where zero indicates the meta-data for a data
item by adding meta-data information to the at- item that is public and 70 indicates the meta-data
tributes of each entity in relational-data bases and for a data item that is top secret (note that other
domains in classes in object-oriented databases. meta data values are also possible and for this
These meta-data information could be the application we have chosen between values zero
value or degree of security, privacy or other to seventy).
related policies for that data item. This can be Now assume that the following domain meta-
demonstrated using a simple relational database data values for these linguistic variable, TP =
top secret, SE = “secret”, CO =“confidential”,

Table 2.

Formulas for calculation triangular fuzzy Formulas for calculation trapezoidal fuzzy
memberships memberships
m A ( x ) = 0 , x < a1 m A ( x ) = 0 , x < a1
x - a1
x - a1 mA ( x ) = , a1 ≤ x ≤ a2
mA ( x ) = , a1 ≤ x ≤ a2 a2 - a1
a2 - a1
m A ( x ) = 1 , a2 ≤ x ≤ a3
a -x
mA ( x ) = 3 , a2 ≤ x ≤ a3 a4 - x
a3 - a2 mA ( x ) = , a3 ≤ x ≤ a4
a4 - a3
m A ( x ) = 0 , x > a3
m A ( x ) = 0 , x > a4

207
 Intelligent Agent Framework for Secure Patient-Doctor Profiling and Profile Matching

Table 3. Metadata values for table customer

Patient Table Meta-data Value base on organization Meta-data Value base on government
policy regulatory policy
PatientID 68 39
Name 64 70
Address 30 60
TelNo 44 68

MC = “mission critical”, NC = “not critical”, PR of the attribute PatientID to fuzzy set security
= “private but not top secret”, PU = “Public”. classification based on meta-data from Table 3
Assume that the linguistic terms describing the can then be calculated as shwon in Figure 3.
meta-data for the attributes of entities in the above Now that the data can be classified and catego-
database has are: TP = [58,..,70], SE = [48,..,60], rized into fuzzy sets (with membership value), a
CO =[37,..,50], MC = [28,..,40], NC = [16,..,30], process for determining precise actions to be ap-
PR = [8,..,20], PU = [0,..,10]. plied must be developed. This task involves writing
Based on the metadata value for each attribute a rule set that provides an action for any data clas-
the membership of that attribute to each data sification that could possibly exist. The formation
classification can be calculated. In the Figure 3 of the rule set is comparable to that of an expert
triangular and trapezoidal fuzzy set was used to system, except that the rules incorporate linguistic
represent the data security classifications (e.g. Data variables with which human are comfortable. We
security classification levels: TP = “top secret”, SE write fuzzy rules as antecedent-consequent pairs
= “secret”, CO =“confidential”, MC = “mission of If-Then statements. For example:
critical”, NC = “not critical”, PR = “private but
not top secret”, PU = “Public”). The membership IF Organizational_Security_Classification is TopSecret and
value of PatientID based on its meta-data can be Government_Security_Classification is Confidential
calculated for all these classification using the Then Level of Encryption is High
formulas in Table 2.
Where x is metadata value for the attribute The overall fuzzy output is derived by apply-
PatientID and α1, α2 and α3 are the lower middle and ing the “max” operation to the qualified fuzzy
upper bound values of the fuzzy set data security outputs each of which is equal to the minimum
classification. The degree of membership value of the firing strength and the output membership

Figure 3. Fuzzy membership of metadata value of PatientID based on

TP SE CO MC NC PR PU
μ(PatientID) 0.8 0 0 0 0 0 0
(a) Organization policy

TP SE CO MC NC PR PU
μ(PatientID) 0 0 0.3 0.16 0 0 0
(b) government regulatory policy

208
Intelligent Agent Framework for Secure Patient-Doctor Profiling and Profile Matching

Figure 4. Center of gravity inference method

A A B
1 2 1

w
1
X Y Z
A A
34 B
2

w
2
X Y Z
y intersec tion Max
x
or min

z (Centroid of area)

function for each rule. Various schemes have been Future trends
proposed to choose the final crisp output based
on the overall fuzzy output. In this article a type RFID in medical environment is an innovative
of inference method called centre of gravity and and applicable idea. Linking RFID’s and wireless
illustrated in Figure 4. technologies will provide the required informa-
Securing medial data seems to be uncom- tion to achieve timely services to patients as fast
plicated, yet the main danger of compromising as possible. It also will pave the way for future
such data comes from the people managing it, paperless hospitals.
e.g. doctors, nurses and other medical staff. For RFID technology has many potential impor-
that, it is noted that even though the transmitted tant applications in hospitals. With the progress
medical data is classified and encrypted, doctors the RFID technology is currently gaining, it seems
have to run application level encryption on their to become a standard as other wireless technolo-
wireless mobile devices in order to protect this gies, and eventually manufacturers building them
important data if the devices gets lost, left behind, in electronic devices; biomedical devices with
robbed, etc. Nevertheless, there is a compromise. reduced cost.
Increasing security through using multiple lay-
ers, and increasing length of encryption keys
decreases the encryption\decryption speed and conclusIon
causes unwanted time delays, whether we were
using application or hardware level of encryption. Managing patients’ data wirelessly can prevent er-
As a result, this could delay medical data sent to rors, enforce standards, make staff more efficient,
doctors or on-line monitoring units. simplify record keeping and improve patient care.
This research in the wireless medical environment

209
 Intelligent Agent Framework for Secure Patient-Doctor Profiling and Profile Matching

introduces new ideas in conjunction to what is Bigus, J. P. and Bigus, J. (1998). Constructing
already available in the RFID technology and Intelligent software agents with Java – A Pro-
wireless networks. Linking both technologies grammers Guide to Smarter Applications, Wiley,
with each other to achieve the research main goal, ISBN: 0-471-19135-3.
delivering patients medical data as fast and secure
Butterfield, R. (2007). “Data classification: A
as possible, to pave the way for future paperless
prerequisite to ILM”, http://www.snwonline.com/
hospitals.
implement/data_class_05-30-05.asp
With the reduction in cost of radio frequency
identification (RFID) technology, it is expected the Cline, J. (2007). “Growing pressure for data
increased use of RFID technology in healthcare classification”, http://www.computerworld.com/
in monitoring patients and assisting in health care action/article.do?articleId=9014071&command
administration. A fuzzy logic based for profile =viewArticleBasic, Last accessed on 22/11/2007
matching and data classification system is also
Do, H. Rahin, E. (2002), “Coma: A System for
developed to improve the application of regula-
Flexible Combination of Schema Matching Ap-
tory data requirements for security and privacy
proaches”, Proceedings of 28th Conference on
of data exchange.
Very Large Databases, Morgan Kaufmann, pp
Finally it should be noted that the world of
610-621USA.
RFIDs could be frustrating. Often potential us-
ers have to wade through product information to Doan, A-H. Lu, y. Lee, T. Han, J (2003), “Profile-
select the right RFID technology for their needs. Based Object Matching for Information Integra-
The terminology, concepts, and uses seem to be tion”, IEEE Intelligent Systems Magazine, pp
vendor driver rather than potential user driver. 54-59. USA.
The RFID vendor websites often describe how
Finkenzeller. K. (1999), RFID Handbook, John
and where to use RFIDs using arcane and unfa-
Wiley and Sons Ltd, USA.
miliar language.
Fuhrer, P. and Guinard, D. (2007). Building a
Smart Hospital using RFID technologies: Use
acKnowledgMent Cases and Implementation. rfidehealth.pdf
Glover. B and Bhatt H. (2006), RFID Essentials,
The author would like to acknowledge the initial
O’Reilly Media, Inc.USA, (ISBN: 10-0596009445)
research work performed on this project at Uni-
versity of Canberra by MIT students. Gruber, T.R. (1993) A translation approach to
portable ontology specifications. Knowledge
Acquisition, 5:199-220.
reFerences
Hedgepeth W. O. (2007). RFID metrics: decision
making tools for today’s supply chains.Boca
Angeles, R. (2007), An empirical study of the
Raton, FL : CRC Press, (ISBN: 9780849379796)
anticipated consumer response to RFID product
item tagging, Industrial Management & Data Kowalke, M. (2006) TMCnet. Wireless Mobility
Systems, Vol. 107 No. 4, pp. 461-583. Blog. (2006). RFID vs. WiFi for Hospital Inven-
tory Tracking Systems. http://blog.tmcnet.com/
Bhuptani, M., & Moradpour, S. (2005). RFID Field
wireless-mobility/rfid-vs-wifi-for-hospital-in-
Guide - Deploying Radio Frequency Identification
ventory-tracking-systems.asp, October 23, 2006.
Systems. NJ: Prentice Hall.

210
Intelligent Agent Framework for Secure Patient-Doctor Profiling and Profile Matching

Lahiri, S. (2005), RFID Sourcebook, IBM Press, Schuster E. W., Allen S. J. and Brock D. L. (2007)
USA, (ISBN: 10-0131851373 ) Last accessed on Global RFID : the value of the EPCglobal network
22/11/2007 for supply chain management. Berlin ; New York
: Springer, (ISBN: 9783540356547).
Lewan, T. (2007). ZDNet Australia. Human RFID
chipping invading the US?. 23 July 2007. Shaalana, K. El-Badryb,M. and Rafeac, A. (2004)
, “A multiagent approach for diagnostic expert
McGraw, G. (2006), “Software Security”, Addi-
systems via the internet”, Expert Systems with
son-Wesley, USA.
Applications, Volume No. 27 Issue No. 1, Elsevier
Messmer, E. (2006). Network World. Hos- Publishing, USA pp1-10.
pital finds a bloody good RFID application.
Shepard S. (2005). RFID : radio frequency
http://www.computerworld.com.au/index.php/
identification, New York : McGraw-Hill, (ISBN:
id;2117167769;fp;4;fpid;18.
0071442995).
Mickey, K. (2004), RFID grew in 2002. Traffic
Tindal, S. (2008). ZDNet Australia.. One in ten RFID
World 5: 20-21.
projects tag humans. http://www.zdnet.com.au/
Odell, J. (Ed.), (2000). Agent Technology, OMG news/hardware/soa/RFID-people-tagging-ben-
Document 00-09-01, OMG Agents interest Group, efits-health-sector-/0,130061702,339285273,00.
September 2000 htm. 21 January 2008.
Pramatari, K.C., Doukidis, G.I. and Kouroutha- Weinstein, R. (2005). RFID: A Technical Over-
nassis, P. (2005), “Towards ‘smarter’ supply and view and Its Application to the Enterprise. IT
demand-chain collaboration practices enabled by Professional Magazine, 7(3), 27-33.Whiting, R.
RFID technology”, in Vervest, P., Van Heck, E., (2004), MIT = RFID + Rx. Information Week
Preiss, K. and Pau, L.F. (Eds), Smart Business 988: 16.
Networks, Springer, Germany, pp. 197-210.
Wooldridge, M. and Jennings, N. (1995). Intel-
Qiu R, Sangwan R. (2005), Toward collabora- ligent software agents: Theory and Practice. The
tive supply chains using RFID. CIO Wisdom II, Knowledge Engineering Review. vol. 10, no 2,
more best practices. New Jersey: Prentice-Hall; pp 115-152, USA.
pp. 127–44.
Zadeh, L. A. (1965). “Fuzzy sets”, Information
and control, Vol. 8. pp 338-352.

211
 Intelligent Agent Framework for Secure Patient-Doctor Profiling and Profile Matching

appendIx

Fuzzy logic data classification Knowledgebase:

IF Organizational_Classification is Top Secret and Government_Classification is Top Secret Then Level of Encryption isl High
IF Organizational_Classification is Top Secret and Government_Classification is Secret Then Level of Encryption is High
IF Organizational_Classification is Top Secret and Government_Classification is Confidential Then Level of Encryption is High
IF Organizational_Classification is Top Secret and Government_Classification is Mission Critical Then Level of Encryp-
tion is High
IF Organizational_Classification is Top Secret and Government_Classification is Not Critical Then Level of Encryption is High
IF Organizational_Classification is Top Secret and Government_Classification is Private but not Top Secret Then Level of
Encryption is High
IF Organizational_Classification is Top Secret and Government_Classification is Public Then Level of Encryption is High
IF Organizational_Classification is Secret and Government_Classification is Top Secret Then Level of Encryption is High
IF Organizational_Classification is Secret and Government_Classification is Secret Then Level of Encryption is High
IF Organizational_Classification is Secret and Government_Classification is Confidential Then Level of Encryption is High
IF Organizational_Classification is Secret and Government_Classification is Mission Critical Then Level of Encryption is High
IF Organizational_Classification is Secret and Government_Classification is Not Critical Then Level of Encryption is High
IF Organizational_Classification is Secret and Government_Classification is Private but not Top Secret Then Level of
Encryption is High
IF Organizational_Classification is Secret and Government_Classification is Public Then Level of Encryption is High
IF Organizational_Classification is Mission Critical and Government_Classification is Top Secret Then Level of Encryp-
tion is High
IF Organizational_Classification is Mission Critical and Government_Classification is Secret Then Level of Encryption is High
IF Organizational_Classification is Mission Critical and Government_Classification is Confidential Then Level of Encryp-
tion is High
IF Organizational_Classification is Mission Critical and Government_Classification is Mission Critical Then Level of
Encryption is High
IF Organizational_Classification is Mission Critical and Government_Classification is Not Critical Then Level of Encryp-
tion is High
IF Organizational_Classification is Mission Critical and Government_Classification is Private but not Top Secret Then
Level of Encryption is High
IF Organizational_Classification is Mission Critical and Government_Classification is Public Then Level of Encryption is High
IF Organizational_Classification is Not Critical and Government_Classification is Top Secret Then Level of Encryption is High
IF Organizational_Classification is Not Critical and Government_Classification is Secret Then Level of Encryption is High
IF Organizational_Classification is Not Critical and Government_Classification is Confidential Then Level of Encryption
is High
IF Organizational_Classification is Not Critical and Government_Classification is Mission Critical Then Level of Encryp-
tion is High
IF Organizational_Classification is Not Critical and Government_Classification is Not Critical Then Level of Encryption
is Medium
IF Organizational_Classification is Not Critical and Government_Classification is Private but not Top Secret Then Level
of Encryption is Medium

212
Intelligent Agent Framework for Secure Patient-Doctor Profiling and Profile Matching

IF Organizational_Classification is Not Critical and Government_Classification is Public Then Level of Encryption is Zero
IF Organizational_Classification is Private but not top secret and Government_Classification is Top Secret Then Level of
Encryption is High
IF Organizational_Classification is Private but not top secret and Government_Classification is Secret Then Level of En-
cryption is High
IF Organizational_Classification is Private but not top secret and Government_Classification is Confidential Then Level
of Encryption is Medium
IF Organizational_Classification is Private but not top secret and Government_Classification is Mission Critical Then Level
of Encryption is Medium
IF Organizational_Classification is Private but not top secret and Government_Classification is Not Critical Then Level of
Encryption is Medium
IF Organizational_Classification is Private but not top secret and Government_Classification is Private but not Top Secret
Then Level of Encryption is Medium
IF Organizational_Classification is Private but not top secret and Government_Classification is Public Then Level of En-
cryption is Medium
IF Organizational_Classification is Public and Government_Classification is Top Secret Then Level of Encryption is High
IF Organizational_Classification is Public and Government_Classification is Secret Then Level of Encryption is High
IF Organizational_Classification is Public and Government_Classification is Confidential Then Level of Encryption is Medium
IF Organizational_Classification is Public and Government_Classification is Mission Critical Then Level of Encryption
is Medium
IF Organizational_Classification is Public and Government_Classification is Not Critical Then Level of Encryption is Zero
IF Organizational_Classification is Public and Government_Classification is Private but not Top Secret Then Level of
Encryption is Medium
IF Organizational_Classification is Public and Government_Classification is Public Then Level of Encryption is Zero

This work was previously published in International Journal of Healthcare Information Systems and Informatics, Vol. 3, Is-
sue 3, edited by J. Tan, pp. 38-57, copyright 2008 by IGI Publishing (an imprint of IGI Global).

213
214

Chapter 15
Towards a Conceptual
Framework of Adopting
Ubiquitous Technology in
Chronic Health Care
Jongtae Yu
Mississippi State University, USA

Chengqi Guo
James Madison University, USA

Mincheol Kim
Jeju National University, South Korea

aBstract
In the advent of pervasive computing technologies, the ubiquitous healthcare information system, or
U-health system, has emerged as an innovative avenue for many healthcare management issues. Drawing
upon practices in healthcare industry and conceptual developments in information systems research,
this paper aims to explain the latent relationships amongst user-oriented factors that lead to individual’s
adoption of the new technology. Specifically, this study focuses on the introduction of chronic disease
U-health system. Using the Ordinary Line Square (OLS) regression analysis, we are able to discover
the insights concerning which constructs affect service subscriber’s behavioral intention of use. Based
on the data collected from over 440 respondents, empirical evidences are presented to support that fac-
tors such as medical conditions, perceived need, consumer behavior, and effort expectancy significantly
influence the formation of usage intention.

IntroductIon nicating each other through wireless network in

actual activities of everyday life” (Weiser, 1993a;
Ubiquitous computing can be defined as the Weiser, 1993b). Some articulate such concept as
“contemplation of today’s computers commu- the embedded computers in walls, refrigerator,
tables, and objects in the surrounding environ-
DOI: 10.4018/978-1-61692-002-9.ch015
ment (Rhodes & Mase, 2006). In other words,

Copyright © 2011, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Towards a Conceptual Framework of Adopting Ubiquitous Technology in Chronic Health Care

computer is expected to become an ubiquitous million (in USD) in 2010 with 7 million subscrib-
resource, much similar to the light with a switch ers in their 30s~40s (Jee et. al., 2005).
and water with a tap. Two functionalities, comput- In summary, we define the U-Health system
ing tasks and telecommunications, are required to as the use of ubiquitous computing technologies
realize the features of ubiquitous computing such to support expeditious and personalized com-
as localized information, localized control, and munications, activities, and transactions between
resource management (Weiser, 1993a; Rhodes a medical service provider and its various stake-
& Mase, 2006). The evolving mobile technol- holders. In the literature review section, Figure
ogy has expanded the applicability of ubiquitous 3 describes the layout of U-health system and its
computing to areas including virtual reality, head associated stakeholders.
mounted display (HMD), wearable computing, Despite the strong potential of the technology,
and smart office room (Weiser, 1993b). however, the studies of ubiquitous healthcare ser-
An agile, responsive, and location-aware vices are not widely conducted. Existing literature
service delivery system is highly desired by the has focused mostly on pure technical concerns or
healthcare management business. Correspond- system development process; whereas manage-
ingly, ubiquitous computing allows patients to rial issues and behavior perspectives of U-health
receive prompt medical care anywhere (home, system application are largely overlooked. In this
office, outdoor, or hospital) and any time (24 hours paper, the authors discuss the issue of applying
/ 7 days), thus improving the service quality and ubiquitous computing technology to healthcare
decreasing the risk of medical treatment failures. management from end user’s perspective. The
For example, a doctor can check the status of a main goal of this paper is to identify and investi-
patient in a real time manner using the sensor gate the factors and their inter-relationships that
which is installed in patient’s home or attached affect end user’s intention of adopting U-Health
to the patient’s body. In case of emergency, the system. For example, one critical factor is effort
sensor can detect changes of the patient’s health expectancy, which refers to the degree to which
condition in the early stage and automatically the user believes it is easy to use the technology.
contact the designated hospital to initiate treatment Theoretically, the effort expectancy is positively
procedure. Moreover, medical expenses have been correlated with behavioral intention of use, name-
continuously surging in the past and present years ly, the easier the user finds to use the product, the
due to aging population and increasing demand more likely s/he will adopt it. A more thorough
for chronic disease management. Facing the soar- articulation of these factors can be found in the
ing healthcare cost, many countries attempt to Research Model and Hypotheses section.
encourage patients to leave hospitals as early as Realizing that the scope of healthcare dis-
possible and introduce medical services that are cussion can be extremely broad, we focus our
portable and effective for treatments. These efforts attention on chronic disease treatment in this
require treatment that delivers remote medical research. A main question to be addressed is what
service to patient’s own residence (Kang & Lee, factors influence end user’s perception about the
2007). Hence, a medical service utilizing ubiqui- U-health system and how these factors are related
tous healthcare technology offers an innovative with each other. Therefore, an investigation of
solution that allows patients to access medical people’s subjective perception is warranted. The
service whenever and wherever. In the sense or authors select South Korea as the place for data
market demand, taking South Korea as example, collection because: first, people in South Korea
the U-Health market is expected to be worth $150 have a high level of concern about health issues,
especially chronic diseases. Largely due to the

215
 Towards a Conceptual Framework of Adopting Ubiquitous Technology in Chronic Health Care

rapidly changing socio-economic structure, more build preliminary ground work for utilizing earlier
than 16% of total population is reported to suffer findings in the area of healthcare management to
from chronic diseases. According to a report by the guide our research activities.
Korean National Insurance Corporation (2007),
South Korea is among one of the top countries prior research in Healthcare
that are mostly concerned with chronic diseases; Management Information systems
second, South Korea has one of the highest mobile
technology penetration rates in the world, thus The evolution of HMIS can be summarized as a
providing a suitable environment for studying transferring process from isolation to integration
ubiquitous healthcare technology. In South Korea, (Briggs, Nunamaker, & Sprague, 2005). Initially
more than 85% of Koreans use at least one cel- installed in stand alone PCs, HMIS took shape
lular phone and subscribe wireless value added among segments of healthcare management
service (Shim, 2007). process, such as drug inventory and account
Regarding the organization of this study, the administration (Tan, 1995). Due to the soaring
authors first examine the current knowledge base in operational cost of healthcare and increasing need
both medical research and supportive technology for cross-boundary collaboration (Bandyopadhyay
aspects. Such literature review process enables us & Schkade, 2000), a true HMIS was introduced
to consolidate theoretical and practical findings so and adopted by organizations including the Health
that a research model is developed to cater for our Maintenance Organization (HMO). Information
core research issue – understanding the perceptions and communication technology provides health-
of chronic disease patient towards U-Health sys- care industry with capabilities to perform tasks that
tem. Next, we present the data analysis results that are deemed impossible in the past. For instance, a
empirically validate the proposed hypotheses. At medical center uses broadband video conferenc-
last, we conclude our findings and contributions. ing technology to enable a group of specialists
to examine the patient, who is located in a far
distanced hospital. Previous studies also indicate
lIterature reVIew that new interaction technologies utilizing human
movements may provide more flexible, naturalistic
It is widely recognized that the IT innovation has interfaces and support the ubiquitous or pervasive
historically and continually benefited healthcare computing paradigm (Abawajy, 2009).
business in numerous aspects. For instance, col- Practical implications – In pervasive comput-
laborative computer system allows healthcare ing environments the challenge is to create intuitive
workers to reengineer and streamline business and user-friendly interfaces. Application domains
processes within medical service life cycle by that may utilize human body movements as input
leveraging software and network applications. are surveyed here and the paper addresses issues
In the case of U-Health system implementation, such as culture, privacy, security and ethics raised
we need to consider both soft and hard factors. by movement of a user’s body-based interaction
Soft factors are subtle factors containing people’s styles.
experiences, feelings and perceptions; whereas Since technology adoption has been consid-
hard factors refer to the instantiations of comput- ered as one of the main driving forces behind the
ing technology (e.g. a HIPPA compliant patient growth of healthcare expenditures (Barros, Pinto,
accounts management system). Following such & Machado, 1999), numerous studies have been
idea, the purpose of reviewing previous literature done to investigate the factors influencing HMIS
is two fold: to establish the scope of this study; to adoption on the organization’s side (Berta, et al.,

216
Towards a Conceptual Framework of Adopting Ubiquitous Technology in Chronic Health Care

2005; Wong & Legnini, 2000). These studies make been set up concerning how PU can be adjusted
contributions in terms of designing and validating to apply technology acceptance theory under dif-
cost-effective procedures that facilitate system ferent research topics. A comprehensive discussion
adoption process and identifying environmental of research model construction is presented in the
factors that influence the success of systems adop- Research Model and Hypotheses section.
tion. On the other hand, the authors found that
extensive attention has been put on physician’s the u-Health system concept and
intention of using invested Information Technol- targeted problem domain
ogy (IT) products. However, further research is
needed to investigate the role of patient, who is Global companies including IBM, Intel and
the ultimate service subscriber, in the process of Google are investing in U-Health market while
healthcare system implementation (Hu, Chau, GE and Philips are leaders in the world’s medi-
Sheng, & Tam, 1999). Another viable lens for cal equipment market. As Table 1 indicates, the
explaining subscriber side information system U-Health sector in medical industry has become
usage is the Technology Adoption Model (TAM), a promising business. Leveraging computing
which was formalized by F.D. Davis in 1989. Davis devices that are available yet invisible in the
argued that the user’s intention of use is essentially physical environment, U-Health system is a built-
determined by two constructs: Perceived Useful- in application combining technologies, methods,
ness (PU) and Perceived Ease of Use (PEOU), as and procedures that aim to monitor, maintain,
seen in Figure 1. and improve individual’s health condition (Park,
Parsimonious and generalizable notwithstand- 2003). One salient feature of U-Health system is
ing, TAM becomes less robust in this research the anytime anywhere accessibility, which allows
context because it fails to address some unique the real time information of patient to be collected
dimensions of U-Health technology adoption. and then transmitted to medical organization for
Therefore, we need modifications (e.g. broken a cure or diagnosis (Yoo, 2006). With its surging
down PU) that customize TAM constructs so as
to accommodate the issue of chronic disease
healthcare management. For instance, PU can be Figure 2. Modifying PU (Davis, 1989)
articulated as the formative relationship between
Health Condition and Health Concern, as shown
in Figure 2. Patient’s awareness of his/her health
condition leads to health concern, which essen-
tially form the evaluation of PU of suggested
medical solutions. In this way, an example has

Figure 1. The TAM model (Davis, 1989)

217
 Towards a Conceptual Framework of Adopting Ubiquitous Technology in Chronic Health Care

Table 1. The movement of overseas companies for U-Health business (Kang & Lee, 2007)

Company U-Health Business Activities

Providing insurance companies and medical service suppliers with various U-Health solutions such as
IBM
remote monitoring and individual health measurement.
Investing on innovative health care technology including software and patents from Global Care Solutions
Microsoft
(GCS) located in Bangkok.
Sold its semiconductor business in 2006 and focused on healthcare and lifestyle market.
Philips Introduced its TV-based custom health care service (Motiva) for old patients who are not familiar with
computer network.
Created Digital Health department to enter digital health market of hospital computerization and home care
service in 2005.
Intel
Announced healthcare solution in joint with LG (Korea) to branch out into home health care service market
2007.
Cooperated with Cleveland Clinic (US) for a program by which patients keep their own medical informa-
Google
tion in Google account in 2008.

penetration rate in global range, personal mobile promising solution that enhances information
device such as cellular phone offers an important dissemination throughout the healthcare service
platform on which the U-Health system concept delivery cycle. A visual description of a typical
can be realized. The challenge, however, lies in U-Health system implementation is presented in
the seamless integration between mobile technol- Figure 3. Using wired and wireless network, U-
ogy and the existing data network infrastructure. Health system has the following advantages
The current global healthcare industry is con- (Korean National Insurance Corporation, 2007):
fronted with numerous issues. For instance,
modern e-health systems incorporate different • Delivering a time-efficient medical service.
healthcare providers in one system. Further, • Automating communications among
medical information is distributed and shared on stakeholders.
one common electronic platform, which contains • Real time detection of patient’s conditions.
large quantity of cross-context communications • Easy portability of patient’s historical
among stakeholders (Deng, Cock & Preneel, medical records and prescriptions.
2009). Since U-Health valuates medical treatments • High level of accessibility and flexibility
and information portability, we need to look mitigating the effects of time and place
deeper in a context oriented perspective. This constraints.
study hence puts a specific focus on the imple-
mentation of U-Health system in South Korea for In Figure 3, a sensor, which can be a patient’s
chronic disease patients, whose number is increas- cellular phone or wearable computing device,
ing steadily and raises health concerns of the measures the health status of patient and transmits
society. Treating chronic disease such as diabetes, the information to U-Health service provider,
tuberculosis, and anemia requires long term treat- who maintains a large collection of patients’
ing efforts and cares from both patient and records in database server. The patient’s status is
medical center. In addition, effective communica- then forwarded to medical personnel in hospital
tion between patients and medical service pro- for diagnose and possible treatment, which can
vider is critical in stabilizing patient’s condition be transmitted back to the patient over the same
(e.g. diabetes). Hence, U-Health system offers a network.

218
Towards a Conceptual Framework of Adopting Ubiquitous Technology in Chronic Health Care

Figure 3.The U-Health System

researcH Model condition is proactive to adopt healthcare service

and HypotHeses or activity that is believed to improve one’s health
condition (Lahiri & Xing, 2004). Further, health
Many of the contemporary health problems, espe- condition is considered a critical starting point in
cially chronic diseases, are mostly related with life the healthcare utilization process (Windmeijer &
style changes such as smoking or unhealthy dietary Santos Silva, 1997). Therefore, it is assumed that
pattern (Riska, 1982; McKeown, 1971). Medical negatively perceived health condition leads to
research model that is based on individualistic- more medical service usage or activity to improve
mechanistic notion of disease argues that an the health status.
individual’s willingness to improve his/her health When people perceive health problems, they
status results in the changes of his/her behavior and are likely to raise their health concerns about
thus promotes health status positively (Cohen & subsequent medical consequences. Such concern,
Cohen, 1978). Such willingness can be generated in turn, can lead to certain activities including
and motivated by the perceived personal health health information seeking, adherence to treat-
condition, namely, the medical concern. Individual ment, and interpretation of symptoms (Uskul &
who has a negative evaluation of his/her health Hynie, 2007). Earlier studies have argued that the

219
 Towards a Conceptual Framework of Adopting Ubiquitous Technology in Chronic Health Care

Figure 4.Mediating Effect of Medical Concern

Internet has become the new alternative source of manien, Elms, & Hallsworth, 2006; Chae, Black,
healthcare information (Madden, 2003; Eysenbach & Heitmeyer, 2006). The capability of identifying
& Köhler, 2003; Morahan-Martin, 2004). Other and satisfying the consumer’s needs basically
research has suggested that patients diagnosed with determines the success of promoting a certain
chronic diseases appear to be active in seeking the product or service (Lahiri & Xing, 2004). Several
medical information. Therefore, a high level of studies (e.g. Junginger et al., 2006; Zhang, Fang,
concern about health problem and consequence & Olivia, 2007) have provided strong evidence
leads to more active medical service usage. Hence, indicating the effects of consumer’s perceived
it is important to specify and validate the mediat- need on the purchase intention. On the other hand,
ing effect of medical concern, as shown in Figure such need is positively correlated with perceived
4, between medical condition and action taking. quality and perceived usefulness of the service or
Following the previous discussion, the authors product. Moreover, users accept new technology
contend that individuals who are actively involved when they expect certain performance improve-
with preventive or restorative healthcare measures ment in completing tasks. If people believe the
are more likely to use the U-health service. Such technology will be useful in increasing their job
argument can be further decomposed into the performance, they are more likely to adopt the
following hypotheses: technology (Davis, 1989; Venkatesh et al., 2003).
Hence, we hypothesize that:
H1: High level of medical concern leads to active
medical activity. H5: High level of perceived need for U-health
H2: Perceived medical condition leads to active system leads to behavioral intention of use.
medical activity.
H3: Perceived medical concern mediates the It has been established in previous IS adoption
relationship between medical condition and literature that the behavioral intention of use is
medical activity. affected, although not exclusively, by Effort Ex-
H4: Active medical activity leads to behavioral pectancy, which is defined as the degree of ease
intention of use. associated with the use of the system (Venkatesh
et al., 2003). The concept of effort expectancy is
Specifying the relation between consumer rooted in earlier constructs such as perceived ease
needs and purchase intention has been a major of use in Technology Adoption Model (TAM)
research topic across various fields (Bhaskaran (Davis, 1989) and perceived level of difficulty
& Hardley, 2002; Odom, Kumar, & Saunders, in Theory of Reasoned Action (TRA) (Ajzen
2002; Burke & Payton, 2006; Kervenoael, Soopra- & Fishbein, 1975). In our research model, we

220
Towards a Conceptual Framework of Adopting Ubiquitous Technology in Chronic Health Care

articulate effort expectancy in two dimensions: repurchase of the services or products (Gupta &
experience and consumer behavior. Several studies Stewart, 1996; Morgan & Rego, 2006). Hence,
have identified how significantly experience (e.g. the hypothesis of the inter-relationship between
learning) is related to job performance (Guido, behavioral intention of U-Health system use
2007; Ngwenyama, Guergachi, & McLaren, and customer recommendation is stated as the
2007). These results strongly support the argument following:
that experienced workers are more efficient and
more productive in performing tasks than less H9: High level of behavioral intention of use leads
experienced ones. Hence, it is inferred that expe- to recommendation.
rienced user of ubiquitous or mobile technology
are more likely to use the medical service offered Based on the above hypotheses, we propose
through U-health system, namely, the less time a our research model in Figure 5. In addition, we
user needs to learn how to use U-Health system summarize the previous hypotheses in Table 2.
indicates a lower level of adoption barrier.
Consumer behavior theory classifies the types
of consumers according to how early they pur- researcH MetHodology
chase the products. Innovators, or early adopters,
are willing to take perceived risks to adopt newly To test the hypotheses, the authors employ the
introduced technology. So it is expected that in- Ordinary Line Square (OLS) regression analysis
dividual who follows the behavior of innovator to pinpoint predictor variables’ contribution in
would adopt the U-health service earlier than the explaining the variances of dependent variable –
follower or late adopter. Some researchers have behavioral intention of use, which is also a media-
identified certain correlation between the early tor variable that describes how recommendation
adoption behavior and business success on the effects will occur. In academia, it has always been a
organizational level. For instance, Hendricks and fundamental dilemma for social researchers when
his colleagues (2007) point out that early adopter attempting to maximize the three dimensions of
firms report higher profitability improvements of research methodology: generalizability, precision,
ERP systems than late ERP adopter firms. More- and realism. In other words, one cannot increase
over, Smith (2006) suggests that large companies one of these three features without reducing the
are inclined to be the early adopter of mobile other one or two (McGrath, 1982). Therefore, it
commerce and thus reap the first mover benefits. is up to researchers to justify their selection of
To sum up, we hypothesized that: research process that has the best goodness of fit
with the research context. In this case, the authors
H6: High level of mobile technology experience determine to focus on the generalizability dimen-
leads to behavioral intention of use. sion mainly because the aim of this research is to
H7: Low level of expected time to learn how to obtain a finding that can be generalized to wider
use U-Health system leads to behavioral inten- population, namely, chronic disease patients. Al-
tion of use. though a certain level of the in-depth knowledge is
H8: Early adopter behavior leads to behavioral lost, it would essentially lead to another research
intention of use. project, such as a qualitative case study, that can
make up for the deficiency.
Finally, the recommendation construct is As mentioned earlier, the research design of
considered in our model because, for managers, this study aims to account for usage intention
it is critical to estimate the future purchase or formulation of U-Health system that provides

221
 Towards a Conceptual Framework of Adopting Ubiquitous Technology in Chronic Health Care

Figure 5. The Research Model

Table 2.The Nine Hypotheses

Independent Variable Dependent Variable Description

High level of medical concern leads to active medi-
H1 Medical Concerns Medical Activities
cal activity.
Perceived medical condition leads to active medical
H2 Medical Conditions Medical Activities
activity.
Perceived medical concern mediates the relationship
H3 Medical Conditions Medical Concerns
between medical condition and medical activity.
Active medical activity leads to behavioral intention
H4 Medical Activities Behavioral Intention
of use.
High level of perceived need for U-health system leads
H5 Perceived Need Behavioral Intention
to behavioral intention of use.
High level of mobile technology experience leads to
H6 Technology Experience Behavioral Intention
behavioral intention of use.
Low level of expected time to learn how to use U-Health
H7 Expected Learning Time Behavioral Intention
system leads to behavioral intention of use.
Early Adoption Be- Early adopter behavior leads to behavioral intention
H8 Behavioral Intention
havior of use.
High level of behavioral intention of use leads to
H9 Behavioral Intention Recommendation
recommendation.

services to chronic patients. However, it is professional, the capability of healthcare system,

generally difficult to directly assess the quality and communication with customer play respec-
of prescription service, especially in the case tive roles in determining the service quality. The
of chronic patient. The expertise of healthcare authors contend that it is not our top priority

222
Towards a Conceptual Framework of Adopting Ubiquitous Technology in Chronic Health Care

Table 3. Test of the influence of medical concern and condition on medical activities

Standardized Coef-
Un-standardized Coefficients
Model ficients t Sig.
B Std. Error Beta
(Constant) 1.214 .204 5.949 .000
medical concern .241 .037 .270 6.541 .000
medical condition .225 .023 .399 9.670 .000
F=73.698, p=0.000, R2=0.249

to validate strengths of U-Health system in the and regular exercising to maintain or to improve
aspect of service quality improvement, but to their health status.
scrutinize the factors influencing user’s percep- To test the mediate effect, in which medical
tion of adopting the system. To administrate the condition leads to medical activity through
data collection process, a broad range of survey medical concern, the authors employ the four-step
was distributed in South Korea, where there is a approach proposed by Baron & Kenny (1986).
high level of concern about chronic disease. As a The first step is to examine the direct effect of
rule of thumb in social research, the satisfactory independent variable (medical condition) on
sample size is 200 effective responses at least. dependent variable (medical activity). The second
In this research the authors were able to obtain step is to test the relation between independent
effective surveys from 447 respondents and thus variable and mediate variable (medical concern).
achieving a good sample size in order to warrant The third step is to conduct regression analysis
the significance of statistical findings. In terms focusing on the inter-relationship between medi-
of respondent’s characteristics, most of feedbacks ate variable and dependent variable. The final step
were collected from people who are aware of or is to conduct regression analysis of the overall
understand chronic disease and have experience effects of independent variable and mediate vari-
in using wireless computing technology. able on dependent variable. The mediate effect is
proved to be significant when the effect of the
data analysis and results coefficient in the first step is not significantly
different from zero and the coefficient in the third
In this section, the results are presented accord- step is significantly greater than zero. If both
ing to the following order: Healthcare Related conditions are satisfied, a fully mediated model
Factors, Perceived Need, Effort Expectancy, and is said to be found (Frazier & Tix, 2004). In this
Recommendation. study, Table 4 indicates that the effect of medical
condition on the medical activity through the
Healthcare Related Factors medical concern (H3) denotes partially instead
of fully mediated relationship. The major reason
As shown in Table 3, the medical concerns (H1) is because the coefficient of medical condition is
and medical conditions (H2) positively affect significantly different from zero while the coef-
medical activities. In other words, individuals ficient of the medical concern is also signifi-
with a high level of medical concern, or poor cantly greater than zero. Hence, it is argued that
medical condition, are more likely to be involved medical concern partially mediates the formative
with medical activities such as taking medicine relationship between perceived health condition
and medical activity.

223
 Towards a Conceptual Framework of Adopting Ubiquitous Technology in Chronic Health Care

The results in Table 5 show that the effect of extreme (high medical activity group) have
medical activities on the behavioral intention (H4) higher concerns and poorer health condition than
is not significant in this case. However, the scat- other groups. Such concerns and conditions would
ter plot, as shown in Figure 6, provides an alterna- lead to the usage of innovative technology that
tive explanation that the relationship between the might be helpful to improve their health status.
two variables is not linear. In the plot, behavioral On the other hand, the individuals in the left ex-
intention (Y-axis) appears to be significant in the treme (low medical activity group) do not give
left and right areas across the line of medical much attention to their health status due to sev-
activities (X-axis). In the middle area of medical eral restrictions. For example, because of high
activities, however, the correlation is scattered work pressure, many employees have a difficult
and loose. time arranging healthcare consulting service on
Such U-shaped regression curve is subject to a regular basis, which serves as a complementary
receiving two transformations in statistical reason for increasing cases of chronic diseases.
analysis: logarithmic form and quadratic form. In Therefore, these people may realize the potential
this study, quadratic model is applied and the advantage of U-health system and thus believe
correlation between medical activities and behav- that U-health system is able to eliminate the re-
ioral intention is significant at LOS.05, as shown strictions of accessing medical services. Indi-
in Table 6 and Figure 7. viduals who are in the middle group may be
It can be inferred from previous discussion satisfied with the current treatments for their health
that individuals in two extreme groups are more status management and find few benefits in seek-
prone to use U-health system. Those in the right

Table 4. Test of the influence of medical condition on medical concerns

Un-standardized Coefficients
Step Model T Sig.
B Std. Error
Medical condition
Step 1 .237 .024 9.779 .000
→ medical activity
Medical condition
Step 2 .051 .030 1.700 .090
→ medical concern
Medical concern
Step 3 .270 .040 6.680 .000
→ medical activity

Medical condition & concern .225 .023 9.670 .000a

Step 4
→ medical activity .241 .037 6.541 .000b
a-medical condition / b-medical concern

Table 5. Test of the influence of medical activities on behavioral intention

Standardized Coef-
Un-standardized Coefficients
Model ficients t Sig.
B Std. Error Beta
(Constant) 156.926 14.136 11.101 .000
medical activities -2.869 4.237 -.032 -.677 .499
F=.458, p=.499

224
Towards a Conceptual Framework of Adopting Ubiquitous Technology in Chronic Health Care

Figure 6. The scatter plot between medical activities and behavioral intention

Table 6. Test of the influence of medical activities on behavioral intention (Quadratic model)

Parameter Estimates
Equation
Constant b1 b2
Quadratic 236.413 -55.900 7.737
F=3.889, p=0.021, R =0.017
2

Figure 7. The scatter plot of quadratic estimation between medical activities and behavioral intention

225
 Towards a Conceptual Framework of Adopting Ubiquitous Technology in Chronic Health Care

ing other solutions due to the economic lock-in to consider that the less learning time is needed
effect. to use the system, the more likely the behavioral
intention of use can be formed. Whereas for H8,
Perceived Need since we use a reverse scale, that is, 1 represents
strong and 5 means weak, for measuring early
Referring back to the literature review, we have adoption behavior, there is negative correlation
found a plethora of theoretical arguments vali- between those two factors. In congruence with
dating and explaining the relationship between the theory, individuals who are early adopters are
customer’s perceived need and purchase attention. more likely to accept U-health service. However,
Such argument is also verified in this research experience of using mobile technology (H6) is not
context. According to the statistical results shown significant at the 0.05 level. It can be attributed
in Table 7, the perceived need (H5) for U-health to the high mobile technology penetration rate in
system positively affects behavioral intention. It South Korea. Since people are already skillful in
strongly supports the fact that individuals having a using mobile device, there is lack of concern in
high demand for U-health system are more likely technology experience that leads to behavioral
to use the service than those who have lower need. intention of U-Health system use.

Effort Expectancy Recommendations

In Table 8, the analysis of effort expectancy in- According to the data analysis, which is shown in
dicates that early adoption behavior (H8) and ex- Table 9, higher behavioral intention leads to higher
pected learning time (H7) for considering adoption recommendation, which encourages the continu-
are negatively connected with behavioral intention ance of system use and reuse through customer
at the 0.05 level of significance. It is relatively easy recommendation after initial adoption behavior.

Table 7. Test of the influence of perceived need on behavioral intention

Standardized Coef-
Unstandardized Coefficients
Model ficients t Sig.
B Std. Error Beta
(Constant) 51.293 16.126 3.181 .002
Perceived Need 20.586 3.229 .323 6.375 .000
F=40.640, p=0.000, R =0.104
2

Table 8. Test of the influence of adoption behavior, experience, and expected time on behavioral intention

Standardized Coef-
Un-standardized Coefficients
Model ficients t Sig.
B Std. Error Beta
(Constant) 311.662 21.742 14.335 .000
Adoption behavior -13.440 6.803 -.126 -1.976 .049
Experience 2.143 1.266 .100 1.693 .092
Expected time -23.756 4.215 -.352 -5.637 .000
F=18.103, p=0.000, R =0.180
2

226
Towards a Conceptual Framework of Adopting Ubiquitous Technology in Chronic Health Care

The results for all hypotheses testing are sum- the early adopters of new technology is estimated
marized in Table 10. An individual would not take to have a lower barrier to embrace the new system.
action until s/he perceives his/her medical condi- Once the behavioral intention is formed, the user
tion, which are supported in H1, H2 and H3. The is likely to recommend the product to others based
impact of medical activities to behavioral intention on the previous positive experience obtained.
of using medical product (e.g. U-health system)
is non-linear. Two groups of individuals: high
medical activity group, those who have higher dIscussIon and conclusIon
concerns and poorer health condition, and low
medical activity group, those who do not give The authors argue that consumer’s intention of
much attention to their health status due to some using U-Health system can be better understood if
restrictions (e.g. tight working schedule), are it is viewed through a multi-faceted framework. In
estimated to have high desire of adopting conve- this paper, we draw upon literature from different
nient health care services enabled by IT. The paradigms including specific medical research and
perceived need of individual leads to positive technology acceptance and use. The paper’s basic
formation of behavioral intention of use. Where- premise is that medical factors, perceived eco-
as people will be reluctant in adopting the product nomic needs, and effort expectancy exert formative
if it takes too much efforts and the learning curve impacts on consumer’s behavioral intention. Such
is too long, which leads to another inference that premise is then broken into a series of hypotheses

Table 9. Test of the influence of behavioral intention on recommendation

Standardized Coef-
Un-standardized Coefficients
Model ficients t Sig.
B Std. Error Beta
(Constant) .511 .035 14.498 .000
Behavioral Intention .001 .000 .314 6.980 .000
F=48.718, p=0.05, R =0.099.
2

Table 10. Summary of Hypotheses Testing

Independent Variable Dependent Variable Hypothesized Effect Support

H1 Medical Concerns Medical Activities Positive Yes1
H2 Medical Conditions Medical Activities Positive Yes
Yes
H3 Medical Conditions Medical Concerns Positive
(Partially)
No
H4 Medical Activities Behavioral Intention Positive
(Non-linear)
H5 Perceived Need Behavioral Intention Positive Yes
H6 Technology Experience Behavioral Intention Positive No
H7 Expected Learning Time Behavioral Intention Negative Yes
Early Adoption Be-
H8 Behavioral Intention Negative2 Yes
havior
H9 Behavioral Intention Recommendation Positive Yes

227
 Towards a Conceptual Framework of Adopting Ubiquitous Technology in Chronic Health Care

in Table 9 extending the discussion to a further reFerences

detailed level. By validating these assumptions,
we are able to obtain more insights in terms of Abawajy, J. H. (2009). Human-computer
how key adoption factors interact with each other. interaction in ubiquitous computing environ-
The statistical findings contend that high level ments. International Journal of Pervasive
of health concern and perceived health condition Computing and Communications, 5(1), 61–77.
positively lead to relevant medical activities in- doi:10.1108/17427370910950311
cluding regular sport exercises and consulting with Ajzen, I., & Fishbein, M. (1973). Attitudinal
doctors. Although the data analysis doesn’t support and normative variables as predictors of specific
direct formative relationship between medical ac- behavior. Journal of Personality and Social Psy-
tivities and actual U-Health system adoption (e.g. chology, 27(1), 41–57. doi:10.1037/h0034440
people who visit doctors might not intend to use
U-Health system), a further investigation on the Bandyopadhyay, K., & Schkade, L. L. (2000).
data indicates that two groups of potential users - Disaster recovery planning by HMOs: theoreti-
people who have high level of health concerns and cal insights. Health Care Management Review,
poor health condition and people who are unable 25(2), 74–85.
to manage their health issues due to limitations
Barros, P. P., Pinto, C. G., & Machado, A. (1999).
such as severe work pressure – are more likely to A signaling theory of excessive technological
use the system. Moreover, as many other studies adoption. Health Care Management Science, 2(2),
have argued, the perceived need or perceived 117–125. doi:10.1023/A:1019075425763
value of the system is found positively related with
behavioral intention of use. In congruence with Berta, W. Teare, G. F., Gilbart, E., Soberman G.,
seminal theories including TAM, the study shows L. Lemieux-Charles, L., Davis, D., & Rappol,
that individuals who are early adopters are more S. (2005). The contingencies of organizational
likely to accept U-health service since the expected learning in long-term care: factors that affect
learning cycle is short, or, the system is easy to innovation adoption. Health Care Management
use. The expected learning time is measured using Review, 30(4), 282–292.
reversed scale hence it has a negative relationship
Bhaskaran, S., & Hardley, F. (2002). Buyer beliefs,
with the behavioral intention of use. Some of the
attitudes and behaviour: foods with therapeutic
limitations in this study lie in the measurements
claims. Journal of Consumer Marketing, 19(7),
development and some insignificant statistical
591–606. doi:10.1108/07363760210451410
findings, which hinder us from better validating
our proposal. Future efforts are required to shed Briggs, R. O., Nunamaker, J. F. Jr, & Sprague,
the light on designing and validating measures that R. H. Jr. (2005). Special issue: crossing boundar-
would fill the voids in this study. Also, since no ies in information systems research. Journal of
single research can maximize the three features of Management Information Systems, 22(4), 7–11.
research process: generalizability, precision, and doi:10.2753/MIS0742-1222220401
realism (McGrath, 1982), a different perspective
Burke, R., & Payton, P. E. (2006). Measuring
can be obtained through a qualitative case study
and managing shoppability: Ten principles to
of U-health system or other innovative health
convert shoppers into buyers. European Retail
care services.
Digest, 50, 66–78.

228
Towards a Conceptual Framework of Adopting Ubiquitous Technology in Chronic Health Care

Chae, M. H., Black, C., & Heitmeyer, J. (2006). Hu, P. J., Chau, P. Y. K., Sheng, O. R. L., & Tam, K.
Pre-purchase and post-purchase satisfaction Y. (1999). Examining the technology acceptance
and fashion involvement of female tennis wear model using physician acceptance of telemedicine
consumers. International Journal of Consumer technology. Journal of Management Information
Studies, 30(1), 25–33. doi:10.1111/j.1470- Systems, 16(20), 91–112.
6431.2005.00434.x
Jee, K. Y. (2005). Healthcare in ubiquitous era.
Cohen, C. I., & Cohen, E. J. (1978). Health educa- Jinhan M&B.
tion: panacea, pernicious or pointless? The New
Junginger, M., de Visser, E., Hjort-Gregersen, K.,
England Journal of Medicine, 28, 718–720.
Koornneef, J., Raven, R., Faaij, A., & Turkenburg,
Davis, F. D. (1989). Perceived usefulness, per- W. (2006). Technological learning in bioenergy
ceived ease of use, and user acceptance of informa- systems. Energy Policy, 34(18), 4024–4041.
tion technology. Management Information Systems doi:10.1016/j.enpol.2005.09.012
Quarterly, 13(3), 319–340. doi:10.2307/249008
Kang, S., & Lee, S. (2007). Economic effect and
Deng, M., Cock, D. D., & Preneel, B. (2009). To- growth strategy of U-Health. SERI (Samsung
wards a cross-context identity management frame- Economic research Institute). Issue Paper.
work in e-health. Online Information Review,
Kervenoael, R. D., Soopramanien, D., Elms, J., &
33(3), 422–442. doi:10.1108/14684520910969880
Hallsworth, A. (2006). Exploring value through
Eysenbach, G., & Köhler, C. (2003). What is the integrated service solutions: the case of e-grocery
prevalence of health-related searches on the World shopping. Managing Service Quality, 16(2),
Wide Web? Qualitative and quantitative analysis 185–202. doi:10.1108/09604520610650646
of search engine queries on the Internet. In Pro-
Korean National Insurance Corporation (2007),
ceedings of the AMIA Symposium (pp. 225-229).
Daily Report, Jan/29.
Frazier, P. A., & Tix, A. P. (2004). Testing modera-
Lahiri, K., & Xing, G. (2004). An econometric
tor and mediator effects in counseling psychol-
analysis of veterans’ healthcare utilization us-
ogy. Journal of Counseling Psychology, 51(1),
ing two-part models. Empirical Economics, 29,
115–134. doi:10.1037/0022-0167.51.1.115
431–449. doi:10.1007/s00181-003-0178-0
Guido, F. (2007). The organizational learn-
Lim, H., Widdows, R., & Jungkun, J. (2006).
ing curve. European Journal of Operational
M-loyalty: winning strategies for mobile carri-
Research, 177(3), 1375–1384. doi:10.1016/j.
ers. Journal of Consumer Marketing, 23(4/5),
ejor.2005.04.009
208–218. doi:10.1108/07363760610674338
Gupta, K., & Stewart, D. W. (1996). Customer
Madden, M. (2003). America’s online pursuits: the
satisfaction and customer behavior: the differential
changing picture of who’s online and what they
role of brand and category expectations. Marketing
do. Washington DC: Pew Internet and American
Letters, 7(3), 249–263. doi:10.1007/BF00435741
Life Project. Retrieved from www.pewinternet.
Hendricks, K. B., Singhal, V., & Stratman, J. K. org/reports/toc.asp?Report=106
(2007). The impact of enterprise systems on cor-
McGrath, J. E. (1982). Judgment Calls in Re-
porate performance: a study of ERP, SCM, and
search. Beverly Hills, CA: Sage Publications, Inc.
CRM system implementations. Journal of Opera-
tions Management, 25(1), 65–82. doi:10.1016/j.
jom.2006.02.002

229
 Towards a Conceptual Framework of Adopting Ubiquitous Technology in Chronic Health Care

McKeown, T. (1971). A Historical Appraisal of the Uskul, A. K., & Hynie, M. (2007). Self-construal
Medical Task. In McLachlan, G., & McKeown, and concerns elicited by imagined and real health
T. (Eds.), Medical History and Medical Care (pp. problems. Journal of Applied Social Psychol-
29–50). London: Oxford University Press. ogy, 37(9), 2156–2189. doi:10.1111/j.1559-
1816.2007.00283.x
Morahan-Martin, J. M. (2004). How internet users
find, evaluate, and use online health information: Venkatesh, V., Morris, M. G., Davis, G., & Davis,
a cross-cultural review. Cyberpsychology & Be- F. D. (2003). User acceptance of information
havior, 7(5), 497–510. technology - toward to unified view. Management
Information Systems Quarterly, 27(3), 425–478.
Morgan, N. A., & Rego, L. L. (2006). The value
of different customer satisfaction and loyalty Weiser, M. (1993a, October). Hot topics: ubiqui-
metrics in predicting business performance. Mar- tous computing. IEEE Computer.
keting Science, 25(5), 426–439. doi:10.1287/
Weiser, M. (1993b). Some computer science issues
mksc.1050.0180
in ubiquitous computing. Communications of the
Ngwenyama, O., Guergachi, A., & McLaren, T. ACM, 36(7), 75–84. doi:10.1145/159544.159617
(2007). Using the learning curve to maximize
Windmeijer, F. A. G., & Silva, J. S. (1997).
IT productivity: a decision analysis model for
Endogeneity in count data models: an applica-
timing software upgrades. International Journal
tion to demand for healthcare. Journal of Ap-
of Production Economics, 105(2), 524–535.
plied Econometrics, 12, 281–294. doi:10.1002/
doi:10.1016/j.ijpe.2006.02.013
(SICI)1099-1255(199705)12:3<281::AID-
Odom, M., Kumar, A., & Saunders, L. (2002). Web JAE436>3.0.CO;2-1
assurance seals: how and why they influence con-
Wong, H. J., & Legnini, M. W. (2000). The diffu-
sumers’decisions. Journal of Information Systems,
sion of decision support systems in healthcare: are
16(2), 231–250. doi:10.2308/jis.2002.16.2.231
we there yet? Journal of Healthcare Management,
Park, K. (2003). Background to Ubiquitous 45(4), 240–254.
Healthcare. No.272 Conference between Industry,
Wong, H. K. (2007). Consumerism and qual-
Academic, Research Centers, Korea.
ity standards for subvented social services: a
Rhodes, B., & Mase, K. (2006). Wearables in 2005. case study of Hong Kong. International Jour-
IEEE Pervasive Computing / IEEE Computer nal of Public Administration, 30(4), 441–458.
Society [and] IEEE Communications Society, doi:10.1080/01900690601158287
5(1), 92–95. doi:10.1109/MPRV.2006.21
Yoo, S. (2006). Ubiquitous Health, Industrial
Riska, E. (1982). Health education and its ideologi- Health. Korea & World Affairs, 5, 13–18.
cal content. Acta Sociologica, 25, 41–46.
Zhang, J., & Fang, X., L. S., & Olivia, R. (2007).
Smith, A. D. (2006). Exploring m-commerce in Online consumer search depth: theories and new
terms of viability, growth and challenges. Interna- findings. Journal of Management Information
tional Journal of Mobile Communications, 4(6), 4. Systems, 23(3), 71–95. doi:10.2753/MIS0742-
1222230304
Tan, J. K. H. (1995). Health management informa-
tion systems: theories, methods, and applications.
Gaithersburg, MD: Aspen Publishers.

230
Towards a Conceptual Framework of Adopting Ubiquitous Technology in Chronic Health Care

endnotes
1
The default level of significance for all
hypotheses testing is.05.
2
Revere scales are used for the measure-
ment.

231
232

Chapter 16
E-Patients Empower Healthcare:
Discovery of Adverse Events
in Online Communities1
Roy Rada
University of Maryland Baltimore County, USA

aBstract
E-patients can empower themselves and improve healthcare. In online communities, patients may discuss
adverse events that are inadequately addressed in the literature. The author as a patient joined various
online patient discussion groups and identified several such adverse events. For each such adverse event,
the patient findings, the medical literature, and the implications are noted. Extracts from the literature
that were provided to the patients were welcomed by the patients. Possible approaches to financially
supporting such activities are sketched.

IntroductIon edge technology and electronic medical records so

that we can cut red tape, prevent medical mistakes,
The President of the United States in an address and help save billions of dollars each year.”
to the nation said (Obama, 2008):
The emphasis on electronic medical records is
“In addition to connecting our libraries and a natural one for a provider-centric nation. How-
schools to the internet, we must also ensure that ever, another benefit from greater digitization and
our hospitals are connected to each other through connectivity of the world in health care matters is
the internet. That is why the economic recovery the emergence of patient power or the e-patient. A
plan I’m proposing will help modernize our health Dutch study showed that patients may use blogs to
care system – and that won’t just save jobs, it will advance some of the principles of Web 2.0, as the
save lives. We will make sure that every doctor’s author noted (Adams, 2008): “enabling patients
office and hospital in this country is using cutting to be more active in documenting and managing
information related to their health experiences”.
DOI: 10.4018/978-1-61692-002-9.ch016 What is the role of the patient in the e-health world?

Copyright © 2011, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
E-Patients Empower Healthcare

In the already classic, oft-cited article “Will the conclusion that patients seek information
Disruptive Innovations Cure Health Care”, the differently and sometimes inefficiently and inef-
authors propose two key steps to improving the fectively but online support tools could improve
health care system (Christensen et al., 2000): that search behavior (Keselman et al., 2008).
Customer Relationship Management systems are
• create and then embrace a system where extensively used in healthcare systems (Calhoun
the clinician’s skill is matched to the dif- et al., 2006). Data mining of web information is
ficulty of the medical problem and an alternative way to learn what consumers think.
• invest less money in high-end, complex This paper explores the means by, and extent
technologies and more in technologies that to, which participants in online patient-patient
simplify complex problems. discussion groups provide useful information
about medical adverse events.An adverse event
The empowerment of patients online fits nicely occurs when some intervention by a healthcare
into this desiderata. Patients have varying exper- provider produces an unwanted reaction. For
tises and common experiences across the globe. instance, radiation treatment for oral cancer can
They can meet with one another and share stories cause obstructive sleep apnea. The literature on
and insights both about their medical conditions adverse events addresses their causes, how to re-
and their doctors. The technology is simple and duce them, and the impact they have on patients,
already well developed for other commercial staff, and health care organizations (Misson, 2001).
purposes. What evidence exists to support this Typically, health care professionals investigate
contention that patients by moving online and adverse events through the medical record (Duff
sharing information about health can help the et al., 2005).
health care system? Many online patient groups are established by
A E-Patient Network with support from the volunteers on free sites, such as groups.yahoo.com
Robert Woods Johnson Foundation has developed (Rada, 2006). However, some healthcare entities
a wikipedia-based white paper titled “e-Patients: maintain patient online discussion groups. For
How they can help us heal healthcare” for which instance, Joslin Diabetes Center runs an online,
the conclusion is this (Ferguson, 2009): diabetes discussion group for the public, and
experts from the Center provide feedback online.
“The creation of optimal health care may depend Kaiser Permanente maintains numerous discus-
on our ability to embrace our first generation of sion groups moderated by Kaiser’s professionals,
e-patients, providing them with the autonomy, but access is restricted to enrollees in the Kaiser
authority, and empowerment they desire and de- Plan. Healthcare professional in online modera-
serve and inviting them to join us in a combined tor roles might address adverse events, among
effort to improve healthcare for everyone. It will other things.
be only by joining forces with these new colleagues Listening to patients is a key to reducing ad-
that we can hope to solve the pervasive problems verse events (Cleary, 2003):
that plague the healthcare system: quality, cost,
access, and consumer satisfaction.” “by relying on the observations and insights of
patients such as Mr. Q., the physicians and staff
Many different aspects of this e-patient revolu- will be able to close the gap between Mr. Q.’s
tion are being explored. For instance, the support experience and what they can achieve.”
that might be provided for patients seeking health
care information online has been addressed with

233
 E-Patients Empower Healthcare

The book “Partnering with Patients to Reduce online groups via ethnography is in many ways
Medical Errors” states (Spath, 2004): “The health easier than studying face-to-face groups (Pac-
care team can only be as strong as its weakest cagnella, 1997). The term netnography has been
link, and unfortunately, the weakest link is often coined by Kozinets to apply to such ethnography.
the recipient of care – the patient.” According to (Kozinets, 2002): “As a method,
Patients in online groups hold a unique and netnography is faster, simpler, and less expensive
valuable position because of their sheer numbers than traditional ethnography and more naturalistic
and an intense focus on their shared illness. Patient and unobtrusive than focus groups or interviews.”
groups may have contact with larger numbers of In the context of this research, an online group
disease-specific patients than many physicians and uses a software system that provides a search-
have the luxury of spending many hours discuss- able archive of previous messages. Members of
ing similarities and differences. After hundreds the group create messages and post them to the
of hours of conversation, patterns can begin to system, and the system in turn distributes these
emerge. These patterns might lead to new insights messages to the group. The system may interface
about adverse events. Norman Scherzer started to a group member via an email client or a web
a patient discussion group to explore issues of site. The online groups noted in this paper include
how cancer patients, such as his wife, might be patients, their significant others, and sometimes
best treated (Ferguson, 2002). Through sharing others who want to help. This population will
information the patients and their significant- be typically represented with the umbrella term
others made discoveries about side effects of the ‘patients’ with its meaning apparent in the context.
treatment for the cancer that were published in a This author joined two cancer patients groups
scholarly journal. as a patient in 2003. Both groups had formally wel-
This paper explores the discovery of adverse comed participants to use personally de-identified
events through online patient groups.The hy- information in the messages for research purposes.
pothesis is that patients share information online, The groups had a total membership of several
that when combined with information from the hundred. The author read the patient messages,
literature, can help identify important gaps in the identified messages of interest, studied relevant
medical literature. More generally, the argument clinical, journal articles, and where appropriate
is that these online groups can be an important shared extracts from the literature with the group.
resource for both patients and healthcare providers. The author identified several cases where the
information needs of the patients led to the dis-
covery of adverse events and gaps in the medical
MetHod literature. For each case, the patient findings, the
medical literature, and the implications are noted.
The author was trained as a medical doctor and Four years after the treatment of his cancer,
became a cancer patient. As doctors become ill the author developed a long-term adverse effect
and see the world from the patient’s side, they of the cancer treatment. This lead the author to
often have useful insights to share about the join a long-term care discussion group (also with
relationship between patients and healthcare hundreds of members) and to discover yet again
providers (Rosenbaum, 1988). Since an online the marvels of what patients can share that medi-
discussion group is self-documenting by nature, cal professionals rarely have the time or ability
the opportunity exists for a participant in a group to share -- namely, the personal experiences of
to review the discussion and to engage in a kind chronic disease suffers who are coping with their
of retrospective ethnographic analysis. Studying life and the medical establishment.

234
E-Patients Empower Healthcare

results care versus the ability to pay for the care

create a kind of adverse event for the patient.
The four cases follow: Again, when extracts from the literature were
shared with the patients, they replied with
1. Patients discussed ways to cope with their messages including a ‘thank you’.
fatigue. In the online discussion groups, over 3. A patient reported severe allergic response
a dozen patients reported signs and symp- to a drug (amifostine) that was first being
toms of Obstructive Sleep Apnea (OSA) in used for cancer patients as a radioprotectant.
relation to this fatigue. The literature reveals The literature at the time suggested that
incomplete information about OSA in head- severe reactions to amifostine treatment
and-neck cancer patients. Two articles pro- were rare: “Amifostine administration was
vide interestingly different perspectives on well tolerated, with a low incidence of side
OSA as a complication of the treatment. In effects” (Antonadou et al., 2002). However,
one article the incidence of OSA was 92% in the patient group included two patients with
treated patients (Friedman et al., 2001), while severe allergic reaction. A year later the re-
in the other article 8% of treated patients sults of a clinical trial were published which
developed OSA (Rombaux et al., 2000). The confirmed what the patients feared (Rades
literature at that time provided no mention of et al., 2004): “Administration of amifostine
a radiated-only patient developing OSA, but during radiotherapy is associated with a high
one of the patients developed OSA after only rate of serious adverse effects.” When a
radiation. The outcomes of these observa- new drug use appears, detecting uncommon
tions were two-fold. On one hand, extracts adverse events may be supported by having
of the literature were shared with the online patient groups monitoring and discussing
group, and the patients expressed gratitude their reactions to their treatments. Some of
for that literature information. Secondly, the the patients in turn took this group informa-
observation of a gap in the literature became tion to their doctors, and helped their doctors
the basis of published, medical literature by appreciate the problem.
this author (Rada, 2005a, 2007).
2. Hyperbaric oxygen treatment for osteora- Four years after treatment that patient began
dionecrosis of the mandible is routine in developing spasms in the muscles. Two years
the United States. A patient in a discussion later this was diagnosed as radiation-induced
group presented his concerns about hyper- neuropathy. The doctors said “progressive and
baric for osteoradionecrosis and said: “Every irreversible”. Furthermore, no treatment existed
dentist that I have seen in San Antonio has and the disease progression was variable. Doctors
recommended hyperbaric oxygen, but does are hardly able to say more than that. However,
anyone know if hyperbaric oxygen is worth patients who are experiencing this problem are
the $50,000 cost?” The patient went to eager to say more. Unfortunately, in meeting with
Mexico and was told hyperbaric oxygen was one’s doctor, the patient gets no access to other
unnecessary. European studies have shown patients with this condition and generally speak-
that hyperbaric oxygen is not appropriate ing no other such patients are locally accessible
(Annane et al., 2004), but the American -- the condition is relatively rare in that it affects
literature defends it (Mendenhall, 2004). about 1% of cancer survivors.
Differences in the standard of care in one The author joined a long-term care survivors
country versus another and the standard of group and discovered that radiation-induced

235
 E-Patients Empower Healthcare

neuropathy was experienced by multiple people categorIes and procedures

in that group. Their fate had been invariably an
unpleasant one. The author’s doctor suggested Each of the preceding adverse events concerned
an experimental treatment which had been tried more than one clinical specialty. The adverse
with two patients elsewhere. The author found a events might be categorized as follows:
few patients in the online group who had tried a
similar treatment unsuccessfully. However, the • A diffuse symptom: Obstructive sleep ap-
opportunity to understand what is happening nea secondary to treatment for head-and-
proved itself a powerful elixir -- “knowing is neck cancer may tend to be overlooked by
half the battle”. otolaryngologists because the symptoms
When patients are confused about an adverse are diffuse and obstructive sleep apnea is
event, the possibility exists that the health pro- often addressed by sleep specialists rather
fession itself is also confused. The patients often than otolaryngologists.
do not share their confusion in a compelling way • Standard of care: Hyperbaric oxygen as
with the healthcare professional. However, in the part of the national standard of care for
patient group, the patients may be comfortable osteoradionecrosis of the mandible is not
to elaborate. From this study, topics in which supported by clinical trials internationally,
patients felt particularly unable to get adequate but the practicing otolaryngologist is not
explanations from their healthcare professionals expected to dispute the national standard
where topics which the healthcare industry had not of care.
adequately addressed and which would require the • Uncommon reaction to new drug: When
coordinated attention of healthcare professionals the drug amifostine was initiated for a new
from different disciplines. purpose, researchers needed further expe-
Patients appreciated receiving extracts of the rience to uncover adverse events.
medical literature that pertained to their questions.
While the information might have improved health The foremost causes of adverse events as re-
outcomes, it also could lead to other positive out- ported by the US Institute of Medicine (Kohn et
comes. For instance, patients could contribute to al., 2000) are technical errors, diagnostic errors,
community awareness initiatives. Or a healthcare failure to prevent injury, and medication errors.
professional might author a scholarly paper about That classification is, however, not necessarily
a particular adverse event. the optimal one for understanding what can be
The majority of the discussion in the online gleaned from patient online groups.
groups was not about the preceding adverse If a provider has decided to support an online
events. Much of the discussion was about emo- discussion group and to provide moderators, then
tional topics,such as a patient reporting the good it might guide moderators relative to the findings
news that the latest checkup with the oncologist of this study. To find evidence of adverse events
revealed no progression of the disease and other that are inadequately appreciated in the literature,
patients congratulating the patient on the good a healthcare professional might:
news. Patients often complained about diffuse
ill effects of the cancer treatment. However, the 1. Join an online discussion group for patients
importance of these groups to the well-being of with a particular chronic disease.
the participants seems obvious, and, at times, 2. Identify a finding that is highlighted by a
discoveries are possible that would seem unlikely patient as a problem. Findings may include
to practically be obtained any other way. symptoms, signs, laboratory or test results,

236
E-Patients Empower Healthcare

observations, or specific events (such as In general, adverse events may be least well under-
hospitalization or receiving a bill). A finding stood where 1) the responsibility for the adverse
is a problem when patient says so. event falls among several medical specialties
3. Review the medical literature to determine and 2) the medical specialists inadequately com-
whether a medical intervention experienced municate with one another.
by the patient might have a causal relation to The data from the online groups leads to quali-
the problematic finding. Relevant PubMed tative results. In online groups, most participants
‘Medical Subject Headings’ are identified, a are typically lurkers (Preece et al., 2004). To obtain
query is posed to PubMed, and full-text cop- accurate incidence data, clinical trials might be
ies of journal articles are retrieved through needed. The online patient information supported
membership in a medical library subscription the identification of a problem. An article in the
program. Temporal relationship, strength Archives of Internal Medicine supports this posi-
of association, biological plausibility, and tion by saying (DeMonaco, 2009):
other relationships contribute to a judgment Postmarketing surveillance and the determina-
of causality (Darden & Rada, 1988). tion of the real-world safety profile of prescription
4. Determine whether the literature provides drugs is arguably flawed. Recent identification
conflicting or unclear guidance. Sometimes of significant adverse effects associated with
the published literature suggests conflicting newly approved prescription drugs support the
algorithms for diagnosis or treatment, and sometimes-held view that a new system needs to
more research is needed to harmonize the be introduced. The present voluntary system has
literature. not provided a sufficient early warning system, and
5. Extract information from the literature and some have called for active systems that probe for
return that information to the group. The potential adverse effects of approved prescription
extract should be clear to the intended audi- drugs. Patient-oriented Web sites may provide an
ence, embedded within a personal context, opportunity to identify potential adverse effects
and made as a reply to recently posted mes- early in a drug’s postmarket history.
sage that has not already received a similar The interest in collecting adverse events from
response. online patient groups is clearly growing.
If a healthcare provider wanted to support its
If an extract from the literature is simply posted employees in online patient discussion groups with
without context or explanation, then the impact, the intent of also helping identify adverse events,
as measured by patient response, is less. Infor- then a proposal to the provider’s Institutional Re-
mation systems can support this work be parsing view Board would be in order. The patients joining
patient messages and semi-automatically linking the group would be provided a consent form that
to relevant citations from PubMed (Rada, 2005b). detailed the conditions, the patient alternatives,
and other components of a proper consent form.
Given that patients had to register to join the
dIscussIon group, their successful registration would only
occur after they noted online that they consented.
If one takes the preceding categorization of adverse While patients might be asked to sign a consent
events from online groups and tries to generalize form, they are not invited to the online groups to
further, one might note that the problems occur get a diagnosis or a treatment. Rather the groups
where the otolaryngologist’s responsibility is support patient-patient interaction, and the patients
blurred because someone else is also responsible. are responsible for the content of the message

237
 E-Patients Empower Healthcare

that they share. If a knowledgeable person brings surance companies reimburse such contributors
extracts from the literature to the discussion, those when their contributions are deemed adequately
extracts cite the original source and are informa- significant. One could worry that such an approach
tional only. Responsibility for taking action based would lead to rampant charlatanism. However,
on the information rests with the patient. peer-peer evaluations in online systems have
A healthcare provider that wants to sponsor an proven remarkably robust (Rada & Hu, 2002).
online discussion group has a range of options, Given the dearth of options for reducing costs
including the option to provide a healthcare pro- but improving quality, ways to involve patients
fessional as a moderator or to have no moderator. themselves more effectively into the health care
One problem with providing a moderator is the solution should be explored.
cost in human effort. Healthcare professionals
have many demands on their time and often do not
see participation in an online patient discussion conclusIon
group as a cost-effective use of their time. For the
typical healthcare provider in the United States, Empowering patients is vital to improving health-
efforts invested in an online discussion group care. One source of information that has been
cannot be billed to a health insurance company on largely overlooked by the healthcare industry
behalf of the patients’ in the group. At least one comes from online patient discussion groups.
healthcare provider continues to support online Online patient groups may provide an opportunity
patient discussion groups because the provider for healthcare providers to both build customer
has found that appreciative patients donate money relationships and explore adverse effects.
to the provider that offsets that provider’s cost The author participated as a patient, though he
of maintaining the discussion group. A financial is also a doctor, in several online patient groups.
cost-benefit analysis that considered a wide-range Patients discussed various types of adverse events,
of factors, such as healthcare professional labor but several types were particularly intriguing
costs, patient loyalty, and patient health outcomes, for the gaps between what the patients needed
would be appropriate before an entity decided to know and what the literature offered. These
how much, if anything, to invest in online patient adverse events have been categorized as involv-
discussion groups. ing a diffuse symptom, a standard of care, and
The world is eager for methods to reduce the uncommon reaction to a drug. The cases are multi-
costs of healthcare without decreasing the quality disciplinary in nature. The gaps in the literature
of healthcare. The advent of the digital, Internet create an opportunity for someone to 1) produce
age is creating new opportunities for people to a synthesis of the literature that highlights the
cooperate. Disruptive innovations are needed gap and 2) publish that synthesis in a scholarly
(Christensen et al., 2000) that lower cost. En- medical journal.
couraging patients to help one another can save As measured by their responses, patients ap-
costs. Might it be possible to create some kind of preciated receiving information from the literature
a financial reward system to encourage patients about their adverse events. A systematic approach
to actively participate? to identifying such adverse events and providing
The sharing of information online is not con- relevant literature to patients is sketched based
sidered a reimbursable event by most American on the experiences of the author. Software can
health insurance entities. A way could be found to support the retrieving of relevant literature, but
identify online participants who were recognized posing the response in the context of the patient’s
by their peers as having expertise and having in- concerns requires human judgment.

238
E-Patients Empower Healthcare

People seeking health-related support are one Darden, L., & Rada, R. (1988). Hypothesis for-
of the most common users of the Internet. They mation via interrelations. In Prieditis, A. (Ed.),
can contribute to the health care of one another by Analogica (pp. 109–128). London: Pitman.
sharing their experiences. Such contributions to
DeMonaco, H. (2009). Patient- and physician-
health care are essentially cost-free to the society
oriented web sites and drug surveillance: Bisphos-
but leverage the power of the patient.
phonates and severe bone, joint, and muscle
pain. Archives of Internal Medicine, 169(12),
1164–1166. doi:10.1001/archinternmed.2009.133
reFerences
Duff, F. L., Daniel, S., Kamendje, B., Le Beux, P.,
Adams, S. (2008). Blog-based applications and & Duvauferrier, R. (2005). Monitoring incident
health information: Two case studies that illustrate report in the healthcare process to improve qual-
important questions for consumer health informat- ity in hospitals. International Journal of Medical
ics (chi) research. International Journal of Medical Informatics, 74(2-4), 111–117. doi:10.1016/j.
Informatics, 1–8. Retrieved from http://dx.doi. ijmedinf.2004.06.007
org/10.1016/j.ijmedinf.2008.06.009.
Ferguson, T. (2002, September 2002). Key
Annane, D., Depondt, J., Aubert, P., Villart, M., concepts in online health: E-patients as medical
Gehanno, P., & Gajdos, P. (2004). Hyperbaric researchers. The Ferguson Report Retrieved Au-
oxygen therapy for radionecrosis of the jaw: A gust 2005, from http://www.fergusonreport.com/
randomized, placebo-controlled, double-blind articles/fr00903.htm
trial from the orn96 study group. Journal of Clini-
cal Oncology, 22(24), 4893–4900. doi:10.1200/ Ferguson, T. (2009). E-patients: How they can
JCO.2004.09.006 help us heal healthcare. Retrieved June 27, 2009,
2009, from http://www.acor.org/epatientswiki/
Antonadou, D., Pepelassi, M., Synodinou, M.,
Puglisi, M., & Throuvalas, N. (2002). Prophylactic Friedman, M., Landsberg, R., Pryor, S., Syed,
use of amifostine to prevent radiochemotherapy- Z., Ibrahim, H., & Caldarelli, D. (2001). The
induced mucositis and xerostomia in head-and- occurrence of sleep-disordered breathing among
neck cancer. International Journal of Radiation patients with head and neck cancer. The Laryngo-
Oncology*Biology*Physics, 52(3), 739-747. scope, 111, 1917–1919. doi:10.1097/00005537-
200111000-00008
Calhoun, J., Banaszak-Hol, J., & Hearld, L. (2006).
Current marketing practices in the nursing home Keselman, A., Browne, A., & Kaufman, D.
sector. Journal of Healthcare Management, 51(3), (2008). Consumer health information seeking
185–200. as hypothesis testing. Journal of the American
Medical Informatics Association, 15(4), 484–495.
Christensen, C., Bohmer, R., & Kenagy, J. (2000). doi:10.1197/jamia.M2449
Will disruptive innovations cure health care.
Harvard Business Review, (September-October): Kohn, L., Corrigan, J., & Donaldson, M. (Eds.).
102–112. (2000). To err is human: Building a safer health
system. Washington: National Academy Press.
Cleary, P. D. (2003). A hospitalization from hell: A
patient’s perspective on quality. Annals of Internal Kozinets, R. (2002). The field behind the screen:
Medicine, 138(1), 33–39. Using netnography for marketing research in
online communications. JMR, Journal of Mar-
keting Research, 39(1), 61–72. doi:10.1509/
jmkr.39.1.61.18935

239
 E-Patients Empower Healthcare

Mendenhall, W. M. (2004). Mandibular osteora- Rada, R. (2007). Sleep and quality of life in
dionecrosis. Journal of Clinical Oncology, 22(24), head and neck neoplasm. In Verster, J. C., Pandi-
4867–4868. doi:10.1200/JCO.2004.09.959 Perumal, S. R., & Streiner, D. L. (Eds.), Sleep and
quality of life in head and neck neoplasm. Totowa,
Misson, J. C. (2001). A review of clinical risk
New Jersey: Humana Press.
management. Journal of Quality in Clinical
Practice, 21, 131–134. doi:10.1046/j.1440- Rada, R., & Hu, K. (2002). Patterns in stu-
1762.2001.00421.x dent–student commenting. IEEE Transactions
on Education, 45(3), 262–267. doi:10.1109/
Obama, B. (2008). Remarks of president-elect
TE.2002.1024619
barack obama: Radio address on the economy.
Retrieved June 27, 2009, from whitehouse.gov Rades, D., Fehlauer, F., Bajrovic, A., Mahlmann,
B., Richter, E., & Alberti, W. (2004). Serious
Paccagnella, L. (1997). Getting the seats of your
adverse effects of amifostine during radiotherapy
pants dirty: Strategies for ethnographic research
in head and neck cancer patients. Radiotherapy
on virtual communities. Journal of Computer-
and Oncology, 70, 261–264. doi:10.1016/j.ra-
Mediated Communication, 3(1). Retrieved from
donc.2003.10.005
http://ascusc.org/jcmc/vol3/issue1/paccagnella.
html. Rombaux, P., Hamoir, M., Plouin-Gaudon, I.,
Liistro, G., Aubert, G., & Rodenstein, D. (2000).
Preece, J., Nonnecke, B., & Andrews, D. (2004).
Obstructive sleep apnea syndrome after recon-
The top five reasons for lurking: Improving com-
structive laryngectomy for glottic carcinoma.
munity experiences for everyone. Computers in
European Archives of Oto-Rhino-Laryngology,
Human Behavior, 20(2), 201–223. doi:10.1016/j.
257(9), 502–506. doi:10.1007/s004050000267
chb.2003.10.015
Rosenbaum, E. E. (1988). A taste of my own
Rada, R. (2005a). Obstructive sleep apnea and
medicine, when the doctor is the patient. New
head and neck neoplasms. Otolaryngology - Head
York: Random House.
and Neck Surgery, 132(5), 794–799. doi:10.1016/j.
otohns.2004.12.002 Spath, P. (2004). Preface. In P. Spath (Ed.),
Partnering with patients to reduce medical er-
Rada, R. (2005b). Software for patient-patient dis-
rors (pp. xix-xxiii). Chicago: American Hospital
cussions. Montreal Conference on E-Technologies
Association Press. 1 This paper borrows heavily
2005, Montreal, Canada (pp. 205-209).
from Roy Rada (2008) “Ethnographic discovery
Rada, R. (2006). Membership and online groups. of adverse events in patient online discussions:
e-Society 2006: IADIS International Conference, customer relationship management” International
Dublin, Ireland (pp 290-293). Journal of Healthcare Information Systems and
Informatics, Vol. 3, Issue 3, pp 77-85

240
 241

Chapter 17
Towards Process-of-Care
Aware Emergency Department
Information Systems:
A Clustering Approach to Activity
Views Elicitation
Andrzej S. Ceglowski
Monash University, Australia

Leonid Churilov
The University of Melbourne, Australia

aBstract

The critical role of emergency departments (EDs) as the first point of contact for ill and injured patients
has presented significant challenges for the elicitation of detailed process models. Patient complexity
has limited the ability of ED information systems (EDIS) in prediction of patient treatment and patient
movement. This article formulates a novel approach to building EDIS Activity Views that paves the
way for EDIS that can predict patient workflow. The resulting Activity View pertains to “what is being
done,” rather than “what experts think is being done.” The approach is based on analysis of data that
is routinely recorded during patient treatment. The practical significance of the proposed approach is
clinically acceptable, verifiable, and statistically valid process-oriented clusters of ED activities that
can be used for targeted process elicitation, thus informing the design of EDIS. Its theoretical signifi-
cance is in providing the new “middle ground” between existing “soft” and “computational” process
elicitation methods.

IntroductIon the design and development of IS (Mertins, Bernus,

& Schmidt, 1998). The requirements definition is
Information system (IS) design principles call for a document that outlines all the needs that users
requirements definition as an intermediate stage in require of the prospective system. The requirements

Copyright © 2011, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 Towards Process-of-Care Aware Emergency Department Information Systems

definition is designed to allow for the translation of 2. what data are relevant as inputs to, or outputs
the physical needs of a process into an automated from these activities (data and output views);
environment. Programmers should be able to work 3. who performs each activity and where it is
from this document without going back to the users performed (an organisational or resource
for clarification. The requirements definition can thus view); and
“be compared to a schematic of a plan or a diagram 4. when and how the process activities are
of how a technical device works” (Langer, 2008). being performed (a control view).
The requirements definition commonly takes
the organisation from a functional view of activi- Information systems designed for hospital
ties (who, in which department, does what) to a emergency departments (ED), also known as
process-oriented view of operations (what hap- emergency rooms (ER) or accident and emer-
pens, when, and where). This shift from functional gency departments (A&E), are commonly called
data structuring to process event recording is best emergency department information systems
described in process models. Process models are (EDIS). EDIS as they exist today mainly address
formalised representations of the activities enacted two aspects of ED operations: (i) providing for
by a human or a machine that are considered im- tracking of patients and (ii) making patient infor-
portant to the achievement of the objective of the mation available to clinicians and administrators
process (Dumas, van der Aalst, & ter Hofstede, (Figure 1).
2005). The process models provide a structured The first function, having patient information
framework for IS specification and design, one available online, promotes efficiency of opera-
example is ARIS, a widely used reference archi- tions through enhanced data entry capabilities,
tecture and methodology (Scheer, 1999), and allow transferability between hospital departments and
configuration of the IS to support or control the locations, and potential for bedside update of
flow of work in the operational process (Rozinat patient records. Electronic health records have
& van der Aalst, 2008). Comprehensive process wide reaching implications for EDIS design that
models combine different views (Poulymenopou- are being addressed through various electronic
lou, Malamateniou, & Vassilacopoulos, 2003; health record initiatives (e.g., OpenEHR in Aus-
Seltsikas, 2001) that describe: tralia; GEHR in Europe; and HL7 in the USA).
The second function of existing EDIS relates to
1. what activities are being performed within patient flow management. Most systems provide
a process and their interactions (an activity a list of patients awaiting treatment along with
view); presenting problem, urgency, and disposition. This

Figure 1. Current EDIS deal with patient management retrospectively. There is an absence of informa-
tion about the pathways future treatment might take and likelihood of patient admission.

P atient m anag em ent (trackin g) C lin ical r ecords s uppo rt

P a tien ts w aiting f or: Labora to ry, Im agin g P a tient sum m aries: eH ealth R ecord s:
 R egistratio n and P harm ac y:  D em ographics  D iagnosis
 T riage  O rd er s tatus  P revious history  P ro c edures
 A R oom  U rgency  T re atm e nt
 A D octor or N urse R esources:  P re senting P roblem  D isposition
 A dm ission  R oo m A vailability  S ta tus  F ollo w -up
 A p ro cedure  E qu ipm ent l ocation

242
Towards Process-of-Care Aware Emergency Department Information Systems

allows clinicians to select patients for treatment ability to logically represent the groups of activi-
and to see when test results are available. Many ties that occur in the course of patient treatment
systems can provide tracking while patients are in within ED through a clear and comprehensive
the ED. Patient tracking is passive with the system activity view constitutes a necessary (although
reporting the location of patients in an effort to not sufficient) condition for moving EDIS to a
assist with efficient use of resources. process aware basis.
Unfortunately, existing EDIS do not have The objective of this article is to: (i) describe
predictive functions. This limits their capacity an activity view that supports the requirements
to provide prediction of workflow (where patients definition of process aware systems and (ii) provide
will go next) and optimisation of throughput a method for the development of this activity view.
(by minimising queues within the system) and The search for clusters of activities within
resource use (by making sure that resources are existing ED patient data records represents a new
aware of the impending need). In this article, middle ground between “soft” (or ethnographic)
we argue that existing EDIS are unable to pro- methods for the elicitation of process models
vide such benefits because every workflow is (Bustard, Oakes, & He, 1999) and “hard” methods
instantiated with every patient presentation (or such as workflow (or process) mining (Business
“instance”). By this, we mean that the EDIS sees Process Management Center, 2003) that arrive
every patient presentation as a unique instance directly at process models from data analysis.
that will follow an undefined sequence of activi- The article’s contribution to practice lies in the
ties and use unknown resources—there are no potential to design process aware EDIS since the
“template” processes programmed into the EDIS activity view developed in this article fits with
that might be followed, even if the treatment is standard IS design methodologies. The formula-
highly reproducible. In making EDIS “process tion in this article of an activity view for EDIS
aware,” the system is able to apply process tem- design is new because of the lack of existing of
plates to patients and so is capable of predicting comprehensive process models for patient treat-
the pathway that a patient might follow through ment in EDs (Djorhan & Churilov, 2003).
the ED, what resources might be required, and
what the overall state of the ED might be at any Background: context and Motivation
time. Process awareness does not imply only that for process aware edIs
the system designers considered processes when
they specified the system requirements, but rather EDIS focus on patient management and data
that the systems allow for explicit definition of the retrieval and storage (Amouh, Gemo, Macq, et
process logic, execution, and monitoring. al., 2005). Patient management aspects of EDIS
If EDIS are going to progress beyond the have functions that prioritise patients according
limitations of existing systems, then work flows to their urgency, allocate beds and rooms, and
need to be defined so that patient instances can register occurrence of key events such as nurse and
be linked to specific treatments. Once this is doctor assessments and patient departure. EDIS
achieved, it will be possible to build higher level data management revolves around the transport
understanding of ED operations. This understand- and storage of electronic patient records and the
ing will allow decision support functions such as ordering and reporting of pathology tests and
prediction of resource need, optimisation of bed imaging investigations. Supplementary hardware
use, and warning of impending blockage. such as wireless devices can track patients and
The principles of IS design described in the enhance the gathering of data for both patient
opening paragraphs clearly indicated that the and data management (Amouh et al., 2005). The

243
 Towards Process-of-Care Aware Emergency Department Information Systems

systems do not forecast patient flow and resource and reroute any step in each of the patient care
use, nor do they facilitate exchange of information processes (Rucker, 2003).
with outside organisations. EDIS functionality This extreme view, while attractive to propo-
supports resource allocation and patient selection nents of workflow systems (systems that automati-
for treatment based on priority (Aronsky, Jones, cally direct work throughout a process), is unlikely
Lanaghan et al., 2007, gives an example of this), to be achieved in the real world. The ED environ-
but it fails to coordinate patient movement and ment is simply too complex, too many decisions
patient treatment (for instance, there is no tracking need to be made by clinicians based on patient
of patients sent for X-rays, estimation of the time observations that are impossible to include in any
they will take at the X-ray location, nor booking of IS, and patient variation is too vast to be included
space at the next resource, such as the plaster cast in a single configurable workflow. However, this
room, that they might require once their X-rays extreme view can guide thinking towards EDIS
have been assessed.). that support real processes and provide avenues
An extreme view of an “ideal” process aware for decision support while remaining sufficiently
EDIS would be one that is able to incorporate flexible to sustain the unique treatment needs of
every step in patient care in which human hand- individual patients. This is impossible without
offs are automated, each step in patient treatment understanding the interaction of activities involved
automatically logged and tracked, and timing and in the processes, as depicted in the explicit process
sequencing of steps analysed for performance aware EDIS/activity view relationship presented
evaluation. Human interactions with networked in Figure 2.
electronic devices such as personal computers, CT Thus, the research problem addressed in this
scanners, lab systems, telephones, IV pumps, and article is to how to elicit an activities view for an
wireless patient tracking tags would be linked to EDIS. The following sections will discuss (i) the
the EDIS for automation of process control. Order methods available for elicitation of activity views
entering by physicians, bar-coding of medication and (ii) the work that has been done to group ED
by nurses, patient registration by clerks, and sur- treatment processes.
gery scheduling for surgeons should be linked
and coordinated automatically. Overall, the EDIS
should be able to sequence, monitor, track, alert,

Figure 2. Process aware EDIS are supported by diverse views: organisational, data, control, activity
and output views. Existing EDIS tend to be derived solely from listings organisational (resource) and
data views and so are only able to encompass a part of ED operations.

P rocess E D tr eatm ent

A w are E D IS F acilitate O perations

E ntity
S upport D epend o n
R elationship
F igure 1

diverse Include R esources

View(s) and d ata e
m

244
Towards Process-of-Care Aware Emergency Department Information Systems

The Elicitation of Activity analysis of logged data such as ARIS-Process

Views and Views of ED Performance Manager (IDS-Scheer, 2004).
Activities Process mining provides an excellent solution to
process elucidation in situations where workflow
Activity View Elicitation is fully automated (“mature” information system
applications) but little process elicitation help is
The most common way in which activities available for less mature “case handling” applica-
are elicited is through interviews with experts tions where process is undefined or much of the
and people who perform the work (Earl, 1994; work is performed by an expert who initiates
Kotonya & Sommerville, 1998; Weerakkody & activities based on the particulars of each job
Currie, 2003). This is frequently termed an eth- (van der Aalst et al., 2003).
nographic approach (Schuler & Namioka, 1993). Detailed event logs are seldom available for
As a technique for building the Activity View, activities that take place outside computer systems.
it is prone to subjective views of the work that Fortunately, “computer external” activities are
may be distorted according to social dynamics often logged in databases for billing and other
unrelated to the work (Rennecker, 2004) and may purposes. Such information may be captured in
encounter situations where the interviewees are batches after the activities have been completed
unable to provide generalized pictures of the work but lack information about sequence or timing
(Gospodarevskaya, Churilov, & Wallace, 2005). of events. The existence of these nonsequential
EDs appear to fall into the latter class. “activity logs” provides an avenue for the iden-
Process mining (van der Aalst et al., 2003) is a tification of patterns of activity. Activity logs of
purely computational technique that is able to ex- this sort are commonly associated with ill-defined
tract feasible “as is” process models directly from processes where experts make complex decisions
workflow data without resorting to interviews. It while performing the work (such as hospital EDs).
has its roots in a data mining idea that associations The binary “event” logs do not have the detail
between variables in a relation can be counted necessary for process mining and the mere volume
and granted some level of confidence (Agrawal, of data is likely to make the logs inscrutable to
Imielinski, & Swami, 1993). Combination of traditional data analysis techniques (past efforts
this concept with inference algorithms (Angluin in this regard are discussed in the next section),
& Smith, 1983) provided a way in which time- but they do provide an avenue for the identification
series data could be mined (Cook & Wolf, 1998) of patterns of activity if non-traditional methods
to retrospectively build a picture of sequences are used. The use of these logs in clustering of
of events in software (Agrawal, Gunopulos, & activities falls between the “soft” ethnographic
Leymann, 1998). This idea has been extended and “hard” process mining that were introduced
so that the branches, loops, and joins common in above (depicted in Figure 3). The clustering
most processes may be inferred from event logs “compromise” approach will be described later
(de Medeiros, van der Aalst, & Weijters, 2003). in this article, but it is first necessary to describe
Process mining requires access to a log that existing views of ED activities in order show to
records the sequence of defined tasks in workflow current understanding of ED processes.
for a large number of cases. Such data is readily
accessible for most work that takes place within Existing Views of ED Activities
or on computer systems, as evidenced by busi-
ness activity management software tools for the There are well over a thousand diagnoses listed
in the Victorian Emergency Data Set guidelines

245
 Towards Process-of-Care Aware Emergency Department Information Systems

Figure 3. An extension of Figure 2, showing that elicitation is achieved through social and technical
elicitation extremes. The middle ground is explored in this paper.

P rocess E D tr eatm ent

A w are E D IS F acilitate O perations

E ntity
S upport D epend F igure 2
R elationship on

diverse R esources
Include and d ata
View(s)

D erived fr om W orkflow R equire S equential

M ining Logs

T he technique
elicitation C lustering R equire Lists of described in
Methods activities this paper

E thnographic R equire D escriptions

(2007), and many other governments have a simi- Hoffenberg, Hill, & Houry, 2001; Walley, 2003).
lar proliferation of diagnoses. It is not feasible to These approaches address the activities related
incorporate separate activity views for every one to patient movement (whether patients are likely
of these diagnoses in a requirements definition to be admitted to hospital or not, or whether they
because this will probably lead to a situation where may be treated in a chair, rather than a bed, for
overhead costs for the control systems surpass the example) but not the detailed activities involved
benefit of efficient coordination of every variant in patient treatment.
(Becker, Kugeler, & Rosemann, 2003). Simpli- Patient movements have been analysed in
fication is necessary to reduce the large number simulation studies (Brailsford, Churilov, & Liew,
of diagnoses and permutations of treatment to a 2003; Mahapatra, Koelling, Patvivatsiri et al.,
number suitable for cost effective implementation 2003; Sinreich & Marmor, 2004), but the models,
in an IS. Some grouping of the activities involved in keeping with the philosophy of discrete event
in patient treatment is needed in order to reduce simulation, tend to imitate the physical move-
the complexity of the requirements definition. ment of patients through the ED, rather than the
Without this grouping, it is difficult to conceive treatment provided to patients. The studies result
systems that can indicate what the next step in in recommendations on how to reduce wait or fa-
patient care may be, where it should take place, cilitate throughput but they fail to provide insight
and who should be responsible for it. into the management of patients in accord with
Groupings and simplifications have been at- their treatment requirements.
tempted many times in the past. Analysts have Clinicians have tried to get a simplified picture
used output data such as “the time treatment of ED operations by grouping patient “cases,”
takes,” and input data such as “the complexity characteristically according to combinations of
of presentation” to group patients (for example, age, urgency of complaint, diagnosis, time in

246
Towards Process-of-Care Aware Emergency Department Information Systems

ED, and outcome of visit linked to cost (Bond, process aware because treatment activities have
Baggoley, Erwich-Nijhout et al., 1998; Cameron, not been functionally related in the activity view,
Baraff, & Sekhon, 1990; Jelinek, 1995). These treatment objectives are not available for the
clinical classification schemes have failed to processes, and supporting applications (bedside
group patients by similarity of treatment because monitoring, for instance) cannot be integrated into
they have used the use of cost as an objective the process. Elucidation of the activity view is
function. Low and high cost attributes trade off, necessary to allow coordination of the resources,
so dissimilar patients end up in similar groups. communication, and technology associated with
Clinical guidelines (Clinical Pathways) exist treatment.
that detail every aspect of patient treatment and Since the activity view is, in its simplest form,
act as checklists for common chronic ailments a collection of objects such as activities, objectives,
(for example, the Action plan for anaphylaxis and software applications, clinical procedures that
available at http://www.medicalobserver.com. are manually recorded (as described at the end of
au/clinicalguidelines. Porter, Cai, Gribbons et al. the previous section) may be used to provide a
(2004) provide an example of decision support for library of activities. Even though there is a wide
a single treatment). Clinical Pathways provide a range of patients and presentations, much of the
notion of a sequence of prescribed activities in- patient treatment related work in EDs is based on
volved in patient treatment that would be ideal for application of a short list of medical procedures
a process aware EDIS. Unfortunately, they only such as patient observation, administration of
cover a small number of narrow treatments so they drugs, and laboratory and imaging investigations.
do not provide the variety of ED treatments that Just 62 procedures are used for reporting ED
would need to be included for predictive decision treatment in Victoria, Australia (Metropolitan
support. A different approach is needed to unravel Health and Aged Care Services Division of the
the complexity of patient treatment in EDs. Such Victorian State Government, 2007). Thirty six
an approach is described in the following sections. procedures account for 99% of all procedures in
Victorian hospitals. Almost 17% are classed as
“Other,” which includes observation of patients
MetHod descrIptIon: FroM by medical staff; 6% are “No procedures”; some
clInIcal procedures to an 10% are drug administrationl; and over 9% X-
actIVIty VIew For a process ray imaging. Other significant procedures are
aware edIs venipuncture, intravenous catheter access in
preparation for infusion of fluid or drugs, and
Mertins et al. (1998) suggest that it is possible to echocardiogram diagnostics (figures derived from
build information systems from a subset of the Victorian Emergency Medical database for 2002).
organisational, data, control, activity, and output An activity view that simply lists all these
views (defined as aspects of process models in procedures would result in processes models
the Introduction). This approach is apparent in based on a vast number of possible permutations
the design of most existing EDIS, where the data of causal linkages between all procedures. Indi-
and organisational views are well represented (for vidual procedures could have associated objec-
exampl, the linking of Figure 1 to the resources tives but it would be difficult to understand how
and data component in Figure 2), but the activity the objectives were related to each other. Some
view is less well represented or not present at all form of simplification is necessary to advance the
(Rucker, 2003). The resulting EDIS’ cannot be activity view beyond this primitive.

247
 Towards Process-of-Care Aware Emergency Department Information Systems

The grouping of procedures into treatment ValIdatIon tHrougH exaMple:

specific clusters is a way of providing this sim- clusterIng clInIcal
plification—procedures are logically linked as procedures to BuIld
activities within treatments. While the grouping an actIVIty VIew
of activities in this way does not provide a true
process view (the activities are not sequenced), it Victorian Emergency Medical Data (VEMD)
significantly reduces the number of permutations made up of de-identified records of all ED presen-
in order to provide a simplified view of ED op- tations across 31 anonymous hospital campuses
erations. It is this simplified view that has eluded was obtained and five similar-sized campuses (by
designers of ED information systems up to now. number of records) were selected for analysis and
Each group of activities, or treatment cluster, may comparison. Each record contained demographic
be scrutinised to determine the most likely (or particulars and details of the visit such as “ap-
most desirable) sequence of activities. parent severity of complaint,” “key time points,”
The grouping of procedures for this activity and “disposition,” plus all medical procedures
view may be achieved through nonparametric performed, but cost data was not available.
clustering. One nonparametric technique is self All cases where patients underwent more than
organising mapping (SOM) (Kohonen, 1995). The one procedure were included for analysis. This was
SOM is a grouping technique that is algorithm generally around 60% of all patient presentations.
driven and relies on data rather than domain- The 13 least-used procedures were omitted from
specific expertise. The objectives of the technique the analyses, as was the “NONE” procedure. These
are to minimise diversity within groups and to exclusions totalled less than 1% of procedures in
maximise differences between groups. The tech- cases where patients had more than one procedure.
nique generally employs large datasets, works Random samples of approximately 10,000 cases
well with many input variables, and produces having Departure Status “Discharged home” were
arbitrarily complex models unlimited by human extracted. Where there were less than 10,000 cases
comprehension (Kennedy, Lee, Van Roy et al., with this departure status at a campus, all cases
1998). were included for analysis.
Self organising maps provide a visual under- The data was saved into a tab delimited format.
standing of patterns in data through a two dimen- Each record had 50 fields—a record identifier and
sional representation of all variables. Records 49 procedures coded as “0” for “not performed” to
that have similar characteristics are adjacent in 1,2,3...16 for repeated applications. The procedure
the map, and dissimilar records are situated at a data was sparse, but the number of records gave
distance determined by degree of dissimilarity. some assurance that patterns of recurring groups
Viscovery SOMine, the software tool used in this of procedures could be found. Self organising
analysis, employs a variant of Kohonen’s Batch- mapping was applied to the data using the same
SOM (Kohonen, 1995), enhanced with a scaling clustering settings. These settings included:
technique for speeding up the learning process
(Eudaptics Software Gmbh, 1999). • The initial map height that gave the initial
The concept of providing a simplified activity number of vertical nodes in the two dimen-
view of treatment through the grouping of proce- sional representation of the data (the number
dures is illustrated in the next section. of horizontal nodes was automatically set
from the map ratio);
• A scaling factor that specified the increase
in horizontal number of nodes as the clus-

248
Towards Process-of-Care Aware Emergency Department Information Systems

tering algorithm progressed through each • 12 Lead ECG (1.0), with Peripheral IV
growth step (the vertical number of nodes catheter and Venipuncture (0.6);
was adjusted according to the desired map • 12 Lead ECG and ECG monitoring (1.0),
ratio); with Venipuncture (0.9);
• The influence radius of the neighbourhood • Peripheral IV catheter and IV drug infusion
interaction of the intermediate maps (the (1.0), with Venipuncture (0.7);
reach of the Gaussian neighbourhood func- • Suture, Steri-strip, glue (1.0) with Dressing
tion). A high intermediate tension “averages” (between 0.6 and 0.9);
the data distribution, while a low tension • Plaster of Paris (1.0) and X-ray (0.9) with
allows adaptation to display finer features. Drug administration (around 0.5);
• The number of iterations and the Wegstein • Splint (1.0), X-Ray (0.8) with Drug admin-
factor (a convergence parameter in the istration (between 0.2 and 0.6);
batch SOM algorithm comparable to the
momentum factor as it is commonly used These clusters each comprise between 2% and
in supervised neural network algorithms). 9% of patients and add up to between 20% and
30% of patients. In addition to these widespread
Between 13 and 27 clusters were identi- clusters, there are typically two to three large
fied across the five data sets. The clusters were clusters at each campus that provide for some 25%
validated through a range of internal measures of of presentations. These large clusters often include
cluster quality. These measures were analogous drug administration, venipuncture and full ward
to traditional indicators of cluster quality such as test (urine) or 12 Lead ECG, but the proportions
the Rand Statistic, Jaccard Coefficient, and the differ between campuses. It is important to note
Folkes and Mallows Index; however, they operated that these clusters represent “core treatment” ac-
on two dimensions, rather than the single figure tivities within the EDs. These treatments employ
of the traditional measures. Maps that displayed resources most frequently
the frequency of records across two dimensional The truncated example in Table 1 indicates that
cluster space, the quantisation error, the proximity the patient was likely to be experiencing breathing
of nodes to neighbours, and curvature of the map difficulties (possibly an asthma attack) and might
through n-dimensional space were scrutinised to have been classed as urgent or nonurgent (note
ensure that the spread of clusters was even, the that this is in direct contrast to previous studies,
clustering error within reasonable limits, and the which have segmented patients by urgency—a
cluster shapes regular. patient movement, rather than patient treatment
This clustering of procedures into prototype orientation). Nebulised medication was provided
“workflows” was verified through discussions to all patients. This treatment was supported with
with a specialist in emergency medicine and X-rays, drugs, venipuncture, electrocardiogram,
found to be clinically sensible. It was concluded peripheral intravenous catheter, and full ward
that the clusters reflected “as is” core ED treat- tests. Two campuses also indicated that intrave-
ments. Similar sets of procedures were apparent nous injections were commonly recorded in this
in clusters across campuses. Typical prototype group. The numbers in Table 1 may be considered
workflows that were common across multiple to represent the likelihood between 0 and 1 of
campuses are (figures in brackets indicate likeli- patients in that cluster undergoing the procedure.
hood of procedure): In the example given in Table 2, it can be seen
that there is much agreement across campuses
regarding the reporting of “head injury observa-

249
 Towards Process-of-Care Aware Emergency Department Information Systems

Table 1. Nebulised medication procedures across three hospital campuses, A, B, and C

Campus A B C
Procedure Description % Patients 2.9% 2.4% 3.6%
Nebulised Medication NEB 1.00 1.00 1.00
X-ray XRAY 0.43 0.26 0.47
Oral/ sublingual/ topical/ rectal drug administration DRUG 0.81 0.89 0.80
Venipuncture VB 0.28 0.21 0.29
12 Lead ECG ECG 0.16 0.14 0.15
Peripheral intravenous catheter (IV access) IV 0.28 0.22 0.24
Full ward test - urine FWT 0.06 0.09 0.06
Intravenous infusion IVS 0.13 0.12

Table 2. Head injury observation across five campuses

Campus

% Patients

HIO

XRAY

IV

IVI

VB

DRS

SUT

ECG

FWT

INF

DRUG

CT

IVS

RBG

SPR
A 2.3% 1.0 0.4 0.5 0.0 0.4 0.1 0.1 0.3 0.2 0.2 0.5 0.1 0.3 0.2 0.0
B 0.9% 1.0 0.4 0.4 0.1 0.3 0.2 0.2 0.1 0.3 0.1 0.5 0.1 0.2 0.1 0.0
C 4.2% 1.0 0.5 0.4 0.0 0.5 0.1 0.0 0.4 0.1 0.2 0.6 0.2 0.2 0.0 0.0
D 3.4% 1.0 0.3 0.1 0.2 0.2 0.1 0.1 0.1 0.1 0.0 0.0 0.0 0.0 0.0 0.0
E 0.6% 1.0 0.4 0.3 0.0 0.4 0.2 0.1 0.3 0.3 0.1 0.4 0.2 0.1 0.5 0.5
Key: HIO = Head Injury Observation; XRAY = X-ray examination; IV = Peripheral intravenous catheter; IVI =IV
drug infusion; VB = Venipuncture; DRS = Dressing; SUT = Suture, steristrip, glue; ECG = 12 lead Electrocardio-
gram; FWT = Full Ward Test of urine; INF = Infusion of fluid (not blood); DRUG = Oral, sublingual, topical, rec-
tal drug administration; CT = Computerised tomography scan; IVS = Intravenous drip; RBG = Test of Random
Blood Glucose; SPR = Spirometry

tion” (HIO), despite the varying percentage of nature of the core treatments (i.e., one procedure
patients (second column of the table). The campus might be used in multiple core treatments) that
in the last row displays anomalous random blood arose for the clustering technique made the tech-
glucose (RBG) and Spirometry (SPR) that may nique far superior (in defining treatments that were
indicate different patient and treatment profiles. clinically realistic) to other methods that were
Similar studies were carried out for admitted attempted such as association mining and CART.
as well as discharged patients and core treatments This section set out to describe the realisation
identified that overlapped the above and expanded of the idea that a simplified activities view could
to reflect the treatment accorded to patients who be developed if procedures involved in patient
were later admitted to hospital. The studies were treatment were clustered into process-oriented
reinforced with text mining exercises that com- groups. Distinct clusters of procedures were
pared patient reason for presentation with their identified in the treatment of patients at a number
treatment and through simulation of treatment of EDs. The grouping was validated and verified
activities. It was also found that the nonexclusive and considered a reasonable “as is” simplification

250
Towards Process-of-Care Aware Emergency Department Information Systems

of ED activities into core treatments (Ceglowski, The core treatments add knowledge about
Churilov, & Wassertheil, 2007, describes simula- common treatment activities that can be further
tion of ED operations based on treatment clusters). explored using various techniques, such as eth-
The implications of these findings are discussed nographic studies or additional data investiga-
in the next section. tions. The control view can be structured once
sequencing and mapping of cause-effect linkages
has been done (possibly through expert clinical
dIscussIon: How clustered input) because the roles and responsibilities may
actIVItIes support systeMs be identified and linked to each activity within
desIgn the treatment. This moves EDIS towards process
aware systems that can not only coordinate patient
In providing core treatments, the previous section treatment and movement, but predict them, too.
has expanded the simple activity view of individual The clustering technique may further be ex-
procedures to one with grouped procedures in a tended through use of techniques such as associa-
handful of frequently applied core treatments. The tion mining to provide more detailed segmenta-
simplification reduced the number of objects from tions of the groups and better understanding of
62 procedures to around 20 core treatments. This the business rules, especially if the data is linked
reduction in complexity facilitates the association to attributes such as urgency, age, and whether the
of activities with other aspects of the activities patient was admitted or not. The insights provided
view such as process goals because each core by such investigations give rise to another, more
treatment represents a process-oriented class of refined, activity view where patient attributes may
treatment. be causally linked to treatment activities, making
The primary benefit of identifying prototype prediction of patient routings feasible.
workflows and linking them to patient attributes
has been the potential for predictive decision
support. Certain core treatments take longer than conclusIon
others take and have a higher rate of hospital ad-
mission associated with them. When many patients This article indicated that existing EDIS suffer
requiring such long duration treatments reside in from inadequate understanding of patient treat-
the ED concurrently, then throughput slows and ment processes that manifest as EDIS that are un-
the ED becomes prone to blockage—regardless able to predict future patient treatment and patient
of its capacity. By linking patient presentation to movement needs. It was noted that development of
core treatments and the time they take, it is pos- an activities view would assist movement towards
sible to update prediction of potential blockage process aware, predictive EDIS. A suggestion was
as each new patient presents at the ED. made that clinical procedures recorded for other
An extension of this high level prediction is purposes could be used to populate the activity
the potential for prediction of future workflow view, but simplification was necessary in order
requirements. Having patient presentation linked to reduce the number of possible process models
to their most likely future treatment allows their that could be deduced. A method for simplification
future workflow to be mapped and planned for, was described that resulted in procedures being
so handoffs can be identified, responsibilities grouped in clinically acceptable core treatments
indicated, and resources specified. In this way that covered 99% of ED treatment operations.
the organisational, data, and output views may The utility of these core treatments in providing
be linked to core treatments. an activity view was discussed.

251
 Towards Process-of-Care Aware Emergency Department Information Systems

This article has described a route to take ex- Agrawal, R., Imielinski, T., & Swami, A. (1993).
isting EDIS (which deal primarily with the man- Mining association rules between sets of items in
agement of historic data such as patient records large databases. In ACM SIGMOD Conference on
and the queue of patients awaiting treatment) to Management of Data, Washington, DC.
event-based systems that can “look forward” to
Amouh, T., Gemo, M., Macq, B., et al. (2005).
predict what pathways patients might take through
Versatile clinical information system design for
the ED. EDIS that are capable of managing
emergency departments. IEEE Transactions on
workflows (predicting the work to be performed
Information Technology in Biomedicine, 9(2),
at each successive stage of treatment) facilitate:
174-183.
• Logistics management, by “booking” re- Angluin, D., & Smith, C.H. (1983). Inductive infer-
sources for when patients will need them; ence: Theory and methods. Computing Surveys,
• Predictive reporting, by linking presenta- 15(3), 237-269.
tion problem and severity with likely length
Aronsky, D., Jones, I., Lanaghan, K., et al. (2007).
of stay, probably treatment, and whether a
Supporting patient care in the emergency depart-
patient might be admitted to a ward; and
ment with a computerized whiteboard system.
• Management decision support, through the
Journal of the American Medical Informatics
ability to assimilate the status of patients in
Association.
the ED with the new presentations awaiting
treatment. Becker, J., Kugeler, M., & Rosemann, M. (2003).
Process management: A guide for the design of
Clustering represents an additional tool in the business processes. Berlin, Heidelberg, New
elicitation of process models for EDIS design. It is York: Springer.
one that is not limited in breadth in the way that
Bond, M., Baggoley, C., Erwich-Nijhout, M., et
ethnographic studies often have to be—all history
al. (1998). Urgency, disposition and age groups:
(as it exists) can be included in the analysis. Nor
A casemix model for emergency departments.
does clustering have the requirement of workflow
Emergency Medicine Australasia, 10(2), 103-110.
mining that detailed logs exist—simple activity
records will suffice to provide a working activities Brailsford, S., Churilov, L., & Liew, S.K. (2003).
view to supplement existing organisational, data, Treating ailing emergency departments with
and input views for EDIS design. Transforming simulation: An integrated perspective. In Inter-
such a supplementation into a truly integrated national Conference on Health Sciences Simula-
process aware EDIS design presents a challenging tion. Orlando, FL: The Society for Modeling and
direction for future research. Simulation International.
Business Process Management (BPM) Center.
(2003). Process mining. Retrieved May 24,
reFerences
2008, from http://tmitwww.tm.tue.nl/research/
processmining/
Agrawal, R., Gunopulos, D., & Leymann, F.
(1998). Mining process models from workflow Bustard, D., Oakes, W.R., & He, Z. (1999). Models
logs. In Sixth International Conference on Ex- to promote effective system change. In IEEE In-
tending Database Technology, Valencia, Spain. ternational Conference on Software Maintenance
(ICSM ‘99).

252
Towards Process-of-Care Aware Emergency Department Information Systems

Cameron, J., Baraff, L., & Sekhon, R. (1990). Hoffenberg, S., Hill, M.B., & Houry, D. (2001).
Case-mix classification for emergency depart- Does sharing process differences reduce patient
ments. Medical Care, 28, 146-158. length of stay in the emergency department?
Annals of Emergency Medicine, 38(5), 533-540.
Ceglowski, A., Churilov, L., & Wassertheil, J.
(2007). Combining data mining and discrete IDS-Scheer. (2004). ARIS process performance
event simulation for a value-added view of a manager. Retrieved May 24, 2008, from http://
hospital emergency department. Journal of the www.ids-scheer.com
Operational Research Society, 58(2), 246-254.
Jelinek, G.A. (1995). Casemix classification of pa-
Cook, J.E., & Wolf, A.L. (1998). Discovering tients attending hospital emergency departments
models of software processes from event-based in Perth, Western Australia: Development and
data. ACM Transactions on Software Engineering evaluation of an urgency-based casemix system.
and Methodology, 7(3), 215-249. Perth, Australia: University of Western Australia.
de Medeiros, A.K.A., van der Aalst, W.M.P., Kennedy, R., Lee, Y., Van Roy, B., et al. (1998).
& Weijters, A.J.M.M. (2003). Workflow min- Solving data mining problems through pattern
ing: Current status and future directions. In R. recognition. Prentice Hall.
Meersman, Z. Tari, & D.C. Schmidt (Eds.), On
Kohonen, T. (1995). Self-organizing maps. Berlin,
the move to meaningful Internet systems 2003:
Germany: Springer.
CoopIS, DOA, and ODBASE (pp. 389-406). Berlin,
Heidelberg: Springer-Verlag. Kotonya, G., & Sommerville, I. (1998). Require-
ments engineering: Processes and techniques.
Djorhan, V., & Churilov, L. (2003). Business
New York: John Wiley.
interactions in an acute emergency department
in Australia: A clinical process modeling per- Langer, A.M. (2008). Analysis and design of
spective. In the 6th Pacific Asia Conference on information systems. London: Springer (pp. 418;
Information Systems (PACIS 2002). Tokyo, Japan: p. 249 illus.).
The Japan Society for Management Information
Mahapatra, S., Koelling, C.P., Patvivatsiri, L.,
(JASMIN).
et al. (2003). Pairing emergency severity index
Dumas, M., van der Aalst, W., & ter Hofstede, 5-level triage data with computer aided system
A. (2005). Process-aware information systems: design to improve emergency department access
Bridging people and software through process and throughput. In The 2003 Winter Simulation
technology. Hoboken, NJ: Wiley-Interscience. Conference. New Orleans, LA: INFORMS.
Earl, M.J. (1994). The new and the old of busi- Mertins, K., Bernus, P., & Schmidt, G. (1998).
ness process redesign. The Journal of Strategic Handbook on architectures of information sys-
Information Systems, 3(1), 5-22. tems. Berlin, Germany: Springer.
Eudaptics Software Gmbh. (1999). Viscovery Metropolitan Health and Aged Care Services
SOMine. Wien: Eudaptics Software Gmbh. Division of the Victorian State Government.
(2007). Victorian emergency minimum dataset
Gospodarevskaya, E., Churilov, L., & Wallace,
(VEMD). Retrieved May 24, 2008, from http://
L. (2005). Modeling the patient care process of
www.health.vic.gov.au/hdss/vemd/2004-05/
an acute care ward in a public hospital: A meth-
manual/index.htm
odological perspective. Hawaii: HICSS.

253
 Towards Process-of-Care Aware Emergency Department Information Systems

Porter, S.C., Cai, Z., Gribbons, W., et al. (2004 Seltsikas, P. (2001). Organizing the information
). The asthma kiosk: A patient-centered technol- management process in process-based organiza-
ogy for collaborative decision support in the tions. In 34th Hawaii Int. Conf. Syst. Sci., Hawaii.
emergency department. Journal of the American
Sinreich, D., & Marmor, Y. (2004). A STAFFING
Medical Informatics Association, 11, 458-467.
emergency. Industrial Engineering, 36(6).
Poulymenopoulou, M., Malamateniou, F., &
van der Aalst, W.M.P., ter Hofstede, A.H.M., &
Vassilacopoulos, G. (2003). Specifying workflow
Weske, M. (2003). Business process manage-
process requirements for an emergency medi-
ment: A survey. In Business Process Manage-
cal service. Journal of Medical Systems, 27(4),
ment (BPM2003) (pp. 1-12). Berlin, Germany:
325-335.
Springer-Verlag.
Rennecker, J. (2004). Updating ethnography to
van der Aalst, W.M.P., van Dongena, B.F., Herbst,
investigate contemporary organisational forms.
J., et al. (2003). Workflow mining: A survey of
In M.E. Whitman & A.B. Woszczynski (Eds.),
issues and approaches. Data & Knowledge En-
The handbook of information systems research
gineering, 47(2), 237-267.
(pp. xi, 349). Hershey, PA: IGI Global.
Walley, P. (2003). Designing the accident and
Rozinat, A., & van der Aalst, W.M.P. (2008).
emergency system: Lessons from manufacturing.
Conformance checking of processes based on
Emerg Med J, 20(2), 126-130.
monitoring real behavior. Information Systems
Frontiers, 33(1), 64-95. Weerakkody, V., & Currie, W. (2003). Integrating
business process reengineering with information
Rucker, D. (2003). Finally, a tool to re-engineer
systems development: Issues and implications. In
health care: The workflow engine. San Francisco,
W.M.P. van der Aalst, A. H.M. ter Hofstede, &
CA: Montgomery Research, Inc.
M. Weske (Eds.), Business process management
Scheer, A.W. (1999). Business process frame- (BPM2003) (pp. 302-320). Berlin, Germany:
works. Berlin, Germany: Springer. Springer-Verlag.
Schuler, D., & Namioka, A. (1993). Participatory
design: Principles and practices. Hillsdale, NJ:
L. Erlbaum Associates.

This work was previously published in International Journal of Healthcare Information Systems and Informatics, Vol. 3, Is-
sue 4, edited by J. Tan, pp. 1-16, copyright 2008 by IGI Publishing (an imprint of IGI Global).

254
 255

Chapter 18
Applying Dynamic Causal
Mining in Health Service
Management
Yi Wang
Nottingham Trent University, UK

aBstract

This article describes an application that illustrates the role of data mining technology in identifying
hidden causal knoledge from health and medical data repositories. Across the health care and medical
enterprises, a wide variety of data is being generated at a rapid rate. Current information technologies
tends to focus on a more statical side of causal knowledge and do not address the dynamic causal knowl-
edge. This article shows that the dynamic causal relation data can be captured for treatment, payment,
operations purposes and administrative directed insights. Accessing this currently unrealized knowledge
potential would enable the delivery of actionable knowledge to medical practitioners, healthcare system
managers, policy planners and even patients to make a significant difference in overall healthcare.

lIterature reVIew specific clinical information and knowledge-based

information.
Medical and Health Management Knowledge management capabilities have
been incorporated in many clinical systems since
Patient record management systems is desired in the 1980s in order to provide a better understand-
clinical settings (Abidi, 2001; Heathfield & Louw, ing and management basis. In the HELP system,
1999; Jackson, 2000). The major reasons include decision logic was stored to allow it to respond
physicians’ significant information needs (Dawes to new data entered (Kuperman, Gardner, &
& Sampson, 2003) and clinical information Pryor, 1991). The SAPHIRE system performs
overload. Hersh (1996) classified textual health automatic indexing of radiology reports by utiliz-
information into two main categories: patient- ing the UMLS Metathesaurus (Hersh, Mailhot,

Copyright © 2011, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 Applying Dynamic Causal Mining in Health Service Management

Arnott-Smith, & Lowe, 2002). The clinical data information. Text mining is used to study of
repository at Columbia-Presbyterian Medical protein interaction network inference (Iossifov
Center (Friedman, Hripcsak, Johnson, Cimino, et al., 2004).
& Clayton, 1990) is a database used for decision System Dynamics has a number of strengths
support (Hripcsak, 1993). Another clinical data that make it especially useful in health care settings
repository is the University of Virginia Health (Roberts & Hirsch, 1976). Some of the System
System (Schubart & Einbinder, 2000). Case based Dynamics work in health care has been done
reasoning also has been proposed in Montani by Pugh-Roberts Associates (Hirsch & Miller,
and Bellazzi (2002). Janetzki, Allen, and Cimino 1974). The other area in which early work took
(2004) use a natural language processing approach place was community mental health (Levin &
to link electronic health records to online infor- Roberts, 1976).
mation resources. The Soundview-Throgs Neck community
examined the forces contributing to a rapid rise
in heroin addiction (Levin, Kirsch, & Roberts,
applyIng data MInIng In 1972). A generic model of ambulatory care was
MedIcal and HealtH developed to analyse economic performance of
ManageMent organization (Hirsch & Bergan, 1973). Another
model developed for the State of Minnesota exam-
Data mining has been used to extract diagnostic ined the factors affecting the success of new orga-
rules from breast cancer data (Kovalerchuk, nizations, called Human Service Boards (Hirsch,
Vityaev, & Ruiz, 2001). Data mining has also Bergan, & Frohman, 1974). Two comprehensive
been applied to clinical databases to identify new System Dynamics models were developed to aid
medical knowledge (Hripcsak, Austin, Alderson, manpower policy formulation in dentistry (Hirsch
& Friedman, 2002; Prather, Lobach, Goodwin, & Killingsworth, 1975) and nursing (Bergan &
Hales, Hage, & Hammond, 1997). Hirsch, 1976).
Dreiseitl, Ohno-Machado, Kittler, Vinterbo, One model reflected the effects of such factors
Billhardt, and Binder (2001) compare five classi- as the “technology gap” between the hospital and
fication algorithms for the diagnosis of pigmented nearby community hospitals (Hirsch, Forsyth,
skin lesions. This is similar to the measurement of Bergan, & Goodman, 1976). A model helped a
algorithmic performances in other areas applica- medical school examine problems in its relation-
tions (Yang & Liu, 1999). For example, Acir and ships with affiliated hospitals and restructure
Guzelis (2004) apply support vector machines those relationships accordingly (Stearns, Bergan,
in automatic spike signal detection in Electro- Roberts, & Cavazos, 1978). Models created for
EncephaloGrams (EEG). Kandaswamy, Kumar, project changes in areas’ populations, health
Ramanathan, Jayaraman, and Malmurugan (2004) care consequences of those changes, and shifts
use artificial neural network to classify lung sound in utilization patterns as a result of changes in
signals into different categories. resources available, insurance coverage, and vari-
Chrisman et al. (2003) incorporated biological ous health policies (Hirsch & Henderson, 1977).
knowledge and expression data using a Bayes- Several modelling efforts, in fact, were designed
ian framework. Imoto et al. (2003) incorporated to stop at the point where a set of causal diagrams
protein-protein interaction, protein-DNA inter- had been completed (Stearns, Bergan, Roberts,
actions, and transcriptional factor binding site & Quigley, 1976).

256
Applying Dynamic Causal Mining in Health Service Management

dynamic causal Mining in the database. By only considering large itemsets

of the previous pass, the number of candidate large
The Dynamic Causal Mining (DCM) algorithm itemsets is significantly reduced.
was discovered in 2005 (Pham, Wang, & Dimov, The idea of process mining was investigated
2005) using only counting algorithm to integrate in contexts of software processes and workflow
with Game Theory. It was extended in 2006 (Agrawal, Gunopulos, & Leymann, 1998). Cook
(Pham, Wang, & Dimov, 2006) with delay and and Wolf (1998) propose methods for process dis-
feedback analysis, and was further improved for covery in case of software sequential processes.
the analysis in Game Theory with Formal Con- Herbst and Karagiannis used a hidden Markov
cept analysis (Wang, 2007). DCM enables the model in the context of workflow management and
generation of dynamic causal rules from data sets other sequential processes (Herbst, 2000; Herbst
by integrating the concepts of Systems Thinking & Karagiannis, 1999, 2000). Maruster, van der
(Senge, Kleiner, Roberts, Ross, & Smith, 1994) Aalst, Weijters, van den Bosch, and Daelemans
and System Dynamics (Forrester, 1961) with (2001) suggest a technique for discovering the
Association mining (Agrewal et al., 1996). The underlying process from hospital data assuming
algorithm can process data sets with both cat- that the workflow log does not contain any noisy
egorical and numerical attributes. Compared with data. A heuristic method that can handle noise is
other Association Mining algorithms, DCM rule presented in Weijters and van der Aalst (2001).
sets are smaller and more dynamically focused.
The pruning is carried out based on polarities. system thinking and system
This reduces the size of the pruned data set and dynamics
still maintains the accuracy of the generated rule
sets. The rules extracted can be joined to create System thinking is based on the belief that the
dynamic policy, which can be simulated through component parts of a system will act differently
software for future decision making. The rest of when isolated from its environment or other parts
this section gives a brief review of Association of the system.
Mining and System Dynamics. Systems thinking is about the interrelated
actions which provide a conceptual framework
association Mining and process or a body of knowledge that makes the pattern
Mining clearer Senge et al. (1994). It is a combination of
many theories such as soft systems approach and
Association mining was discovered by Agrewal, system theory (Coyle, 1996). Systems thinking
Mannila, Srikant, Toivonen, and Inkeri (1996). seeks to explore things as wholes, through pat-
It was further improved in various ways, such terns of interrelated actions.
as in speed (Agrewal et al., 1996; Cheung, Han, System Dynamics can be defined as “a qualita-
Ng, & Fu, 1996) and with parallelism (Zaki, Par- tive and quantitative approach to describe model
thasarathy, Ogihara, & Li, 1997) to find interest- and design structures for managed systems in
ing associations and/or correlation relationships order to understand how delays, feedback, and
among large sets of data items. It shows attributes interrelationships among attributes influence the
value conditions that occur frequently together in behaviour of the systems over time” (Coyle, 1996),
a given dataset. It generates the candidate itemsets or “the purpose of the model is to solve a problem,
by joining the large itemsets of the previous pass not simply to model a system. The model must
and deleting those subsets which are small in the simplify the system to a point where the model
previous pass without considering the transactions replicates a specific problem”(Sterman, 1994).

257
 Applying Dynamic Causal Mining in Health Service Management

System dynamics is a tool to visualize and revision of the models. The stages are depicted
understand such patterns of dynamic complexity, in Figure 1.
which is built up from a set of system archetypes
based on principles in System thinking (Sterman, Stage 1: Problem definition. In this phase, the
2000). System dynamics visualizes complex problem is identified and the key variables
systems through causal loop diagrams. A causal are given. Also, the time horizon is defined
loop diagram consists of a few basic shapes, that so that the cause and effects can be identified.
together describe the action modelled. Stage 2: Data preparation. Data are collected
System dynamics addresses two types of from various sources and a homogeneous
behaviour, sympathetic and antipathetic (Pham data source is created to eliminate the rep-
et al., 2005). Sympathetic behaviour indicates resentation and encoding differences.
an initial quantity of a target attributes starts to Stage 3: Data mining. This stage involves trans-
grow, and the rate of growth increases. Antipa- forming data into rules. This thesis suggests
thetic behaviour indicates an initial quantity of using DCM as a data mining tool. The details
a target attributes starts either above or below a of DCM are explained in the following sec-
goal level and over time moves toward the goal. tions and next article.
Stage 4: Policy formulation. Policies are groups
of the rules extracted by mining techniques.
BasIc concepts oF Policies improve the understanding of the
dynaMIc causal MInIng system. The interactions of different policies
must also be considered since the impact of
This section provides a general framework for combined policies is usually not the sum of
DCM. It is an iterative and continual process of their impacts alone. These interactions may
mining rules, formulating policies, testing, and reinforce each other or have an opposite ef-

Figure 1. DCM process

Model Simulation

Policy Formulation

Data mining

Data preparation

Problem definition

258
Applying Dynamic Causal Mining in Health Service Management

fect. The policy can be used for behaviour In order to carry out DCM, the time stamps are
simulation to predict the future outcome. summarised or partitioned into equal-sized time
Stage 5: Model Simulation. This stage tests the stamps. A time stamp is useful for describing and
accuracy of the policies. The policies will prescribing changes to the systems and objects.
predict results for new cases so the manag-
ers can alter the policy to improve future data
behaviour of the system. It is necessary to
capture the appropriate data and generate a Definition 2. A dynamic attribute is the change
prediction in real time, so that a decision can or the difference between two attribute values
be made directly, quickly, and accurately. with consecutive time stamps. The two types of
value do not have the same nature. Let D denote
a data set which contains a set of n records with
dataset attributes {A1, A2, A3,… Am}, where each attribute
is of a unique type (for example; cost of medi-
To find dynamic causality among a set of attributes cine, treatment volume, inventory volume, etc).
means to identify correlation and interdependen- Each attribute is associated with a time stamp ti,
cies between them. The DCM algorithm is a way where i ={1,2,3,…n}. Let Dnew be a new database
of describing the state of a target system as it constructed from D such that dynamic attribute
evolves in time. It discovers dynamic causality ∆Am,∆t in Dnew is given by:
i
in a data set by matching the dynamic behaviour
between separated attributes. ∆Am,∆t = Am,t - Am,t (2)
i i+1 i

time stamp where m identifies the attribute of interest.

The classical Association Mining algorithms
Definition 1: A dynamic time stamp is created from can be applied only to data in the original form
two time stamps. Consider two time stamps ti and (attribute form), for example, in the market basket
ti+1. The dynamic time stamp ∆ti is equal to the problem (Agrawal et al., 1993) the focus is on the
difference between two consecutive time stamps. items of each purchase. On the other hand, DCM
is interested in the dynamic changes between
∆ti = ti+1 – ti (1) data. To apply DCM, the records are arranged in
a temporal sequence (t = 1, 2 ,…, n). Definition 2
Time stamps are used for identifying the is only for numerical attributes and the causality
range of variables. The size of each time stamp between categorical attributes in D can be identi-
is selected by the specific need and may vary in fied by examining the differences of correspond-
different situations. For instance, a time stamp ing changes in attribute values. An example of
for an increase in production may be in the order such a database is shown in Table 1. A1 and A2
of months, while for a change in a cell may be in represent attributes, such as from a tax database
the order of milliseconds. The time stamp also like income and tax.
can help to determine how detailed the variables
need to be. The attribute may increase or decrease Definition 3. In the case of categorical attributes,
dramatically if the time stamp is in the order the dynamic causal attributes can be identified by
of seconds, however it may be assumed to be joining the polarities of corresponding changes
constant if the time stamp is in the order of years. in attribute values. Let Dnew be a new data set
All the time stamps should be of uniform length.

259
 Applying Dynamic Causal Mining in Health Service Management

constructed from D such that attribute ∆Am,∆t in Table 1. Original database D

i
Dnew is given by:
Time A1 A2

∆Am,∆t = join(Am,t , Am,t ) (3) 1 9 2

i i+1 i
2 17 3
where join is a function combining Am,t and Am,t . 3 10 12
i+1 i
For example, ( t=1, Am,t = Red) and (t=2, Am,t = 4 4 16
i+1 i
Blue) then ( ∆t =1, ∆Am,∆t = RedBlue). 5 7 24
i

The attribute is gathered first and the dynamic

attribute is derived from the attribute. The dy- Definition 6. A support is the ratio of records of a
namic attribute identifies the significant relation- certain polarity combination over the total number
ship between the dynamics of the attribute. of records in the dynamic attributes. Three sup-
ports are applied in DCM; sympathetic support,
antipathetic support, and single support. For data
MeasureMents set Dnew and any two attributes ∆A1,∆t and ∆A2,∆t
i i
the three kinds of supports are defined as follows:
Since the input of the DCM algorithm can be quite
large, it is important to prune away the redundant Sympathetic Support (∆A1,∆t , ∆A2,∆t ) = freq (+, + )
i i
attributes. n
(4a)
Definition 4. A polarity indicates the direction of or
a change of an attribute. There are three types of freq ( -, -)
polarity (+, -, 0); where + indicates an increase, (4b)
n
- indicates a decrease, and 0 indicates neutrality,
that is, no change at all. Antipathetic Support (∆A1,∆t , ∆A2,∆t ) = freq (+, -)
i i
n
Definition 5. A polarity combination is a joint (5.a)
set of two or more polarities. The simultaneous or
presence of combinations (+,+) and (-,-) indi- freq ( -, +)
cates sympathetic changes and will produce a (5.b)
n
sympathetic rule. The simultaneous presence of
combinations (+,-) and (-,+) indicates antipa- Single Attribute Support (∆Am,∆t ) = freq ( )
i
thetic changes and produces antipathetic rules. n
There are four different combinations of polarity (6.a)
(+,-) (antipathetic negative), (+,+) (sympathetic
positive), (-,+) (antipathetic positive), and (-,-)
(sympathetic positive) to indicate the degree of Table 2. Derived database Dnew
causality.
∆t ∆A1 ∆A2
This differs from the classical causal loops ∆t1 +8 +1
relation which has only + and -, due to a simultane- ∆t2 -7 +9
ous increase of an attribute set not automatically ∆t3 -6 +4
leading to a simultaneous decrease of the same set. ∆t4 +3 +8

260
Applying Dynamic Causal Mining in Health Service Management

Figure 2. Graph of two dynamic attributes

Dynamic attribute value

Time stamps

or support. The support is the frequency of occur-

rences of attribute sets that support a rule.
freq ( )
(6.b)
n Figure 2 shows the results of the comparison
or between two dynamic attributes. The X-axis repre-
sents the time stamp and the Y-axis represents the
freq (0)
(7.c) value of the dynamic attributes. Figure 3 shows
n the graphical representation of two dynamic attri-
butes, where the polarity combination is indicated.
Where freq (+,+) is a function counting the
number of times where an increase in ∆A1,∆t is Definition 7. A support level is the value thresh-
i
associated with a simultaneous increase in ∆A2,∆t , old for each dynamic attribute. Every record in
i
freq (-,-) is a function counting the number of a dynamic attribute must have an absolute value
times where an decrease in ∆A1,∆t is associated larger or equal to the support level in order to
i
with a simultaneous decrease in ∆A2,∆t , freq be considered as candidate for the dynamic rule.
i
(-,+) is a function counting the number of times
where an decrease in ∆A1,∆t is associated with a For a given support level, a positive and nega-
i
simultaneous increase in ∆A2,∆t , and freq (+,-) is a tive value of the support level can be then drawn
i
function counting the number of times where an as shown in Figure 3. The support is the occur-
decrease in ∆A1,∆t is associated with a simultane- rence of the polarity combination above the value
i
ous decrease in ∆A2,∆t . of the support level.
i
All supports relate to the frequencies of the
occurring patterns. For a given user specified Definition 8. A frequent dynamic set is a pair
support, the problem of DCM is to find all rules of dynamic attributes which contain a polarity
where the support is greater than the user defined combination with frequency occurrence above a
user-defined support threshold.

261
 Applying Dynamic Causal Mining in Health Service Management

Figure 3. Illustration of polarity combination

Value
sympathetic antipathetic
* * *
Support
level
*

*
0 time
stamp
- Support
level
* *
*
antipathetic
sympathetic

Table 3. Derived database Dnew with arrows Table 4. Counting result

indicating support counting direction
(+,+) (-,-) (+,-,) (-,+,)
∆t ∆A1 ∆A2
Supports(ΔA1, ΔA2) 2/4 0 0 2/4
∆t1 +8 +1
∆t2 -7 +9
∆t3 -6 +4 an attribute is equal to the total number of time
∆t4 +3 +8 stamps. Thus it makes the confidence equal to the
support. Instead, multiple supports are introduced
to further reduce the running time and size of the
Theorem 1. If a pair of dynamic attributes relevant rules.
(∆A1, ∆A2) is infrequent, then either one individiual
dynamic attribute is infrequent or both dynamic
attributes are infrequent. rule representatIon
Proof: If a dynamic attribute set is frequent,
then this indicates that both the dynami attributes A dynamic causal rule consists of variables con-
are above the user-defined threshhold. nected by arrows denoting the causal influences
The above theorem is a consequence of Defini- among the attributes. Figure 4 shows an example
tion 7 for the frequent dynamic set. This obser- of the notation. Two attributes, A1 and A2, are linked
vation forms the basis of the pruning strategy in by a causal arrow. Each causal link is assigned
the search procedure for frequent dynamic sets, a polarity and the link indicates the direction of
which has been leveraged in many Association the change.
Mining algorithms (Zaki, 2000), that only the In System Dynamics, a symbol x →+y can be
single dynamic attribute found to be frequent interpreted as δy/δx > 0 and x →-y can be inter-
needs to be extended as candidate for the rule. preted as δy/δx < 0. This analogy is applied in
DCM and the dynamic causal rules produced by
Theorem 2. Confidence measure is not useful DCM can form causal diagrams, which will be
in DCM analysis. used to simulate future behaviour.
Proof: The confidence measure is not used
here because the total numbers of records in

262
Applying Dynamic Causal Mining in Health Service Management

Figure 4. Notation of a dynamic causal rule

Link polarity

+,-
Variable A1 Variable A2

Attribute Attribute
Causal link

Definition 9. A dynamic causal rule is derived direction. An antipathetic rule represents an

from a frequent dynamic attribute set. A dynamic adjustment to achieve a certain goal or objective.
causal rule can be either strong or weak. A weak It indicates a system attempting to change from
rule is a set of attributes with polarity that partially its current state to a goal state. This implies that
fulfils equation (4), (5), or (6). A strong rule is a if the current state is above the goal state, then
set of attributes with polarity that completely fulfils the system forces it down. If the current state is
equation (4), (5), or (6). There are two types of below the goal state, the system pushes it up. An
strong rule, sympathetic and antipathetic. antipathetic rule provides useful stability but
resists external changes.
Figure 2 shows that variables A1 and A2 are
causally dependent. If any two variables A1 and dynamic policy
A2 are truly causally related, then a change of A1
causes a change of A2. This article focuses on the Definition 10. A dynamic policy consists of one
discovery of the causality with no time delay, or more dynamic causal rules. In a dynamic
which means that attributes are in same time policy with several dynamic causal rules, each
period or interval. This implies that attributes rule should share at least one dynamic attribute
should occur within the same time stamp. with other rules.

Theorem 3. The support for a strong rule is less Definition 11. A dynamic policy is single if it
than or equal to the half of the total time stamps. consists only of one rule and if that rule does not
Proof: Each dynamic attribute can have only share any common attribute with other rules in
one type of polarity at one time stamp. If the oc- a dynamic policy.
currence of one polarity is huge, the occurrence
of the other two polarities will diminish. Follow- Given a single dynamic policy where an attri-
ing the definition of strong rules, the occurrence bute A1 is dynamically causally related to another
of the polarities + and – have both to be highly attribute A2, then such a rule can be represented
frequent. Since support indicates the times each as the following function.
polarity pair occurred over the total time stamp,
the support cannot be more than ½ of the total ∆A2 = k∆A1 (8)
time stamps.
where
A sympathetic rule causes an increase or A2,t
decrease in the output of a target system. It re- kt = (9)
A1,t
inforces a change with more change in the same

263
 Applying Dynamic Causal Mining in Health Service Management

Given A1new indicating new values added in T = ( 1, 2, 3, ….t) and Ct = A2,t* A3,t *A4,t
A1, a ∆A1new can be calculated based on A1new, and
based on equation 8 a ∆A2new can be calculated. If IllustratIVe exaMple
given a value A2,t=0 a new set of A2 can be derived,
where A2,t=2 = A2,t=1 + Anew , A2,t=3 = A2,t=2 + Anew • Stage 1: To establish a dynamic data set.
2,t 1 2,t 2
, and so forth. The dynamic set is calculated based on the
dataset illustrated in Table 5 and Equation
Definition 12. A dynamic policy is serial if each 1. Table 6 shows the dynamic set after the
rule shares only one attribute with one other rule subtraction. A1 ,A2 … An could represent
in a dynamic policy. A serial dynamic policy is treatment volume, doctor capacity, and so
open if each attribute in the policy has only one forth, which all varies through time.
causal link, in other words, if there is a start and • Stage 2: To prune the dynamic data set based
an ending attribute. A serial dynamic policy is on the specified support. Pruning is carried
closed if there is no start or ending attribute. out to remove columns (attributes) where
the level of support is below the minimum
Given an open serial dynamic policy where set. In this example, the support is set to 2,
an attribute A1 is dynamically causally related to which means columns with two or more 0s
another attribute A2, and A2 is dynamically caus- are removed (value 0 indicates no dynamics
ally related to another attribute A3, then such a in the attribute at the corresponding time
rule can be represented as the following function. stamp and more occasionally the set value
of 0 indicates the attribute is not dynamic).
A2,T = k1,T A1,T and A3,T = k2,T A2,T= k2,T k1,T A1,T Table 7 illustrates the “pruned” dynamic
(10) data set.
• Stage 3: To create dynamic rules, the dy-
where namic supports are counted and calculated
A2,t according to Equations (4), (5), and (6). Table
A3,t
T = ( 1, 2, 3, ….t), k1,t = and k2,t = 8 shows the dynamic supports for the pairs
A1,t A2,t of attributes in Table 7. In this example, the
(11) user-specified support is set to 0.3, which

Definition 15. A dynamic policy is complex if the

policy consists of a mixture of open and closed Table 5. Input dataset
serial dynamic policies.
ΔA1 ΔA2 ΔA3 ΔA4 ΔA5 ΔA6 ΔA7
For a complex or closed dynamic policy there +8 +1 0 +6 0 0 0
are attributes which are connected with more than -7 +9 0 -1 0 0 -7
one attribute. For instance, A1 is connected with -6 +4 0 -5 0 0 -4
A2, A3, and A4. A2, A3 and A4 are not connected. +3 +8 0 -8 0 -97 +8
Then A1 can be represented with: -1 -6 0 -3 0 97 -5
0 0 0 -6 0 0 -1
A2,T* A3,T *A4,T = CT
+5 +3 0 0 0 0 +5
+9 -8 0 -4 0 0 -7
where
+1 -5 0 +7 -15 0 +6

264
Applying Dynamic Causal Mining in Health Service Management

means any attribute pair with dynamic sup- of A1,new and A2,new and it is clear that the two plots
port with value larger than or equal to 0.3 is are causally related.
a dynamic rule. Table 7 shows how the DCM
rules are generated. Note that (F1&F7) is the
only strong sympathetic rule because when MInIng algorItHM
one of the attributes increases its value, the
other will automatically increase its value problem Formulation
and vice versa.
Let D denote a database which contains a set
Table 8 shows the attribute pair A1 and A7 and of n records with attributes {A1, A2, A3,… An.},
dynamic attributes ∆A1 and ∆A7.The strokes indi- where each attribute is of a unique type (sale
cate redundant attributes and redundant dynamic price, production quantity, inventory volume, etc).
attributes. Given a new attribute A1,new as shown in Each attribute is linked to a time stamp t. To ap-
Table 9, from it ∆A1,new and ∆A2,new can be calculated ply DCM, the records are arranged in a temporal
as shown in Figure 6. Figure 4.3 shows the plots sequence (t = 1, 2,…, n). The causality between
attributes in D can be identified by examining the
Table 6. Dynamic dataset polarities of corresponding changes in attribute
values. Let Dnew be a new data set constructed
Time A1 A2 A3 A4 A5 A6 A7 from D. A generalized dynamic association rule
1 9 2 10 22 20 100 13 is an implication of the form A1 →p A2, where
2 17 3 10 28 20 100 13 A1 ⊂ D, A1 ⊂ D, A1∩ A2= f and p is the polarity.
3 10 12 10 27 20 100 6 The implementation of the DCM algorithm
4 4 16 10 22 20 100 2 must support the following operations:
5 7 24 10 14 20 3 10
6 6 18 10 11 20 100 5
7 6 18 10 5 20 100 4
Table 8. Dynamic support
8 11 21 10 5 20 100 9
9 20 13 10 1 20 100 2 (+,+) (-,-) (-,+) (+,-)
10 21 8 10 8 5 100 8
F1&F2 0.3 0.1 0.2 0.2

F1&F4 0.2 0.3 0 0.2

Table 7. Pruned dataset F1&F7 0.3 0.3 0 0.1

Δ A1 Δ A2 Δ A4 Δ A7 F2&F4 0.1 0.2 0.1 0.3

+8 +1 +6 0 F2&F7 0.2 0.2 0.1 0.2

-7 +9 -1 -7 F4&F7 0.1 0.5 0.1 0
-6 +4 -5 -4
+3 +8 -8 +8
-1 -6 -3 -5 Table 9. The rules generated
0 0 -6 -1
+5 +3 0 +5 (+,+) (F1&F2), (F1&F7)

+9 -9 -4 -7 (-,-) (F1&F4), (F1&F7), (F4&F7)

+1 -5 +7 +6 (+,-) (F2&F4)

265
 Applying Dynamic Causal Mining in Health Service Management

Figure 5. Plot for A1,new and A2,new

dynamic causal behaviour for new values

30
20
attribute value

10
A1,new
0
A2,new
-10 1 2 3 4 5 6 7

-20
-30
timestamps

1. To add new attributes. The process time would be n2-n, where n is the
2. To maintain a counter for each polarity with number of attributes. This becomes a huge prob-
respect to every dynamic value set. While lem if n becomes too large. It is obvious that the
making a pass, one dynamic set is read at a task becomes much simpler if the size of n could
time and the polarity count of candidates sup- be reduced before the search. Table 14 shows the
ported by the dynamic sets is incremented. pruned percentage of the total database based on
The counting process must be very fast as the support. The support level is set to 0.
it is the bottleneck of the whole process.

algorithm description experIMent

DCM makes two passes over the data as shown in data preparation
Figure 6 and Figure 7. In the first pass, the support
of individual attributes is counted and the frequent The overall aim is to identify hidden dynamic
attributes are determined. The dynamic values are changes. It was taken from a local hospital The
used for generating new potentially frequent sets original data was given as shown in Table 11. The
and the actual support of these sets is counted dur- only data of interest are the data with changes,
ing the pass over the data. In subsequent passes, for example sale amounts of a medicine, the time
the algorithm initializes with dynamic value sets stamp, and so forth. The rest of the static data,
based on dynamic values found to be frequent in such as the weight and the cost of the product,
the previous pass. After the second of the passes, can be removed. This data consists of real life
the causal rules are determined and they become data. This dataset contains 65,536 attributes of
the candidates for the dynamic policy. In the metal manufacturing, with eight records in each
DCM process, the main goal is to find the strong attribute.
dynamic causal rule in order to form a policy. After cleaning the data, the dynamic attributes
It also represents a filtering process that prunes are found as shown in Figure 16. The dynamic
away static attributes, which reduces the size of attribute is calculated by finding the difference
the data set for further mining. between sales amounts in one month and sales
amounts in the previous month.

266
Applying Dynamic Causal Mining in Health Service Management

Figure 6. The steps of DCM

Part 1: – Preprocessing: Removal of the “least” causal data from database

Part 2: – Mining: Formation of a rule set that covers all training examples with
minimum number of rules
Part 3: – Checking: Check if an attribute pair is self contradicting (sympathetic
and antipathetic at the same time)

Input: The original database, the values of the pruning threshold for the neutral,
sympathetic and antipathetic supports.
Output: Dynamic sets
Step 1: Check the nature of the attributes in the original database (numerical or
categorical). Initialize a new database with dynamic attributes based on
the attributes and time stamps from original database.

Step 2: Initialize a counter for each of the three polarities.

Step3: Prune away all the dynamic attributes with supports above the input
thresholds.

Figure 7. The checking step of DCM

Input: The mined database, the values of the pruning threshold for the
supports of the polarity combinations.

Output: Dynamic sets

Step 1. Check weather a rule is self-contradictory (a rule is both

sympathetic and antipathetic).
Step 2. If step 1 returns true then
Retrieve the attribute pair form the preprocessed database
Step Initialize a counter that includes polarity combination
Step 4. For the pair of attributes
Count the occurrence of polarity combination with two
records each time.
Prune away the pairs if the counted support is below
the input threshold.

In the next step, the neutral attributes are time stamps, it will be pruned. In this case, 429
pruned. The idea of pruning is to remove redundant attributes remain for the rule generation.
dynamic attributes; thus fewer sets of attributes In this experiment, dynamic sets are compared
are required when generating rules. The first based on a simultaneous time stamp. Then the
pruning is based on the single attribute support. support of sympathetic and antipathetic rules for
In this case, the single attribute support is defined each dynamic set is calculated. The support is
to be 0.5, which means that if an attribute with used as the threshold to eliminate unsatisfactory
polarity +, -, or 0 occurs in more than half of total dynamic sets and to obtain the rules from the

267
 Applying Dynamic Causal Mining in Health Service Management

Table 10. Pruned results

Single Support
Data set
0.05 0.10 0.15 0.20 0.25 0.30 0.35
Adult 5% 20% 27% 74% 100% 100% 100%
Bank 11% 20% 60% 94% 100% 100% 100%
Cystine 5% 33% 70% 100% 100% 100% 100%
Market basket 6% 10% 50% 86% 100% 100% 100%
Mclosom 1% 13% 38% 72% 90% 100% 100%
ASW 1% 8% 40% 68% 70% 97% 100%
Weka-base 6% 25% 52% 86% 100% 100% 100%

satisfactory sets. In Table 12 the left most colum ing support will lead to exponential growth of
indicates the codes for each attributes. The top the rules. As the support reaches 0.05 or 5, as it
row indicates the times stamps and the rest of indicates on the figure, the number of rules is 630.
table is dynamic attributes. Most of these rules are redundant and have no
The algorithm was run based on the procedures meaning due to the low support. Figure 9 shows
described in previous sections. Figure 8 shows the rule plot with support equal to average value,
the plot of sympathetic and antipathetic support. where the +support = the average of all positive
The x-axis represents the support and the y- records and –support = the average of all negative
axis represents the number of rules. This database records. The number of rules has decreased by
shows that there are more sympathetic rules than applying the support level.
antipathetic rules. The figure shows that increas-

Table 11. Original data sets

268
Applying Dynamic Causal Mining in Health Service Management

Table 12. Dynamic attributes

Figure 8. Rule plot with support level = 0

support level = 0

700

600
number of rules rules

500

400
Sympathetic rules
300 Antipathetic rules

200

100

0
0.11 >0.10 >0.09 >0.08 >0.07 >0.06 >0.05
support

Table 13 shows the extracted strong rules dIscussIon

with support level equal to average value and
support larger than 0.08. C15276179, A04004004 Apriori provides some form of causal informa-
…… are the code for each attributes. There are tion, that is, suggesting a possible direction of
only dynamic pairs so there is no need to do the causation between two attributes, but there is no
simulation basis to conclude that the arrow indicates direct

269
 Applying Dynamic Causal Mining in Health Service Management

Figure 9. Rule plot with support level = frequent

support level = average vale

350
number of rules

300

250

200
Sympathetic rules
150 Antipathetic rules

100

50

0
>0.11 >0.10 >0.09 >0.08 >0.07 >0.06 >0.05
support

Table 13. Result generated by the algorithm

Strong rules Support

Sympathetic
{C15276179, F030008} 0,093
{J08008008, F060010} 0,089
{A04004004, A05005005} 0,086
{A05005006, C10251104} 0,084
{A04004004, F100020} 0,082

Antipathetic
{A05005008, C15276179} 0.092
{C10251104, F070010} 0.083
{A05005008, F030008} 0.082

or even indirect causation. The DCM algorithm, pathetic or antipathetic). A genuine causality such
on the other hand, shows causality between at- as A1 →+ A2 provides useful information because
tributes. Thus, where association rule generation it indicates that the relationship from A1 to A2 is
techniques find surface associations, causal infer- strictly sympathetic causal.
ence algorithms identify the structure underlying The rules extracted by DCM can be simulated
such associations. by using software to model the future behaviour.
Each type of relationship generated by the The rules extracted by association algorithm can-
DCM algorithm provides additional informa- not be simulated.
tion. The DCM algorithm finds four kinds of
relationships, each of which deepens the user’s
understanding of their target system by construct- conclusIon and suMMary
ing the possible models. For example, A1 →+ A2
provides more information than A1→ A2 because This article provides insight into how to use dy-
the latter indicates that A1 coexists with A2. The namic causal mining approaches to medical and
condition of the rule is not stated (whether sym- health management research can be addressed

270
Applying Dynamic Causal Mining in Health Service Management

through management strategies specifically de- Bergan, T., & Hirsch, G.B. (1976). Analysis and
signed to respond. Planning for improved distribution of Nursing
The algorithm shows opportunities for more personnel and services. A national model of sup-
extensive examination of dynamics between ply, demand, and distribution. Cambridge, MA:
attributes as well as produces a framework for Pugh-Roberts Associates, Inc.
such analysis. A mining framework will provide
Cheung, D., Han, J., Ng, V.T., & Fu, A.W. (1996).
medical and health management researchers, as
Fast distributed algorithm for mining association
well as researchers interested in other complex
rules. In International Conference on Parallel
social phenomena, with insights into determining
and Distributed Information Systems, Tokyo,
the best research design given a specific context.
Japan (pp. 31-42).
This article addresses important and challeng-
ing issues of causal analysis within e-government Cook, J.E., & Wolf, A.L. (1998, November). Event-
and implementation of causality extraction and based detection of concurrency. In Proceedings
construction algorithms for medical and health of the Sixth International Symposium on the
management data. This article also studies the Foundations of Software Engineering (FSE-6),
computational costs of causal, and implements Orlando, FL (pp. 35-45).
a multipruning approach for building models
Coyle, R.G. (1996). System dynamics modelling: A
that incurs minimal computational cost while
practical approach. London: Chapman and Hall.
maintaining accuracy and unsupervised causality
extraction algorithms to reduce the reliance on Dawes, M., & Sampson, U. (2003). Knowledge
medical and health management data. management in clinical practice: Asystematic
review of information seeking behavior in physi-
cians. International Journal of Medical Informat-
reFerences ics, 71, 9-15.
Dreiseitl, S., Ohno-Machado, L., Kittler, H.,
Abidi, S.S.R. (2001). Knowledge management in
Vinterbo, S., Billhardt, H., & Binder, M. (2001).
healthcare: Towards knowledgedriven decision-
A comparison of machine learning methods for
support services. International Journal of Medical
the diagnosis of pigmented skin lesions. Journal
Informatics, 63, 5-18.
of Biomedical Informatics, 34, 28-36.
Acir, N., & Guzelis, C. (2004). Automatic spike
Flood, R. (1999). Rethinking the fifth discipline.
detection in EEG by a two-stageprocedure based
Learning within the unknowable. London: Ruth-
on support vector machines. Computers in Biology
ledge.
and Medicine,34(7), 561-575.
Forrest, S., Hofmeyr, S.A., Somayaji, A., &
Agrawal, R., Gunopulos, D., & Leymann, F.
Longstaff, T.A. (1996). A sense of self for unix
(1998). Mining process models from workflow
processes. In 1996 IEEE Symposium on Secu-
logs. In Sixth International Conference on Ex-
rity and Privacy (pp. 120-128). IEEE Computer
tended Database Technology (pp. 469-483).
Society. Building the virtual state. Information
Agrawal, R., Mannila, H., Srikant, R., Toivonen, technology and institutional change. Washington,
H., & Inkeri, A. (1996). Fast discovery of associa- DC: Brookings Institution Press.
tion rules, advances in knowledge discovery and
Forrester, J.W. (1961, July-August). Industrial
data mining. The Association for the Advance-
Dynamics: A Major Breakthrough for Decision
ment of Artificial Intelligence and the MIT Press
Makers. Harvard Business Review, 37-66.
(pp. 307-328).

271
 Applying Dynamic Causal Mining in Health Service Management

Friedman, C., Hripcsak, G., Johnson, S.B., Hirsch, G.B., & Bergan, T.A. (1973). Simulat-
Cimino, J.J., & Clayton, P.D. (1990). Ageneral- ing ambulatory care systems: The relationship
ized relational schema for an integrated clinical between structure and behaviour. In Proceedings
patient database. In Proceedings of the 14th An- of the Summer Computer Simulation Conference,
nual Symposium on Computer Applications in Montreal, Quebec, Canada (pp. 877-884).
Medical Care (pp. 335-339).
Hirsch, G.B., Bergan, T.A., & Frohman, A.L.
Heathfield, H., & Louw, G. (1999). New challenges (1974). Organizational change in the provision
for clinical informatics: Knowledge management of human services. In Proceedings of the Sum-
tools. Health Informatics Journal, 5(2), 67-73. mer Computer Simulation Conference, Houston,
Texas.
Herbst, J. (2000). A machine learning approach
to workflow management. In 11th European Con- Hirsch, G.B., Forsyth, B.R., Bergan, T.A., &
ference on Machine Learning. Berlin, Germany: Goodman, M.R. (1976). Projecting a teaching
Springer (LNCS 1810, pp. 183-194). hospital’s future utilization: A dynamic simula-
tion approach. Journal of Medical Education,
Herbst, J., & Karagiannis, D. (1999, August),
51(11), 937-939.
An inductive approach to the acquisition and
adaptation of workflow models. In M. Ibrahim & Hirsch, G.B., & Henderson, J.D. (1977). A
B. Drabble (Eds.), Proceedings of the IJCAI’99 simulation model for “what if?” questions in
Workshop on Intelligent Workflow and Process plan development. American Journal of Health
Management: The New Frontier for AI in Busi- Planning, 2(1), 40-46.
ness (pp. 52-57). Stockholm, Sweden.
Hirsch, G.B., & Killingsworth, W.R. (1975). A
Herbst, J., & Karagiannis, D. (2000). Integrating new framework for projecting dental manpower
machine learning and workflow management to requirements. Inquiry, 12(2), 126-142.
support acquisition and adaptation of workflow
Hirsch, G.B., & Miller, S. (1974). Evaluating
models. International Journal of Intelligent Sys-
HMO policies with a computer simulation model.
tems in Accounting, Finance and Management,
Medical Care, 12(8), 668-681.
9, 67-92.
Jackson, J.R. (2000). The urgent call for knowl-
Hersh, W., Mailhot, M., Arnott-Smith, C., &
edge management in medicine. The Physician
Lowe, H. (2002). Selective automated indexing
Executive, 26(1), 28-31.
of findings and diagnoses in radiology reports.
Journal of Biomedical Informatics, 34, 262-273. Janetzki, V., Allen, M., & Cimino, J.J. (2004).
Using natural language processing to link from
Hripcsak, G. (1993). Monitoring the monitor:
medical text to online information resources. In
Automated statistical tracking of a clinical event
Proceedings of Medinfo (pp. 1665).
monitor. Computers and Biomedical Research,
26(5), 449-466. Kandaswamy, A., Kumar, C.S., Ramanathan,
R.P., Jayaraman, R., & Malmurugan, N. (2004).
Hripcsak, G., Austin, J.H., Alderson, P.O., &
Neural classification of lung sounds using wavelet
Friedman, C. (2002). Use of natural language
coefficients. Computers in Biology and Medicine,
processing to translate clinical information from
34, 523-537.
a database of 889,921 chest radiographic reports.
Radiology, 224(1), 157-163.

272
Applying Dynamic Causal Mining in Health Service Management

Kovalerchuk, B., Vityaev, E., & Ruiz, J.F. (2001). Prather, J.C., Lobach, D.F., Goodwin, L.K., Hales,
Consistent and complete data and expert mining J.W., Hage, M.L., & Hammond, W.E. (1997).
in medicine. In K.J. Cios (Ed.), Medical data Medical data mining: Knowledge discovery in
mining and knowledge discovery. New York: a clinical data warehouse. In Proceedings of the
Physica-Verlag. AMIA Annual Symposium Fall 1997 (pp. 101-105).
Kuperman, G.J., Gardner, R.M., & Pryor, T.A. Richardson, G.P. (1996). System dynamics. In
(1991). The HELP system. New York: Springer- S.I. Gass & C.M. Harris (Eds.), Encyclopedia of
Verlag. operations research and management science
(pp. 656-660). Boston, MA: Kluwer Academic
Levin, G., Hirsch, G.B., & Roberts, E. (1972).
Publishers.
Narcotics and the community: A system simula-
tion. American Journal of Public Health, 62, 8-6. Richardson, G.P. (1999). Citation for winner of the
1999 Jay W. Forrester Award: Jac Vennix. System
Levin, G., & Roberts, E.B. (1976). The dynam-
Dynamics Review, 15 4, 375-377.
ics of human service delivery. Cambridge, MA:
Ballinger Publishing Company. Richardson, G.P., & Andersen, D.F. (1995). Team-
work in group model building. System Dynamics
Maruster, L., van der Aalst, W.M.P., Weijters,
Review, 11(2), 113-137.
T., van den Bosch, A., & Daelemans, W. (2001,
October 25-26). Automated discovery of work- Roberts, E.B., & Hirsch, G.B. (1976, Winter).
flow models from hospital data. In B. Kröse et al. Strategic modelling for health care managers.
(Eds.), Proceedings 13th Belgium- Netherlands Health Care Management Review, 69-77.
Conference on Artificial Intelligence (BNAIC’01),
Schubart, J.R., & Einbinder, J.S. (2000). Evalu-
Amsterdam, The Netherlands (pp. 183-190).
ation of a data warehouse in an academic health
Montani, S., & Bellazzi, R. (2002). Supporting sciences center. International Journal of Medical
decisions in medical applications: The knowledge Informatics, 60, 319-333.
management perspective. International Journal
Senge, P., Kleiner, A., Roberts, C., Ross, R., &
of Medical Informatics, 68, 79-90.
Smith, B. (1994). The fifth discipline fieldbook.
Pham, D.T., Wang, Y., & Dimov, S. (2005). Intel- New York: Doubleday
ligfent manufacturing strategy selection. In Proc
Stearns, N.S., Bergan, T.A., Roberts, E.B., &
1st International Virtual Conference on Intelligent
Cavazos, L.F. (1978). A systems intervention for
Production Machines and Systems (pp. 312-318).
improving medical school-hospital interrelation-
Oxford: Elsevier.
ships. Journal of Medical Education, 53, 464-472.
Pham, D.T., Wang, Y., & Dimov, S. (2006).
Stearns, N.S, Bergan, T.A., Roberts E.B., &
Incorporating delay and feedback in intelligent
Quigley, L. (1976, Fall). Systems intervention:
manufacturing strategy selection. In Proc 2nd
New help for hospitals. Health Care Management
International Virtual Conference on Intelligent
Review, 1, 9-18.
Production Machines and Systems (pp. 246-252).
Oxford: Elsevier. Sterman, J.D. (1994). Learning in and about
complex systems. System Dynamics Review,
10(2-3), 291-330.

273
 Applying Dynamic Causal Mining in Health Service Management

Sterman, J. D. (2000). Business dynamics: Sys- Weijters, T., & van der Aalst, W.M.P. (2001, Octo-
tems thinking and modelling for a complex world. ber 25-26). Process mining: Discovering workflow
Boston, MA: Irwin McGraw-Hill. models from event-based data. In B. Kröse et al.
(Eds.), Proceedings 13th Belgium-Netherlands
Yang, Y., & Liu, X. (1999). A re-examination of
Conference on Artificial Intelligence (BNAIC’01),
text categorization methods. In Proceedings of
Amsterdam, The Netherlands (pp. 283-290).
the 22nd Annual International ACM Conference
on Research andDevelopment in Information Zaki, M.J., Parthasarathy, S., Ogihara, M., &
Retrieval (SIGIR’99) (pp. 42-49). Li, W. (1997). New parallel algorithms for fast
discovery of association rules. Data Mining and
Wang, Y. (2007). Integration of data mining with
Knowledge Discovery, 1(4), 343-37.
game theory. Journal of International Federation
for Information Processing, 207/2006, 275-280.
Boston, MA: Springer

This work was previously published in International Journal of Healthcare Information Systems and Informatics, Vol. 3, Is-
sue 4, edited by J. Tan, pp. 17-38, copyright 2008 by IGI Publishing (an imprint of IGI Global).

274
 275

Chapter 19
Application of Unified Modelling
Language (UML) to the
Modelling of Health Care
Systems:
An Introduction and Literature Survey

Christos Vasilakis
University College London, UK

Dorota Lecnzarowicz
University of Westminster, UK

Chooi Lee
Kingston Hospital, UK

aBstract

The unified modelling language (UML) comprises a set of tools for documenting the analysis of a sys-
tem. Although UML is generally used to describe and evaluate the functioning of complex systems, the
extent of its application to the health care domain is unknown. The purpose of this article is to survey
the literature on the application of UML tools to the analysis and modelling of health care systems. We
first introduce four of the most common UML diagrammatic tools, namely use case, activity, state, and
class diagrams. We use a simplified surgical care service as an example to illustrate the concepts and
notation of each diagrammatic tool. We then present the results of the literature survey on the applica-
tion of UML tools in health care. The survey revealed that although UML tools have been employed
in modelling different aspects of health care systems, there is little systematic evidence of the benefits.

Copyright © 2011, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 Application of Unified Modelling Language (UML) to the Modelling of Health Care Systems

IntroductIon an exception and most studies have used UML

without an evaluative component. We conclude
Health care systems are known to be complex and, with a brief discussion of the results.
as a result, difficult to analyse and re-engineer
(Berwick, 2005). Health system engineers often
rely on computer modelling and simulation to as- uMl dIagraMMatIc tools
sist with the analysis of existing systems and the
pretesting of suggested changes. To this extend, UML 2.0 has 13 types of diagrams, which can
a variety of software engineering techniques and be categorised hierarchically as follows (Object
tools have been employed (Jun, 2007). Examples Management Group, 2005):
include data flow diagram (Pohjonen et al., 1994),
state transition diagram (Mehta, Haluck, Frecker, • Structure diagrams used to represent the
& Snyder, 2002), entity relationship diagram (Kalli elements of the system being modelled.
et al., 1992), integrated definition or IDEF (Hoff- They include class, component, composite
man, 1997), and more recently, Unified Modelling structure, deployment, object, and package
Language, commonly known as UML (Object diagrams.
Management Group, 2005). • Behaviour diagrams that allow the repre-
UML provides a comprehensive set of tools sentation of what happens in the modelled
that can be used for documenting the analysis of system in the activity, state, and use case
a system and for developing model requirements. diagrams.
UML diagrams are graphical depictions that • Interaction diagrams, a subset of behaviour
demonstrate the flow of events within the system diagrams, that allow the representation of the
(Object Management Group, 2005). Depending control and data flow among the elements of
on the perspective chosen for the study (e.g., ac- the system being modelled. These are com-
tor oriented, activity oriented), different tools are munication, interaction overview, sequence,
available to the analyst. Due to its versatility and and timing diagrams.
the ability to analyse systems from different per-
spectives, UML is said to be effective in describing We briefly introduce here the four UML
and evaluating the functioning of complex systems diagrammatic tools that appear in the surveyed
such as health care (Kumarapeli, De Lusignan, literature, namely, use cases and use case dia-
Ellis, & Jones, 2007). However, there seems to gram, activity, state, and class diagram. A full
be very little systematic evidence on its benefits. description of the concepts and syntax of UML
The focus of the article is to review the lit- diagrams is beyond the scope of this article. A
erature on the application of UML tools to the plethora of user guides and technical notes are
analysis and modelling of health care systems. To available on the subject, with the monograph by
this end, we first briefly introduce four of the most Ambler (2004) a particularly useful introduction.
common UML diagrammatic tools, namely use We illustrate the basic concepts and notation
case, activity, state, and class diagrams. We use of each diagrammatic tool by presenting simple
a simplified surgical care service as an example models of a simplified care process of surgical
to illustrate the notation and concepts of each consultation with a patient in an outpatient clinic.
diagrammatic tool. Next, we present the results In general, physicians refer patients for surgical
of the literature survey on the application of consultation if they believe the underlying health
UML in health. The survey revealed that studies problem is amenable to surgical intervention.
of the benefits of UML to health evaluation are Following the referral, the outpatient clinic books

276
Application of Unified Modelling Language (UML) to the Modelling of Health Care Systems

Figure 1. UML use case diagram of a simplified surgical care service (surgical consultation in outpa-
tient clinic)

Surgical Consultation
Order tests
«extends»
«extends»
Book appointment

Take samples

Register on surgical waiting list

Nurse
«extends»
Educate on operation

Decide on treatment
Patient requiring surgical consultation
«uses»

Surgeon

«uses» Evaluate symptoms and test results

Report on samples
«uses»

Patient
Analyse samples

Lab technician

the patient an appointment with the surgeon and ments of a system (Object Management Group,
also arranges for samples to be taken if further 2005). Each use case provides one or more
diagnostic tests are required. At the consultation, scenarios that convey how a specific part of the
the surgeon assesses the need for an operation by system interacts with the users (called actors) to
evaluating symptoms and test results. Following a achieve a business goal or function. There is no
decision to operate, the patient’s name is registered standard format for detailing use cases but some
on a prioritised surgical wait list so that appropriate tabular layout is commonly used.
time can be booked at the operating theatre of a The UML use case diagram, on the other hand,
hospital. The patient may also be educated about allows the graphical representation of a set of
the operation by a specialist nurse. If an operation use cases. The UML standard sets out a specific
is not deemed suitable, then the patient may be graphical notation (Object Management Group,
further referred for medical treatment. 2005). Use cases and UML use case diagrams
not only provide clarity in terms of actors and
use cases and uMl use case sequence of steps involved in the event but also
diagrams serve as a useful tool to present details of the ac-
tor’s progression in the system.
In software and system engineering, a use case is Figure 1 shows a use case diagram depicting
a technique for capturing the functional require- the process of surgical consultation with a patient

277
 Application of Unified Modelling Language (UML) to the Modelling of Health Care Systems

in outpatient clinic. Actors are represented by lowing the referral to the outpatient clinic, booking
stick figures and use cases by ovals. Associations the appointment and ordering of the diagnostic
between actors and use cases are depicted by tests happen in parallel (for the purposes of this
edges. The “uses” arrow points to a use case that is example). Depending on the outcome of the con-
always invoked, while the “extends” to a use case sultation, the patient’s name is registered on the
that is conditionally invoked. In the example, the surgical waiting list if operation is deemed neces-
surgeon always evaluates the patient’s symptoms sary and the patient may receive education about
and the test results. The outpatient clinic nurse may the operation, otherwise the patient is referred for
order additional tests when booking the appoint- medical treatment.
ment, which in turn, may require taking samples
(e.g., blood) from the patient. The actor “Patient uMl state diagram
requiring surgical consultation” is a type of the
generic actor “Patient”, an example of the com- The UML state diagram is essentially a Harel
mon construct of specialisation/generalisation. (1987) Statechart with standardized notation
that can describe any system that is (or can be
uMl activity diagram conceptualised as) reactive, from computer pro-
grams to business processes. In this context, a
The purpose of the activity diagram is to depict reactive system—as opposed to a transformational
the procedural flow of actions that are part of a system—is a system that constantly responds to
larger activity (Object Management Group, 2005). internal and external stimuli by changing states
In projects in which use cases are generated, or by performing some action.
activity diagrams can model a specific use case Like state machines, a UML state diagram
at a more detailed level. Activity diagrams can includes state-transition diagrams that represent
be also used independent of use cases for model- the operations of a system through discrete states
ling a function, such as admission to hospital or and transitions from one state to another. In ad-
discharge procedure. They can also be used to dition, state diagrams include notions of state
model system functions, such as computerised hierarchy, parallelism, and event broadcasting
physician order systems, and complete patient (Sobolev, Harel, Vasilakis, & Levy, 2008).
pathways, such as from admission to hospital In UML state diagrams, rectangles represent
to discharge. Activity diagrams also allow the states and arrows represent transitions. An arrow
depiction of parallel activities that often occur may have a transition label that controls the transi-
in health systems. tion. The label includes the events that trigger the
In the UML activity diagram, which is based transition, and the condition that needs to be true
on the semantics of Petri nets, each activity is for the transition to occur in square brackets. The
represented by a rounded rectangle. An arrow actions associated with the transition also appear
represents the transition from one activity to on the labels following the forward slash. Draw-
another. The starting point is represented by a ing states inside other states represents hierarchy.
filled-in circle and the endpoint by a bull’s-eye. Dashed rectangles symbolise parallel states.
Activities enclosed within parallel bars happen Figure 3 shows a UML state diagram of patient
at the same time. Diamond shaped objects denote states in the example of surgical care service. Fol-
a decision mandated by conditions stated in the lowing the referral to clinic, the initial substates
brackets above the arrows. called “pending” of parallel states “appointment”
Figure 2 shows a UML activity diagram that and “diagnostic tests” are activated. When the
models the exemplar surgical care process. Fol- event “make booking” is fired and if there are

278
Application of Unified Modelling Language (UML) to the Modelling of Health Care Systems

Figure 2. UML activity diagram of the care process of a simplified surgical care service (surgical con-
sultation in outpatient clinic)

Referral to clinic

Booking of appointment with surgeon Ordering of diagnostic tests

Consultation with surgeon

[Operation is deemed necessary ] [Otherwise]

Referral for medical treatment

Registration on surgical waiting list Education about operation

available slots in the clinic, the patient is consid- butes and methods, and the relationships between
ered to have the appointment booked. Once the the classes (Object Management Group, 2005).
treatment has been decided and depending on the A class, indicated by a rectangle, can be thought
outcome, the patient state transitions to “surgical of as a blueprint for defining similar objects. Each
waiting list” and “education,” or to “waiting for object is an instance of a class and encapsulates
medical treatment.” both state, in terms of attributes, and behaviour,
in terms of methods. Attributes (or properties)
uMl class diagram are shown in an optional compartment below the
class name. Each attribute is shown with at least
A class diagram describes the static structure of a its name, and optionally with its type, initial value,
system by showing the system’s classes, their attri- and other information. The class methods (or op-
erations) appear in a second optional compartment.

279
 Application of Unified Modelling Language (UML) to the Modelling of Health Care Systems

Figure 3. UML state diagram of patient states in a simplified surgical care service (surgical consulta-
tion in outpatient clinic)
Patient at surgical consultation in outpatient clinic

Waiting for referral to clinic

referral to clinic

Appointment Diagnostic tests

Pending Pending

make booking [ slots available] take samples

Samples taken
Booked

patient presents at clinic

At consultation

treatment decided

[operation deemed necessary ] [otherwise]

Surgical waiting list E ducation

Pending Pending

educate Waiting for medical treatment

register

On waiting list Educated

book operation referral for medical treatment

Each method is shown with at least its name, and In the example shown in Figure 4, a patient
additionally with its parameters and return type. may have none, one, or more referrals (on differ-
The association between two classes is indicated ent dates). Each referral has a priority, is made to
by a line. The number of objects participating in the named surgeon, and the appointment slot is
the association, known as multiplicity, is given updated once it has been scheduled. Patient names
by an optional notation at each end of the line are placed on the surgical list of a surgeon follow-
(“0..1” if none or only one object participates in ing the consultation. A surgeon may be associated
the association, “1” exactly one, “0..*” zero or with no patients or many. A class diagram may
more, “1..*” at least one). convey a lot more information that is omitted here
in the interests of brevity.

280
Application of Unified Modelling Language (UML) to the Modelling of Health Care Systems

Figure 4. UML class diagram of a simplified surgical care service (surgical consultation in outpatient
clinic)
patient
-name
 placed  is made for
-dob
-gender
1 1
+add patient ()
+remove patient ()

1
has
0..* 0..*
0..*
surgical waiting list referral
diagnostic test
-patient name -patient name
-type of operation -test -date
-priority -patient name -priority
-surgeon name -surgeon name -surgeon
-placement date -result -appointment date
-removal date -date ordered
+add referral ()
-date of result
+add patient name () +remove referral ()
+remove patient name () +add order () +update appointment date ()
+audit waiting list() +update result ()

0..* 0..*
0..*  orders

surgeon
 operates on -name consults w ith
-specialty
1 +add surgeon () 1
+remove surgeon ()

application of uMl tools to Health of the benefits of using UML in the analysis of
care health care systems.
The retrieved literature can broadly be classi-
For the literature survey, we searched the medical fied into three categories according to the appli-
literature for articles that demonstrate the appli- cation domain: modelling health care processes,
cation of UML tools to the broad area of health evaluating and modelling clinical guidelines,
care. We included all articles that demonstrated and evaluating and generating requirements of
the application of any of use case and UML use information systems in health care. We now
case diagram, UML activity diagram, UML state briefly review each collected article according
diagram, and UML class diagram. Eighteen pa- to the these categories.
pers were identified to be relevant to this review,
found via published (as listed in Pubmed) and grey Modelling Health care processes
literature search. The literature survey identified
use cases and use case diagrams (11 papers), and Recent work by Jun (2007) and Jun, Ward, and
activity diagrams (13) as the most common UML Clarkson (2005) aimed at providing the effective
tools of the four included in this survey used in application of various modelling methods to health
health system analysis. Some studies have reported care with the end goal of enabling professionals
on the use of class diagrams (3) but none on the and managers to understand care processes more
use of UML state diagrams. The search yielded clearly, manage risk, and as a result improve patient
very little evidence of a systematic evaluation safety. Among the modelling methods evaluated
was the swim lane activity diagram, which is a

281
 Application of Unified Modelling Language (UML) to the Modelling of Health Care Systems

variation of UML activity diagram where activi- be made. However, despite the self critique of
ties are grouped according to the actors involved or the study, the authors did not evaluate the UML
in a single thread. Three case studies were used to activity and class diagram benefits and drawbacks
illustrate the methods: patient discharge, diabetic as a means to study health care processes.
care process in GP practise, and prostate cancer Spyrou, Bamidis, Pappas, and Maglaveras
diagnosis process in the hospital. The findings of (2005) proposed an extension of UML to model
the case studies were evaluated against benchmark processes in the health care domain using work-
goals of enhancing understanding and validated flow modelling techniques. The work presented
via key user review of findings. in the article extends the UML activity diagram
Cruz, Gramaxo Silva, Soares, Oliveira, Ser- to support workflow characteristics as well as
rano, and Paulo Cunha (2002) reported on the standardised clinical documents that are handled
experience of the using UML use case diagrams by the processes. The extended notation was then
as a tool to optimise hospital processes. The dia- used to model the flow of patients in a regional
grams were developed at two levels of abstraction: health system. No evaluation of the benefits was
a global use case diagram to represent the main reported.
processes of the care services examined, and Lyalin and Williams (2005) aimed at improv-
detailed use case diagrams to study parts of the ing the way cancer registration and other processes
system seen as critical bottlenecks. Activity dia- are described through enhancements in the nota-
grams were to be developed in subsequent project tion of UML activity diagrams. The article illus-
phases to model dynamic concepts of each detailed trates a UML activity diagram used to describe
use case. Although Cruz et al. (2002) concluded the process of cancer registration and which was
that UML helped in communication, discussion, enhanced by allowing the depiction of timeline,
and validation of the different steps of the project, duration for individual activities, responsibilities
there was no evaluation of these benefits. for individual activities, and descriptive text. The
Goossen et al. (2004) looked at the feasibility authors claim that this provides for clarification
of mapping and modelling of nursing care process of the process of cancer registration and can
information to some international standards. They broaden its understanding among different spe-
represented the nursing care process as a dynamic cialists. Lyalin and Williams (2005) included an
sequence of phases, each containing information extensive description of benefits and weaknesses
specific to the activities of the phase, and used of activity diagrams and a comprehensive activ-
UML to represent this domain knowledge in ity diagram of death clearance process at cancer
models. A UML activity diagram was developed registry. The authors conclude that the enhance-
as a model of a generic nursing care process. After ments add value to the tool and cite the positive
creating a structural model of the information col- response they received after using the enhanced
lected at each stage of the nursing process, various UML activity diagrams in a cancer registry best
working groups mapped that information to other practices development workshop.
standards as a means of validation. An activity Similarly, Saboor, Ammenwerth, Wurz, and
diagram of a generic nursing process was also Chimiak-Opoka (2005) aimed at improving UML
developed as a problem solving approach to patient activity diagram by developing and testing a
care and a UML class diagram enhanced this view process modelling method which included details
and demonstrated further the care process from a of clinical processes necessary for a systematic
nursing perspective. The authors concluded that and even semiautomatic quality assessment. The
their study produced a good model of the nursing premise of the authors’ work was that UML is
care process but that improvements could still only a descriptive tool that does not allow for

282
Application of Unified Modelling Language (UML) to the Modelling of Health Care Systems

comprehensive quality assessment. Subsequently, technology. However the authors provided no

Saboor et al. (2005) proposed a new modelling systematic evaluation of the benefits of employing
method based on the UML activity diagram with UML tools in health systems, but rather focused
extra notations to allow for evaluation. The method on the utility of UML activity diagram in general.
was validated by modelling various versions of the
process of ordering a radiological examination. It evaluating and generating
was suggested that further evaluation of improve- requirements of Information
ments on the UML activity diagram were needed. systems

evaluating and Modelling clinical Maij, Toussaint, Kalshoven, Poerschke, and

guidelines Zwetsloot-Schonk (2002) looked at the problem of
alignment between information and communica-
Sutton, Taylor, and Earle (2006) developed a tion technology (ICT) infrastructure and business
computerised system to allow hypertensive processes in health care organisations. The paper
patients to be monitored and assessed without investigated whether the combination of Dynamic
visiting their family doctors. The Web-based Essential Modelling of Organisations, that is a
system, created using PROforma, made recom- business process modelling methodology, with
mendations for continued monitoring and for UML can solve the problem. It used the example
changes in medication. PROforma is a language of a screening case study on the management of
that allows clinical guideline to be expressed preoperative centres and focused on developing
in a computer-interpretable manner. The study an efficient information system. Maij et al. (2002)
concluded that PROforma proved adequate as a used UML use cases to derive the functionality
language for the implementation of the clinical of the information systems. It also provided a use
reasoning but lacked notational convenience. case diagram and description for a transaction at
Hence, UML activity diagrams were employed the preoperative centre. The authors concluded
instead to create the models that were used during that the combination of the two techniques is use-
the knowledge acquisition and analysis phases ful in aligning business processes and functional
of the project. Sutton et al. (2006) also reported features of ICT infrastructure and should help the
on the application of UML activity diagram to end-user to develop a better understanding with
represent the clinical guidelines in the manage- regards to the relationship between the two areas.
ment of hypertensive patients. The authors praised Although the paper did include a brief discussion
the notational convenient of UML but did not of UML, it did not provide a systematic assess-
systematically evaluate its benefits. ment of its utility in health care.
In similar fashion, Hederman, Smutek, Lee, Bakken, and John (2006) briefly reported
Wade, and Knape (2002) compared a technique on the use of UML tools (use case, activity, and
for representing and sharing clinical guidelines sequence diagrams) to store and present the func-
(GLIF) with UML activity diagram. The authors tional requirements of a handheld-based decision
concluded that there are clear potential benefits in support system for morbid obesity screening and
using a mainstream modelling language such as management. The authors stated that UML is use-
UML as opposed to a specific clinical guideline ful in depicting processes related to management
representation technique such as GLIF. The po- of clinical based guidelines, facilitating discussion
tential benefits include availability of modelling and agreement in developing data model, and in
tools, the ability to transfer between modelling aiding the design of Web-based prototype.
tools, and to automate via business workflow

283
 Application of Unified Modelling Language (UML) to the Modelling of Health Care Systems

Lunn, Sixsmith, Lindsay, and Vaarama (2003) Aggarwal (2002) highlighted the benefits of
reported using UML activity diagram to model the UML in specifying, visualising, constructing,
various processes of monitoring care and use documenting, and communicating the model of a
case diagrams to generate requirements for the health care information system. It illustrated the
development of an information system intended usage of use cases and use case diagram, activ-
to support planning in the provision of elderly ity and class diagram by employing simplified
care services. The study is a good example of the examples of a nurse submitting a blood-count
application of UML tools in health but provides order, a physician order system, and of an emer-
little guidance in terms of actual benefits and gency room.
weaknesses. Ganguly and Ray (2000) discussed the de-
Weber et al. (2001) aimed at developing a tool velopment of a methodology for the design of
to support clinical trial centres in developing trial interoperable telemedicine systems based on
specific modules for the computer-based docu- UML. Their research focused on the feasibility
mentation system of paediatric oncology. The of the development of agent-based interoperable
research carried out an object-oriented business telemedicine systems and used the example tele-
process analysis for a clinical trial conducted electrocardiography in the case study. Among
at a German hospital. The results comprised a the tools suggested for system design, Ganguly
comprehensive business process model consisting and Ray (2000) used a UML use case diagram to
of UML diagrams and use case specifications, describe a distributed electrocardiogram system.
which included use case diagrams (“manage Finally, Hoo, Wong, Laxer, Knowlton, and
trial,” “plan trial,” “conduct trial,” “document Wan (2000) had as an objective to develop soft-
course of therapy,” and documentation view of the ware that facilitates more efficient and effective
latter) and an example of use case specification. utilisation of medical images and associated data
Weber et al. (2001) concluded by recommending in biomedical research. The area of focus was
the use of object-oriented analysis in the context assisting clinicians in presurgical evaluation of
of therapeutic trials but did not carry out the an patients with medically refractory epilepsy as an
evaluation. example. The authors drafted use cases to sum-
LeBozec, Jaulent, Zapletal, and Degoulet marize operational scenarios of clinicians using
(1998) described a UML approach to the designing the system and used UML class diagrams to
of a case-based medical imaging retrieval system describe object-oriented concepts of the system.
for pathologists. The authors created UML use
case and class diagrams to illustrate the steps of
the case-based reasoning systems methodology dIscussIon
used to develop sound knowledge systems in
pathology. The diagrams were used to visualize In this article, we introduced four common UML
the relevant objects and to evaluate the model diagrammatic tools (use case, activity, state, and
before implementation, and included use case class diagrams) and used original models of a
diagram of the image retrieval system and use simplified example of surgical service to illus-
case with corresponding scenarios chart. The trate their usage. We also presented the findings
authors concluded that, although further evalu- of a literature survey on the application of these
ation is required, UML seems to be a promising tools to the modelling of health care systems. The
formalism improving the communication between survey identified a number of articles in which
the developers and users.

284
Application of Unified Modelling Language (UML) to the Modelling of Health Care Systems

UML tools were used but very limited systematic conclusIon

evaluation of their benefits.
One notable exception is the recent work by It is apparent from this literature survey that UML
Jun (2007) where a variety of modelling methods, has a role in the analysis of health care systems.
including a variation of UML activity diagrams, There are clear benefits, especially in terms of
was evaluated for their utility in modelling health clarity of communication and repeatability, if a
systems. Jun (2007) identified that only a limited standardised and rigorous notation is employed
number of modelling methods have been con- broadly. However, the application of UML to the
sidered or evaluated for purpose of use in health modelling of health care systems is probably not
care settings. Following systematic evaluation as prevalent as in other application domains, at
by health care professionals, Jun (2007) con- least as it is documented in the medical literature.
cluded that there is no single method preferred Therefore, it is essential to conduct a thorough
by all users or applicable to all areas but there evaluation of the use and potential benefits of
is a strong case for using a variety of modelling UML in a health care context if more wide spread
techniques in enhancing the understanding of care application is to be recommended.
process among practitioners. This work is, to our
knowledge, the only comprehensive evaluation
of different process mapping tools in health care acKnowledgMent
and provided a clear insight to the benefits those
methods can offer to the system, practitioners, and This work was partially supported by an award
the patient. Similar evaluation may also be needed from the Strategic Promotion of Ageing Re-
for all UML tools before firm recommendations search Capacity (SPARC) initiative. The authors
can be made as to their applicability. acknowledge the support they received from the
Despite the obvious analytical applicability, Harrow School of Computer Science, University
also pointed out by Jun (2007), state diagrams of Westminster, as well as the generous advice
have not been used in the analysis of health care they received from Prof. Peter Millard, Prof. Peter
systems. The closest case is in an analysis of Lansley, and Dr. Elia El-Darzi.
biological systems by Roux-Rouquie, Caritey,
Gaubert, and Rosenthal (2004), where the utility
of UML state diagrams to describe and specify reFerences
biological systems and processes was examined.
Roux-Rouquie et al. (2004) mapped biological Aggarwal, V. (2002). The application of the unified
concepts to UML ones and presented state dia- modeling language in object-oriented analysis of
grams with states and substates of the active and healthcare information systems. Journal of Medi-
inactive enzymes, and with concurrent substates cal Systems, 26, 383-397.
at molecular and phenotype levels. Outside the
UML notation, the original Statecharts notation Ambler, S. (2004). The object primer: Agile
that forms the basis of the UML state diagram modeling-driven development with UML 2.0.
notation has recently been used in the modelling Cambridge: Cambridge University Press.
and simulation of a cardiac surgical service (So- Berwick, D.M. (2005). The John Eisenberg lecture:
bolev et al., 2008; Vasilakis, Sobolev, Kuramoto, Health services research as a citizen in improve-
& Levy, 2007). ment. Health Services Research, 40, 317-336.

285
 Application of Unified Modelling Language (UML) to the Modelling of Health Care Systems

Cruz, I.A., Gramaxo Silva, A., Soares, P., Oliveira, Jun, G.C.T., Ward, J.R., & Clarkson, P.J. (2005).
I., Serrano, J.A. V., & Paulo Cunha, J. (2002). Mapping the healthcare process in order to design
Modelling the hospital into the future with UML. for patient safety. In International Conference on
International Journal of Healthcare Technology Engineering Design (ICED 05) (pp. 353-364).
and Management, 4, 193-204.
Kalli, S., Ambroso, C., Gregory, R., Heikela, A.,
Ganguly, P., & Ray, P. (2000). A methodology Ilomaki, A., Leaning, M., et al. (1992). Inform:
for the development of software agent based Conceptual modelling of intensive care informa-
interoperable telemedicine systems: A tele- tion systems. International Journal of Clinical
electrocardiography perspective. Telemedicine Monitoring and Computing, 9, 85-94.
Journal, 6, 283-294.
Kumarapeli, P., De Lusignan, S., Ellis, T., & Jones,
Goossen, W., Ozbolt, J., Coenen, A., Park, A., B. (2007). Using Unified Modelling Language
Mead, C., Ehnfors, M., et al. (2004). Development (UML) as a process-modelling technique for
of a provisional domain model for the nursing clinical-research process improvement. Medical
process for use within the health level 7 reference Informatics and the Internet in Medicine, 32,
information model. Journal of the American Medi- 51-64.
cal Informatics Association: JAMIA, 11, 186-194.
LeBozec, C., Jaulent, M.C., Zapletal, E., & De-
Harel, D. (1987). Statecharts: A visual formal- goulet, P. (1998). Unified modeling language and
ism for complex systems. Science of Computer design of a case-based retrieval system in medical
Programming, 8, 231-274. imaging. In AMIA Annual Symposium Proceed-
ings (pp. 887-891).
Hederman, L., Smutek, D., Wade, V., & Knape, T.
(2002). Representing clinical guidelines in UMl: Lee, N.J., Bakken, S., & John, R. (2006). Func-
A comparative study. Studies in Health Technol- tional requirements specification and data mod-
ogy and Informatics, 90, 471-477. eling for a PDA-based decision support system
for the screening and management of obesity. In
Hoffman, K.J. (1997). Demystifying mental health
AMIA Annual Symposium Proceedings (pp. 1002).
information needs through integrated definition
(IDEF) activity and data modeling. Journal of Lunn, K., Sixsmith, A., Lindsay, A., & Vaarama,
the American Medical Association, Symposium.l, M. (2003). Traceability in requirements through
111-115. process modelling, applied to social care applica-
tions. Information and Software Technology, 45,
Hoo, K.S., Wong, S.T.C., Laxer, K.D., Knowlton,
1045-1052.
R.C., & Wan, C. (2000). Design patterns in medical
imaging information systems. Proc. SPIE Medical Lyalin, D., & Williams, W. (2005). Modeling
Imaging, 3980, 256-263. cancer registration processes with an enhanced
activity diagram. Methods of Information in
Jun, G.C.T. (2007). Design for patient safety: A
Medicine, 44, 11-13.
systematic evaluation of preocess modelling ap-
praoches for healthcare system safety improve- Maij, E., Toussaint, P.J., Kalshoven, M., Poerschke,
ment, PhD Thesis. Cambridge: University of M., & Zwetsloot-Schonk, J.H. (2002). Use cases
Cambridge, Department of Engineering. and DEMO: Aligning functional features of ICT-
infrastructure to business processes. International
Journal of Medical Informatics, 65, 179-191.

286
Application of Unified Modelling Language (UML) to the Modelling of Health Care Systems

Mehta, N.Y., Haluck, R.S., Frecker, M.I., & Snyder, Spyrou, S., Bamidis, P., Pappas, K., & Maglaveras,
A.J. (2002). Sequence and task analysis of instru- N. (2005). Extending UML activity diagrams for
ment use in common laparoscopic procedures. workflow modelling with clinical documents in
Surgical Endoscopy, 16, 280-285. regional health information systems. In Proceed-
ings of MIE2005 (pp. 1160-1165).
Object Management Group. (2005). Unified
Modeling Language: Superstructure. Version Sutton, D.R., Taylor, P., & Earle, K. (2006). Evalu-
2.0. OMG Inc. ation of PROforma as a language for implementing
medical guidelines in a practical context. BMC
Pohjonen, H., Bondestam, S., Karp, P., Kinnunen,
Medical Informatics and Decision Making, 6, 20.
J., Korhola, O., Sakari, P., et al. (1994). Quality as-
surance of the bedside chest radiography process. Vasilakis, C., Sobolev, B., Kuramoto, L., & Levy,
Acta Radiologica, 35, 235-243. A. R. (2007). A simulation study of scheduling
clinic appointments in surgical care: Individual-
Roux-Rouquie, M., Caritey, N., Gaubert, L., &
surgeon vs. pooled lists. Journal of the Operational
Rosenthal, S. (2004). Using the Unified Model-
Research Society, 58, 202-211.
ling Language (UML) to guide the systemic
description of biological processes and systems. Weber, R., Knaup, P., Knietig, T., Hauz, R.,
Biosystems, 75, 3-14. Merzweiler, A., Mludek, V., et al. (2001). Object-
oriented business process analysis of the coop-
Saboor, S., Ammenwerth, E., Wurz, M., &
erative soft tissue sarcoma trial of the German
Chimiak-Opoka, J. (2005). MedFlow - improving
Society for Paediatric Oncology and Haematol-
modelling and assessment of clinical processes.
ogy (GPOH). In Proceedings of the 10th World
Studies in Health Technology and Informatics,
Congress on Medical Informatics (MEDINFO
116, 521-526.
2001) (pp. 58-62).
Sobolev, B., Harel, D., Vasilakis, C., & Levy,
A.R. (2008). Using the statecharts paradigm for
simulation of patient flow in surgical care. Health
Care Management Science, 11, 79-86.

This work was previously published in International Journal of Healthcare Information Systems and Informatics, Vol. 3, Is-
sue 4, edited by J. Tan, pp. 39-52, copyright 2008 by IGI Publishing (an imprint of IGI Global).

287
288

Chapter 20
TreeWorks:
Advances in Scalable Decision Trees
Paul Harper
Cardiff University, UK

Evandro Leite Jr.

University of Southampton, UK

aBstract

Decision trees are hierarchical, sequential classification structures that recursively partition the set of
observations (data) and are used to represent rules underlying the observations. This article describes
the development of TreeWorks, a tool that enhances existing decision tree theory and overcomes some
of the common limitations such as scalability and the ability to handle large databases. We present a
heuristic that allows TreeWorks to cope with observation sets that contain several distinct values of
categorical data, as well as the ability to handle very large datasets by overcoming issues with computer
main memory. Furthermore, our tool incorporates a number of useful features such as the ability to move
data across terminal nodes, allowing for the construction of trees combining statistical accuracy with
expert opinion. Finally, we discuss ways that decision trees can be combined with Operational Research
health care models, for more effective and efficient planning and management of health care processes.

IntroductIon easier access to large storage capacity media; and

advances in data collection tools. The Internet
Since the second half of the 20th Century, an and all of its related services like the World Wide
upsurge of electronic data has been taking place Web, e-mail, and online databases as a global
worldwide. Studies such as Frawley, Piatestsky- information system have flooded humanity with
Shapiro, and Matheus (1991) show that the amount a tremendous amount of data and information.
of data is doubling each year. Contributing factors With the continuous growth in size and com-
for the data explosion include the widespread use plexity of information, there is an urgent need for
of computer systems for nearly any commercial, a new generation of computational theories and
financial, governmental, or research activity; tools to assist us in extracting useful information

Copyright © 2011, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
TreeWorks

(knowledge) from the rapidly growing volumes tion tree, but considers in his studies only datasets
of digital data. These theories and tools are the that could fit in main memory which were rather
subject of the fields of data mining and knowl- limited in size. Methods for partitioning the da-
edge discovery in databases (KDD). The need to taset such that each subset fits in main memory
understand data is extremely important for busi- are considered by Chan and Stolfo (1993). Even
ness, government, and research. Examples are though this method makes classification of large
numerous and varied and include applications in datasets possible, their studies show that the qual-
optimising market shares by increasing competi- ity of the resulting decision tree is not as good
tive advantage using knowledge extracted from as if the classifier had used all the available data.
sales transactions (Rygielski, Wang, & Yen, 2002), Most existing methods for automatic construction
maximising customer retention (Edelstein, 1998), of classification trees utilise greedy heuristics,
predicting the size of television audiences (Brach- choosing locally optimal splits to divide the data
man, Khabaza, Kloesgen, Piatetsky-Shapiro, & at each level. Unfortunately these locally optimal
Simoudis, 1996), and improving ovarian cancer values cannot be obtained quickly when a large
detection (Li et al., 2004). dataset is analysed.
Decision trees are one such data mining tech- In this article, we present TreeWorks, a CART
nique that learn from data and generate models tool that enhances existing decision tree theory
containing explicit rule-like relationships among and overcomes some of these common decision
the variables. Decision tree algorithms begin with tree limitations. Whilst the primary focus of our
the entire training set of data, split into two or research, as presented here, is on scalable deci-
more subsets until the split size reaches an ap- sion trees, the resulting TreeWorks tool is highly
propriate level. The entire modelling process can practical and user-friendly and has already found
be visualised in a tree structure. This structure considerable application by the UK National
maps observations between dependent and inde- Health Service (NHS). The NHS handles millions
pendent variables. An arc between two nodes in of patient records each year and TreeWorks has
the tree represents a partition of the parent node assisted the NHS Information Centre with the
into child nodes. All observations follow a path redesign of Healthcare Resource Groups (HRGs).
from the root (initial node) and are assigned to HRGs, which are similar to DRGs as used in the
a leaf (terminal node) based on splitting criteria U.S., are standard groupings of clinically similar
(values of the independent variables). The two treatments which use common levels of healthcare
best-known and most widely used decision tree resource, and are fundamental for standardising
algorithms are Classification and Regression Trees healthcare commissioning across the country as
(CART) and C4.5 (a successor of ID3). CART was part of the UK Government’s policy of Payment
developed by the statisticians Breiman, Friedman, by Results (PbR) (Department of Health, 2008).
Olshen, and Stone (1984), and C4.5 was developed In this article, we also highlight ways in which
by Quinlan, a computer scientist in the field of decision trees can support Operational Research-
machine learning (Quinlan, 1993). ers building health care models. A particular
The original methods to grow decision trees feature of health care processes is the inherent
are not ideal for handling some of the features variation and uncertainty in treating individuals.
that are present in modern-day data sets such as Homogeneity leads to increased certainty in indi-
categorical variables with many distinct values and vidual patient predictions (resource consumption,
the ability to handle extremely large datasets. In outcomes, pathways, etc.), which in turn results in
order to help overcome such issues, Catlett (1991) the potential for more effective and efficient plan-
proposed sampling at each node of the classifica- ning and management of health care processes.

289
 TreeWorks

Thus decision trees can play a useful role here where p(i0) = s(i0)/s(i) and p(i1) = s(i1)/s(i) are the
and the benefits of a combined data mining and proportion of records assigned to node i0 and i1
modelling approach are illustrated. respectively. The tree is grown recursively: at
In the next section, a brief overview of deci- each step we choose which node to split and the
sion trees (both for classification and regression) variable to split on in order to maximise the gain,
is provided together with methods for minimising that is, minimise the impurity. We continue to do
impurity of trees. This is followed by a section this until all nodes are pure enough, according to
on our approach for handling large datasets. A some stopping rule.
heuristic for handling categorical variables is Two examples of ways to evaluate the homo-
then presented and evaluated, and some other geneity of categorical data are the Gini index of
features of TreeWorks are discussed. Finally we diversity and the information entropy. Gini is
discuss the benefits of a combined decision tree- based on squared probabilities of membership
modelling approach. for each target category in the node. It reaches
its minimum (zero) when all cases in the node
classification and regression trees fall into a single target category. Suppose that a
variable y takes on values in 1, 2, . . . , m, and let
Based on the type of the dependent variable, deci- f(i, j) be the frequency of value j in node i. That
sion trees can be divided into classification trees is, f(i, j) is the proportion of records assigned to
(categorical dependent variable) and regression node i for which y = j.
trees (continuous dependent variable). TreeWorks m
is a CART tool that is able to handle both clas- I G (i ) = 1 - ∑ f (i, j ) 2
sification and regression problems. CART works j =1

on the principle of binary splitting of data based

on improving gain (or improving the purity of the Information entropy is based on the concept
tree into more homogeneous nodes). A number of entropy as used in information theory, and
of purity measures are available both for classi- takes the form:
fication and regression problems. These include m
information gain, gini index of diversity, gain I E (i ) = -∑ f (i, j ) log f (i, j )
ratio criterion, MaxDif, and Generalized Gini. j =1

The interested reader may consult Berzal et al.

(2003) for a comprehensive review. Whilst classification trees classify objects
Classification trees are suitable when the into discrete classes, regression trees are used
dependent variable is categorical, for example when the class is continuous. A function y(x1,
in predicting patient survival or outcome in the x2, . . . , xn) of n continuous or discrete attributes
healthcare domain. The nodes are split on values can be implemented. Values of the independent
of the independent variables by minimising purity variables are split by a measure for continuous
measures for categorical data. Let s(i) be the size variable such as least square deviance (LS). Let
of node i, and suppose we have a measure for the y( j) be the value of y for record j and let be the
impurity I(i) of node i, then the gain in purity mean value of y over node i, then:
made by splitting node i into nodes i0 and i1 is:

Gain(i, i0,i1) = I(i) - (I(i0)p(i0)+I(i1)p(i1) IV (i ) =

∑ j∈i
( y ( j ) - y (i)) 2
s (i ) - 1

290
TreeWorks

Other methods used to build regression trees Figure 1. Row referencing framework for han-
include the least absolute deviation (LAD) which dling data
minimises the differences between the values and
the median of the dependent variable. Studies car-
ried out to compare LS and LAD regression trees,
such as Torgo (1999), reveal that both models have
different preference biases that can be considered
useful depending on the application. LAD trees
tend to produce predictions that, on average, are
more accurate. However, these trees do commit
large errors more often than LS trees.

Handling large datasets

There exist a number of algorithms to construct

decision trees. Most algorithms in the machine
learning and statistics community are main mem-
ory algorithms, even though today’s databases are
in general much larger than main memory. There
have been several approaches to dealing with large
databases, and the interested reader is referred to are executed by a DBMS (Database Management
Gehrke, Ramakrishnan, and Ganti (2000) for a System). In order to evaluate purity functions, the
review of approaches. One particular approach algorithm will examine notably smaller groups
is to group into similar subgroups each ordered of data called the Attribute-Value-Class-set, or
attribute and run the algorithm on the grouped AVC-set, and Attribute-Value-Square-set, or
data. However all grouping methods for classifica- AVS-set. The AVC-set and the AVS-set, which are
tion that take the class label into account assume respectively used for explaining categorical and
that the database fits into main memory (Fayyad numerical data, are extracted using SQL queries
& Irani, 1993; Maass, 1994; Quinlan, 1993). We and are subsequently used to measure the value
follow the same principle as described in Gehrke of the purity functions. The size of the AVC and
et al. (2000) that “the scalable versions of the al- AVS sets depends only upon the number of distinct
gorithms produce exactly the same decision tree values of the independent variable being tested,
as if sufficient main memory were available to run thus most of the complex and time-consuming
the original algorithm on the complete database process for accessing data is undertaken internally
in main memory.” Therefore, below we describe by the database management system. Unlike the
our approach that does not sample the data for RainForest framework, TreeWorks uses row ref-
gaining speed but instead is a heuristic that uses erences for the node’s data, linking to data rows
all the available data for the split. available in all the sets rather than replicating
The TreeWorks software utilizes a variation on a dataset for the node. The implemented varia-
the RainForest framework (Gehrke et al., 2000) tion of RainForest AVC (which is illustrated in
that enables large blocks of data to be analysed. Figure 1) is:
The basic concept revolves around proposing a
model that does not implement the data storage and SELECT ALL
access methods on a low level, as these functions AllData.Xi, AllData.Y,

291
 TreeWorks

COUNT(*) a proposed Heuristic for

FROM classification problems
AllData,
WHERE Categorical splits have frequently presented a
AllData.Id = IndexesNode∂.Id problem to classification trees due to the fact that
and Fn categories cannot be ordered and the number of
GROUP BY possible partitions to be tested explodes at the
AllData.Xi, AllData.Y rate of 2m-1− 1, where m is the number of values
belonging to a given independent variable, which
Where: is NP-hard. In our studies, using a PC Intel Core2
Duo @2.13GHz with 2GB of RAM, the time
AllData is the table that represents all data avail- in seconds, s, to evaluate just one split with a
able for training the model, categorical variable containing m distinct values
IndexesNode∂ represents a node’s data table con- and using the Gini purity function, was found to
taining only one column that links to rows be s=0.00185e0.569m. This relationship is shown in
of AllData which should be the records that Figure 2. With just 28 distinct values, this would
belongs to IndexesNode∂. In Figure 1 this take over 4 hours just to explore one possible split.
can be any of root node, child1 or child2, With many categorical variables to consider at
and so forth. each level in the tree, clearly this could be very
Y is the dependent variable. time-consuming. With 30 distinct values it would
X1, X2, X3, and so forth, are the independent take approximately 13 hours for the full-search and
variables. with 35 more than 5,000 days. Yet in our review
Fn corresponds to the table where the node train- of commercially available CART packages, all
ing data is stored. In our version, it is the appeared to use the full-search algorithm, with
join between the main database table that some well-know tools complaining of memory-
contains the training cases and the one space limitations and others simply crashing.
column table which contains the symbolic We propose a heuristic that is straightforward
links to the training case rows. to implement and in tests has shown to be suf-
ficiently accurate. The heuristic is based on the
Although this variation requires the joining of notion that the inclusion of values will either
two tables, an aspect that takes some extra time increase or reduce the purity of a split, thus we
to create the AVC set, it saves considerable time can recursively select a value of the independent
during the split step where only one column will variable to be analysed as a split candidate. We
need to be populated with data. then include all of the rows with this value in
TreeWorks utilises the PosgreSQL (Momjian, the left node and compute the node purity, then
2001) database management system. There were remove these row from the left and try adding
many reasons for this choice, such as PosgreSQL’s them to the right node and recompute the purity,
excellent performance and scalability, and cross- finally choosing the side that has the better gain
platform functionality with versions for both based on the chosen split criteria. This procedure
Unix and Windows computers enabling it to be is divided into two parts. Part (1) is a set of steps
used in a dedicated database server or on a user’s executed on every occasion and is where the
workstation. majority of the improvement occurs, and part
(2) contains optional steps that can be executed
in order to try and achieve further improvement.

292
TreeWorks

Figure 2. Relationship between number of distinct categorical values and CPU time

The algorithm for a parent node p to be split into part (2)

a left node (left), and a right node (right), by using
the independent variable i, is described bellow: i. Execute all the steps of part one.
ii. For every element in left test if the gain will
part (1)
improve if it is moved to right.
iii.For every element in right test if the gain
i. Sort the values of i contained in p by its
will improve if it is moved to left.
lexicographical order.
iv. Use that combination of values and impurity
ii. Assign the data cases that contain first value
value for the split.
of i to left.
End of part two
iii. Assign the data cases that contain the second
value of i to right.
iv. From the third to last values, try putting the evaluation of the Heuristic
values in left or right and assign the value
to the side that minimises the impurity of Once again using a PC Intel Core2 Duo @2.13GHz
the split. with 2GB of RAM, the time in seconds, s, to
v. Three last combinations need to be tested: evaluate just one split with a categorical variable
a. Remove the first value from left and put containing m distinct values using the heuristic
in right. was found to be s=0.000000195m2.8+0.18. Even
b. Remove the second value from right and with m = 4,000 the split takes less than 40 minutes.
send to left. One clear advantage of the heuristic search
c. Put back the first value to left. algorithm is that it permits the construction of
vi. Use the smallest impurity combination from all trees including some categorical variables that
the steps iv) or v). End of part one would otherwise have not been possible using the
full-search with 2m-1-1 combinations. In our tests

293
 TreeWorks

on various size datasets, TreeWorks has never run and costs. Independent variables include such
out of memory and has always been able to produce fields as patient source (emergency or elective),
trees in reasonable time even for datasets with NHS Trust (or hospital), current version of the
categorical variables containing many thousands HRG, age, sex and procedure code (operation
of distinct values. In TreeWorks, the user defines type) where applicable. Due to sensitivity of
the threshold value of m, above which the heuristic the data, we are unable here to provide explicit
will be implemented. Categorical variables with names of variables in our results. For experiment
distinct values below this threshold will be split 3, we use datasets that are freely available from
using the full-search algorithm. UCI Machine Learning Repository, School of
We have compared the accuracy of the full- Information and Computer Science, University
search and heuristic partial-search algorithms. of California (Asuncion & Newman, 2007)
The heuristic performed extremely well on all
tests with only a few tests where there were small experiment 1
differences in gain between the two approaches.
Three such experiments are detailed below. For We created 10 data samples, with each sample
experiments 1 and 2, we use a large datasets sup- containing 20% of the overall dataset with
plied by the NHS Information Centre as part of data randomly sampled without reposition. We
the redesign of HRGs. The dataset contains over tested all 10 datasets using both the full-search
100,000 records for hospital inpatient stays in the and our developed heuristic, and recorded the
UK and more than 50 variables. The dependent gain for one split using 6 different categorical
variables of interest include patient length of stay independent variables. Each chosen categorical

Table 1(a). Gain values obtained from the full-search.

Variable Set 1 2 3 4 5 6 7 8 9 10
A 35.02 37.28 34.67 35.14 33.88 37.96 35.07 35.68 37.42 37.40
B 8.353 6.372 5.683 7.096 5.385 7.545 8.243 8.603 7.281 7.771
C 7.945 6.431 5.559 6.636 4.909 7.959 8.141 7.969 8.101 7.078
D 86.27 84.98 84.98 87.18 84.20 85.46 86.74 85.23 86.79 84.88
E 35.12 37.84 34.71 35.38 34.00 38.45 35.14 36.18 37.50 37.68
F 0.920 1.140 1.529 1.553 0.813 1.082 0.913 2.395 1.673 1.159

Table 1(b). Gain values obtained from the heuristic search. †denotes a difference in gain compared to
Table 1(a).
Variable Set 1 2 3 4 5 6 7 8 9 10
A 35.02 37.28 34.67 35.14 33.88 37.96 35.07 35.68 37.42 37.40
B 8.353 6.372 5.683 7.096 5.385 †7.478 8.243 8.603 7.281 †7.613
C 7.945 6.431 5.559 6.636 4.909 7.959 8.141 7.969 8.101 7.078
D 86.27 84.98 84.98 87.18 84.20 85.46 86.74 85.23 86.79 84.88
E 35.12 37.84 34.71 35.38 34.00 38.45 35.14 36.18 37.50 37.68
F 0.920 1.140 1.529 †1.531 0.813 1.082 0.913 2.395 1.673 1.159

294
TreeWorks

variable contained different numbers of distinct independent variables. presents results from the
values but in this first experiment were chosen comparison and shows CPU time and gain. We
to be of reasonable size (less than 30) such that have used three different NHS datasets covering
the full-search was possible. In Table 1, sections intensive care, HRG data, and inpatient admission
(a) and (b) show results from the full-search and data. Fifteen tests and 30 comparisons across the
heuristic (implemented with both part (1) and (2) three different datasets have been performed. For
of the algorithm) respectively. some tests, such as with those with 3,754 distinct
There were only three occasions where the values of the independent variable, it was not
gain values differed, marked by † in (b) of Table possible to evaluate the gain for the full-search
1. Overall the heuristic achieved the optimal gain since the CPU time would have been in the order
in 57 of the 60 tests (95% of the time) and with an of days, months, or even years, and in such cases
overall gain accuracy, measured using the absolute we can only estimate what the CPU might have
difference in gain values, of 99.8%. Chi-squared been based on the exponential relationship as
tests showed that there was no evidence to reject described earlier.
the hypothesis that the gains were the same for In summarising the performance of the heu-
the full and partial searches. ristic, there are clear benefits in CPU time with
the partial-search at most taking 1.2 hours when
experiment 2 handling in excess of 115,000 cases and 3,754
distinct values of the independent variable (test
The second experiment evaluates the performance N). With 4 and 37 distinct values of the depen-
of the heuristic, with a focus on both accuracy and dent and independent variables respectively for
CPU time, for a varying number of data cases and 115,000 cases, TreeWorks takes less than one
distinct values for both categorical dependent and second whereas a full search is estimated to take

Table 2. Comparison of CPU time between full-search and heuristic for experiment 2. (e) indicates an
estimated CPU time and (†) denotes a difference in gain between the full-search and heuristic.

Number of distinct values Full-search Heuristic

Test Dependent Independent Cases
CPU Time Gain CPU Time Gain
Variable Variable
A 2 23 1332 253.906 6.266 0.016 6.266
B 2 19 768 11.828 2.715 0.016 2.666 (†)
C 2 4 564 0.016 0.706 0.016 0.706
D 2 7 602 0.016 0.524 0.015 0.524
E 3 20 2582 50.384 6.238 0.017 6.238
F 3 23 2582 422.528 5.996 0.018 5.996
G 3 26 2582 3984 7.026 0.019 7.026
H 3 28 2582 11729 9.815 0.320 9.815
I 7 20 2582 90.7 9.019 0.017 9.019
J 37 4 80815 0.156 19.108 0.158 19.108
K 37 4 115451 0.294 19.12 0.201 19.12
L 4 37 115451 10808496 (e) N/A 0.212 22.565
M 37 563 115451 2.3E+166 (e) N/A 42.723 20.533
N 37 3754 115451 8.7E+1038 (e) N/A 4305.9 20.536
O 4 3754 115451 3.2E+1038 (e) N/A 3758.8 30.319

295
 TreeWorks

125 days (Test L). For those tests that that could be a of Table 3. Tests 1 – 4 use the Adult Data Set
compared, in only one (test B) was the heuristic (http://archive.ics.uci.edu/ml/datasets/Adult) and
not optimal with a discrepancy of less than 2%. tests 5 – 9 the Census-Income (KDD) Data Set
The overall accuracy of the heuristic was greater (http://archive.ics.uci.edu/ml/datasets/Census-
than 99.95%. It is also worthy of note that some Income+%28KDD%29). In all tests, TreeWorks,
of the more significant and valuable split gains in is used to predict whether an individual’s income
the region of 20% to 30% were achieved in tests exceeds $50,000 per annum. Varying number
L, M, N, and O, which would have never been of independent variables and cases are included
possible with the full-search. across the range of tests we have performed and
are shown in section b of Table 3. The KDD dataset
experiment 3 contains 299,285 instances (records) and for tests
5 through 8, we sample 5% of records to allow
The third set of experiments makes use of available a quicker comparison between the heuristic and
extensive datasets from UCI Machine Learning the full search but varied the selected independent
Repository, University of California (Asuncion variable and hence associated number of distinct
& Newman, 2007). We repeat the same testing values (from 9 to 42). However in test 9, we use
structure as experiment 2, focusing on both ac- all 299,285 instances.
curacy and CPU time. Results are given in section A full search gain was possible in 8 of the
9 tests. For each of these 8 tests, the heuristic

Table 3a. Comparison of CPU time between full-search and heuristic for experiment 3.

Number of distinct values Full-search Heuristic

Test Dependent Independent Cases
CPU Time Gain CPU Time Gain
Variable Variable
1 2 9 48842 0.092 2.101 0.075 2.101
2 2 16 48842 1.429 10.704 0.077 10.704
3 2 15 48842 0.778 9.479 0.079 9.479
4 2 42 48842 Not possible N/A 0.093 0.764
5 2 17 14964 3.631 11.92 0.557 11.92
6 2 24 14964 648.845 6.295 0.522 6.295
7 2 15 14964 1.3 12.821 0.553 12.821
8 2 28 14964 8877.493 4.632 0.595 4.632
9 2 24 299285 677.945 6.438 1.055 6.438

Table 3b. Summary of independent variables used in experiment 3.

Test Dataset Independent Variable

1 Adult workclass
2 Adult education
3 Adult occupation
4 Adult native_country
5 Census-income education-detailed
6 Census-income major_industry_code
7 Census-income major_occupation_code
8 Census-income household_and_family_stats
9 Census-income major_industry_code

296
TreeWorks

achieved the same gain as the full search. In test personnel. The resulting tool allows for new HRG
9, which contained the full 299,285 instances of groups to be agreed using a combination of the
the census-income data, TreeWorks computed CART statistical algorithm combined with expert
the same gain as the full search in only 1 second opinion. For example, it was often the case that a
compared to over 11 minutes for the full search. purely statistical tree with optimal gain (reduction
We conclude that given the accuracy and CPU in variance) did not make complete clinical sense,
time across a range of different datasets and ex- and subsequently could not be considered in its
periments, both health care and nonhealth care entirety as an HRG. Consequently, it was neces-
related, that TreeWorks is sufficiently accurate sary to move some surgical procedures and patient
and responsive for real-time business applications clusters from one node to another. TreeWorks
handling realistic large sized databases. allows for this to be conducted in real-time and
reports on the consequence on the reduction in
additional treework Features variance before the data is moved. An illustrative
screen-shot is shown in Figure 3. The move data
In addition to the ability to handle large databases form allows the user to create a series of SQL-
and many categorical values, TreeWorks incorpo- type commands to select the data to move. For
rates a number of other features, some of which example, in Figure 3, data for emergency patients
are mentioned here. Many of these features were aged 16 and under are to be moved from node 1 to
based on requests from users of TreeWorks, such 2. The user can preview the selected data rows and
as those from staff at the NHS Information Centre, reduction in gain prior to confirming the move.
which further enhance the tool and permit it to Other well-known software does not permit for
be both powerful and highly user-friendly and this level of interaction.
interactive. Features include: Model validation is also important, and Tree-
Works allows the user to select the percentage
• Provision of tree model validation (train- split between the training and testing datasets. The
ing and testing datasets) with user-chosen two samples are randomly selected using pseudo-
percentage for training set. random numbers based on Knuth’s subtractive
• Possibility of moving data across terminal random number generator algorithm (Knuth 1981).
nodes. Figure 4 illustrates TreeWork’s model validation.
• Description of the node paths. A summary of each node (with terminal nodes
• Visualisation of the data at each node with highlighted) provides statistics such as number
standard database functionality. of cases, average value and standard deviation.
• Support of missing values in the data. The final columns show whether the node has
• Allowing the user to interact with the model, passed or failed a t-test which is used to compare
choosing the split variable, split values and the training and testing models for each node.
growing and pruning sub-trees.
• User-friendly wizards guiding the user decision trees for Health care
through the data handling processes. Modelling

The interactive nature of the tool was par- Operational Research methods are widely used for
ticularly important in the redesign of HRGs, health care modeling and have found considerable
since it was a preferred NHS approach to build application (Brandeau, Sainfort, & Pierskalla,
classification and regression trees interactively 2004; Ozcan, 2005). Health care modelling,
with teams of expert consultants and medical however, is beset with many challenges (Harper

297
 TreeWorks

Figure 3. Moving data between nodes

Figure 4. Terminal node model validation

& Pitt, 2004). A particular feature is the inherent 2005). Homogeneity leads to increased certainty
variation and uncertainty in treating individuals. in individual patient predictions (resource con-
For example, length of stay in hospital or the infec- sumption, outcomes, pathways, etc.), which in
tious period for a given disease typically varies turn results in the potential for more effective
from patient to patient. From both a clinical and and efficient planning and management of health
operational perspective, it is desirable to be able care processes.
to understand and capture this variability (Harper,

298
TreeWorks

In designing and building health care models, we construct decision trees. The chosen dependent
there are a number of approaches when consider- variable would be of relevance to the nature of
ing how to capture patient variability: the patient-based model, such as dwelling time in
a particular state of the patient pathway or prob-
• Ignore variability: build deterministic mod- ability of transition from one state to another. All
els. Essentially here we have one patient patients will be assigned to one of the terminal
group (all the available data) and are using nodes in the decision tree. Each terminal node
average values. becomes a unique patient group in the model.
• Resample all individuals. In this model we Here we take the model to represent the individual
recreate every observed individual to exactly patient pathways, such as movements through a
recreate real-life. This is time-consuming health service provider or transitions through a
and still lacks the ability to provide insight natural history of a disease. Each individual pa-
for future predictions, such as case-mix or tient that enters the model will belong to a patient
demand for services. group. Dwelling times and other parameters in
• Build a stochastic model with one “generic” the model will be taken from the group of which
patient group. In this model we define one the patient is a member. To capture any dynamic
distribution for each parameter in the model. effects, we may decide to create multiple decision
We use all the available date to define each trees for different parts of the model, and re-assign
distribution, thus we are sampling indi- patients to groups as appropriate
viduals from the entire possible range of A combined data mining and modelling ap-
(observed) values. proach has been adopted for various studies by
• Create patient groups. Each patient group the authors. These include hospital capacities
will have their own set of parameters, dis- (Harper, 2002), intensive care (Costa, Ridely,
tributions, care-pathways, and so forth. Shahani, Harper, de Senna, & Nielsen, 2003),
diabetic retinopathy (Harper, Sayyad, de Senna,
The benefit of the last approach is that we are Shahani, Yajnik, & Shelgika, 2003) and screen-
able to construct clinically and/or statistically ing for chlamydia (Evenden, Harper, Brailsford,
meaningful patient groups that we can then use & Harindra, 2005). To illustrate the concept, for
as patient groups in developed models in order modelling hospital resource capacities, Harper
to capture variability. As we create more groups, (2002), developed a discrete event simulation
naturally we capture more of the variability and to capture individual patient pathways through
increase information content. However, what hospital and monitor corresponding resource
typically happens is that we reach a point when needs. The challenge was to adequately handle
creating more patient groups does not lead to a the variability such as length of stay, operating
further significant capture of variability or in- times and workforce needs. Decision trees were
creased understanding. This is similar to Pareto’s constructed to define patient groups and fit dis-
principle (80-20 rule). If we pursue the patient tributions for various parameters in the model.
grouping approach, then we need to know how We were able to work with hospital clinicians and
many groups to create and group definitions managers to create groups that were both statis-
(e.g., for hospital length of stay we might create tically and clinically meaningful. For example,
groups using indictors such as age, sex, elective in one hospital it was possible to categorize all
or emergency, speciality, etc.). in-patients into 15 patient groups that were then
We suggest therefore that a combined data fed into the simulation model. Hospital managers
mining and modelling approach is beneficial. First could then change any of the parameter values

299
 TreeWorks

for any of the 15 groups for scenario analysis, There are a number of future TreeWork en-
such as a reduction in length of stay or change hancements proposed. These include the ability
in workforce needs by skill-mix of staff for that to produce hybrid trees that automatically select
patient group. Future research will examine a the best splitting function (for example from Gini
framework for formally interfacing TreeWorks index of diversity or information entropy) at each
with a developed simulation shell to facilitate the level to create the overall best tree. Also we wish
combined data mining and operational research to explore the incorporation of “fuzzy” splitting
modelling approach. rules. Rule-based trees, with “crisp” splitting
criteria such as CART, can occasionally seem
overly harsh. In reality, individual membership
conclusIon could well be in different terminal nodes and
each new case may be assigned node membership
This article presents TreeWorks, a classification probabilities. Finally, future research will also
and regression tree tool that overcomes some of examine a framework for interfacing TreeWorks
the common limitations of decision tree software. with a developed simulation modelling shell to
We suggest a heuristic for allowing decision trees facilitate a combined decision tree and operational
to overcome issues of scalability and handle research modelling approach.
observation sets that contains several distinct
values of categorical data. Experimentation has
demonstrated that the heuristic nearly always reFerences
achieves the optimal gain and for practical pur-
poses is sufficiently accurate. The distinct advan- Asuncion, A., & Newman, D.J. (2007). UCI ma-
tage is the CPU time and the ability to produce chine learning repository. Irvine, CA: University
models in reasonable times even for situations of California, School of Information and Computer
with thousands of distinct values of categorical Science. Retrieved May 26, 2008, from http://
data. Indeed some of the best gains (reductions in www.ics.uci.edu/~mlearn/MLRepository.html
variance) were found in splits which would not be
Berzal, F., Juan-Carlos, C., Cuenca, F., & Martín-
possible to consider using the standard full-search
Bautista, M.J. (2008). On the quest for easy-to-
approach. All the commercial software that we
understand splitting rules. Data and Knowledge
have reviewed implements the full-search and so
Engineering, 44(1), 31-48.
are severely limited. With the continuous growth
in size and complexity of information, there is Brachman, R.J., Khabaza, T., Kloesgen, W.,
an urgent need for a new generation of tools to Piatetsky-Shapiro, G., & Simoudis, E. (1996).
assist in extracting knowledge from the rapidly Mining business databases. Communications
growing volumes of digital data. We believe that of the Association for Computing Machinery,
TreeWorks represent a new chapter in enhancing 39(11), 42-48.
the scalability of decision trees, and the tool has
Brandeau, M.L., Sainfort, F., & Pierskalla, W.P.
already found considerable application in han-
(2004). Operations research and health care.
dling millions of healthcare records by the UK
Boston, MA: Kluwer.
National Health Service. Further features of our
tool increase the ease of use and functionality, Breiman, L., Friedman, J., Olshen, R., & Stone,
such as the ability to combine the CART algo- C. (1984). Classification and regression trees.
rithm with expert opinion by moving data across Belmont, CA: Wadsworth International Group.
terminal nodes.

300
TreeWorks

Catlett, J. (1991). Megainduction: Machine learn- Harper, P.R. (2002). A framework for operational
ing on very large databases. Unpublished doctoral modelling of hospital resources. Health Care
dissertation, Basser Department of Computer Management Science, 5(3), 165-173.
Science, University of Sydney.
Harper, P.R. (2005). A review and comparison
Chan, P.K., & Stolfo, S.J. (1993). Meta-learning of classification algorithms for medical decision
for multistrategy and parallel learning. In Pro- making. Health Policy, 71, 315-331.
ceedings of the Second International Workshop
Harper, P.R., & Pitt, M. (2004). On the challenges
on Multistrategy Learning.
of health care modelling and a proposed project
Costa, A.X., Ridely, S.A., Shahani, A.K., Harper, life-cycle for successful implementation. Journal
P.R., de Senna, V., & Nielsen, M.S. (2003). Math- of the Operational Research Society, 55, 657-661.
ematical modelling and simulation for planning
Harper, P.R., Sayyad, M.G., de Senna, V., Shahani,
critical care capacities. Anaesthesia, 58, 320-327.
A.K., Yajnik, C.S., & Shelgika, K.M. (2003). A
Department of Health. (2008). Payment by results systems modelling approach for the prevention
(PbR). Retrieved May 26, 2008, from www.dh.gov. and treatment of diabetic retinopathy. European
uk/en/Policyandguidance/Organisationpolicy/ Journal of Operational Research, 150, 81-91.
Financeandplanning/NHSFinancialReforms/
Li, Tang, Wu, Gong, Gruidl, Zou, et al. (2004).
index.htm.
Data mining techniques for cancer detection using
Edelstein, H. (1998). Data mining: Myth and real- serum proteomic profiling. Artificial Intelligence
ity. GIGA Information Group Telepresentation. in Medicine, 32(2), 71-83.
Evenden, D., Harper, P.R., Brailsford, S.C., & Maass, W. (1994). Efficient agnostic pac-learning
Harindra, V. (2005). Improving the cost-effec- with simple hypothesis.
tiveness of chlamydia screening with targeted
Momjian, B. (2001). PostgreSQL: Introduction
screening strategies. Journal of the Operational
and concepts. Addison-Wesley Longman Pub-
Research Society, 57(12), 1400-1412.
lishing Co., Inc.
Fayyad, U.M., & Irani, K. (1993). Multi-interval
Ozcan, Y.A. (2005). Quantitative methods in
discretization of continuous-valued attributes
health care management. San Francisco, CA:
for classification learning. In Proceedings of the
Jossey-Bass (Wiley).
International Joint Conferences on Artificial
Intelligence 1993, San Francisco, CA. Quinlan, J.R. (1993). C4.5: Programs for machine
learning. San Mateu: Morgan Kaufmann series
Frawley, Piatestsky-Shapiro, & Matheus (1991).
in machine learning.
Knowledge discovery in databases: An overview.
Menlo Park, CA: AAAI Press. Rygielski, C., Wang, J.-C., & Yen, D.C. (2002).
Data mining techniques for customer relationship
Gehrke, J., Ramakrishnan, R., & Ganti, V. (2000).
management. Technology in Society, 24, 483-502.
RainForest - a framework for fast decision tree
construction of large datasets. Data Mining and
Knowledge Discovery, 4(2-3), 127-162.

This work was previously published in International Journal of Healthcare Information Systems and Informatics, Vol. 3, Is-
sue 4, edited by J. Tan, pp. 53-68, copyright 2008 by IGI Publishing (an imprint of IGI Global).

301
302

Chapter 21
Ambient Intelligence and
Pervasive Architecture Designed
within the EPI-MEDICS Personal
ECG Monitor
Hussein Atoui
Université de Lyon and INSERM, France

David Télisson
Université de Lyon and INSERM, France

Jocelyne Fayn
Université de Lyon and INSERM, France

Paul Rubel
Université de Lyon and INSERM, France

aBstract

Recent years have witnessed a growing interest in developing personalized and nonhospital based
care systems to improve the management of cardiac care. The EPI-MEDICS project has designed an
intelligent, portable Personal ECG Monitor (PEM) embedding an advanced decision making system.
We present two of the ambient intelligence models embedded in the PEM: the neural-network based
ischemia detection module and the Bayesian-network risk stratification module. Ischemia detection was
expanded to take into account the patient ECG, clinical data, and medical history. The neural-network
ECG interpretation module and the Bayesian-network risk factors module collaborate through a fuzzy-
logic-based layer. We also present two telemedicine solutions that we have designed and in which the
PEM is integrated. The first telemedical architecture was created to allow the collection of medical
data and their transmission between healthcare providers to get an expert opinion. The second one is
intended for improving healthcare in old people’s homes.

Copyright © 2011, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Ambient Intelligence and Pervasive Architecture Designed within the EPI-MEDICS Personal ECG Monitor

IntroductIon This approach has thus proved to be very im-

practical for patients with infrequent symptoms
Recent years have witnessed a growing inter- such as arrhythmias and ischemia that represent
est in developing personalized and nonhospital 85% of the cardiac diseased patients, and would be
based care systems to improve the management very expensive if adopted for every citizen at risk.
of cardiac care (Axisa, Schmitt, Gehin, Del- The European project EPI-MEDICS has de-
homme, McAdams, & Dittmar, 2005; Campbell signed a solution based on the interpretation of
et al., 2005; Kerkenbush & Lasome, 2003). The ECG derived cardiological syndromes and devel-
reason behind such interest is due to the fact that oped a friendly and easy-to-use, cost-effective
cardiovascular diseases now represent the leading intelligent personal ECG monitor (PEM) (Rubel
cause of mortality in Europe and reducing the et al, 2004; Rubel et al., 2005). The device (Figure
time before hospitalization is crucial to reducing 1) is capable of recording a simplified 4-electrode,
cardiac morbidity and mortality (McMurray & professional quality 3-lead ECG, to derive the
Rankin, 1994; Task Force Report, 1998). missing 5 leads (V1, V3… V6) of the standard
Event recorders and transtelephonic ECG 12-lead ECG (Atoui, Fayn, & Rubel, 2004), to
recorders are thus increasingly used to improve store the derived 12-lead ECG according to the
decision making in the prehospital phase. How- SCP-ECG standard (EN 1064, 2007), to analyze
ever, such systems are usually unable to capture and interpret the recorded ECG, to detect arrhyth-
transient ECG events such as infrequent ar- mias and ischemia or acute myocardial infarction,
rhythmias or ischemic episodes. In addition, all and to send an alarm message to the appropriate
these systems require setting up new information health care providers.
technology infrastructures and medical services To develop the PEM software platform, we
and need skilled personnel to interpret the ECG were confronted to a variety of problems related
and make decisions for the patient care. to the system intelligence and functioning: from
recording and storing the ECG according to the

Figure 1. The personal ECG monitor (PEM) device allows for early detection of arrhythmia and ischemia
in the pre-hospital phase and thus for better and more adapted treatment

303
Ambient Intelligence and Pervasive Architecture Designed within the EPI-MEDICS Personal ECG Monitor

SCP-ECG standard, to analyzing the recorded for handling any communication problem with
ECG and combining the result of interpreta- the contacted health care providers.
tion with the patient’s risk factors in a unique Arrhythmia detection and ischemia diagnosis
risk score, to finally triggering and managing are based on the comparison of the last recorded
an appropriate alarm message according to the ECG with a reference ECG. Arrhythmia detec-
calculated risk score. tion is rule based. For the diagnosis of ischemia,
The objective of this article is to present an a multi-expert decision making module based on
overview of the current intelligence aspects em- artificial neural network committees has been
bedded in the PEM prototypes in terms of services designed (Rubel et al., 2004).
and scenarios and to describe some of the latest Different levels of alarms are forwarded to
ambient intelligence and pervasive solutions that the relevant health care providers depending on
have been designed to: the severity of the cardiac event (Rubel et al.,
2005). Major alarms (acute ischemia/infarction,
• allow the PEM user to record a good quality severe arrhythmia) are automatically transmitted
ECG without any professional assistance to the nearest emergency call center. In case of a
using a reduced electrode set in a simplified medium alarm level (suspicion of ischemia and/
positioning and to derive the missing leads or atypical arrhythmia), all information is sent
to generate an artificial standard 12-lead to and temporarily stored on an alarm server
ECG as accurate and as similar as possible that automatically sends an SMS to the attend-
to the original 12-lead ECG. ing health professional (cardiologist or general
• adapt the decision making process within the practitioner) stored in the patient’s contact list
system to the specificity of the user by using of the PEM Card (Figure 2). In case of a minor
advanced neural network-based decision- alarm (small ECG changes), the PEM displays a
making methods, taking into account the short message inviting the user to report about
serial ECG measurements and the patient the message at the occasion of one of his next
risk factors and clinical data. visits to his cardiologist or attending physician.

Subsequently, we present a prospect of the a Multiclassifier Multimodule

deployment scheme of these solutions in terms of decision Making
modules and their integration within the PEM tele-
medical platform and tele-expertise applications. The use of a neural network-based classifier
to analyze the recorded ECG and to trigger an
appropriate alarm according to risk thresholds
eMBedded IntellIgence In an (major, medium, minor, or no alarm thresholds)
e-cardIology perVasIVe allows an optimal discrimination between the
MultI-actor enVIronMent different alarm levels by simply moving upward
or downward these thresholds.
The EPI-MEDICS project developed an intelligent However, there is a major interest in reduc-
pervasive PEM for the early detection of patients ing the rate of false-positives while maintaining
and citizens at cardiac risk. Decision making the highest possible sensitivity and specificity in
embedded in the PEM is performed at 4 different order to reduce the rate of false alarms that could
levels: detection, diagnosis, generation of alarms, generate expensive and unnecessary costs. One
and intelligent management of the alarm messages possible way to enhance the diagnostic accuracy
of the PEM device consists of including a patient’s

304
Ambient Intelligence and Pervasive Architecture Designed within the EPI-MEDICS Personal ECG Monitor

Figure 2. The EPI-MEDICS medium alarm level tele-expertise architecture (from Rubel et al., 2005)

risk stratification, based on the patient medical (i.e., infarction) knowing the symptoms or signs
history stored in an electronic health record in the (cholesterol rate, age, and so forth). The choice of
PEM smartcard, in the decision-making process. using Bayesian networks to calculate a risk score
Consequently, merging the ECG interpretation upon the patient’s risk factors is justified by the
with the user (patient/citizen) cluster of risk fac- capacity of Bayesian networks both to model the
tors can be of a great benefit since it would help uncertainty inherent in medical reasoning and to
to better quantify the patient’s risk for develop- make decisions based on incomplete data (Lagor,
ing cardiac complications and thus to recognize Aronsky, Fiszman, & Haug, 2001).
whether the clinical situation is changing and We configured and assessed the accuracy of
turning into a cardiac disease that could quickly the Bayesian network over the INDANA database
turn a healthy citizen into a patient at high risk which is a very large collection of individual-
of major cardiac events or of even sudden death patient data with at least a six year follow-up for
because of life-threatening arrhythmias, cardiac cardiovascular events and deaths: myocardial
ischemia or myocardial infarction. infarction, stroke, or cardiovascular death (Gueyf-
fier et al., 1995). The Bayesian network uses the
risk Factors Quantification using patient data (age, gender, diabetes, and so forth) to
Bayesian networks predict the risk of cardiovascular event (Figure 3).
The assessment over the INDANA database
For risk factors quantification, to merge quantita- of the Bayesian network in comparison to logistic
tive and qualitative data into one decision process, regression and discriminant analysis has shown a
we have designed a solution based on Bayesian clear advantage of more than 5% in terms of Area
networks to calculate the probability of the event Under the Roc Curve (AUC) when compared with

305
Ambient Intelligence and Pervasive Architecture Designed within the EPI-MEDICS Personal ECG Monitor

these two conventional statistical methods (Atoui, • Once successfully recorded and stored in
Fayn, Gueyffier, & Rubel, 2006). the PEM smartcard, the ECG is interpreted
by the neural-network module.
collaboration scheme • If the neural-network module output passes
the high risk threshold, the ECG is consid-
The collaboration between the ECG interpreta- ered as abnormal, and the system will trigger
tion module and the risk factors module has to be a major alarm sequence.
smooth to ensure a more accurate outcome. For • If the output falls within the medium or
this reason, a fuzzy-logic-based layer to control minor risk interval, the ECG is considered
the dialogue between both modules can be helpful as suspect, and the ECG classifier output is
in a flexible decision-support system as the PEM. combined with a second output issued by the
The choice of fuzzy-logic is due to the fact that Bayesian module to produce a global risk
such a technology allows an efficient collaboration score that will in turn trigger the appropriate
between classifiers and thus enhances the global alarm level: major, medium, minor.
decision-making process. The overall decision
scenario operates as follows (Figure 4): The determination of the different risk thresh-
olds and of the fuzzy logic rules are for the mo-
• A patient, suffering from fatigue or any other ment being based on human expert considerations
common symptoms known to precede a pos- about the desired sensitivities and specificities
sible cardiac event, uses the PEM device to that can be deduced from the different Receiver
acquire an ECG. Operating Characteristic (ROC) curves. Indeed,

Figure 3. Schematic representation of the bayesian network used to estimate the risk of cardiovascular
outcome (from Atoui et al., 2006)

306
Ambient Intelligence and Pervasive Architecture Designed within the EPI-MEDICS Personal ECG Monitor

Figure 4. Schematic representation of the overall decision making platform that will be embedded in
the final PEM devices for detecting ischemia after adding the bayesian and the fuzzy logic modules in
the decision-making process

the development of an optimized fuzzy-logic developed during the EPI-MEDICS project, where
interaction layer and the validation of the global healthcare providers might request an expert
decision making process would require a very opinion about a patient’s health record in almost
large database that contains not only serial 12- real time, no matter the degree of intelligence
lead and PEM 3-leads ECGs but also exhaustive that has been embedded in the ECG recording
electronic health records of the patients of the devices. One typical example is a GP who wants
database. Such a PEM-specific database is cur- to have an expert advice from a cardiologist for the
rently under construction through clinical trials ECG of his patient. Another example is a patient
conducted both by health professionals and by the who is admitted in an emergency department
patients in emergency centers, coronary care units, of a general hospital. The interpretation of the
cardiology clinics, offices of cardiologists and ECG tracing may be not clear for the nonexpert
GPs, home care, and ambulatory care situations. medical staff. A solution is to send the ECG to a
cardiologist at the nearest cardiology hospital for
an expert opinion.
exaMples oF peM IntegratIon To allow these transactions between the
In teleMedIcal applIcatIons requester (GP, emergency doctor, paramedics,
etc.) and the recipient (emergency call center,
tele-expertise architecture competence center, expert), a client-server type
adapted to e-cardiology architecture has been designed (Figure 5) (Télis-
son, Fayn, & Rubel, 2004). This architecture is
In medical applications, there are numerous based on TCP/IP Internet technology and on
situations in addition to the selfcare scenarios the XML metalanguage for data representation,
storage, and communication. It is composed of

307
Ambient Intelligence and Pervasive Architecture Designed within the EPI-MEDICS Personal ECG Monitor

Figure 5. The tele-expertise architecture. A web server provides intelligent management of messages
exchanges and a set of distributed client applications for sending and receiving XML tele-expertise
requests via HTTPS (from Télisson et al., 2004)

a main application installed on a Web server forth). The telerequester may use a standard digital
that provides an intelligent management of the ECG recorder, a PEM, or any other personal ECG
messages exchanges and a set of distributed cli- monitor with Bluetooth and GPRS transmission
ent applications for sending and receiving the capabilities.
tele-expertise requests via HTTPS (HTTP with This infrastructure has been experimented
a Secure Socket Layer). The requests are repre- within the Lyon area in a setting that includes
sented in the XML meta-\language. A message an emergency department, regional and general
contains information about the communicating hospitals, the Cardiology Hospital of Lyon, pa-
parties, the Electronic Health Record (EHR), and tients homes, and the informatics department of
any attached file like SCP-ECGs. The architecture the Lyons hospitals which hosts the servers. To
is designed to support the pervasive paradigm. In facilitate the recording of digital ECGs and their
order to guarantee the reception of the message automatic transmission, in several situations
in due time, the expert should be able to receive we have used the EPI-MEDICS Personal ECG
anywhere different kinds of notifications by means Monitor (PEM). Preliminary evaluation results
of different types of devices that depend on the have shown the potential of such an architecture
available technologies (Pager, SMS, PDA, and so to improve patient care by allowing to complete

308
Ambient Intelligence and Pervasive Architecture Designed within the EPI-MEDICS Personal ECG Monitor

the global medical patient record and to guarantee eoconferencing system with the possibility to
the traceability of the exchanges. remotely control the position and the zooming of
the camera, and of several medical acquisition de-
shared care telemedical solution vices: tensiometer, a standard portable SCP-ECG
dedicated to elderly patient nursing compliant 12-lead interpreting electrocardiograph
services: the tele-nurse project or a PEM device from the EPI-MEDICS project,
oxymeter, glucometer, thermometer, digital
In modern societies, the number of elderly popu- stethoscope, and a weight scale. Most of these de-
lations is steadily growing and the mean age at vices automatically transmit their measurements
which the retirees are entering elderly nursing wireless via Bluetooth or an industry standard
homes is also steadily increasing. The paramedical radio link to the laptop PC that is embedded in
staff is thus increasingly confronted with medical the trolley which sends the collected data to the
requests that usually result, because of the ab- server.
sence of medical in-house support, in sending the Two platforms were deployed in two French
patient to the hospital for security reasons. It is elderly homes, one equipped with a PEM device
thus felt that the development of a medical trolley and the other with a portable 12-Lead ECG re-
equipped with several medical instruments and corder. The main evaluation result was a high level
video-telephony allowing remote decision making of acceptance of the Tele-Nurse system concept.
can reduce the number of hospitalizations.
The goal of the Tele-Nurse project (Télisson,
Fayn, Placide, Rubel, & Comet, 2006) is to set conclusIon
up a remote decision support system connecting
the elderly patients’ rooms or apartments in the In this article, we described some of the latest ambi-
nursing homes to a PC based, easy to use and to ent intelligence and pervasive solutions that have
deploy clinical workstation enabling the remote been designed and are being embedded in the PEM
GP to provide quicker and more effective actions device, and, more specifically, the ANN-based
or treatments in case of disease episodes leading ECG interpretation and the Bayesian-network
to emergency situations such as a heart attack. risk factors modules and their integration into
The medical trolley allows data transmission the overall PEM telemedical platform.
and information exchange, via an intelligent At this stage of development, and after con-
server, between the doctor and the nurse. The doc- ducting several clinical trials by both patients and
tor can thus establish a diagnosis within minutes health professionals, the PEM and the associated
after he has been called and remotely decide the software tools were judged extremely easy-to-use
actions to be taken, which can instantly be car- and user-friendly and have allowed the detection
ried out by the nurse. To meet these targets, the of several arrhythmia-based cases (Fayn, Restier,
system must allow real time information sharing Li, Chevalier, & Rubel, 2006; Rubel et al., 2005).
as well as bilateral medical data update, such as, The capability of the PEM to detect acute infarc-
for example, the recording and the upload of an tion in self-care situations remains, however, to
ECG by the nurse whilst the remote GP is keying be demonstrated because no such event occurred
in the clinical signs. during the clinical trials.
The medical trolley includes useful equipment Ongoing works are intended to create and de-
allowing the medical staff to collect information. ploy a palette of predictors for a variety of profiles
It consists of specific, high quality audio-visual representative of the PEM potential user (male ≥
equipment functioning like an enhanced vid- 45 years, female ≥ 55 years, and so forth). The next

309
Ambient Intelligence and Pervasive Architecture Designed within the EPI-MEDICS Personal ECG Monitor

Figure 6. The global architecture of the Tele-Nurse project. The medical trolley is equipped with a high
performance visioconferencing system (Visadom), a laptop and several medical devices (PEM, oxym-
eter, etc.). The remote PC based GP workstation is staffed with two screens, one for the display of the
Electronic Health Record, the other for the video conferencing (from Télisson et al., 2006)

step is to validate the global decision platform after Axisa, F., Schmitt, P., Gehin, C., Delhomme, G.,
deploying the newly developed decision support McAdams, E., & Dittmar, A. (2005). Flexible
tools (the Bayesian network and the fuzzy alarm technologies and smart clothing for citizen medi-
generator), and work on ways to make the system cine, home healthcare, and disease prevention.
self-improving on both the embedded intelligence IEEE Transactions on Information Technology
level (adapted to the patient medical history) and in Biomedicine, 9(3), 325-336.
the communication level, in order to reduce the
Campbell, P., Patterson, J., Cromer, D., Wall, K.,
number of false positives and/or false negatives.
Adams, G., Albano, A., et al. (2005). Prehospital
triage of acute myocardial infarction: Wireless
transmission of electrocardiograms to the on-call
reFerences
cardiologist via a handheld computer. Journal of
Electrocardiology, 38(4), 300-309.
Atoui, H., Fayn, J., Gueyffier, F., & Rubel, P.
(2006). Cardiovascular risk stratificationin deci- EN 1064:2005+A1:2007. (2007) Medical in-
sion support systems: a probabilistic approach. formatics - standard communication protocol
Application to p-health. In Proceedings of the - computer-assisted electrocardiography (SCP-
Computers in Cardiology 2006 Conference (Vol. ECG). Brussels. Retrieved May 26, 2008, from
33, pp. 281-284). http://www.cen.eu/
Atoui, H., Fayn, J., & Rubel, P. (2004). A neural Fayn, J., Restier, L., Li, B., Chevalier, P., &
network approach for patient-specific 12-lead ECG Rubel, P. (2006, September 2-6). Assessment of
synthesis in patient monitoring environments. In the benefit of pervasive self-care ECG recording
Proceedings of the IEEE Computers in Cardiology in arrhythmogenic patients. World Congress of
2004 Conference (Vol. 31, pp. 161-164). Cardiology 2006, Barcelona, Spain. European
Heart Journal, 27(Suppl. 1), 140.

310
Ambient Intelligence and Pervasive Architecture Designed within the EPI-MEDICS Personal ECG Monitor

Gueyffier, F., Boutitie, F., Boissel, J., Pocock, Rubel, P., Fayn, J., Simon-Chautemps, L., Atoui,
S., Coope, J., Cutler, J., et al. (1995). INDANA: H., Ohlsson, M., Télisson, D., et al. (2004). New
A meta-analysis on individual patient data in paradigms in telemedicine: Ambient intelligence,
hypertension. Protocol and preliminary results. wearable, pervasive and personalized. Studies in
Therapie, 50(4), 353-362. Health Technology and Informatics, 108, 123-132.
Kerkenbush, N.L., & Lasome, C.E. (2003). The Task Force Report. (1998). Recommendations of
emerging role of electronic diaries in the manage- a task force of the European society of cardiology
ment of diabetes mellitus. AACN Clinical Issues: and the European resuscitation council on the
Advanced Practice in Acute and Critical Care, prehospital management of acute heart attacks.
14, 371-378. European Heart Journal, 19, 1140-1164.
Lagor, C., Aronsky, D., Fiszman, M., & Haug, Télisson, D., Fayn, J., Placide, J., Rubel, P., &
P. (2001). Automatic identification of patients Comet, B. (2006). A new shared care telemedi-
eligible for a pneumonia guideline: Comparing cal solution dedicated to elderly patient nursing
the diagnostic accuracy of two decision support services for remote GP decision support. In Pro-
models. In Proceedings of the 10th World Con- ceedings of the Computers in Cardiology 2006
gress on Medical Informatics, MEDINFO (Vol. Conference (Vol. 33, pp. 229-232).
10, pp. 493-497).
Télisson, D., Fayn, J., & Rubel, P. (2004). De-
McMurray, J., & Rankin, R. (1994). Cardiology- sign of a tele-expertise architecture adapted to
I: Treatment of myocardial infarction, unstable pervasive multi-actor environments. Applica-
angina, and angina pectoris. British Medical tion to e-cardiology. In Proceedings of the IEEE
Journal, 309(6965), 1343-1350. Computers in Cardiology 2004 Conference (Vol.
31, pp. 749-752).
Rubel, P., Fayn, J., Nollo, G., Assanelli, D., Li, B.,
Restier, L., et al. (2005). Toward personal ehealth
in cardiology. Results from the EPI-MEDICS
telemedicine project. Journal of Electrocardiol-
ogy, 38(4), 100-106.

This work was previously published in International Journal of Healthcare Information Systems and Informatics, Vol. 3, Is-
sue 4, edited by J. Tan, pp. 53-68, copyright 2008 by IGI Publishing (an imprint of IGI Global).

311
312

Compilation of References

Abawajy, J. H. (2009). Human-computer interaction in Agrawal, R., Gunopulos, D., & Leymann, F. (1998).
ubiquitous computing environments. International Jour- Mining process models from workflow logs. In Sixth
nal of Pervasive Computing and Communications, 5(1), International Conference on Extended Database Tech-
61–77. doi:10.1108/17427370910950311 nology (pp. 469-483).

Abidi, S. S. R. (2001). Knowledge management in health- Agrawal, R., Imielinski, T., & Swami, A. (1993). Mining
care: Towards knowledgedriven decision-support services. association rules between sets of items in large databases.
International Journal of Medical Informatics, 63, 5–18. In ACM SIGMOD Conference on Management of Data,
Washington, DC.
ACGME Outcome Project. (2007). Accreditation Coun-
cil for Graduate Medical Education. Retrieved January Ajzen, I., & Fishbein, M. (1973). Attitudinal and norma-
29, 2008, from http://www.acgme.org/outcome/comp/ tive variables as predictors of specific behavior. Journal
compFull.asp. of Personality and Social Psychology, 27(1), 41–57.
doi:10.1037/h0034440
Acir, N., & Guzelis, C. (2004). Automatic spike detec-
tion in EEG by a two-stageprocedure based on support Alberta, G. P. C. (2003). Oipc stakeholder survey 2003—
vector machines. Computers in Biology and Medicine, highlights report. Office of the Information and Privacy
34(7), 561–575. Commissioner of Alberta. Tech. Rep.

Adams, S. (2008). Blog-based applications and health in- Al-Fedaghi, S. (2005a). How to calculate the information
formation: Two case studies that illustrate important ques- privacy. In Proceedings of the Third Annual Conference
tions for consumer health informatics (chi) research. Inter- on Privacy, Security and Trust: How to Calculate Infor-
national Journal of Medical Informatics, 1–8. Retrieved mation Privacy, St. Andrews, New Brunswick, Canada.
from http://dx.doi.org/10.1016/j.ijmedinf.2008.06.009. Retrieved January 29, 2008, from http://www.lib.unb.ca/
Texts/PST/2005/pdf/fedaghi.pdf
Aggarwal, V. (2002). The application of the unified mod-
eling language in object-oriented analysis of healthcare Al-Fedaghi, S. (2005b). Privacy as a base for confiden-
information systems. Journal of Medical Systems, 26, tiality. In Proceedings of the Fourth Workshop on the
383–397. Economics of Information Security, Harvard University,
Cambridge, MA. Retrieved January 29, 2008, from http://
Agrawal, R., Mannila, H., Srikant, R., Toivonen, H., &
infosecon.net/workshop/schedule.php
Inkeri, A. (1996). Fast discovery of association rules,
advances in knowledge discovery and data mining (pp. Al-Fedaghi, S. (2006, June 20-23). Aspects of personal
307–328). The Association for the Advancement of Ar- information theory. In Proceedings of the Seventh Annual
tificial Intelligence and the MIT Press. IEEE Information Assurance Workshop (IEEE-IAW):
Aspects of Personal Information Theory. West Point, NY:
United States Military Academy.

Copyright © 2011, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Compilation of References

Al-Fedaghi, S. (2007, July 12-14). When reinventing Anderson, J., & Aydin, C. (Eds.). (2005). Evaluating the
principles is necessary. In Proceedings of the Seventh organizational impact of healthcare information systems
International Computer Ethics Conference, University (2nd ed.). Springer.
of San Diego, CA.
Anderson, M., Banker, R., & Ravindran, S. (2003). The
Alkan, A., Koklukaya, E., & Subasi, A. (2005). Automatic new productivity paradox. Communications of the ACM,
seizure detection in EEG using logistic regression and ar- 46(3), 91–94. doi:10.1145/636772.636776
tificial neural network. Journal of Neuroscience Methods,
Angeles, R. (2007). An empirical study of the anticipated
148(2), 167–176. doi:10.1016/j.jneumeth.2005.04.009
consumer response to RFID product item tagging. Indus-
Allians för Sverige. (2006, August 23). Election pro- trial Management & Data Systems, 107(4), 461–583.
gram [Valmanifest]. Retrieved January 29, 2008, from
Angluin, D., & Smith, C. H. (1983). Inductive inference:
www.maktskifte06.se/fileadmin/Upload/pdf/Valmani-
Theory and methods. Computing Surveys, 15(3), 237–269.
fest_2006.pdf
Annane, D., Depondt, J., Aubert, P., Villart, M., Gehanno,
Alpsan, D., Towsey, M., Ozdamar, O., Tsoi, A. C., &
P., & Gajdos, P. (2004). Hyperbaric oxygen therapy for ra-
Ghista, D. N. (1995). Efficacy of modified backpro-
dionecrosis of the jaw: A randomized, placebo-controlled,
pogation and optimization methods on a real-world
double-blind trial from the orn96 study group. Journal
medical application. Neural Networks, 8(6), 945–962.
of Clinical Oncology, 22(24), 4893–4900. doi:10.1200/
doi:10.1016/0893-6080(95)00034-W
JCO.2004.09.006
Al-Qirim, N. (2007). Championing telemedicine adoption
Antman, E., Lau, J., Kupelnick, B., Mosteller, F., & Chalm-
and utilization healthcare organizations in New Zealand.
ers, T. (1992). A comparison of results of meta-analyses
International Journal of Medical Informatics, 76, 42–54.
of randomized control trials and recommendations of
doi:10.1016/j.ijmedinf.2006.02.001
clinical experts. Journal of the American Medical As-
Ambler, S. (2004). The object primer: Agile modeling- sociation, 268, 240–248.
driven development with UML 2.0. Cambridge: Cambridge
Antonadou, D., Pepelassi, M., Synodinou, M., Puglisi, M.,
University Press.
& Throuvalas, N. (2002). Prophylactic use of amifostine
Amenssisa, J., & Dabi, S. (2003). District-based tele- to prevent radiochemotherapy-induced mucositis and xe-
medicine project in Ethiopia. Addis Ababa, Ethiopia: rostomia in head-and-neck cancer. International Journal
Ministry of Health. of Radiation Oncology*Biology*Physics, 52(3), 739-747.

American Library Association. (1989). Presidential Com- Anttiroiko, A. (2004). Introduction to democratic e-gov-
mittee on Information Literacy - Final Report. Retrieved ernance. In Mälkiä, M., Anttiroiko, A.-V., & Savolainen,
from http://www.ala.org/ala/acrl/acrlpubs/whitepapers/ R. (Eds.), E-transformation in governance: New direc-
presidential.htm. tions in government and politics (pp. 22–49). Hershey,
PA: Idea Group.
Amouh, T., Gemo, M., & Macq, B. (2005). Versatile
clinical information system design for emergency depart- Apache Jakarta Project. (2007). Retrieved from http://
ments. IEEE Transactions on Information Technology in jakarta.apache.org
Biomedicine, 9(2), 174–183.
Ari, S., & Saha, G. (2008). Classification of heart sounds
Anderson, D. R., Whitmer, R. W., & Goetzel, R. Z. (2000). using empirical mode decomposition based features.
Health enhancement research organization (HERO) re- International Journal of Medical Engineering and Infor-
search committee. The relationship between modifiable matics, 1(1), 91–108. doi:10.1504/IJMEI.2008.019473
health risks and group-level health care expenditures.
American Journal of Health Promotion, 15, 45–52.

313
 Compilation of References

Aronsky, D., Jones, I., Lanaghan, K., et al. (2007). Sup- Bandyopadhyay, K., & Schkade, L. L. (2000). Disaster
porting patient care in the emergency department with a recovery planning by HMOs: theoretical insights. Health
computerized whiteboard system. Journal of the American Care Management Review, 25(2), 74–85.
Medical Informatics Association.
Barros, P. P., Pinto, C. G., & Machado, A. (1999). A
Aronson, A. (2001). Effective mapping of biomedical text signaling theory of excessive technological adoption.
to the UMLS Metathesaurus: The MetaMap program. Health Care Management Science, 2(2), 117–125.
Proceedings of the 2001 AMIA Symposium, pp. 17-21. doi:10.1023/A:1019075425763

Ash, A. S., Zhao, Y., Ellis, R. P., & Schlein, K. M. (2001). Bates, D. (2005). Physicians and ambulatory electronic
Finding future high-cost cases: Comparing prior cost ver- health records. Health Affairs, 24(5), 1180–1189.
sus diagnosis-based methods. Health Services Research,
Baxt, W. G. (1991). Use of an artificial neural network for
36, 194–206.
the diagnosis of myocardial infarction. Annals of Internal
Asuncion, A., & Newman, D. J. (2007). UCI machine Medicine, 115(11), 843–848.
learning repository. Irvine, CA: University of California,
Baxt, W. G., Shofer, F. S., Sites, F. D., & Hollander, J. E.
School of Information and Computer Science. Retrieved
(2002). A neural computational aid to the diagnosis of acute
May 26, 2008, from http://www.ics.uci.edu/~mlearn/
myocardial infarction. Annals of Emergency Medicine,
MLRepository.html
39(4), 366–373. doi:10.1067/mem.2002.122705
ATA. (2009). Americal Telemedicine Association. Re-
Bazeley, P., & Richards, L. (2000). The NVivo qualita-
trieved from http://www.atmeda.org
tive project book. London; Thousand Oaks, CA: Sage
Atoui, H., Fayn, J., & Rubel, P. (2004). A neural network Publications.
approach for patient-specific 12-lead ECG synthesis in
Becker, J., Kugeler, M., & Rosemann, M. (2003). Process
patient monitoring environments. In Proceedings of the
management: A guide for the design of business processes.
IEEE Computers in Cardiology 2004 Conference (Vol.
Berlin, Heidelberg, New York: Springer.
31, pp. 161-164).
Bennett, C. J. (2001, June 19). What government should
Atoui, H., Fayn, J., Gueyffier, F., & Rubel, P. (2006).
know about privacy: A foundation paper. In Paper pre-
Cardiovascular risk stratificationin decision support sys-
pared for the Information Technology Executive Lead-
tems: a probabilistic approach. Application to p-health.
ership Council’s Privacy Conference (Revised August
In Proceedings of the Computers in Cardiology 2006
1, 2001). Retrieve January 29, 2008, from http://www.
Conference (Vol. 33, pp. 281-284).
accessandprivacy.gov.on.ca/english/pub/wgskap.doc
Axisa, F., Schmitt, P., Gehin, C., Delhomme, G., McAd-
Berg, M. (2001). Implementing information systems in
ams, E., & Dittmar, A. (2005). Flexible technologies and
health care organizations: myths and challenges. Interna-
smart clothing for citizen medicine, home healthcare, and
tional Journal of Medical Informatics, 64(2/3), 143–156.
disease prevention. IEEE Transactions on Information
doi:10.1016/S1386-5056(01)00200-3
Technology in Biomedicine, 9(3), 325–336.
Bergan, T., & Hirsch, G. B. (1976). Analysis and Plan-
Baggot, R. (2005). A funny thing happened on the way
ning for improved distribution of Nursing personnel and
to the forum? Reforming patient and public involvement
services. A national model of supply, demand, and distri-
in the NHS in England. Public Administration, 83(3),
bution. Cambridge, MA: Pugh-Roberts Associates, Inc.
533–551.
Berghout, B. M., Eminovic, N., de Keizer, N. F., & Bir-
Baker, G. B. (2002). Integrating technology and disease
nie, E. (2007). Evaluation of general practitioner’s time
management—the challenges. Healthplan Magazine,
investment during a store-and-forward teledermatology
43, 60–66.
consultation. International Journal of Medical Informat-
ics, 76, 384–391. doi:10.1016/j.ijmedinf.2007.04.004

314
Compilation of References

Berta, W. Teare, G. F., Gilbart, E., Soberman G., L. Boendermaker, P. M., Schuling, J., Meyboom-de-Jong,
Lemieux-Charles, L., Davis, D., & Rappol, S. (2005). B., Zwierstra, R. P., & Metz, J. C. M. (2000). What are
The contingencies of organizational learning in long-term the characteristics of the competent general practitioner
care: factors that affect innovation adoption. Health Care trainer? Family Practice, 17, 547–553.
Management Review, 30(4), 282–292.
Bond, M., Baggoley, C., & Erwich-Nijhout, M. (1998).
Berwick, D. M. (2005). The John Eisenberg lecture: Urgency, disposition and age groups: A casemix model
Health services research as a citizen in improvement. for emergency departments. Emergency Medicine Aus-
Health Services Research, 40, 317–336. tralasia, 10(2), 103–110.

Berzal, F., Juan-Carlos, C., Cuenca, F., & Martín-Bautista, Borich, G. (1996). Effective teaching methods (3rd ed.).
M. J. (2008). On the quest for easy-to-understand splitting Englewood Cliffs, NJ: Merrill/Prentice Hall.
rules. Data & Knowledge Engineering, 44(1), 31–48.
Boshier, R. (1990). Socio-psychological factors in elec-
Bhaskaran, S., & Hardley, F. (2002). Buyer beliefs, at- tronic networking. International Journal of Lifelong
titudes and behaviour: foods with therapeutic claims. Education, 9(1), 49–64.
Journal of Consumer Marketing, 19(7), 591–606.
Bowen, J. L. (2006). Medical education: Educational
doi:10.1108/07363760210451410
strategies to promote clinical diagnostic reasoning. The
Bhuptani, M., & Moradpour, S. (2005). RFID Field Guide New England Journal of Medicine, 355, 2217–2225.
- Deploying Radio Frequency Identification Systems. NJ:
Brachman, R. J., Khabaza, T., Kloesgen, W., Piatetsky-
Prentice Hall.
Shapiro, G., & Simoudis, E. (1996). Mining business
Bhutan Department of Information Technology. (2007). databases. Communications of the ACM, 39(11), 42–48.
Bhutan’s Journey Towards Open Source. Paper presented
Brailsford, S., Churilov, L., & Liew, S. K. (2003). Treat-
at the DebConf7, Edinburgh, Scotland. https://penta.
ing ailing emergency departments with simulation: An
debconf.org/~joerg/events/25.en.html
integrated perspective. In International Conference on
Biermann, E., Dietrich, W., Rihl, J., & Standl, E. (2002). Health Sciences Simulation. Orlando, FL: The Society
Are there time and cost savings by using telemanage- for Modeling and Simulation International.
ment for patients on intensified insulin therapy: A
Brandeau, M. L., Sainfort, F., & Pierskalla, W. P. (2004).
randomised, controlled trial. Computer Methods and
Operations research and health care. Boston, MA: Kluwer.
Programs in Biomedicine, 69, 137–146. doi:10.1016/
S0169-2607(02)00037-8 Breiman, L., Friedman, J., Olshen, R., & Stone, C. (1984).
Classification and regression trees. Belmont, CA: Wad-
Bigus, J. P., & Bigus, J. (1998). Constructing Intelligent
sworth International Group.
software agents with Java – A Programmers Guide to
Smarter Applications, Wiley, ISBN: 0-471-19135-3. Brender, J. (2006). Handbook of evaluation methods for
health informatics. Elsevier.
Bishop, C. (1995). Neural networks for pattern recogni-
tion. Oxford University Press. Brent, E., Slusarz, P., & Thompson, A. (2002). Qualrus:
The intelligent qualitative analysis program (manual).
Bobrie, G., Postel-Vinay, N., Delonca, J., & Corvol, P.
Columbia, MO: IdeaWorks.
(2007). Self-Measurement and Self-Titration in Hyperten-
sion. American Journal of Hypertension, 20, 1314–1320. Briggs, R. O., Nunamaker, J. F. Jr, & Sprague, R. H. Jr.
doi:10.1016/j.amjhyper.2007.08.011 (2005). Special issue: crossing boundaries in information
systems research. Journal of Management Information
Boendermaker, P. M., Conradi, M. H., Schuling, J.,
Systems, 22(4), 7–11. doi:10.2753/MIS0742-1222220401
Meyboom-de-Jong, B., Swierstra, R. P., & Metz, J. C.
M. (2003)... Advances in Health Sciences Education :
Theory and Practice, 8, 111–116.

315
 Compilation of References

Brown, S., Lincoln, M., Groen, P., & Kolodner, R. (2003). Cain, M. (2006). A Two-step estimation applicaton in
VistA—U.S. Department of Veterans Affairs national- health care management evaluating the effects of choice
scale HIS. International Journal of Medical Informatics, of public health plan on health outcomes. Journal of
69(2-3), 135–156. Academy of Business and Economics, 6(2), 65–72.

Brynjolfsson, E. (1993). The productivity paradox of Cain, M. (2007). Self-selection and evaluation of health
information technology. Communications of the ACM, care programs. Journal of Academy of Business and
36(12), 67–77. doi:10.1145/163298.163309 Economics, 7(3), 82–89.

Brynjolfsson, E., & Hitt, L. (1996). Paradox lost? Firm Caine, R. N., & Caine, G. (1991). Making connections:
level evidence on the returns to information systems spend- Teaching and the human brain. Alexandria, VA: Asso-
ing. Management Science, 42(4), 541–558. doi:10.1287/ ciation for Supervision and Curriculum Development.
mnsc.42.4.541
Calhoun, J., Banaszak-Hol, J., & Hearld, L. (2006). Current
Buchanan, D. (1997). The limitations and opportuni- marketing practices in the nursing home sector. Journal
ties of business process re-engineering in a politicized of Healthcare Management, 51(3), 185–200.
organizational climate. Human Relations, 50, 51–72.
Cameron, J., Baraff, L., & Sekhon, R. (1990). Case-mix
doi:10.1177/001872679705000103
classification for emergency departments. Medical Care,
Buescher, P. A., & Horton, S. J. (2002). Prenatal WIC 28, 146–158.
participation in relation to low birth weight and Med-
Campbell, P., Patterson, J., Cromer, D., Wall, K., Adams,
icaid infant costs in North Carolina-An update. Center
G., & Albano, A. (2005). Prehospital triage of acute
for Health Informatics and Statistics Studies, 122, 1–9.
myocardial infarction: Wireless transmission of electro-
Buescher, P. A., Roth, M. S., & Williams, D. (1991). cardiograms to the on-call cardiologist via a handheld
An evaluation of the impact of maternity care coordi- computer. Journal of Electrocardiology, 38(4), 300–309.
nation on Medicaid birth outcomes in North Carolina.
Capek, P., Frank, S., Gerdt, S., & Shields, D. (2005). A
American Journal of Public Health, 81(12), 1625–1629.
history of IBM’s open source involvement and strategy.
doi:10.2105/AJPH.81.12.1625
IBM Systems Journal, 44(2), 249–248.
Burke, R., & Payton, P. E. (2006). Measuring and manag-
Carlson, C. (2003). Information overload, retrieval strate-
ing shoppability: Ten principles to convert shoppers into
gies and Internet user empowerment. In: L., Haddon (Ed.),
buyers. European Retail Digest, 50, 66–78.
The Good, the Bad and the Irrelevant (COST 269) 1(1)
Burkle, T., Ammenwerth, E., Prokosch, H., & Dudeck, J. (pp. 169-173). Helsinki, Finland.
(2001). Evaluation of clinical information systems. What
Carnall, D. (2000). Medical software’s free future. BMJ
can be evaluated and what cannot? Journal of Evaluation
(Clinical Research Ed.), 321(7267), 976.
in Clinical Practice, 7(4), 373–385.
Carr, N. (2003). IT doesn’t matter. Harvard Business
Business Process Management (BPM) Center. (2003).
Review, 81(5), 41–49.
Process mining. Retrieved May 24, 2008, from http://
tmitwww.tm.tue.nl/research/processmining/ Carr, N. (2004). Does IT matter?: Information technology
and the corrosion of competitive advantage. Boston, MA:
Bustard, D., Oakes, W. R., & He, Z. (1999). Models to
Harvard Business School Press.
promote effective system change. In IEEE International
Conference on Software Maintenance (ICSM ‘99). Catlett, J. (1991). Megainduction: Machine learning on
very large databases. Unpublished doctoral dissertation,
Butterfield, R. (2007). “Data classification: A prereq-
Basser Department of Computer Science, University of
uisite to ILM”, http://www.snwonline.com/implement/
Sydney.
data_class_05-30-05.asp

316
Compilation of References

CEBM. (2007). Why the sudden interest in EBM? Re- Chen, P., Huang, D., Zhao, X., & Li, X. (2008). Predicting
trieved from http://www.cebm.utoronto.ca/intro/interest. contact map using Radial Basis Function Neural Network
htm. and Conformational Energy Function. International Jour-
nal of Bioinformatics Research and Applications, 4(2),
Ceglowski, A., Churilov, L., & Wassertheil, J. (2007).
123–136. doi:10.1504/IJBRA.2008.018340
Combining data mining and discrete event simulation
for a value-added view of a hospital emergency depart- Chesney, T., Penny, K., Oakley, P., Davis, S., Maffulli, N.,
ment. The Journal of the Operational Research Society, & Templeton, J. (2006). Data Mining Medical Information:
58(2), 246–254. Should Artificial Neural Networks Be Used To Analyze
Trauma Audit Data. International Journal of Healthcare
Chae, Y. M., Lee, J. H., Ho, S. H., Kim, H. J., Jun, K. H.,
Information Systems and Informatics, 1(2), 51–64.
& Won, J. U. (2001). Patient satisfaction with telemedi-
cine in home health services for the elderly. International Cheung, D., Han, J., Ng, V. T., & Fu, A. W. (1996). Fast
Journal of Medical Informatics, 61, 167–173. doi:10.1016/ distributed algorithm for mining association rules. In
S1386-5056(01)00139-3 International Conference on Parallel and Distributed
Information Systems, Tokyo, Japan (pp. 31-42).
Chae, M. H., Black, C., & Heitmeyer, J. (2006). Pre-
purchase and post-purchase satisfaction and fashion Chin, J., Diehl, V., & Norman, K. (1988). Development of
involvement of female tennis wear consumers. Inter- an instrument measuring user satisfaction of the human-
national Journal of Consumer Studies, 30(1), 25–33. computer interface. Paper presented at the CHI, New Ork.
doi:10.1111/j.1470-6431.2005.00434.x
Chiu, R., & Kuo, C. (2006). A decision support model
Chan, P. K., & Stolfo, S. J. (1993). Meta-learning for of demand forecast for national immunization vaccines.
multistrategy and parallel learning. In Proceedings of International Journal of Electronic Healthcare, 2(1),
the Second International Workshop on Multistrategy 92–104. doi:10.1504/IJEH.2006.008699
Learning.
Christensen, C., Bohmer, R., & Kenagy, J. (2000). Will
Chang, I. C., Hwang, H. G., Hung, M. C., Lin, M. H., disruptive innovations cure health care. Harvard Business
& Yen, D. C. (2007). Factors affecting the adoption of Review, (September-October): 102–112.
electronic signature: Executives’ perspective of hospital
Clarke, J. (2006). Consumers, clients or citizens? Politics,
information department. Decision Support Systems, 44(1),
policy and practice in reform in social care. European
350–359. doi:10.1016/j.dss.2007.04.006
Societies, 8(3), 423–442.
Charmaz, K. (2006). Constructing grounded theory.
Clarke, L. L., Miller, M. K., & Mahan, C. S. (1993). The
London; Thousand Oaks, CA: Sage Publications.
Effectiveness of Florida’s Improved Pregnancy Outcome
Chau, P., & Tam, K. (1997). Factors affecting the adoption Program. Journal of Health Care for the Poor and Un-
of open systems: An exploratory study. MIS Quarterly: derserved, 4(2), 117–132.
Management Information Systems, 21(1), 1–20.
Cleary, P. D. (2003). A hospitalization from hell: A pa-
Chen, I. J., Yang, K. F., & Tang, F. I. (2008). Apply- tient’s perspective on quality. Annals of Internal Medicine,
ing the technology acceptance model to explore public 138(1), 33–39.
health nurses’ intentions towards web-based learning:
Cline, J. (2007). “Growing pressure for data classifica-
A cross-sectional questionnaire survey. International
tion”, http://www.computerworld.com/action/article.do
Journal of Nursing Studies, 45, 869–878. doi:10.1016/j.
?articleId=9014071&command=viewArticleBasic, Last
ijnurstu.2006.11.011
accessed on 22/11/2007

Cohen, C. I., & Cohen, E. J. (1978). Health education:

panacea, pernicious or pointless? The New England
Journal of Medicine, 28, 718–720.

317
 Compilation of References

Computing in the Physician’s Practice. (2000). Retrieved Creswell, J. (1994). Research design: Qualitative &
from http://www.harrisinteractive.com/harris_poll/index. quantitative approaches. Thousand Oaks, CA: Sage
asp?PID=58. Publications.

Cook, J. E., & Wolf, A. L. (1998). Discovering models Creswell, J. (2005). Educational research: Planning,
of software processes from event-based data. ACM conducting, and evaluating quantitative and qualitative
Transactions on Software Engineering and Methodology, research (2nd ed.). Upper Saddle River, NJ: Merrill.
7(3), 215–249.
Crosby, P. B. (1979). Quality is free. New York: McGraw-
Cook, J. E., & Wolf, A. L. (1998, November). Event- Hill.
based detection of concurrency. In Proceedings of the
Cruz, I. A., Gramaxo Silva, A., Soares, P., Oliveira, I.,
Sixth International Symposium on the Foundations of
Serrano, J. A. V., & Paulo Cunha, J. (2002). Modelling the
Software Engineering (FSE-6), Orlando, FL (pp. 35-45).
hospital into the future with UML. International Journal
Costa, A. X., Ridely, S. A., Shahani, A. K., Harper, P. of Healthcare Technology and Management, 4, 193–204.
R., de Senna, V., & Nielsen, M. S. (2003). Mathemati-
Dalmadge, C. L., & Cain, M. (2008). Utilizing Neural
cal modelling and simulation for planning critical care
Networks o Assess Community-Level Vulnerability to
capacities. Anaesthesia, 58, 320–327.
the Methamphetamine Epidemic. Issues in Information
Cote, M. J., & Tucker, S. L. (2001). Four methodologies Systems, 9(2), 258–266.
to improve health care demand forecasting. Healthcare
Darden, L., & Rada, R. (1988). Hypothesis formation
Financial Management, 5, 54–58.
via interrelations. In Prieditis, A. (Ed.), Analogica (pp.
Côté, L., & Leclère, H. (2000). How clinical teachers 109–128). London: Pitman.
perceive the doctor-patient relationship and themselves
Darmstadt, G., Bhutta, Z., Cousens, S., Adam, T., Walker,
as role models. Academic Medicine, 75, 1117–1124.
N., & Berni, L. d. (2005). Evidence-based, cost-effective
Coulter, A., Entwistle, V., & Gilbert, D. (1999). Sharing interventions: how many newborn babies can we save?
decisions with patients: Is the information good enough? Lancet, 365(9463), 977–988. doi:10.1016/S0140-
BMJ (Clinical Research Ed.), 318, 318–322. 6736(05)71088-6

Cousins, M. S., Shickle, L. M., & Bander, J. A. (2002). An Das, A., Ben-Menachem, T., Cooper, G. S., Chak, A.,
introduction to predictive modeling for disease manage- Sivak, M. V. Jr, Gonet, J. A., & Wong, R. C. (2003).
ment risk stratification. Disease Management, 5, 157–167. Predicition of outcome in acute lower-gastrointestinal
hemorrahage based on and artificial neural network:
Cox, S. S., & Swanson, M. S. (2002). Identification of
Internal and external validation of a predictive model.
teaching excellence in operating room and clinic settings.
Lancet, 362(9392), 1261–1266. doi:10.1016/S0140-
American Journal of Surgery, 183, 251–255.
6736(03)14568-0
Coyle, R. G. (1996). System dynamics modelling: A practi-
Davis, D., Thomson, M., Oxman, A., & Haynes, R. (1997).
cal approach. London: Chapman and Hall.
Changing physician performance: A systematic review of
Crawford, A. G., Fuhr, J. P., Clarke, J., & Hubbs, B. (2005). the effect of continuing medical education strategies. Jour-
Comparative Effectiveness of Total Population versus nal of the American Medical Association, 274, 700–705.
Disease-Specific Neural Network Models in Prediciting
Davis, F. D. (1989). Perceived usefulness, perceived ease
Medical Costs. Disease Management, 8(5), 277–287.
of use, and user acceptance of information technology.
doi:10.1089/dis.2005.8.277
Management Information Systems Quarterly, 13(3),
Crawford, W. (2006). Why aren’t ebooks more successful? 319–340. doi:10.2307/249008
Retrieved from http://www.econtentmag.com/Articles/
ArticleReader.aspx?ArticleID=18144.

318
Compilation of References

Dawes, M., & Sampson, U. (2003). Knowledge manage- Deodhar, J. (2002). Telemedicine by email--experience
ment in clinical practice: Asystematic review of informa- in neonatal care at a primary care facility in rural India.
tion seeking behavior in physicians. International Journal Journal of Telemedicine and Telecare, 8(Suppl 2), 20–21.
of Medical Informatics, 71, 9–15. doi:10.1258/135763302320301867

de Medeiros, A. K. A., van der Aalst, W. M. P., & Weijters, Department of Health. (2003). Choice of hospitals: Guid-
A. J. M. M. (2003). Workflow mining: Current status and ance for PCTs, NHS Trusts and SHAs on offering patient
future directions. In Meersman, R., Tari, Z., & Schmidt, choice of where they are treated. London: Department
D. C. (Eds.), On the move to meaningful Internet systems of Health.
2003: CoopIS, DOA, and ODBASE (pp. 389–406). Berlin,
Department of Health. (2008). Payment by results (PbR).
Heidelberg: Springer-Verlag.
Retrieved May 26, 2008, from www.dh.gov.uk/en/Poli-
Dedrick, J., & West, J. (2003). Why firms adopt open cyandguidance/Organisationpolicy/Financeandplanning/
source platforms: A grounded theory of innovation and NHSFinancialReforms/index.htm.
standards adoption. Paper presented at the Proceedings of
Despont-Gros, C., Mueller, H., & Lovis, C. (2005).
the Workshop on Standard Making: A Critical Research
Evaluating user interactions with clinical information
Frontier for Information Systems. Seattle, WA.
systems: A model based on human-computer interactions
Dedrick, J., & West, J. (2004). An exploratory study into models. Journal of Biomedical Informatics, 38, 244–255.
open source platform adoption. Paper presented at the
Devaraj, S., & Kohli, R. (2000). Information technology
Proceedings of the 37th Hawaii International Conference
payoff in the health-care industry: a longitudinal study.
on System Sciences.
Journal of Management Information Systems, 16(4),
Degoulet, P., Fieschi, M., Jaulent, M.-C., & Ménard, J. 41–67.
(2004). Patient empowerment, cybermedicine, and citizen
Directive, E. U. (1995, October 24). Directive 95/46/EC of
education. In Nelson, R., & Ball, M. (Eds.), Consumer
the European Parliament and of the council. Retrieved Jan-
informatics. Applications and strategies in cyber health
uary 29, 2008, from http://eur-lex.europa.eu/LexUriServ/
care (pp. 120–128). Berlin: Springer.
LexUriServ.do?uri=CELEX:31995L0046:EN:HTML
DeLano, W. (2005). The case for open source software in
Djorhan, V., & Churilov, L. (2003). Business interac-
drug discovery. Drug Discovery Today, 10(3).
tions in an acute emergency department in Australia: A
DeLone, W., & McLean, E. (2003). The DeLone and clinical process modeling perspective. In the 6th Pacific
McLean model of information systems success: A ten-year Asia Conference on Information Systems (PACIS 2002).
update. Journal of Management Information Systems, Tokyo, Japan: The Japan Society for Management Infor-
19(4), 9–30. mation (JASMIN).

Deming, W. E. (1991). Out of crisis. Cambridge, MA: Do, H. Rahin, E. (2002), “Coma: A System for Flexible
Cambribge University Press. Combination of Schema Matching Approaches”, Proceed-
ings of 28th Conference on Very Large Databases, Morgan
DeMonaco, H. (2009). Patient- and physician-oriented
Kaufmann, pp 610-621USA.
web sites and drug surveillance: Bisphosphonates and
severe bone, joint, and muscle pain. Archives of Internal Doan, A.-H. Lu, y. Lee, T. Han, J (2003), “Profile-Based
Medicine, 169(12), 1164–1166. doi:10.1001/archin- Object Matching for Information Integration”, IEEE Intel-
ternmed.2009.133 ligent Systems Magazine, pp 54-59. USA.

Deng, M., Cock, D. D., & Preneel, B. (2009). Towards Documentation, C. (2007). Retrieved from www.cisco.
a cross-context identity management framework in com
e-health. Online Information Review, 33(3), 422–442.
doi:10.1108/14684520910969880

319
 Compilation of References

Dreiseitl, S., Ohno-Machado, L., Kittler, H., Vinterbo, Erickson, B., Langer, S., & Nagy, P. (2005). The role of
S., Billhardt, H., & Binder, M. (2001). A comparison of open source software in innovation and standardization
machine learning methods for the diagnosis of pigmented in radiology. Journal of the American College of Radiol-
skin lesions. Journal of Biomedical Informatics, 34, 28–36. ogy, 2(11), 927–931.

Duff, F. L., Daniel, S., Kamendje, B., Le Beux, P., & Ethiopian Telecommunications Corporation. (2007).
Duvauferrier, R. (2005). Monitoring incident report in www.telecom.net.et
the healthcare process to improve quality in hospitals.
Eudaptics Software Gmbh. (1999). Viscovery SOMine.
International Journal of Medical Informatics, 74(2-4),
Wien: Eudaptics Software Gmbh.
111–117. doi:10.1016/j.ijmedinf.2004.06.007
Evans, J., Hwang, Y., & Nagarajan, N. (1997). Cost re-
Dumas, M., van der Aalst, W., & ter Hofstede, A. (2005).
duction and process reengineering in hospitals. Journal
Process-aware information systems: Bridging people
of Cost Management, 11(3), 20–27.
and software through process technology. Hoboken, NJ:
Wiley-Interscience. Evenden, D., Harper, P. R., Brailsford, S. C., & Harindra,
V. (2005). Improving the cost-effectiveness of chlamydia
Earl, M. J. (1994). The new and the old of business
screening with targeted screening strategies. The Journal
process redesign. The Journal of Strategic Information
of the Operational Research Society, 57(12), 1400–1412.
Systems, 3(1), 5–22.
Eyler, A., Mayer, J., Rafii, R., Housemann, R., Brownson,
Edelstein, H. (1998). Data mining: Myth and reality.
R., & King, A. (1999). Key informant surveys as a tool
GIGA Information Group Telepresentation.
to implement and evaluate physical activity interventions
Eftekhar, B., Mohammad, K., Ardebili, H. E., Ghodsi, in the community. Health Education Research, 14(2),
M., & Keabchi, E. (2005). Comparison of artificial neural 289–298.
network and logistic regression models for prediction of
Eysenbach, G. (2004). Improving the quality of Web
mortality in head trauma based on clinical data. BMC
surveys: The checklist for reporting results of Internet
Medical Informatics and Decision Making, 5(3), 1–8.
e-surveys (CHERRIES). Journal of Medical Internet
Elmasri, R. (2000). Fundamentals of database systems Research, 6(3), e34.
(3rd ed.). Addison Wesley.
Eysenbach, G. (2001). What is e-health? Journal of Medi-
Elwyn, G., & Edwards, A. (2001). Evidence-based patient cal Internet Research, 3(2), e20. doi:10.2196/jmir.3.2.e20
choice? In A. Edwards & G. Elwyn, G. (Eds.), Evidence-
Eysenbach, G., & Köhler, C. (2003). What is the preva-
based patient choice. Inevitable or impossible? (pp. 3-18).
lence of health-related searches on the World Wide Web?
Oxford University Press.
Qualitative and quantitative analysis of search engine
Elzubeir, M. A., & Rizk, D. E. E. (2001). Identifying queries on the Internet. In Proceedings of the AMIA
characteristics that students, interns and residents look Symposium (pp. 225-229).
for in their role models. Medical Education, 35, 272–277.
Fayn, J., Restier, L., Li, B., Chevalier, P., & Rubel, P. (2006,
EN 1064:2005+A1:2007. (2007) Medical informatics September 2-6). Assessment of the benefit of pervasive
- standard communication protocol - computer-assisted self-care ECG recording in arrhythmogenic patients.
electrocardiography (SCP-ECG). Brussels. Retrieved World Congress of Cardiology 2006, Barcelona, Spain.
May 26, 2008, from http://www.cen.eu/ European Heart Journal, 27(Suppl. 1), 140.

Epstein, R. M. (2007). Assessment in medical education. Fayyad, U. M., & Irani, K. (1993). Multi-interval discreti-
The New England Journal of Medicine, 356, 387–396. zation of continuous-valued attributes for classification
learning. In Proceedings of the International Joint Confer-
ences on Artificial Intelligence 1993, San Francisco, CA.

320
Compilation of References

Ferguson, T. (2002, September 2002). Key concepts in Forrester, J. W. (1961, July-August). Industrial Dynamics:
online health: E-patients as medical researchers. The A Major Breakthrough for Decision Makers. Harvard
Ferguson Report Retrieved August 2005, from http:// Business Review, ▪▪▪, 37–66.
www.fergusonreport.com/articles/fr00903.htm
Foughty, T. (1999). Accomplishing health improvement
Ferguson, T. (2009). E-patients: How they can help us through diabetes disease management. Drug Benefit
heal healthcare. Retrieved June 27, 2009, 2009, from Trends, 12, 43–45.
http://www.acor.org/epatientswiki/
FountainJ.E. (2001).
Ferreira, F.L., Peres, Bota, D., Bross, A., Melot, C., &
Frawley, Piatestsky-Shapiro, & Matheus (1991). Knowl-
Vincent, J.L. (2001). Serial evaluation of the SOFA score
edge discovery in databases: An overview. Menlo Park,
to predict outcome in critically ill patients. Journal of the
CA: AAAI Press.
American Medical Association, 286, 1754–1758.
Frazier, P. A., & Tix, A. P. (2004). Testing mod-
Fichman, R., & Kemerer, C. (1997). Object technol-
erator and mediator effects in counseling psychology.
ogy and reuse: Lessons from early adopters. Computer,
Journal of Counseling Psychology, 51(1), 115–134.
30(10), 47–59.
doi:10.1037/0022-0167.51.1.115
Fink, M. (2003). The business and economics of Linux and
Fridman, S. (2000). Doctors lag when it comes to
open source. Upper Saddle River, NJ: Prentice Hall PTR.
computer use - Industry trend or event. Retrieved from
Finkenzeller. K. (1999), RFID Handbook, John Wiley http://www.findarticles.com/p/articles/mi_m0HDN/
and Sons Ltd, USA. is_2000_March_28/ai_60907019.

Fischer-Hübner, S., & Ott, A. (1998, October 5-8). From Friedman, C., & Wyatt, J. (1997). Evaluation methods in
a formal privacy model to its implementation. In Proceed- medical informatics. New York: Springer.
ings of the 21st National Information Systems Security
Friedman, M., Landsberg, R., Pryor, S., Syed, Z., Ibra-
Conference, Arlington, VA. Retrieved January 29, 2008,
him, H., & Caldarelli, D. (2001). The occurrence of
from http://www.cs.kau.se/~simone/niss98.pdf
sleep-disordered breathing among patients with head
Fitzgerald, & Kenny. (2004). Developing an informa- and neck cancer. The Laryngoscope, 111, 1917–1919.
tion systems infrastructure with open source software. doi:10.1097/00005537-200111000-00008
Software, IEEE, 21(1), 50-55.
Friedman, C., Hripcsak, G., Johnson, S. B., Cimino, J. J.,
Fitzgerald. (2004). A critical look at open source. Com- & Clayton, P. D. (1990). Ageneralized relational schema
puter, 37(7), 92-94. for an integrated clinical patient database. In Proceedings
of the 14th Annual Symposium on Computer Applications
Flood, R. (1999). Rethinking the fifth discipline. Learning
in Medical Care (pp. 335-339).
within the unknowable. London: Ruthledge.
Fuhrer, P. and Guinard, D. (2007). Building a Smart
Forducey, P., Kaur, K., Scheiderman-Miller, C., & Tan,
Hospital using RFID technologies: Use Cases and Imple-
J. (2005). Surveying the e-health landscape: Cases and
mentation. rfidehealth.pdf
applications. In Tan, J. (Ed.), E-health care information
systems. An introduction for students and professionals Gadd, C., & Penrod, L. (2001). Assessing physician
(pp. 203–224). San Fransisco, CA: Jossey Bass. attitudes regarding use of an outpatient EMR: A longi-
tudinal, multi-practice study. Proceedings/AMIA Annual
Forrest, S., Hofmeyr, S. A., Somayaji, A., & Longstaff,
Symposium, 194-198.
T. A. (1996). A sense of self for unix processes. In 1996
IEEE Symposium on Security and Privacy (pp. 120-128).
IEEE Computer Society. Building the virtual state. Infor-
mation technology and institutional change. Washington,
DC: Brookings Institution Press.

321
 Compilation of References

Gagnona, M. P., Godinb, G., Gagnéb, C., Fortina, J. P., Glenton, C., Paulsen, E., & Oxman, A. (2005). Portals to
Lamothec, L., Reinharza, D., & Cloutierd, A. (2003). An wonderland: Health portals lead to confusing information
adaptation of the theory of interpersonal behaviour to about the effects of health care. Medical Informatics and
the study of telemedicine adoption by physicians. Inter- Decision Making, 5(7).
national Journal of Medical Informatics, 71, 103–115.
Glover, B., & Bhatt, H. (2006). RFID Essentials, O’Reilly
doi:10.1016/S1386-5056(03)00094-7
Media, Inc (pp. 10–0596009445). USA: ISBN.
Galpottage, A., & Norris, A. (2005). Patient consent
Glynn, E., Fitzgerald, B., & Exton, C. (2005). Commercial
principles and guidelines for e-consent: A New Zealand
adoption of open source software: An empirical study.
perspective. Health Informatics Journal, 11(1), 5-18.
Paper presented at the Proceedings of the 2005 Interna-
Retrieved January 29, 2008, from http://jhi.sagepub.com/
tional Symposium on Empirical Software Engineering
cgi/reprint/11/1/5
(ISESE’05). Noosa Heads, QLD, Australia.
Gamm, L., Barsukiewicz, C., Dansky, K., & Vasey, J.
Goetzel, R. Z., Anderson, D. R., & Whitmer, R. W.
Pre- and post-control model research on end-users’ sat-
(1998). The relationship between modifiable health risks
isfaction with an electronic medical record: Preliminary
and health care expenditures: An analysis of the multi-
results. Paper presented at the AMIA.
employer HERO health risk and cost database. Journal of
Ganguly, P., & Ray, P. (2000). A methodology for the Occupational and Environmental Medicine, 40, 843–854.
development of software agent based interoperable tele-
Goldberg, L. (1993). The structure of phenotypic per-
medicine systems: A tele-electrocardiography perspective.
sonality traits. The American Psychologist, 48, 26–34.
Telemedicine Journal, 6, 283–294.
Goldman, R., & Gabriel, R. (2005). Innovation happens
Gatignon, H., & Robertson, T. (1989). Technology dif-
elsewhere: Open source as business strategy. Amsterdam;
fusion: An empirical test of competitive effects. Journal
Boston, MA: Morgan Kaufmann.
of Marketing, 53(1), 35–49.
Goldratt, E. M. (1990). Theory of constrains. Croton-on-
Gehrke, J., Ramakrishnan, R., & Ganti, V. (2000). Rain-
Hudson. NY: North River Press.
Forest - a framework for fast decision tree construction
of large datasets. Data Mining and Knowledge Discovery, Goldsmith, D., & Safran, C. (2004). Collaborative
4(2-3), 127–162. healthware. In Nelson, R., & Ball, M. (Eds.), Consumer
informatics. Applications and strategies in cyber health
Gehshan, S., Belnap, D., & Snyder, A. (2009). Medicaid
care (pp. 9–19). Berlin: Springer.
Outreach and Enrollment for Pregnant Women: What Is
the State of the Art?Washington, DC: Urban Institute. Good, T., & Brophy, J. (2000). Looking in classrooms
(8th ed.). New York: Longman.
Giles, J. (2005). Internet encyclopedias go head to head. Re-
trieved from http://www.nature.com/news/2005/051212/ Goode, S. (2005). Something for nothing: Management
full/438900a.html. rejection of open source software in Australia’s top firms.
Information & Management, 42(5), 669–681.
Glaser, B. (1978). Theoretical sensitivity: Advances in
the methodology of grounded theory. Mill Valley, CA: Google. (2007). Search returned http://www.google.com/
Sociology Press. search?hl=en&q=GI+Motility.
Glaser, B. (1999). The future of grounded theory. Qualita- Goossen, W., Ozbolt, J., Coenen, A., Park, A., Mead, C.,
tive Health Research, 9(6), 836–845. & Ehnfors, M. (2004). Development of a provisional
domain model for the nursing process for use within the
Glaser, B., & Strauss, A. (1967). The discovery of grounded
health level 7 reference information model. Journal of
theory; strategies for qualitative research. Chicago, IL:
the American Medical Informatics Association: JAMIA,
Aldine Publishing Company.
11, 186–194.

322
Compilation of References

Gore, A. (1992). Infrastructure for the global village. Haile-Mariam, A. (2002). Renaissance: Strategies for
Scientific American, 265, 150–153. ICT Development in Ethiopia. M.Sc Thesis, School of
Engineering Postgraduate Engineering Program.
Gospodarevskaya, E., Churilov, L., & Wallace, L. (2005).
Modeling the patient care process of an acute care ward Hall, M. (2007). Core servlet and Java server pages.
in a public hospital: A methodological perspective. Sun Microsystems Press, Retrieved from http://www.
Hawaii: HICSS. coreservlets.com.

Goss, E., & Vozikis, G. S. (2002). Improving Health Hammer, M., & Champy, J. (1993). Reengineering the
Care Organizational Management Through Neural Net- corporation - A manifesto for business revolution. New
work Learning. Health Care Management Science, 5(3), York: Harper Business.
221–227. doi:10.1023/A:1019760901191
Hanss, R., Buttgereit, B., Tonner, P. H., Bein, B., Schlep-
Gottlieb, R., Rogers, J., & Rainey, K. (2002). Using cur- pers, A., & Steinfath, M. (2005). Overlapping induction
riculum evaluation as a basis for clinical faculty develop- of anesthesia: An analysis of benefits and costs. Anes-
ment. College Student Journal, 36(2), 279+. Retrieved thesiology, 103(2), 391–400. doi:10.1097/00000542-
January 29, 2008, from Questia database, http://www. 200508000-00023
questia.com/PM.qst?a=o&d=5000799148
Harel, D. (1987). Statecharts: A visual formalism for
Grana, J., Preston, S., McDermott, P. D., & Hanchak, N. complex systems. Science of Computer Programming,
A. (1997). The use of administrative data to risk-stratify 8, 231–274.
asthmatic patients. American Journal of Medical Qual-
Harno, K., Arajärvi, E., Paavola, T., Carlson, C., & Ar-
ity, 12, 113–119.
nala, I. (2001). Clinical effectiveness and cost analysis
Greener, I. (2003). Patient choice in the NHS: The view of patient referral by videoconferencing in orthopaedics.
from economic sociology. Social Theory & Health, 1, Journal of Telemedicine and Telecare, 7(4), 219–225.
72–89. doi:10.1258/1357633011936435

Gruber, T. R. (1993). A translation approach to portable on- Harno, K., Paavola, T., Carlson, C., & Viikinkoski,
tology specifications. Knowledge Acquisition, 5, 199–220. P. (2000). Patient referral by telemedicine: effective-
ness and cost analysis of an intranet system. Jour-
Gueyffier, F., Boutitie, F., Boissel, J., Pocock, S., Coope,
nal of Telemedicine and Telecare, 6(6), 320–329.
J., & Cutler, J. (1995). INDANA: A meta-analysis on
doi:10.1258/1357633001935996
individual patient data in hypertension. Protocol and
preliminary results. Therapie, 50(4), 353–362. Harper, P. R. (2002). A framework for operational mod-
elling of hospital resources. Health Care Management
Guido, F. (2007). The organizational learning curve.
Science, 5(3), 165–173.
European Journal of Operational Research, 177(3),
1375–1384. doi:10.1016/j.ejor.2005.04.009 Harper, P. R. (2005). A review and comparison of clas-
sification algorithms for medical decision making. Health
Gupta, K., & Stewart, D. W. (1996). Customer satisfaction
Policy (Amsterdam), 71, 315–331.
and customer behavior: the differential role of brand and
category expectations. Marketing Letters, 7(3), 249–263. Harper, P. R., & Pitt, M. (2004). On the challenges of
doi:10.1007/BF00435741 health care modelling and a proposed project life-cycle
for successful implementation. The Journal of the Op-
Guyatt, G., Cairns, J., Churchill, D., Cook, D., Haynes,
erational Research Society, 55, 657–661.
B., & Hirsch, J. (1992). Evidence-Based Medicine Work-
ing Group: Evidence-based medicine. A new approach to
teaching the practice of medicine. Journal of the American
Medical Association, 268, 2420–2425.

323
 Compilation of References

Harper, P. R., Sayyad, M. G., de Senna, V., Shahani, A. Hendricks, K. B., Singhal, V., & Stratman, J. K. (2007). The
K., Yajnik, C. S., & Shelgika, K. M. (2003). A systems impact of enterprise systems on corporate performance:
modelling approach for the prevention and treatment of a study of ERP, SCM, and CRM system implementa-
diabetic retinopathy. European Journal of Operational tions. Journal of Operations Management, 25(1), 65–82.
Research, 150, 81–91. doi:10.1016/j.jom.2006.02.002

Harris, K. (2003). How do patients choose physicians? Herbst, J., & Karagiannis, D. (2000). Integrating machine
Enrolees in employment-related health plans. Health learning and workflow management to support acquisi-
Services Research, 38(2), 711–732. tion and adaptation of workflow models. International
Journal of Intelligent Systems in Accounting Finance &
Harrison, S., Dowswell, G., & Milewa, T. (2002). Guest
Management, 9, 67–92.
editoral: Public and user “involvement” in the UK National
Health Service. Health & Social Care in the Community, Herbst, J. (2000). A machine learning approach to work-
10(2), 63–66. flow management. In 11th European Conference on
Machine Learning. Berlin, Germany: Springer (LNCS
Hawking, S. (2001). The universe in a nutshell. New
1810, pp. 183-194).
York, NY: Bantam Books.
Herbst, J., & Karagiannis, D. (1999, August), An inductive
Haynes, R. (1993). Where’s the meat in clinical journals?
approach to the acquisition and adaptation of workflow
[editorial]. ACP Journal Club, 119, A22–A23.
models. In M. Ibrahim & B. Drabble (Eds.), Proceedings
Hazen, G. (2000). Preference factoring for stochastic of the IJCAI’99 Workshop on Intelligent Workflow and
trees. Management Science, 46(3), 389–403. Process Management: The New Frontier for AI in Busi-
ness (pp. 52-57). Stockholm, Sweden.
Hazen, G. B. (1992). Stochastic trees: A new technique
for temporal medical decision modeling. Medical Deci- Herman, A. A., & Berendes, H. W. (1996). The evalua-
sion Making, 13, 227–236. tion of the effectiveness of a community-based enriched
model of prenatal intervention project in the District of
He, Q., & Antón, A. I. (2003, June 16-17). A framework Columbia. Health Services Research, 33(5), 609–621.
for modeling privacy requirements in role engineering.
In Proceedings of the International Workshop on Re- Hersh, W., Mailhot, M., Arnott-Smith, C., & Lowe, H.
quirements Engineering for Software Quality (REFSQ (2002). Selective automated indexing of findings and
2003), Klagenfurt / Velden, Austria. Retrieved January diagnoses in radiology reports. Journal of Biomedical
29, 2008, from http://crinfo.univ-paris1.fr/REFSQ/03/ Informatics, 34, 262–273.
papers/P14-He.pdf
Heskett, L. (1983). Shouldice Hospital Ltd, Rev. 1/88.
Heathfield, H., & Louw, G. (1999). New challenges Boston, MA: Harvard Business School.
for clinical informatics: Knowledge management tools.
Heymann, D. L., & Rodier, G. (2004). Global Surveil-
Health Informatics Journal, 5(2), 67–73.
lance, National Surveillance, and SARS. Emerging
Hederman, L., Smutek, D., Wade, V., & Knape, T. (2002). Infectious Diseases Journal, 10(2). Retrieved 1st De-
Representing clinical guidelines in UMl: A comparative cember 2006, from http://origin.cdc.gov/ncidod/EID/
study. Studies in Health Technology and Informatics, vol10no2/03-1038.htm
90, 471–477.
HHS. (2003). Summary of the HIPAA privacy rule. U.S.
Hedgepeth, W. O. (2007). RFID metrics: decision mak- Department of Health & Human Services. Retrieved
ing tools for today’s supply chains.Boca Raton, FL: CRC January 29, 2008, from http://www.hhs.gov/ocr/priva-
Press, (ISBN: 9780849379796) cysummary.pdf

Heller, C., & Arozullah, A. (2001). Implementing change:

It’s as hard as it looks. Disease Management & Health
Outcomes, 9, 551–563.

324
Compilation of References

Hier, D., Rothschild, A., LeMaistre, A., & Keeler, J. Hoffenberg, S., Hill, M. B., & Houry, D. (2001). Does
(2005). Differing faculty and housestaff acceptance of sharing process differences reduce patient length of stay
an electronic health record one-year after implementa- in the emergency department? Annals of Emergency
tion. International Journal of Medical Informatics, 74, Medicine, 38(5), 533–540.
657–662.
Hoffman, K. J. (1997). Demystifying mental health
Hillestad, R., Bigelow, J., Bower, A., Girosi, F., Meili, information needs through integrated definition (IDEF)
R., & Scoville, R. (2005). Can electronic medical record activity and data modeling. Journal of the American
systems transform healthcare? Potential health benefits, Medical Association, Symposium.l, 111-115.
savings and costs. Health Affairs, 24(5), 1103–1117.
Hogarth, M., & Turner, S. (2005). A study of clinical
HIMSS. (2006). HIMSS bylaws. Retrieved June 1, 2006, related open source software projects. Paper presented at
from http://www.himss.org/content/files/himss_bylaws. the American Medical Informatics Association Conference
pdf. (AMIA). Washington, D.C.

HIPAA. (1996, August 21). Health Insurance Portability Holck, J., Larsen, M., & Pedersen, M. (2005). Manage-
and Accountability Act of 1996, Public Law 104-191, rial and technical barriers to the adoption of open source
104th Congress. Retrieved January 29, 2008, from http:// software. Berlin, Heidelberg: Springer.
aspe.hhs.gov/admnsimp/pl104191.htm
Hoo, K. S., Wong, S. T. C., Laxer, K. D., Knowlton, R.
Hippel, E. v. (2001). Innovation by User Communities: C., & Wan, C. (2000). Design patterns in medical imag-
Learning from Open-Source Software. MIT Sloan Man- ing information systems. Proc. SPIE Medical Imaging,
agement Review, 42(4), 82. 3980, 256-263.

Hirsch, G. B., Forsyth, B. R., Bergan, T. A., & Goodman, Hornik, K., Stinchcombe, M., & White, H. (1989). Multi-
M. R. (1976). Projecting a teaching hospital’s future layer feedforward networks are universal approximations.
utilization: A dynamic simulation approach. Journal of Neural Networks, 2(5), 359–366. doi:10.1016/0893-
Medical Education, 51(11), 937–939. 6080(89)90020-8

Hirsch, G. B., & Henderson, J. D. (1977). A simulation Horsch, A., & Balbach, T. (1999). Telemedicine informa-
model for “what if?” questions in plan development. tion systems. IEEE Transactions on Information Technol-
American Journal of Health Planning, 2(1), 40–46. ogy in Biomedicine, 3, 166–175.

Hirsch, G. B., & Killingsworth, W. R. (1975). A new Howgill, M. (1998). Health care consumerism, the in-
framework for projecting dental manpower requirements. formation explosion, and branding: Why ’tis better to
Inquiry, 12(2), 126–142. be the cowboy than the cow. Managed Care Quarterly,
6(4), 33–43.
Hirsch, G. B., & Miller, S. (1974). Evaluating HMO poli-
cies with a computer simulation model. Medical Care, Hripcsak, G. (1993). Monitoring the monitor: Automated
12(8), 668–681. statistical tracking of a clinical event monitor. Comput-
ers and Biomedical Research, an International Journal,
Hirsch, G. B., & Bergan, T. A. (1973). Simulating ambu-
26(5), 449–466.
latory care systems: The relationship between structure
and behaviour. In Proceedings of the Summer Computer Hripcsak, G., Austin, J. H., Alderson, P. O., & Friedman,
Simulation Conference, Montreal, Quebec, Canada (pp. C. (2002). Use of natural language processing to translate
877-884). clinical information from a database of 889,921 chest
radiographic reports. Radiology, 224(1), 157–163.
Hirsch, G. B., Bergan, T. A., & Frohman, A. L. (1974).
Organizational change in the provision of human services.
In Proceedings of the Summer Computer Simulation
Conference, Houston, Texas.

325
 Compilation of References

Hu, P. J., Chau, P. Y. K., Sheng, O. R. L., & Tam, K. Y. Islam, MT, H. Y., Waxman R, Bhuiyan AB. (2006). Imple-
(1999). Examining the technology acceptance model using mentation of emergency obstetric care training in Ban-
physician acceptance of telemedicine technology. Journal gladesh: lessons learned. Reproductive Health Matters,
of Management Information Systems, 16(20), 91–112. 14(27), 61–72. doi:10.1016/S0968-8080(06)27229-X

Huff, S., & Munro, M. (1985). Information technology Jackson, J. R. (2000). The urgent call for knowledge man-
assessment and adoption: A field study. Management agement in medicine. Physician Executive, 26(1), 28–31.
Information Systems Quarterly, (December): 327–338.
Jakobovits, R. M., Rosse, C., & Brinkley, J. F. (2002).
Huffman, M. (2005). Implementing outcome-based WIRM An Open Source Toolkit for Building Biomedi-
homecare: A workbook of OBQI, care pathways and cal Web Applications. Journal of the American Medical
disease management. Jones and Bartlett. Informatics Association, 9(6), 557–570. doi:10.1197/
jamia.M1138
Hunter, D. J., & Fairfield, G. (1997). Managed care: Dis-
ease management. British Medical Journal, 315, 50–53. Janetzki, V., Allen, M., & Cimino, J. J. (2004). Using
natural language processing to link from medical text to
IDS-Scheer. (2004). ARIS process performance manager.
online information resources. In Proceedings of Medinfo
Retrieved May 24, 2008, from http://www.ids-scheer.com
(pp. 1665).
Iluyemi, A., Fitch, C. J., & Parry, D. T. (2007, April
Java Server Pages Documentation. (2007). Retrieved
18-20th 2007). Mobile and Wireless Technologies for
from http://java.sun.com/products/jsp.
e-Health Provision in Africa: Opportunities for GSM-
Based Patient Centric Services. Paper presented at the Jee, K. Y. (2005). Healthcare in ubiquitous era. Jinhan
Med-e-Tel Luxembourg International Telecommunica- M&B.
tions Union. (2006). Bhutan to be Testbed for ITU’s
Jelinek, G. A. (1995). Casemix classification of patients
E-Post Venture with Universal Postal Union. Retrieved
attending hospital emergency departments in Perth,
20th April 2006, from http://www.itu.int/newsarchive/
Western Australia: Development and evaluation of an
press_releases/2002/10.html
urgency-based casemix system. Perth, Australia: Univer-
IMDB. (2007). How/where do you get your information? sity of Western Australia.
How accurate/reliable is it? Retrieved from http://imdb.
Jokipii, P., Kalliovalkama, K., & Paavola, T. (2006).
com/help/show_leaf?infosource.
Tekonivelpotilaan hoitoprosessin uudistaminen. [Finn-
Institute of Medicine. (2001). Crossing the quality chasm: ish Medical Journal]. Suomen Lääkärilehti, 61(14),
A new health system for the 21st century. Washington, 1626–1627.
DC: National Academy Press, Committee on Quality of
Joos, D., Chen, Q., Jirjis, J., & Johnson, K. (2006). An
Health Care in America.
electronic medical record in primary care: Impact on
Internet, A. (2007a). Three-month traffic statistics for satisfaction, work efficiency and clinic processes. Paper
wikipedia.org. Retrieved from http://www.alexa.com/ presented at the AMIA, Washington, DC.
data/details/main?q=&url=wikipedia.org.
Josefsson, U. (2005). Coping with illness online: The
Internet, A. (2007b). Top Sites United States. Retrieved case of patients’ online communities. The Information
from http://www.alexa.com/site/ds/top_sites?cc=US&ts_ Society, 21(2), 143–153.
mode=country&lang=none.
Jun, G. C. T. (2007). Design for patient safety: A systematic
Irby, D. M. (1994). What clinical teachers in medicine evaluation of preocess modelling appraoches for health-
need to know. Academic Medicine, 69, 333–342. care system safety improvement, PhD Thesis. Cambridge:
University of Cambridge, Department of Engineering.
Irby, D. M., Ramsey, P. G., Gillmore, G. M., & Schaad,
D. (1991). Characteristics of effective clinical teachers of
ambulatory care medicine. Academic Medicine, 66, 54–55.

326
Compilation of References

Jun, G. C. T., Ward, J. R., & Clarkson, P. J. (2005). Map- Kasper, D., Fauci, A., Longo, D., Braunwald, E., Hauser,
ping the healthcare process in order to design for patient S., & Jameson, J. (2005). Harrison’s principles of internal
safety. In International Conference on Engineering Design medicine (16th ed.). New York, NY: McGraw-Hill.
(ICED 05) (pp. 353-364).
Kaufman, D. R., Patel, V. L., Hilliman, C., Morin, P. C.,
Junginger, M., de Visser, E., Hjort-Gregersen, K., Koorn- Pevzner, J., & Weinstock, R. S. (2003). Usability in the
neef, J., Raven, R., Faaij, A., & Turkenburg, W. (2006). real world: assessing medical information technologies in
Technological learning in bioenergy systems. Energy Pol- patients’ homes. Journal of Biomedical Informatics, 36,
icy, 34(18), 4024–4041. doi:10.1016/j.enpol.2005.09.012 45–60. doi:10.1016/S1532-0464(03)00056-X

Kahen, G., & Sayers, B. M. (1997). Health-care technology Kaur, H., & Wasan, S. K. (2006). Empirical study on
transfer: Expert and information systems for developing Applications in Data Mining Techniques in Healthcare.
countries. Methods of Information in Medicine, 36, 69–78. Journal of Computer Science, 2(2), 194–200. doi:10.3844/
jcssp.2006.194.200
Kalli, S., Ambroso, C., Gregory, R., Heikela, A., Ilomaki,
A., & Leaning, M. (1992). Inform: Conceptual model- Kayango, M., Esquiche, E., Luna, M. R., Frias, G., Vega-
ling of intensive care information systems. International Centeno, L., & Bailey, P. (2006). Strengthening emergency
Journal of Clinical Monitoring and Computing, 9, 85–94. obstetric care in Ayacucho, Peru. International Journal
of Gynaecology and Obstetrics: the Official Organ of the
Kandaswamy, A., Kumar, C. S., Ramanathan, R. P., Jayara-
International Federation of Gynaecology and Obstetrics,
man, R., & Malmurugan, N. (2004). Neural classification
92, 299–307. doi:10.1016/j.ijgo.2005.12.005
of lung sounds using wavelet coefficients. Computers in
Biology and Medicine, 34, 523–537. Kennedy, R., Lee, Y., & Van Roy, B. (1998). Solving
data mining problems through pattern recognition.
Kang, S., & Lee, S. (2007). Economic effect and growth
Prentice Hall.
strategy of U-Health. SERI (Samsung Economic research
Institute). Issue Paper. Ker, J. S. (2003). Attributes of trainers for postgraduate
training in general surgery–a national consensus. The
Kantor, G., Wilson, W., & Midgley, A. (2003). Open
Surgeon, 1, 215–220.
source software and the primary care EMR. Journal of the
American Medical Informatics Association, 10(6), 616. Kerkenbush, N. L., & Lasome, C. E. (2003). The emerging
role of electronic diaries in the management of diabetes
Kantor, G. S., Wilson, W. D., & Midgley, A. (2003).
mellitus. AACN Clinical Issues: Advanced Practice in
Open-source software and the primary care EMR. Journal
Acute and Critical Care, 14, 371–378.
of the American Medical Informatics Association, 10(6),
616. doi:10.1197/jamia.M1403 Kervenoael, R. D., Soopramanien, D., Elms, J., &
Hallsworth, A. (2006). Exploring value through inte-
Karahanna, E., Straub, D. W., & Chervany, N. L. (1999).
grated service solutions: the case of e-grocery shop-
Information technology adoption across time: a cross-
ping. Managing Service Quality, 16(2), 185–202.
sectional comparison of pre-adoption and post-adoption
doi:10.1108/09604520610650646
beliefs. Management Information Systems Quarterly, 23,
183–213. doi:10.2307/249751 Keselman, A., Browne, A., & Kaufman, D. (2008). Con-
sumer health information seeking as hypothesis testing.
Karlsson, L. (2003). Conflict or harmony? Individual
Journal of the American Medical Informatics Association,
rights in Swedish and British health care [Konflikt eller
15(4), 484–495. doi:10.1197/jamia.M2449
harmoni? Individuella rättigheter och ansvarsutkrävande
i svensk och brittisk sjukvård]. Doctoral dissertation, Kiernan, M., Kraemer, H. C., Winkleby, M. A., King, A.
Göteborg: CEFOS, Göteborgs universitet. C., & Taylor, C. B. (2001). Do logistic regression and
signal detection identify different subgroups at risk?
Implications for the design of tailored interventions.
Psychological Methods, 6, 35–48.

327
 Compilation of References

Kirigia, J., Seddoh, A., Gatwiri, D., Muthuri, L., & Sed- Kudyba, S., & Hoptroff, R. (2001). Data mining and
doh, J. (2005). E-health: Determinants, opportunities, business intelligence: A guide to productivity. Hershey,
challenges and the way forward for countries in the WHO PA: Idea Group.
African region. BMC Public Health, 5, 137.
Kudyba, S., Hamar, G. H., & Gandy, W. M. (2006). Utilis-
Kirigia, J., Seddoh, A., Gatwiri, D., Muthuri, L., & Sed- ing neural network applications to enhance efficiency in
doh, J. (2005). E-health: Determinants, opportunities, the healthcare industry: prediciting populations of future
challenges and the way forward for countries in the chronic illness. International Journal of Business Intel-
WHO African Region. BMC Public Health, 5(1), 137. ligence and Data Mining, 1(4), 371–383. doi:10.1504/
doi:10.1186/1471-2458-5-137 IJBIDM.2006.010780

Kohn, L., Corrigan, J., & Donaldson, M. (Eds.). (2000). To Kumar, A. (2005). Ontology-driven access to biomedical
err is human: Building a safer health system. Washington: information (ODABI). Undergraduate Thesis, University
National Academy Press. of Arizona.

Kohonen, T. (1995). Self-organizing maps. Berlin, Ger- Kumarapeli, P., De Lusignan, S., Ellis, T., & Jones, B.
many: Springer. (2007). Using Unified Modelling Language (UML) as a
process-modelling technique for clinical-research process
Korean National Insurance Corporation (2007), Daily
improvement. Medical Informatics and the Internet in
Report, Jan/29.
Medicine, 32, 51–64.
Korenbrot, C. C., Gill, A., & Patterson, E. (1995). Evalua-
Kuperman, G. J., Gardner, R. M., & Pryor, T. A. (1991).
tion of California’s statewide implementation of enhanced
The HELP system. New York: Springer-Verlag.
perinatal services as Medicaid benefits. Public Health
Reports (Washington, D.C.), 110(2), 125–133. Kvale, S. (1996). Interviews: An introduction to qualita-
tive research interviewing. Thousand Oaks, CA: Sage.
Kotonya, G., & Sommerville, I. (1998). Requirements
engineering: Processes and techniques. New York: John Kwak, N. K., McCarthy, K. J., & Parker, G. E. (1997).
Wiley. A human resource planning model for hospital/medi-
cal technologists: An analytic hierarchy process ap-
Kovalerchuk, B., Vityaev, E., & Ruiz, J. F. (2001). Con-
proach. Journal of Medical Systems, 21(3), 173–187.
sistent and complete data and expert mining in medicine.
doi:10.1023/A:1022812322966
In Cios, K. J. (Ed.), Medical data mining and knowledge
discovery. New York: Physica-Verlag. Lagor, C., Aronsky, D., Fiszman, M., & Haug, P. (2001).
Automatic identification of patients eligible for a pneumo-
Kowalke, M. (2006) TMCnet. Wireless Mobility Blog.
nia guideline: Comparing the diagnostic accuracy of two
(2006). RFID vs. WiFi for Hospital Inventory Tracking
decision support models. In Proceedings of the 10th World
Systems. http://blog.tmcnet.com/wireless-mobility/rfid-
Congress on Medical Informatics. Medinfo, 10, 493–497.
vs-wifi-for-hospital-inventory-tracking-systems.asp,
October 23, 2006. Lahiri, K., & Xing, G. (2004). An econometric analysis
of veterans’ healthcare utilization using two-part mod-
Kozinets, R. (2002). The field behind the screen: Using
els. Empirical Economics, 29, 431–449. doi:10.1007/
netnography for marketing research in online commu-
s00181-003-0178-0
nications. JMR, Journal of Marketing Research, 39(1),
61–72. doi:10.1509/jmkr.39.1.61.18935 Lahiri, S. (2005), RFID Sourcebook, IBM Press, USA,
(ISBN: 10-0131851373) Last accessed on 22/11/2007
Kudyba, S., Hamer, B., & Gandy, W. (2005, December).
Enhancing efficiency in the health care industry. Com- Landro, L. (2002, January 4). Technology programs make
munications of the ACM, 28(12), 107–110. disease management easier. The Wall Street Journal,
Health Journal.

328
Compilation of References

Landstingsförbundet. (2000a). Recommendation about Leutzinger, J. A., Ozminkowski, R. J., & Dunn, R. L.
choice in health care. [Rekommendation om valmöj- (2000). Projecting future medical care costs using four
ligheter inom hälso- och sjukvården]. Dnr Lf924/00. scenarios of lifestyle risk rates. American Journal of
Stockholm: Landstingsförbundet. Health Promotion, 15, 35–44.

Landstingsförbundet. (2000b). Waiting times in care on Levin, G., Hirsch, G. B., & Roberts, E. (1972). Narcot-
the Internet [Väntetider i vården på Internet. Pressmed- ics and the community: A system simulation. American
delande] 2000-04-10. Stockholm: Landstingsförbundet. Journal of Public Health, 62, 8–6.

Langer, A. M. (2008). Analysis and design of informa- Levin, G., & Roberts, E. B. (1976). The dynamics of
tion systems. London: Springer (pp. 418; p. 249 illus.). human service delivery. Cambridge, MA: Ballinger
Publishing Company.
Larson, P. A., & Janower, M. L. (2005). The Nighthawk:
Bird of Paradise or Albatross? Journal of the American Lewan, T. (2007). ZDNet Australia. Human RFID chip-
College of Radiology, 2(12), 967–970. doi:10.1016/j. ping invading the US?. 23 July 2007.
jacr.2005.08.002
Li, T., Wu, G., & Gruidl, Z. (2004). Data mining techniques
Law, B. (1999). Career-learning space: newDOTS think- for cancer detection using serum proteomic profiling.
ing for careers education. British Journal of Guidance & Artificial Intelligence in Medicine, 32(2), 71–83.
Counselling, 27(1), 35–55.
Likourezos, A., Chalfin, D., Murphy, D., Sommer, B.,
Le Grand, J., & Bartlett, W. (1993). Introduction. In Le Darcy, K., & Davidson, S. (2004). Physician and nurse
Grand, J., & Bartlett, W. (Eds.), Quasi-markets and social satisfaction with an electronic medical record system.
policy (pp. 1–12). London: Macmillan. The Journal of Emergency Medicine, 27(4), 419–424.

LeBozec, C., Jaulent, M. C., Zapletal, E., & Degoulet, Lim, H., Widdows, R., & Jungkun, J. (2006). M-
P. (1998). Unified modeling language and design of a loyalty: winning strategies for mobile carriers.
case-based retrieval system in medical imaging. In AMIA Journal of Consumer Marketing, 23(4/5), 208–218.
Annual Symposium Proceedings (pp. 887-891). doi:10.1108/07363760610674338

LeCompte, M. D., & Preissle, J. (1993). Ethnography and Lindberg, D. (2003). NIH: Moving research from the
qualitative design in educational research (2nd ed.). San bench to the bedside. Presentation to the Subcommittee
Diego, CA: Academic Press. on Health.

Lee, B., & Menon, N. (2000). Information technology Linder, J., Schnipper, J., Tsurikova, R., Melnikas, A.,
value through different normative lenses. Journal of Volk, L., & Middleton, B. (2006). Barriers to electronic
Management Information Systems, 16(4), 99–119. health record use during patient visits. Paper presented
at the AMIA, Washington, DC.
Lee, N. J., Bakken, S., & John, R. (2006). Functional
requirements specification and data modeling for a Lisboa, P. J., & Taktak, A. F. G. (2006). The use of
PDA-based decision support system for the screening artificial neural networks in decision support in cancer:
and management of obesity. In AMIA Annual Symposium A systematic review. Neural Networks, 19, 408–415.
Proceedings (pp. 1002). doi:10.1016/j.neunet.2005.10.007

Leider, H. L., & Krizan, K. (2001). Disease management— Littlejohns, P., Wyatt, J., & Garvican, L. (2003). Evaluating
a great concept, but can you implement it? Disease computerised health information systems: hard lessons
Management, 4, 111–119. still to be learnt. British Medical Journal, 326, 860–863.
doi:10.1136/bmj.326.7394.860
Lempp, H., & Seale, C. (2004). The hidden curriculum
in undergraduate medical education: Qualitative study of
medical students’ perceptions of teaching. British Medical
Journal, 320, 770–773.

329
 Compilation of References

Littlejohns, P., Wyatt, J. C., & Garvican, L. (2003). Martin, J. L., Norris, B. J., Murphy, E., & Crowe, J. A.
Evaluating computerised health information systems: (2008). Medical device development: The challenge
hard lessons still to be learnt. British Medical Journal, for ergonomics. Applied Ergonomics, 39, 271–283.
326(7394), 860–863. doi:10.1136/bmj.326.7394.860 doi:10.1016/j.apergo.2007.10.002

Longman, P. (2007). Best care anywhere: How America’s Maruster, L., van der Aalst, W. M. P., Weijters, T., van
most socialized healthcare system became its finest: What den Bosch, A., & Daelemans, W. (2001, October 25-26).
the VA miracle means for you and your country. Sausalito, Automated discovery of workflow models from hospital
CA: PoliPointPress. data. In B. Kröse et al. (Eds.), Proceedings 13th Bel-
gium- Netherlands Conference on Artificial Intelligence
Lunn, K., Sixsmith, A., Lindsay, A., & Vaarama, M.
(BNAIC’01), Amsterdam, The Netherlands (pp. 183-190).
(2003). Traceability in requirements through process
modelling, applied to social care applications. Information Mascarenhas, L., Eliot, B. W., & MacKenzie, I. Z.
and Software Technology, 45, 1045–1052. (1992). A comparison of perinatal outcome, antenatal
and intrapartum care between England and Wales, and
Lyalin, D., & Williams, W. (2005). Modeling cancer
France. British Journal of Obstetrics and Gynaecology,
registration processes with an enhanced activity diagram.
99(12), 955–958.
Methods of Information in Medicine, 44, 11–13.
McDonald, C., Schadow, G., Barnes, M., Dexter, P., Over-
Lyman, P., & Varian, H. (2003). How much information?
hage, J., & Mamlin, B. (2003). Open source software in
Retrieved from http://www2.sims.berkeley.edu/research/
medical informatics—why, how and what. International
projects/how-much-info-2003/index.htm.
Journal of Medical Informatics, 69(2-3), 175–184.
Maass, W. (1994). Efficient agnostic pac-learning with
McDonald, C. J., Schadow, G., Barnes, M., Dexter, P.,
simple hypothesis.
Overhage, J. M., & Mamlin, B. (2003). Open Source
Madden, M. (2003). America’s online pursuits: the chang- software in medical informatics--why, how and what.
ing picture of who’s online and what they do. Washington International Journal of Medical Informatics, 69(2-3),
DC: Pew Internet and American Life Project. Retrieved 175–184. doi:10.1016/S1386-5056(02)00104-1
from www.pewinternet.org/reports/toc.asp?Report=106
McGrath, J. E. (1982). Judgment Calls in Research.
Magrabi, F., Lovell, N. H., & Celle, B. G. (1999). A web- Beverly Hills, CA: Sage Publications, Inc.
based approach for electrocardiogram monitoring in the
McGraw, G. (2006). Software Security. USA: Addison-
home. International Journal of Medical Informatics, 54,
Wesley.
1145–1153. doi:10.1016/S1386-5056(98)00177-4
McKeown, T. (1971). A Historical Appraisal of the Medical
Mahapatra, S., Koelling, C. P., Patvivatsiri, L., et al. (2003).
Task. In McLachlan, G., & McKeown, T. (Eds.), Medical
Pairing emergency severity index 5-level triage data with
History and Medical Care (pp. 29–50). London: Oxford
computer aided system design to improve emergency
University Press.
department access and throughput. In The 2003 Winter
Simulation Conference. New Orleans, LA: INFORMS. McLaughlin, C., Yang, S., & Van Dierdonck, R. (1995).
Professional service organizations and focus. Management
Maij, E., Toussaint, P. J., Kalshoven, M., Poerschke,
Science, 14(7), 1185–1193.
M., & Zwetsloot-Schonk, J. H. (2002). Use cases and
DEMO: Aligning functional features of ICT-infrastructure McMurray, J., & Rankin, R. (1994). Cardiology-I: Treat-
to business processes. International Journal of Medical ment of myocardial infarction, unstable angina, and angina
Informatics, 65, 179–191. pectoris. British Medical Journal, 309(6965), 1343–1350.

Mann, C., & Steward, F. (2000). Internet communication Mechaca, L. (2007). Goodbye, Axim. Message
and qualitative research: A handbook for researching posted to http://direct2dell.com/one2one/ar-
online. Thousand Oaks, CA: Sage. chive/2007/04/11/11397.aspx.

330
Compilation of References

Medsker, L., & Liebowitz, J. (1994). Design and Devel- Montani, S., & Bellazzi, R. (2002). Supporting decisions
opment of Expert Systems and Neural Networks. New in medical applications: The knowledge management
York: McMillan. perspective. International Journal of Medical Informat-
ics, 68, 79–90.
Mehta, N. Y., Haluck, R. S., Frecker, M. I., & Snyder, A.
J. (2002). Sequence and task analysis of instrument use in Morahan-Martin, J. M. (2004). How internet users find,
common laparoscopic procedures. Surgical Endoscopy, evaluate, and use online health information: a cross-cul-
16, 280–285. tural review. Cyberpsychology & Behavior, 7(5), 497–510.

Mendenhall, W. M. (2004). Mandibular osteoradionecro- Morgan, N. A., & Rego, L. L. (2006). The value of differ-
sis. Journal of Clinical Oncology, 22(24), 4867–4868. ent customer satisfaction and loyalty metrics in predicting
doi:10.1200/JCO.2004.09.959 business performance. Marketing Science, 25(5), 426–439.
doi:10.1287/mksc.1050.0180
Menges, R., & Austin, A. (2001). Teaching in higher edu-
cation. In Richardson, V. (Ed.), Handbook of research on Morrison, E. H., Hitchcock, M. A., Harthill, M., Boker, J.
teaching (4th ed.). Washington, DC: American Educational R., & Masunaga, H. (2005). The online clinical teaching
Research Association. perception inventory: A “snapshot” of medical teachers.
Family Medicine, 37, 48–53.
Mertins, K., Bernus, P., & Schmidt, G. (1998). Handbook
on architectures of information systems. Berlin, Germany: Muney, A. M. (2002). Evidence-based medicine needs to
Springer. be promoted more vigorously. Managed Care (Langhorne,
Pa.), 11, 10–12.
Messmer, E. (2006). Network World. Hospital finds a
bloody good RFID application. http://www.computer- Mureo, M., & Rice, R. (2006). The Internet and health care.
world.com.au/index.php/id;2117167769;fp;4;fpid;18. Theory, research, and practice. NJ: Lawrence Erlbaum.

Metropolitan Health and Aged Care Services Division Murray, J. P. (1999). Faculty development in a na-
of the Victorian State Government. (2007). Victorian tional sample of community colleges. Community
emergency minimum dataset (VEMD). Retrieved May College Review, 27(3), 47. Retrieved January 29, 2008,
24, 2008, from http://www.health.vic.gov.au/hdss/ from Questia database: http://www.questia.com/
vemd/2004-05/manual/index.htm PM.qst?a=o&d=5001886124

Mickey, K. (2004). RFID grew in 2002. Traffic World, MySQL. (2007). http://www.mysql.com.
5, 20–21.
MySQL Connector/J. (2007). Retrieved from http://www.
Miller, R., & Sim, I. (2004). Physicians’ use of electronic mysql.com/products/connector-j/.
medical records: Barriers and solutions. Health Affairs,
Nagy, P. (2007). Open source in imaging informatics.
23(2), 116–126.
Journal of Digital Imaging, 20(0), 1–10.
Misson, J. C. (2001). A review of clinical risk manage-
Najumnissa, D. & ShenbagaDevi, S. (2008). Intelligent
ment. Journal of Quality in Clinical Practice, 21, 131–134.
identification and classification of epileptic seizures using
doi:10.1046/j.1440-1762.2001.00421.x
wavelet transform. International Journal of Biomedical
Mockus, A., Fielding, R. T., & Herbsleb, J. (2000). A case Engineering and Technology, 1(3), 293–314. doi:10.1504/
study of open source software development: the Apache IJBET.2008.016963
server. Paper presented at the Software Engineering, 2000.
Nason, C. S., Alexander, G. R., Pass, M. A., & Bolland, J.
Proceedings of the 2000 International Conference on.
M. (2001). An evaluation of a Medicaid managed maternity
Momjian, B. (2001). PostgreSQL: Introduction and program: The impact of comprehensive care coordination
concepts. Addison-Wesley Longman Publishing Co., Inc. on utilization and pregnancy outcome. Journal of Health
and Human Resources Administration, 24(3/4), 493–521.

331
 Compilation of References

Neilson, J. P. (1998). Ultrasound for fetal assessment Odell, J. (Ed.). (2000). Agent Technology, OMG Docu-
in early pregnancy. Cochrane Database of Systematic ment 00-09-01, OMG Agents interest. Group, (Septem-
Reviews, 4. ber): 2000.

Nelson, R., & Ball, M. (2004). Consumer informatics. Odom, M., Kumar, A., & Saunders, L. (2002). Web
Applications and strategies in cyber health care. Berlin: assurance seals: how and why they influence consum-
Springer. ers’ decisions. Journal of Information Systems, 16(2),
231–250. doi:10.2308/jis.2002.16.2.231
Ngwenyama, O., Guergachi, A., & McLaren, T. (2007).
Using the learning curve to maximize IT productivity: a OECD. (1980). Guidelines on the protection of privacy
decision analysis model for timing software upgrades. and transborder flows of personal data. Retrieved January
International Journal of Production Economics, 105(2), 29, 2008, from http://www.oecd.org/document/18/0,234
524–535. doi:10.1016/j.ijpe.2006.02.013 0,en_2649_34255_1815186_1_1_1_1,00.html

Norén, L., & Ranerup, A. (2005). The Internet Web Oh, H., Rizo, C., Enkin, M., & Jadad, A. (2005). What is
portal as an enrolment device. In Czarniawska, B., & e-health (3): A systematic review of published definitions.
Hernes, T. (Eds.), Actor-network theory and organizing Journal of Medical Internet Research, 7(1), article e1.
(pp. 188–207). Malmö: Liber and Copenhagen Business
Okoroh, M. I., Ilozor, B. D., & Gombera, P. (2007). An
School Press.
artificial Neural Networks approach in managing health-
Norris, J. (2004). Mission-critical development with care. International Journal of Electronic Healthcare, 3(3),
open source software: Lessons learned. Software, IEEE, 279–292. doi:10.1504/IJEH.2007.014548
21(1), 42–49.
Oliver, J., & Srivastava, S. (1999). The Big Five tax-
Nydick, R. L., & Hill, R. P. (1992). Using the analytic onomy: History, measurement, and theoretical perspec-
hierarchy process to structure the supplier selection proce- tives. In Pervin, L., & John, O. P. (Eds.), Handbook of
dure. Journal of Purchasing and Materials Management, personality: Theory and research (2nd ed.). New York:
25(2), 31–36. The Guilford Press.

O’Connell, R., Cho, C., Shah, N., Brown, K., & Shiff- Oliver, J. (1993, August). Shocking to the core. Manage-
man, R. (2004a). Take note: Differential EHR satisfaction ment Today, 18–22.
with two implementations under one roof. Journal of the
OLPC. (2006). One laptop per Child. Retrieved 1st De-
American Medical Informatics Association, 11(1), 43–49.
cember 2006, 2006, from http://laptop.org/
O’Connell, R., Cho, C., Shah, N., Brown, K., & Shiff-
Oser, F., & Baeriswyl, F. (2001). Choreographies of teach-
man, R. (2004b). Take note: Differential EHR satisfaction
ing: Bridging instruction to learning. In Richardson, V.
with two implementations under one roof. Journal of the
(Ed.), Handbook of research on teaching (4th ed.). Wash-
American Medical Informatics Association, 11(1), 43–49.
ington, DC: American Educational Research Association.
O’Connor, P. J., Sperl-Hillen, J. M., Pronk, N. P., & Murray,
Østergren, K. (2004). The institutional construction of
T. (2001). Primary care clinic-based chronic disease care.
consumerism. A study of implementing quality indica-
Features of successful programs. Disease Management
tors (Working paper 15). Bergen: Stein Rokkan Centre
and HealthCare Outcomes, 9, 691–698.
for Social Studies.
Obama, B. (2008). Remarks of president-elect barack
Ozcan, Y. A. (2005). Quantitative methods in health care
obama: Radio address on the economy. Retrieved June
management. San Francisco, CA: Jossey-Bass (Wiley).
27, 2009, from whitehouse.gov
Paavola, T. (2007). Justifying IT system investments. Inter-
Object Management Group. (2005). Unified Modeling
national Journal of Healthcare Technology and Manage-
Language: Superstructure. Version 2.0. OMG Inc.
ment, 8(1/2), 97–106. doi:10.1504/IJHTM.2007.012104

332
Compilation of References

Paavola, T. (2008). Exploring IT system benefits in health Pehlivanli, A. C., Ibrikci, T., & Ersoy, O. K. (2008).
care. Tampere University of Technology Publication 756. Consensual classification of drug and nondrug com-
Tampere: Tampereen Yliopistopaino. pounds. International Journal of Computational Biol-
ogy and Drug Design, 1(3), 224–234. doi:10.1504/
Paccagnella, L. (1997). Getting the seats of your pants
IJCBDD.2008.021416
dirty: Strategies for ethnographic research on virtual com-
munities. Journal of Computer-Mediated Communication, Penrod, L., & Gadd, C. (2001). Attitudes of academic-
3(1). Retrieved from http://ascusc.org/jcmc/vol3/issue1/ based and community-based physicians regarding EMR
paccagnella.html. use during outpatient encounters. Proceedings/AMIA
Annual Symposium, 528-532.
Pagliari, C., Sloan, D., Gregor, P., Sullivan, F., Detmer,
D., & Kahan, P. J. (2005). What Is eHealth (4): A Scoping Perednia, D., & Allen, A. (1995). E-medicine technology
Exercise to Map the Field. Journal of Medical Internet and clinical applications. Journal of the American Medical
Research, 7(1), e9. doi:10.2196/jmir.7.1.e9 Association, 273(6), 483–488.

Pal, A., Mbarika, V. W. A., Cobb-Payton, F., Datta, P., & Perry, T. (2007), December). Identification and prediction
McCoy, S. (2005). Telemedicine diffusion in a develop- of chronic conditions for health plan members using data
ing Country:The case of India (march 2004). Informa- mining techniques: In data mining methods and applica-
tion Technology in Biomedicine. IEEE Transactions on, tions. Taylor Francis.
9(1), 59–65.
Pham, D. T., Wang, Y., & Dimov, S. (2005). Intelligfent
Palier, B. (2005). How to understand the differences be- manufacturing strategy selection. In Proc 1st International
tween various health care systems [Att förstå skillnadena Virtual Conference on Intelligent Production Machines
prestation mellan olika hälso- och sjukvårdssystem]. and Systems (pp. 312-318). Oxford: Elsevier.
Stockholm: Sveriges kommuner och Landsting.
Pham, D. T., Wang, Y., & Dimov, S. (2006). Incorporating
Paneth, N. S. (1995). The Problem of Low Birth Weight. delay and feedback in intelligent manufacturing strategy
The Future of Children: Low Birth Weight, 5(1), 21–34. selection. In Proc 2nd International Virtual Conference
on Intelligent Production Machines and Systems (pp.
Park, K. (2003). Background to Ubiquitous Healthcare.
246-252). Oxford: Elsevier.
No.272 Conference between Industry, Academic, Re-
search Centers, Korea. Pinsky, L. E., Monson, D., & Irby, D. M. (1998). How
excellent teachers are made: Reflecting on success to
Parry, E. C., Sood, R., & Parry, D. T. (2006). Investigation
improve teaching. Advances in Health Science Educa-
of optimization techniques to prepare ultrasound images
tion, 3, 207–215.
for electronic transfer [Abstract]. Ultrasound in Obstetrics
and Gynecology, 28(4), 487-488(482). Pizziferri, L., Kittler, A., Volk, L., Honour, M., Gupta, S.,
& Wang, S. (2005). Primary care physician time utilization
Patton, M. (2002). Qualitative research and evaluation
before and after implementation of an electronic medi-
methods (3rd ed.). Thousand Oaks, CA: Sage Publications.
cal record: A time-motion study. Journal of Biomedical
Paukert, J. L., & Richards, B. F. (2000). How medical Informatics, 38, 176–188.
students and residents describe the roles and characteristics
Pohjonen, H., Bondestam, S., Karp, P., Kinnunen, J.,
of their influential clinical teachers. Academic Medicine,
Korhola, O., & Sakari, P. (1994). Quality assurance of
75, 843–845.
the bedside chest radiography process. Acta Radiologica,
Peckham, S. (2002). Book review. Health & Social Care 35, 235–243.
in the Community, 10(6), 520–522.

Pedersen, K., Christiansen, T., & Bech, M. (2005). The

Danish health care system: Evolution–not revolution–in
a decentralized system. Health Economics, 14, 41–57.

333
 Compilation of References

Poonguzhali, S., & Ravindran, G. (2007). Evaluation of Rada, R. (2005a). Obstructive sleep apnea and head and
feature extraction methods for classification of liver ab- neck neoplasms. Otolaryngology - Head and Neck Sur-
normalities in ultrasound images. International Journal of gery, 132(5), 794–799. doi:10.1016/j.otohns.2004.12.002
Biomedical Engineering and Technology, 1(2), 134–143.
Rada, R., & Hu, K. (2002). Patterns in student–student
doi:10.1504/IJBET.2007.015856
commenting. IEEE Transactions on Education, 45(3),
Porter, S. C., Cai, Z., & Gribbons, W. (2004). The asthma 262–267. doi:10.1109/TE.2002.1024619
kiosk: A patient-centered technology for collaborative
Rada, R. (2007). Sleep and quality of life in head and
decision support in the emergency department. Journal
neck neoplasm. In Verster, J. C., Pandi-Perumal, S. R.,
of the American Medical Informatics Association, 11,
& Streiner, D. L. (Eds.), Sleep and quality of life in head
458–467.
and neck neoplasm. Totowa, New Jersey: Humana Press.
Poulymenopoulou, M., Malamateniou, F., & Vassi-
Rada, R. (2005b). Software for patient-patient discussions.
lacopoulos, G. (2003). Specifying workflow process
Montreal Conference on E-Technologies 2005, Montreal,
requirements for an emergency medical service. Journal
Canada (pp. 205-209).
of Medical Systems, 27(4), 325–335.
Rada, R. (2006). Membership and online groups. e-Society
Pramatari, K. C., Doukidis, G. I., & Kourouthanassis,
2006: IADIS International Conference, Dublin, Ireland
P. (2005). Towards ‘smarter’ supply and demand-chain
(pp 290-293).
collaboration practices enabled by RFID technology. In
Vervest, P., Van Heck, E., Preiss, K., & Pau, L. F. (Eds.), Rades, D., Fehlauer, F., Bajrovic, A., Mahlmann, B.,
Smart Business Networks (pp. 197–210). Germany: Richter, E., & Alberti, W. (2004). Serious adverse effects
Springer. of amifostine during radiotherapy in head and neck can-
cer patients. Radiotherapy and Oncology, 70, 261–264.
Prather, J. C., Lobach, D. F., Goodwin, L. K., Hales, J.
doi:10.1016/j.radonc.2003.10.005
W., Hage, M. L., & Hammond, W. E. (1997). Medical
data mining: Knowledge discovery in a clinical data ware- Raento, M., Oulasvirta, A., Petit, R., & Toivonen, H.
house. In Proceedings of the AMIA Annual Symposium (2005). ContextPhone: A Prototyping Platform for
Fall 1997 (pp. 101-105). Context-Aware Mobile Applications, 4, 51-59.
Preece, J., Nonnecke, B., & Andrews, D. (2004). The top Raghupathi, W. (2006). Designing clinical decision
five reasons for lurking: Improving community experi- support systems in health care: A systemic approach.
ences for everyone. Computers in Human Behavior, 20(2), International Journal of Healthcare Information Systems
201–223. doi:10.1016/j.chb.2003.10.015 and Informatics, 2(1).
Prichard, K. W., & Sawyer, R. M. (Eds.). (1994). Hand- Raghupathi, W., & Gao, W. (2007). An eclipse-based
book of college teaching: Theory and applications. development approach to health information technol-
Westport, CT: Greenwood Press. Retrieved January 29, ogy. International Journal of Electronic Healthcare,
2008, from Questia database: http://www.questia.com/ 3(4), 433–452.
PM.qst?a=o&d=9413558
Ranerup, A. (2006). Electronic markets connecting
Qiu, R., & Sangwan, R. (2005). Toward collaborative citizens to pension reform. Electronic Markets, 16(4),
supply chains using RFID. CIO Wisdom II, more best 282–291.
practices (pp. 127–144). New Jersey: Prentice-Hall.
Rennecker, J. (2004). Updating ethnography to investigate
Quinlan, J.R. (1993). C4.5: Programs for machine learn- contemporary organisational forms. In M.E. Whitman &
ing. San Mateu: Morgan Kaufmann series in machine A.B. Woszczynski (Eds.), The handbook of information
learning. systems research (pp. xi, 349). Hershey, PA: IGI Global.

334
Compilation of References

Rhodes, B., & Mase, K. (2006). Wearables in 2005. IEEE Rosenbaum, E. E. (1988). A taste of my own medicine,
Pervasive Computing / IEEE Computer Society [and] when the doctor is the patient. New York: Random House.
IEEE Communications Society, 5(1), 92–95. doi:10.1109/
Roux-Rouquie, M., Caritey, N., Gaubert, L., & Rosenthal,
MPRV.2006.21
S. (2004). Using the Unified Modelling Language (UML)
Rialle, V., Lamy, J. B., Noury, N., & Bajolle, L. (2003). to guide the systemic description of biological processes
Telemonitoring of patients at home: a software agent ap- and systems. Bio Systems, 75, 3–14.
proach. Computer Methods and Programs in Biomedicine,
Rozinat, A., & van der Aalst, W. M. P. (2008). Confor-
72, 257–268. doi:10.1016/S0169-2607(02)00161-X
mance checking of processes based on monitoring real
Richardson, G. P. (1999). Citation for winner of the 1999 behavior. Information Systems Frontiers, 33(1), 64–95.
Jay W. Forrester Award: Jac Vennix. System Dynamics
Rubel, P., Fayn, J., Nollo, G., Assanelli, D., Li, B., &
Review, 15(4), 375–377.
Restier, L. (2005). Toward personal ehealth in cardiol-
Richardson, G. P., & Andersen, D. F. (1995). Teamwork ogy. Results from the EPI-MEDICS telemedicine project.
in group model building. System Dynamics Review, Journal of Electrocardiology, 38(4), 100–106.
11(2), 113–137.
Rubel, P., Fayn, J., Simon-Chautemps, L., Atoui, H.,
Richardson, G. P. (1996). System dynamics. In Gass, S. Ohlsson, M., & Télisson, D. (2004). New paradigms in
I., & Harris, C. M. (Eds.), Encyclopedia of operations telemedicine: Ambient intelligence, wearable, pervasive
research and management science (pp. 656–660). Boston, and personalized. Studies in Health Technology and
MA: Kluwer Academic Publishers. Informatics, 108, 123–132.

Richardson, V. (2001). (Ed.). Handbook of research on Rucker, D. (2003). Finally, a tool to re-engineer health
teaching (4th ed). Washington, DC: American Educational care: The workflow engine. San Francisco, CA: Mont-
Research Association. gomery Research, Inc.

Ridinger, M. H., & Rice, J. J. (2000). Predictive model- Russell, S. (2004). The economic burden of illness for
ing points way to future risk status. Health Management households in developing countries: a review of studies
Technology, 21, 10–12. focusing on malaria, tuberculosis, and human immuno-
deficiency virus/acquired immunodeficiency syndrome.
Riska, E. (1982). Health education and its ideological
Am J Trop Med Hyg, 71(2_suppl), 147-155.
content. Acta Sociologica, 25, 41–46.
Rygh & Mörland. (2006). Hunting for good quality indica-
Roberts, E. B., & Hirsch, G. B. (1976, Winter). Strategic
tors [Jakten på de gode kvalitetsindikatorerne]. Tidsskrift
modelling for health care managers. Health Care Manage-
for Den Norske Laegeforening, 21, 2822–2825.
ment Review, ▪▪▪, 69–77.
Rygielski, C., Wang, J.-C., & Yen, D. C. (2002). Data
Romano, J. A., Lam, D. M., Moses, G. R., Gilbert, G.
mining techniques for customer relationship management.
R., & Marchessault, R. (2006). The Future of Military
Technology in Society, 24, 483–502.
Medicine Has Not Arrived Yet, but We Can See It from
Here. Telemedicine and e-Health, 12(4), 417-425. Saaty, T. L. (1977). A scaling method for priorities in
hierarchical structures. Journal of Mathematical Psychol-
Rombaux, P., Hamoir, M., Plouin-Gaudon, I., Liistro, G.,
ogy, 15, 234–281. doi:10.1016/0022-2496(77)90033-5
Aubert, G., & Rodenstein, D. (2000). Obstructive sleep
apnea syndrome after reconstructive laryngectomy for Saaty, T. L. (1996). The analytic hierarchy process.
glottic carcinoma. European Archives of Oto-Rhino-Lar- Pittsburgh: RWS Publications.
yngology, 257(9), 502–506. doi:10.1007/s004050000267
Saboor, S., Ammenwerth, E., Wurz, M., & Chimiak-
Ropka, M. E. (2002). Symptom status and functional Opoka, J. (2005). MedFlow - improving modelling
status outcomes: Humanistic outcomes in obesity disease and assessment of clinical processes. Studies in Health
management. Obesity Research, 10, 42S–49S. Technology and Informatics, 116, 521–526.

335
 Compilation of References

Sackett, D. (1997). Using evidence-based medicine to help Scheer, A. W. (1999). Business process frameworks.
physicians keep up-to-date. The Journal for the Serials Berlin, Germany: Springer.
Community, 9, 178–181.
Schubart, J. R., & Einbinder, J. S. (2000). Evaluation of a
Sackett, D., & Straus, S. (1998). Finding and applying data warehouse in an academic health sciences center. In-
evidence during clinical rounds: The evidence cart. Journal ternational Journal of Medical Informatics, 60, 319–333.
of the American Medical Association, 280, 1336–1338.
Schuler, D., & Namioka, A. (1993). Participatory design:
Samanta, B., & Nataraj, C. (2008). Automated diagnosis Principles and practices. Hillsdale, NJ: L. Erlbaum As-
of cardiac state in healthcare systems using computa- sociates.
tional intelligence. International Journal of Services
Schulman, E. F., Sheriff, D. J., & Momany, E. (1997).
Operations and Informatics, 3(2), 162–177. doi:10.1504/
Primary Case Management and Birth Outcomes in the
IJSOI.2008.019331
Iowa Medicaid Program. American Journal of Public
Sandberg, W. S., Daily, B., Egan, M., Stahl, J. E., Health, 87(1), 80–84. doi:10.2105/AJPH.87.1.80
Goldman, J. M., Wiklund, R. A., & Rattner, D. (2005).
Schuster, E. W., Allen, S. J., & Brock, D. L. (2007) Global
Deliberate perioperative systems design improves operat-
RFID: the value of the EPCglobal network for supply
ing room throughput. Anesthesiology, 103(2), 406–418.
chain management. Berlin; New York: Springer, (ISBN:
doi:10.1097/00000542-200508000-00025
9783540356547).
Sandelowski, M. (1995). Focus on qualitative methods:
Schwartz, B. (2004). The paradox of choice. New York,
Sample size in qualitative research. Research in Nursing
NY: HarperCollins Publishers.
& Health, 18, 179–183.
Schwethelm, B., Margolis, L. H., Miller, C., & Smith,
Sanger, L. (2001). Let’s make a wiki. Message posted
S. (1989). Risk Status and Pregnancy Outcome among
to http://web.archive.org/web/20030414014355/http://
Medicaid Recipients. American Journal of Preventive
www.nupedia.com/pipermail/nupedia-l/2001-Janu-
Medicine, 5(3), 157–163.
ary/000676.html.
Scott, J., Rundall, T., Vogt, T., & Hsu, J. (2005). Kaiser
Sarnikar, S. & Gupta, A. (2007). A context-specific
Permanente’s experience of implementing an electronic
mediating schema approach for information exchange
medical record: A qualitative study. BMJ (Clinical Re-
between heterogeneous hospital systems. Forthcoming
search Ed.), 331, 1313–1316.
in International Journal of Healthcare Technology and
Management. Seaman, C. B. (1999). Qualiative methods in empircal
studies of software engineering. Software Engineering.
Sarnikar, S., Zhao, J., & Gupta, A. (2005). Medical in-
IEEE Transactions on, 25(4), 557–572.
formation filtering using content-based and rule-based
profiles. Proceedings of the AIS Americas Conference Searle, N. S., Hatem, C. J., Perkowski, L., & Wilkerson,
on Information Systems (AMCIS 2005), Omaha, NE. L. (2006). Why invest in an educational fellowship pro-
gram? Academic Medicine, 81, 936–940.
Scarsbrook, A. (2007). Open source software for radi-
ologists: A primer. Clinical Radiology, 62(2), 120–130. Seidel, J. (1991). Method and madness in the application
of computer technology to qualitative data analysis. In
Schafer, D. R., & Dillman, D. A. (1998). Development of
Fielding, N., & Lee, R. M. (Eds.), Using computers in
a standard e-mail methodology: Results of an experiment.
qualitative research (pp. 107–116). London: Sage.
Public Opinion Quarterly, 62, 378–397.
Seltsikas, P. (2001). Organizing the information manage-
Schaffer, R., & Thomson, H. (1992). Successful change
ment process in process-based organizations. In 34th
programs begin with results. Harvard Business Review,
Hawaii Int. Conf. Syst. Sci., Hawaii.
22(1), 80–89.

336
Compilation of References

Senge, P. (1994). The fifth discipline: The art and practice SKL & Socialstyrelsen. (2007). Open comparisons of
of the learning organization. New York, NY: Doubleday/ quality and efficiency in health care. A comparison
Currency. between county councils in 2007 [Öppna jämförelser av
hälso- och sjukvårdens kvalitet och efektivitet. Jämförelser
Senge, P., Kleiner, A., Roberts, C., Ross, R., & Smith,
melan landsting 2007]. Stockholm: SKL, Socialstyrelsen.
B. (1994). The fifth discipline fieldbook. New York:
Doubleday. Slaughter, L. A., Harper, B. D., & Norman, K. L. (1994).
Assessing the equivalence of the paper and online formats
Seomun, G., Chang, S. O., Lee, S. J., Kim, I. A., & Park,
of the QUIS 5.5. In Proceedings of Mid Atlantic Human
S. (2006). A Prediction Model for Patient Classification ac-
Factors Conference, (pp. 87-91)Washington, DC.
cording to Nursing Need: Using Data Mining Techniques.
In Park, H., Murray, P., & Delaney, C. (Eds.), Consumer- Slawson, D., Hauck, F., Strayer, S., & Rollins, L. (2007).
Centered Computer-Supported Care for Healthy People. What is information master? Retrieved from http://www.
Fairfax, VA: IOS Press. healthsystem.virginia.edu/internet/familymed/docs/
info_mastery.cfm#Information.
Shaalana, K. El-Badryb,M. and Rafeac, A. (2004), “A
multiagent approach for diagnostic expert systems via Smith, A. D. (2006). Exploring m-commerce in terms of
the internet”, Expert Systems with Applications, Volume viability, growth and challenges. International Journal
No. 27 Issue No. 1, Elsevier Publishing, USA pp1-10. of Mobile Communications, 4(6), 4.

Sharma, R. (2005). Automatic integration of text docu- Smith-Daniels, V., Schweikhart, S., & Smith-Daniels, D.
ments in the medical domain. Undergraduate Thesis, (1988). Capacity management in health care services:
University of Arizona. Review and future research directions. Decision Sciences,
19, 889–919.
Shepard, S. (2005). RFID: radio frequency identification,
New York: McGraw-Hill, (ISBN: 0071442995). Sobolev, B., Harel, D., Vasilakis, C., & Levy, A. R.
(2008). Using the statecharts paradigm for simulation of
Silber, D. (2004). The case for e-health. In Iakovidis, T.,
patient flow in surgical care. Health Care Management
Wilson, P., & Healy, J. (Eds.), E-health (pp. 3–27). The
Science, 11, 79–86.
Netherlands: IOS Press.
Socialdepartementet, S. K. L. Socialstyrelsen. (2007).
Simon, J. S., Rundall, T. G., & Shortell, S. M. (2007).
National IT strategy in health care. Where do we stand?
Adoption of Order Entry with Decision Support for
Report 2007 [Nationell IT-strategi för vård och omsorg.
Chronic Care by Physician Organizations. Journal of the
Var står vi idag? Lägesrapport 2007]. Stockholm: So-
American Medical Informatics Association, 14, 432–439.
cialdepartementet, SKL, Socialstyrelsen.
doi:10.1197/jamia.M2271
Socialstyrelsen. (2001). A comprehensive view of qual-
Sinreich, D., & Marmor, Y. (2004). A STAFFING emer-
ity indicators in health care. A proposal by the National
gency. Industrial Engineering (American Institute of
Swedish Board of Health and Welfare [Övergripande
Industrial Engineers), 36(6).
kvalitetsindikatorer inom hälso- och sjukvården] Stock-
Sittig, D., Kuperman, G., & Fiskio, J. (1999). Evaluating holm: Socialstyrelsens förslag.
physician satisfaction regarding user interactions with
Solberg, L. I., Reger, L. A., & Pearson, T. L. (1997). Us-
an electronic medical record system. Paper presented
ing continuous quality improvement to improve diabetes
at the AMIA.
care in populations: The IDEAL model: Improving care
SKL & Socialstyrelsen. (2006). Open comparisons of for diabetics through empowerment, active collaborations
quality and efficiency in health care. A comparison and leadership. The Joint Commission Journal on Quality
between county councils in 2007 [Öppna jämförelser av Improvement, 23, 581–592.
hälso- och sjukvårdens kvalitet och efektivitet. Jämförelser
melan landsting 2007]. Stockholm: SKL, Socialstyrelsen.

337
 Compilation of References

Sosial og helsedepartementet. (1998). Law about patient Strauss, A., & Corbin, J. (1998). Basics of qualitative
rights [Lov om patientretttigheter]. Ot prpnr 12 (1998-99). research: Techniques and procedures for developing
Oslo: Sosial- og helsedepartementet. grounded theory (2nd ed.). Thousand Oaks, CA: Sage
Publications.
Sosial-og helsedirektoratet. (2004). Law on patient rights.
Instructions [Lov om pasientrettigheter. Rundskriv]. Subasi, A., & Ercelebi, E. (2005). Classification of EEG
IS12/2004. Oslo: Sosial- og helsedirektoratet. signals using neural network and logistic regression.
Computer Methods and Programs in Biomedicine, 78(2),
SOU. (2005). Difficult waters? Premium pension savings
87–99. doi:10.1016/j.cmpb.2004.10.009
on course [Svårnavigerat? Premiepensionssparande på rätt
kurs?] [Swedish Government Official Reports. Stockholm: Sutton, D. R., Taylor, P., & Earle, K. (2006). Evaluation
Fritzes.]. SOU, 2005, 8. of PROforma as a language for implementing medical
guidelines in a practical context. BMC Medical Informat-
Spath, P. (2004). Preface. In P. Spath (Ed.), Partnering
ics and Decision Making, 6, 20.
with patients to reduce medical errors (pp. xix-xxiii). Chi-
cago: American Hospital Association Press. 1 This paper Swinfen, P., Swinfen, R., Youngberry, K., & Wootton,
borrows heavily from Roy Rada (2008) “Ethnographic R. (2005). Low-cost telemedicine in Iraq: an analysis of
discovery of adverse events in patient online discussions: referrals in the first 15 months. Telemedicine and Telecare,
customer relationship management” International Journal 11(Suppl 2), 113.
of Healthcare Information Systems and Informatics, Vol.
Swingler, K. (1996). Applying neural networks: A practi-
3, Issue 3, pp 77-85
cal guide. New York: Academic Press.
Spyrou, S., Bamidis, P., Pappas, K., & Maglaveras, N.
Tait, A. (1999). Face-to-face and at a distance: The
(2005). Extending UML activity diagrams for workflow
mediation of guidance and counselling through new
modelling with clinical documents in regional health infor-
technologies. British Journal of Guidance & Counsel-
mation systems. In. Proceedings, MIE2005, 1160–1165.
ling, 27(1), 113–122.
Stalk, G., & Hout, T. (1990). Competing against time -
Tam, M. C. Y., & Tummala, V. M. R. (2001). An Appli-
how time-based competition is reshaping global markets.
cation of the AHP in vendor selection of a telecommu-
New York: The Free Press.
nications system. Omega, 29(2), 171–182. doi:10.1016/
Stearns, N. S., Bergan, T. A., Roberts, E. B., & Cavazos, S0305-0483(00)00039-6
L. F. (1978). A systems intervention for improving medi-
Tan, J. (2001). Health management information systems:
cal school-hospital interrelationships. Journal of Medical
Methods and practical applications (2nd ed.). Gaithers-
Education, 53, 464–472.
burg, MD: Aspen Publishers, Inc.
Stearns, N. S., Bergan, T. A., Roberts, E. B., & Quigley, L.
Tan, J. (2005). E-health care information systems. An
(1976, Fall). Systems intervention: New help for hospitals.
introduction for students and professionals. San Fransisco,
Health Care Management Review, 1, 9–18.
CA: Jossey Bass.
Sterman, J. D. (1994). Learning in and about complex
Tan, J. K. H. (1995). Health management information
systems. System Dynamics Review, 10(2-3), 291–330.
systems: theories, methods, and applications. Gaithers-
Sterman, J. D. (2000). Business dynamics: Systems think- burg, MD: Aspen Publishers.
ing and modelling for a complex world. Boston, MA:
Tan, J., Kifle, M., Mbarika, V., & Okoli, C. (2005). E-
Irwin McGraw-Hill.
Medicine in developed and developing countries. In Tan,
Strassmann, P. (1990). The business value of computers. J. (Ed.), E-healthcare information systems. Jossey-Bass
New Canaan, CT: The Information Economic Press. (A Wiley Imprint).

Tanenbaum, A. (2004). Computer networks (4th ed.).

Prentice Hall, Inc.

338
Compilation of References

Tashakkori, A., & Teddlie, C. (1998). Mixed methodol- Tovey, D. (2006). The informed patient. Singapore Medi-
ogy: Combining qualitative and quantitative approaches. cal Journal, 47(9), 741–745.
Thousand Oaks, CA: Sage Publications.
Tufte, E. (2001). The visual display of quantitative infor-
Task Force Report. (1998). Recommendations of a task mation. Cheshire, CT: Graphics Press.
force of the European society of cardiology and the
Tung, F. C., Chang, S. C., & Chou, C. M. (2008). An exten-
European resuscitation council on the prehospital man-
sion of trust and TAM model with IDT in the adoption of
agement of acute heart attacks. European Heart Journal,
the electronic logistics information system in HIS in the
19, 1140-1164.
medical industry. International Journal of Medical Infor-
Télisson, D., Fayn, J., & Rubel, P. (2004). Design of a matics, 77, 324–335. doi:10.1016/j.ijmedinf.2007.06.006
tele-expertise architecture adapted to pervasive multi-actor
Turri, J. J. (1988). Program eases decision making. Health
environments. Application to e-cardiology. In Proceedings
Progress (Saint Louis, Mo.), 69, 40–44.
of the IEEE Computers in Cardiology 2004 Conference
(Vol. 31, pp. 749-752). Uskul, A. K., & Hynie, M. (2007). Self-construal and
concerns elicited by imagined and real health problems.
Télisson, D., Fayn, J., Placide, J., Rubel, P., & Comet, B.
Journal of Applied Social Psychology, 37(9), 2156–2189.
(2006). A new shared care telemedical solution dedicated
doi:10.1111/j.1559-1816.2007.00283.x
to elderly patient nursing services for remote GP decision
support. In Proceedings of the Computers in Cardiology Vaidya, O. S., & Kumar, S. (2006). Analytic hierarchy
2006 Conference (Vol. 33, pp. 229-232). process: An overview of applications. European Jour-
nal of Operational Research, 169, 1–29. doi:10.1016/j.
The Simputer Trust. (2000). Simputer(TM) - Welcome. Re-
ejor.2004.04.028
trieved December 1, 2006, from http://www.simputer.org/
Valdes, I., Kibbe, D., Tolleson, G., Kunik, M., & Petersen,
Tindal, S. (2008). ZDNet Australia. One in ten RFID
L. (2004). Barriers to proliferation of electronic medical
projects tag humans. http://www.zdnet.com.au/news/
records. Informatics in Primary Care, 12(1), 3–9.
hardware/soa/RFID-people-tagging-benefits-health-sec-
tor-/0,130061702,339285273,00.htm. 21 January 2008. van den Brink, J. L., Moorman, P. W., de Boer, M. F.,
Pruyn, J. F. A., Verwoerd, C. D. A., & van Bemmel, J.
Tiruneh, M. (2006). ETC, broadband network infrastruc-
H. (2005). Involving the patient: A prospective study on
ture for e-health. ICT-H-2006 Workshop, Addis Ababa,
use, appreciation and effectiveness of an information
Ethiopia.
system in head and neck cancer care. International Jour-
Tomasi, E. F., Luiz Augusto and Maia, Maria de Fatima nal of Medical Informatics, 74, 839–849. doi:10.1016/j.
Santos. (2004). Health information technology in primary ijmedinf.2005.03.021
health care in developing countries: a literature review. Bul-
van der Aalst, W. M. P., ter Hofstede, A. H. M., & Weske,
letin of the World Health Organization, 82(11), 867–874.
M. (2003). Business process management: A survey. In
Topacan, U., Basoglu, A. N., & Daim, T. U. “Exploring Business Process Management (BPM2003) (pp. 1–12).
the Success Factors of Health Information Service Adop- Berlin, Germany: Springer-Verlag.
tion. In Proceedings of Portland International Conference
van der Aalst, W. M. P., van Dongena, B. F., & Herbst,
for Management of Engineering and Technology’08 (pp.
J. (2003). Workflow mining: A survey of issues and
2453-2461).
approaches. Data & Knowledge Engineering, 47(2),
Torkki, P. M., Marjamaa, R. A., Torkki, M. I., Kallio, P. 237–267.
E., & Kirvela, O. A. (2005). Use of anesthesia induction
Van Latum, F., Van Solingen, R., Oivo, M., Hoisl, B.,
rooms can increase the number of urgent orthopedic
Rombach, D., & Ruhe, G. (1998). Adopting GQM-based
cases completed within 7 hours. Anesthesiology, 103(2),
measurement in an industrial environment. Software,
401–405. doi:10.1097/00000542-200508000-00024
IEEE, 15(1), 78–86.

339
 Compilation of References

Vanhecke, T., Barnes, M., Zimmerman, J., & Shoichet, Walley, P. (2003). Designing the accident and emergency
S. (2006). PubMed vs. HighWire Press: A head-to-head system: Lessons from manufacturing. Emergency Medi-
comparison of two medical literature search engines. cine Journal, 20(2), 126–130.
[Epub ahead of print] Computers in Biology and Medicine.
Wang, Y. (2007). Integration of data mining with game
Vasilakis, C., Sobolev, B., Kuramoto, L., & Levy, A. R. theory. [Boston, MA: Springer]. Journal of Interna-
(2007). A simulation study of scheduling clinic appoint- tional Federation for Information Processing, 207/2006,
ments in surgical care: Individual-surgeon vs. pooled 275–280.
lists. The Journal of the Operational Research Society,
Wang, H., Zheng, H., & Hu, J. (2008). Poisson approach
58, 202–211.
to clustering analysis of regulatory sequence. International
Venkatesh, V., & David, F. D. (2000). A theoretical exten- Journal of Computational Biology and Drug Design, 1(2),
sion of the technology acceptance model- Four longitu- 141–157. doi:10.1504/IJCBDD.2008.020206
dinal field studies. Management Science, 46, 186–204.
Waring, T., & Maddocks, P. (2005). Open source software
doi:10.1287/mnsc.46.2.186.11926
implementation in the UK public sector: Evidence from
Venkatesh, V., Morris, M. G., Davis, G., & Davis, F. D. the field and implications for the future. International
(2003). User acceptance of information technology - to- Journal of Information Management, 25(5), 411–428.
ward to unified view. Management Information Systems
Wear, D., Kokinova, M., Keck-McNulty, C., & Aultman,
Quarterly, 27(3), 425–478.
J. (2005). Pimping: Perspectives of 4th year medical stu-
Vrangbæk, K., Østergren, K., Winblad-Spångberg, U., dents. Teaching and Learning in Medicine, 17, 184–191.
& Okkels, B. (2006). Patients’ reactions to free choice
Weber, R. P. (1990). Basic content analysis (2nd ed.).
of hospitals in Norway, Denmark and Sweden. Bergen:
Newbury Park, CA.
Norwegian School of Economics and Business Admin-
istration. Weber, R., Knaup, P., Knietig, T., Hauz, R., Merzweiler,
A., Mludek, V., et al. (2001). Object-oriented business
Vrangbæk, K., & Østergren, K. (2004). The introduction
process analysis of the cooperative soft tissue sarcoma
of choice in Scandinavian hospital systems. Arguments
trial of the German Society for Paediatric Oncology and
and policy processes in the Danish and the Norwegian
Haematology (GPOH). In Proceedings of the 10th World
case (Working paper 5). Bergen: Stein Rokkan Centre
Congress on Medical Informatics (MEDINFO 2001)
for Social Studies.
(pp. 58-62).
Wager, K., Lee, F., & Glaser, J. (2005). Managing
Weerakkody, V., & Currie, W. (2003). Integrating business
healthcare information systems: A practical approach for
process reengineering with information systems develop-
healthcare executives. San Francisco, CA: Jossey-Bass.
ment: Issues and implications. In van der Aalst, W. M.
Wager, K., Lee, F., White, A., Ward, D., & Ornstein, S. P., ter Hofstede, A. H. M., & Weske, M. (Eds.), Business
(2000). Impact of an electronic medical record system on process management (BPM2003) (pp. 302–320). Berlin,
community-based primary care practices. The Journal of Germany: Springer-Verlag.
the American Board of Family Practice, 13(5), 338–348.
Weick, K., & Sutcliffe, K. (2003). Hospitals as cultures
Wagner, E. H., Austin, B. T., & Korff, M. V. (1996). of entrapment. California Management Review, 45(2),
Improving outcomes in chronic illness. Managed Care 73–84.
(Langhorne, Pa.), 4, 12–25.
Weijters, T., & van der Aalst, W. M. P. (2001, October
Walczak, S., Brimhall, B., & Lefkowitz, J. (2006). Non- 25-26). Process mining: Discovering workflow models
parametric decision support systems in medical diagnosis: from event-based data. In B. Kröse et al. (Eds.), Proceed-
Modeling pulmonary embolism. International Journal of ings 13th Belgium-Netherlands Conference on Artificial
Healthcare Information Systems and Informatics, (1): 2. Intelligence (BNAIC’01), Amsterdam, The Netherlands
(pp. 283-290).

340
Compilation of References

Weinstein, R. (2005). RFID: A Technical Overview and Windows, S. (2005). Version 14.0.2005. Chicago: SPSS
Its Application to the Enterprise. IT Professional Maga- Inc.
zine, 7(3), 27-33.Whiting, R. (2004), MIT = RFID + Rx.
Wong, H. J., & Legnini, M. W. (2000). The diffusion of
Information Week, 988, 16.
decision support systems in healthcare: are we there yet?
Weiser, M. (1993b). Some computer science issues in Journal of Healthcare Management, 45(4), 240–254.
ubiquitous computing. Communications of the ACM,
Wong, H. K. (2007). Consumerism and quality standards
36(7), 75–84. doi:10.1145/159544.159617
for subvented social services: a case study of Hong Kong.
Weiser, M. (1993a, October). Hot topics: ubiquitous International Journal of Public Administration, 30(4),
computing. IEEE Computer. 441–458. doi:10.1080/01900690601158287

Westin, A. F., & Gelder, V. (2005, August). Building Wooldridge, M., & Jennings, N. (1995). Intelligent soft-
privacy by design in health data systems. Center for ware agents: Theory and Practice. [USA.]. The Knowledge
Social and Legal Research: A report by The Program on Engineering Review, 10(2), 115–152.
Information Technology, Health Records and Privacy of
Wooton, R. (2001). Telemedicine and developing coun-
the Center for Social and Legal Research. Retrieved Janu-
tries–successful implementation will require a shared
ary 29, 2008, from http://www.amia.org/inside/initiatives/
approach. Telemedicine and Telecare, 7(Suppl 1), 1–6.
healthdata/2006/ehrrept9-6-05_westin.pdf
doi:10.1258/1357633011936589
White, H. (1990). Connectionist nonparametric regression:
Wootton, R. (2001). Recent Advances: Telemedicine.
multilayer feedforward networks can learn arbitrary map-
British Medical Journal, 323, 557–560.
pings. Neural Networks, 5(6), 535–549. doi:10.1016/0893-
6080(90)90004-5 World Health Organization. (2006a). Estimates of child
and adult mortality and life expectancy at birth by country.
Wikipedia. (2007a). Information retrieval. Retrieved
Retrieved January 17, 2007, from http://www.who.int/
from http://en.wikipedia.org/wiki/Information_retrieval.
healthinfo/statistics/mortlifeexpectancy/en/index.html
Wikipedia. (2007b). Wikipedia. Retrieved from http://
World Health Organization. (2006b, 6th June 2006).
en.wikipedia.org/wiki/Wikipedia.
Introduction to the ‘Making Pregnancy Safer’ initiative.
Willems, D. (2001). Balancing rationalities: Gatekeep- Retrieved October 1, 2007, from http://w3.whosea.org/
ing in health care. Journal of Medical Ethics, 27, 25–29. pregnancy/introf.htm

Wimmer, M. (2002). Integrated service modelling for Wright, D. (1998). Telemedicine and developing countries.
online onestop government. Electronic Markets, 12(3), A report of study group 2 of the ITU development sector.
149–156. J Telemedicine Telecare.

Winblad-Spångberg, U. (2003). From decision to practice: Xiaohaua, H. (2005). Temporal rule induction for clini-
The role of the physician in implementing patient choice in cal outcome analysis. International Journal of Business
Swedish health care [Från beslut till verklighet: Läkarnas Intelligence and Data Mining, 1(1), 122–136.
roll vid implementeringen av valfrihetsreformer i hälso-
Yackel, T. (2001). How the open source development
och sjukvården]. Uppsala: Institutionen för folkhälso- och
model can improve medical software. Medinfo, 10, 68–72.
vårdvetenskap, Uppsala Universitet.
Yang, B. H., & Rhee, S. (2000). Development of the
Windmeijer, F. A. G., & Silva, J. S. (1997). Endogene-
ring sensor for healthcare automation. Robotics and
ity in count data models: an application to demand for
Autonomous Systems, 30, 273–281. doi:10.1016/S0921-
healthcare. Journal of Applied Econometrics, 12, 281–294.
8890(99)00092-5
doi:10.1002/(SICI)1099-1255(199705)12:3<281::AID-
JAE436>3.0.CO;2-1

341
 Compilation of References

Yang, Y., & Liu, X. (1999). A re-examination of text cat- Zaki, M. J., Parthasarathy, S., Ogihara, M., & Li, W.
egorization methods. In Proceedings of the 22nd Annual (1997). New parallel algorithms for fast discovery of as-
International ACM Conference on Research andDevelop- sociation rules. Data Mining and Knowledge Discovery,
ment in Information Retrieval (SIGIR’99) (pp. 42-49). 1(4), 343–37.

Yellowlees, P. M., Marks, S. L., Hogarth, M., & Turner, S. Zeller, A., & Krinke, J. (2005). Essential open source
(2008). Standards-Based, Open-Source Electronic Health toolset: Programming with Eclipse, JUnit, CVS, Bug-
Record Systems: A Desirable Future for the U.S. Health zilla, Ant, Tcl/Tk and more. Chichester, England: John
Industry. Telemedicine and e-Health, 14(3), 284-288. Wiley & Sons.

Yin, R.K. (1999). Enhancing the quality of case studies in Zhang, J., & Fang, X., L. S., & Olivia, R. (2007). Online
health service research. Health Service Research, 34(5), consumer search depth: theories and new findings. Jour-
Part II, 1209-1224. nal of Management Information Systems, 23(3), 71–95.
doi:10.2753/MIS0742-1222230304
Yoo, S. (2006). Ubiquitous Health, Industrial Health.
Korea & World Affairs, 5, 13–18. Zurada, J., & Lonial, S. (2005). Comparison Of The
Performance Of Several Data Mining Methods For Bad
Zadeh, L. A. (1965). Fuzzy sets. Information and Control,
Debt Recovery In The Healthcare Industry. The Journal
8, 338–352.
of Applied Business Research, 21(2), 37–53.

342
 343

About the Contributors

Joseph K. H. Tan, PhD, is the Wayne C. Fox Professor of E-Business Innovation & E-Health, Mc-
Master University, Hamilton, Ontario, Canada. Previously, Dr. Tan had served as Professor and Head,
Information System and Manufacturing (ISM) Department, Wayne State University, and as Acting
Chair of the Masters in Health Administration (MHA) Program, Department of Healthcare & Epidemi-
ology, Faculty of Medicine, the University of British Columbia. Currently, as the Editor-in-Chief, the
International Journal of Healthcare Information Systems & Informatics (IJHISI), Professor Tan sits on
various journal advisory and editorial boards as well as on numerous organizing committees for local,
national, and international meetings and conferences. Professor Tan is well published and his research,
which has enjoyed significant support in the last 21 years from local, national, and international funding
agencies and other sources, has also been widely cited and applied across a number of major disciplines,
including health care informatics and clinical decision support, health technology management research,
human processing of graphical representations, ergonomics, health administration education, telehealth,
mobile health, and e-health promotion programming. His hobbies include writing and editing books,
book chapters, and journal articles; working on collaborative grant projects; engaging in philosophical
discussions with colleagues and peers; and reading his son’s work.

***

Xiangyang Li, PhD, is currently Associate Professor with the Department of Industrial and Manu-
facturing Systems Engineering of the University of Michigan – Dearborn. He received a Ph.D. degree
with research in information security from Arizona State University, a M.S. degree in systems simulation
from the Chinese Academy of Aerospace Administration, and a B.S. degree in automatic control from
Northeastern University, China. Dr. Li’s research interests include health system engineering, quality
and security of information systems, knowledge discovery and engineering, human machine studies,
and system modelling and simulation. He has been focusing on applying system and multidisciplinary
approaches to the improvement and management of complex enterprises. His research has been exten-
sively published in a variety of journals and conferences including knowledge engineering, information
assurance, human computer interaction, design, etc. Dr. Li has been well recognized in a set of projects
involving national and regional health service institutions, federal agencies, and industries. He is the
senior member of the Institute of Industrial Engineers, the member of IEEE, the Association for Com-
puting Machinery, and the Chinese Association for Systems Simulation, and Academic Advocate to
the Information Systems Audit and Control Association (ISACA). Aside the research and educational
activities, he enjoys sports, reading and travelling.
Copyright © 2011, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 About the Contributors

Yung-wen Liu, Ph.D., is an Assistant Professor in the Department of Industrial and Manufacturing
Systems Engineering. Dr. Liu has been working on healthcare research/projects for eight years. The
research/projects include OR capacity and patient flow evaluation, pelvic pain studies, sexual satisfac-
tion analysis for treated prostate cancer patients, multi-staged disease progress modeling, and racial/
ethnic disparity in healthcare quality and costs. The methods applied to the research/projects include
statistical analysis, simulation, meta-analysis, cost-effectiveness analysis, stochastic and optimization
modeling. The projects were done/ are being worked through the collaboration with Mercy Hospital
of Buffalo, University of Washington School of Medicine, Seattle Cancer Care Alliance, Henry Ford
Health System, and John D. Dingell VA Medical Center. Some research results were published in the
journals such as Journal of Urology, Clinical Journal of Pain, and Healthcare Issue in IIE Transactions.
His research in disease progress modeling won the best student paper award in 2006 Society for Health
System Annual Conference. He teaches statistics and simulation courses and helps students on health-
care related senior design projects at the University of Michigan-Dearborn.

Umit Topacan is a MA student in the department of Management Information Systems at Bogazici

University, Istanbul, Turkey. He received his BS in Computer Education and Educational Technologies
from the same university. His research interests include technology management, information technol-
ogy adoption, health information services and service development. He has work experiences both of
academia as a teaching and research assistantship; and private sector as IT consultant.

Nuri Basoglu is an Associate Professor in Department of Management Information Systems, Bogazici

University, Istanbul, Turkey. His research interests are socio-technical aspects of IS, customer-focused
product development, information technology adaptation and wireless service design, Intelligent adap-
tive human computer interfaces, information systems strategies. He has published articles in journals
such as ‘Technology Forecasting and Social Change’, ‘Journal of High Technology Management’ and
‘Technology in Society’, ‘International Journal of Services Sciences’. Dr. Basoglu received his BS in
Industrial Engineering, Bogazici University in Turkey, MS and Ph.D. in Business Administration,
Istanbul University.

Tugrul Daim is an Associate Professor of Engineering and Technology Management at Portland

State University. He is published in many journals including “Technology in Society”, “Technology
Forecasting and Social Change”, “Int’l J of Innovation and Technology Management”, “Technology
Analysis and Strategic Management”, and “Technovation”. Dr.Daim received his BS in Mechanical En-
gineering from Bogazici University in Turkey, MS in Mechanical Engineering from Lehigh University
in Pennsylvania, another MS in Engineering Management from Portland State University and a Ph.D.
in Systems Science-Engineering Management from Portland State University.

George E. Heilman is an Associate Professor of Management Information Systems at Winston-

Salem State University’s School of Business and Economics, where he teaches information technology
and accounting courses in the undergraduate, MBA and MHA programs. He holds undergraduate
degrees from Purdue University, masters degrees in business administration and public affairs from
Indiana University, and a Ph.D. in Computer Information Systems from the University of Arkansas. He
has published a number of discipline-based and pedagogical articles and book chapters dealing with a
variety of technology, finance and healthcare issues.

344
About the Contributors

Monica Cain is Associate Professor of Economics at Winston-Salem State University’s School of

Business and Economics. She did her graduate work at Wayne State University, where she earned an
M.S. in Community Health Services from the School of Medicine in 1995 and a Ph.D. in Economics
from the College of Liberal Arts in 2002. Dr. Cain teaches undergraduate economics and graduate
healthcare management courses. She has published several articles in the areas of community health
and public healthcare management.

Russell S. Morton is an Associate Professor of Management Information Systems (MIS) at Winston-

Salem State University. He received a PhD in MIS from the University of Kentucky in 1996. He started
at WSSU in the fall of 1997 after teaching one year at the University of Indiana Southeast in New Albany,
Indiana. He received a teaching award from the University of Kentucky and the first annual Business
and Economics Award for Outstanding Achievement in Teaching at WSSU. His research has appeared
in the International Journal of Electronic Business, the International Journal of Healthcare Information
Systems and Informatics and the Journal of Computer Information Systems. He is the State Volunteer
Chair for The Rocky Mountain Elk Foundation, a national habitat conservation organization, and serves
on the board of the Huntsville Historic Preservation Society. Dr. Morton received a BSBA with emphasis
in Labor Relations and MBA with emphasis in Information Systems from the University of Colorado
at Denver. He was a long time employee of Caterpillar Tractor Company in Denver, Colorado before
leaving to pursue his PhD.

Teemu Paavola, PhD, has worked in teleoperator business development, in academic life, and as a
member of technology strategy group of Sonera Plc. He is currently Managing Director and CEO of LifeIT
Plc, a consulting company owned by a Finnish health district and private parties, such as TietoEnator
Plc. Dr. Paavola holds a research fellow position at Seinajoki Central Hospital, and he has contributed
to numerous articles on information technology management, including the textbooks Tietotekniikan
linkki liiketoimintaan (1999), Linking IT to Business - A Tale of Discovering IT Benefits (2001), Clinical
Knowledge Management (2005), Managing Worldwide Operations and Communications with Informa-
tion Technology (2007), Encyclopedia of Portal Technologies and Applications (2007), Exploring IT
System Benefits in Health Care (2008) and Medical Informatics (2009).

Peter Stone is Professor of Maternal Fetal Medicine in the University of Auckland. His postgradu-
ate training was in Britain, gaining a Doctor of Medicine based on Doppler studies in fetal growth
restriction from the University of Bristol. After working in Wellington at the University of Otago for
11 years, where he set up the maternal fetal medicine service he moved to Auckland in 1998. He has
been involved in Obstetrics , Maternal and Fetal Medicine and Women’s Health throughout his profes-
sional career. He has been a member of a number of Ministerial advisory groups most recently on the
screening advisory groups for HIV and Down Syndrome as well as being a member of the National
Screening Advisory Group for the Director General of Health He is part of the ISTAR group which
brought Mifepristone into New Zealand. He is a councillor for the RANZCOG and is currently Chair
of the New Zealand Training and Accreditation Committee of RANZCOG. Research interests include
fetal welfare assessment, ultrasound studies of the cervix in pregnancy, and early pregnancy develop-
ment including implantation and trophoblast deportation. Other research interests include teaching
quality improvement. Currently he is developing an ultrasound teaching programme for the Pacific in
association with the RANZCOG and the Pacific Women’s Health Research Development unit set up in
his Department in Middlemore Hospital.

345
 About the Contributors

Emma Parry isClinical Director of MFM at Auckland District Health Board. Her main interests
over the last few years have been induction of labour rates and the techniques used to achieve labour
induction. HerMD thesis is titled ‘Induction of labour: How, why and when?’. She has also been in-
volved in looking at Obstetricians views on induction of labour and caesarean section. She is a trained
sub-specialist in Maternal-fetal medicine. Her clinical interests focus around high risk pregnancy and
complex multiple pregnancy. She is leading a team of subspecialists in New Zealand setting up a MFM
Network.

David Parry is a Senior Lecturer in the Auckland University of Technology School of Computing
and Mathematical Sciences New Zealand. His PhD thesis was concerned with the use of fuzzy ontolo-
gies for medical information retrieval. He holds degrees from Imperial College and St. Bartholomew’s
Medical College, London, Auckland University of Technology and the University of Otago, New Zea-
land. His research interests include internet-based knowledge management and the semantic web, health
informatics, the use of Radio Frequency ID in healthcare and information retrieval.

Phurb Dorji is consultant in charge of the perinatal service Jigme Dorji Wangchuck National Referral
Hospital, Thimphu, Bhutan. He has extensive experience in the development of services for the care
of pregnant women in developing nations and a continuing interest in telemedicine. He wasrecently an
invited speaker at the Asia and Oceania Federation of obstetrics and Gynaecology Conference.

Jongtae Yu is a doctoral student in Business Information Systems at Mississippi State University.

He holds a Masters degree in Marketing from the University of Alabama and MBA degree from Pu-
kyong National University in South Korea. He worked as assistant researcher at KIOS research center
in South Korea and was responsible for an economic evaluation and cost-benefit analysis for financial
losses caused by damages from public development projects. His research interests include health care
management information systems, telecommunications, and cross cultural global information systems.
His current research projects include the identification of inhibitors and enablers in adopting digital
multimedia broadcasting (DMB) and the adoption of ubiquitous banking services, which is expected
to be the future banking service.

Chengqi Guo is an assistant professor of Computer Information Systems at the University of West
Alabama, Livingston Alabama. He is also a doctoral student of Business Information Systems in the
College of Business & Industry at Mississippi State University in Starkville, MS. He holds a Masters
degree in Operations & Management Information Systems from Northern Illinois University. He has
years of industrial experience as a business consultant and IS analyst working for large state owned
company in China and Fortune 500 firm in United States. His research interests lie in virtual ventures,
business telecommunication, mobile commerce, healthcare information systems, and cross cultural
research. Some of his current research projects include ubiquitous banking service adoption, online
social network, Web 2.0, and RFID implementation.

346
About the Contributors

Mincheol Kim is an associate professor in the Department of Management Information Systems

at Cheju National University in South Korea. He was awarded a PhD degree in Operation Research
& Management Information Systems from Korea University and a Masters degree in Health Policy &
Management from Seoul National University in South Korea. He worked as researcher in Marketing
Research Team at SK Telecom in South Korea and was responsible for marketing research and demand
forecasting in Telecommunications Industry. His current research interests include ubiquitous health-
care, telecommunications and healthcare management. His current research projects include the entry
strategy for telecommunication company and adoption of ubiquitous banking services.

Roy Rada is a Professor of Information Systems at the University of Maryland Baltimore County.
Previously, he was Boeing Distinguished Professor of Software Engineering at Washington State Uni-
versity, Editor of Index Medicus at the National Library of Medicine, and Professor of Computer Science
at the University of Liverpool. Rada has worked as a consultant on computer-supported diagnosis in
pathology and radiology, led a team developing medical informatics standards, developed online training
material for doctors, and consulted with insurance companies and hospital networks about compliance
with government regulations related to information systems. Rada’s educational credentials include a
Ph.D. from University of Illinois in Computer Science and a M.D. from Baylor College of Medicine.
He has authored hundreds of scientific papers. His first journal article appeared in 1979 in “Computers
and Biomedical Research” and described a novel coding system for medical problem statements. Since
having been treated for cancer in 2003, he has devoted considerable effort to the role of information
systems, particularly patient online support groups, in health care.

347
348

Index

A B
accident and emergency departments (A&E) basic e-Medicine service (BEMS) 44, 45, 46,
242 47, 48, 49, 50, 51, 52
active server pages (ASP) 46 Basic Health Unit (BHU) 167
activity diagram 278, 279, 281, 282, 283, 284 Bayesian networks 305, 310
Act on Specialized Medical Care 95 big five model 156
acute ischemia , 302, 303, 304, 305, xv binary splitting 290
acute physiology and chronic health evaluation body mass index (BMI) 106
(APACHE) 85, 163, 167, 173, 178, 181, broadband multimedia network (BMN) 36, 37,
183 41, 42, 43, 44
administrative support 56 business process reengineering (BPR) 94, 96
agent language mediated activity model
(ALMA) 200 C
American Health Information Management As- capacity limits 131
sociation 67, 71 categorical dependent variable 290
American Library Association 15, 32 causality 259, 260, 263, 265, 270, 271
American Telemedicine Association 1 Centre for Evidence Based Medicine 16
anaesthetization stages 97 class diagram 276, 279, 280, 281, 282, 284
analytical hierarchy process (AHP) 1, 2, 3, 4, classification and regression trees (CART) 289,
5, 8, 9, 10 290, 292, 293, 300
antipathetic behaviour 258 clinical data , 302, xv
application program interfaces (API) 46 clinical pathways 247
Area Under the Roc Curve (AUC) 305 command response protocol 203
arrhythmia 303, 304, 305 composite linked birth file 87
artificial intelligence (AI) 157 computer-base patient records system (CBPR)
artificial-joint surgery 95, 97, 98, 99 177
artificial neural networks (ANN) 84, 85, 86, computerised physician order systems 278
88, 90, 91, 92 computerized physician order entry (CPOE)
association mining 257, 259, 262 68, 76, 82
Association of County Councils in Sweden Consumer Health Informatics 126, 128, 129,
130, 131, 136 140
asynchronous transfer mode (ATM) 36 continuous dependent variable 290
attendings 153, 154, 156 cost-benefit analysis 238
attribute-value-class-set (AVC-set) 291, 292 customer relationship management (CRM) 127
attribute-value-square-set (AVS-set) 291

Copyright © 2011, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Index

D e-patient network 233

e-patients 232, 233, 239
database management system (DBMS) 45, 291 EPI-MEDICS project , 302, 303, 304, 305,
data mining 114, 256, 258, 289, 290, 299, 300 307, 308, xiv, 309
decision-support 55 ethnography 232, 234
decision trees , 288, 289, 290, 291, xiv, 299, EU Privacy Directive 116
300 evidence-based medicine (EBM) 15, 32, 33
diabetes 106, 107, 110, 111 extensible markup language (XML) 307, 308
diffuse symptom 236, 238
digital data network (DDN) 41, 42 F
disease management , 102, 103, 104, x, 106,
107, 108, 109, 110, 111 Free Hospital Choice 132, 135
dynamic causal mining (DCM) 257, 258, 259, fuzzy logic , 200, 211, 212, 302, 306, 307, xv
260, 261, 262, 265, 266, 267, 269, 270 fuzzy rule 200
dynamic causal rule 257, 262, 263, 266 G
E Gastrointestinal Motility Online , vii, 14, 19,
ebusiness 163 21, 22, 23, 24, 25, 26, 27, 28, 29, 32
e-commerce systems 198 general practitioner (GP) 132, 134
economic development 162 GHA 107, 108
ED information systems (EDIS) , 241, 242, global assistance for medical equipment
243, 244, 245, xiii, 247, 248, 251, 252 (GAME) 171, 172
efficient healthcare 232 Global Care Solutions (GCS) 218
e-health 126, 128, 129, 140, 142, 143, 162, global outbreak alert and response network
163, 165, 166, 170, 171, 172, 173 (GORAN) 171, 172
electrocardiogram (ECG) , 2, 10, 302, 303, Google 17, 18, 30, 31, 33
304, 305, 306, 307, 308, 309, xv Gore, Al 16
electronic books (eBooks) 23 graphical user interface (GUI) 44, 50
electronic healthcare record (EHR) 67, 69, 73, H
74, 76, 79, 80, 178, 308, 310
electronic health data capture 55 hardware 205, 209, 211
electronic logistic information system 2 Hawking, Stephen 16
electronic mail (e-mail) 57, 146, 148, 160, 170, head injury observation (HIO) 249, 250
171, 180, 181, 234, 288 head mounted display (HMD) 215
electronic medical record (EMR) , 54, 55, 56, Healthcare Information and Management Sys-
57, 58, 59, 60, 61, 62, 63, 64, viii, 177, tems Society (HIMSS) 67, 70, 73, 180,
178, 179, 183, 193 192
emergency department information systems healthcare information system (HIS) 162, 163,
(EDIS) 69 164, 214
emergency departments (ED) , 241, 242, 243, healthcare management 214, 215, 216, 217
244, 245, 246, xiii, 247, 248, 250, 251, healthcare management information systems
252 (HMIS) 216
emergency obstetric care (EMOC) 167, 173 healthcare resource groups (HRG) 289, 294,
emergency rooms (ER) 242 297
empowerment of patients 233 Health Enhancement Research Organization
enterprise resource planning (ERP) 74 (HERO) 107, 110

349
 Index

health information service (HIS) 2, 6, 8, 10 IT productivity paradox 96

Health Insurance Portability and Accountability
Act (HIPPA) , 74, 75, 79, 112, 113, x, J
118, 119, 120, 121, 122, 124, 125, x java server pages (JSP) 46, 52
Health Maintenance Organization (HMO) 103, Jigme Dorji Wanchuck National Referral Hos-
106, 216 pital (JDWNRH) 162, 167, 172
health management 270, 271 Joint Commission on Accreditation of Health-
health personal information (HPI) 120, 121 care Organizations (JCAHO) 67, 70
health status , 102, 104, x, 107, 108, 109
heart disease 106, 107 K
heuristics 289
knowledge discovery in databases (KDD) 289,
Hippocratic Oath 67
296
human computer interface (HCI) 49
hypertension 106, 107 L
I least absolute deviation (LAD) 291
least square deviance (LS) 290, 291
INDANA database 305
lipid disorders 106, 107
information and communication technology
local area network (LAN) 36, 37, 38, 39, 40,
(ICT) 2, 8, 36, 37, 38, 41, 42, 43, 51, 52,
41, 50, 51, 203
53, 283, 286
LOGIT 85
information entropy 290
information exchange 68 M
information systems (IS) 241, 175, 177, 180,
183, 185, 188, 192, 241, 242, 243, 244, machine learning 289, 291, 300, 301
246 magnetic resonance imaging vendor 2
information technology (IT) , 36, 38, 65, 66, managed care organizations (MCO) , 102, 103,
67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 104, x, 105, 107, 108, 109
78, 79, 80, 81, 82, 83, 84, 94, 95, 96, 99, maternity care coordination (MCC) 84, 86, 87,
100, 101, 109, 126, 127, 128, 169, 140, 88, 89, 90, 91
141, 143, 168, 170, 172, 176, ix, 177, Medicaid 84, 86, 87, 90, 91, 92, 93
178, 179, 180, 181, 182, 183, 184, 185, medical adverse events 232, 233
186, 187, 188, 189, 190, 191, ix, 217, medical errors , 65, 66, ix, 68, 73, 74, 75, 77
227, 230 medical imaging retrieval system 284
Institutional Review Board 57, 148, 237 medical subject headings (MeSH) 17, 27
integrated definition (IDEF) 276, 286 Medical University of South Carolina (MUSC)
integrated services digital network (ISDN) 36, 54, 55, 56, 57, 59, 60, 61, 62, 63
41 metadata 207, 208
intelligent agent framework , xii Ministry of Social Affairs and Health 97
intelligent agents , 195, 198, 199, xii, 200, 204
intelligent monitoring systems 2
N
intelligent software agents 197 National Health Information Network (NHIN)
International Telecommunications Union (ITU) 76
37, 53, 168, 170, 173 Nature Publishing Group (NPG) 20, 22, 30
internet , 65, 72, 73, 77, 82, 83, 146, 147, 181, netnography 234, 239
157, 180, 186, 192, 206, ix neural networks 84, 85, 86, 88, 90, 91, 92, 93,
intranet 72, 206 304, 310

350
Index

neurons 88 population management 109

NeuroShell2 88, 89, 90 Practice Partner Patient Record 55
new productivity paradox 96 preferred provider organization (PPO) 103
new public management (NPM) 127 prescription writer 56
nonidentifiable information (NII) 114 private market 130
North Carolina State Center for Health Statis- process management 94
tics (SCHS) 87 process thinking 95, 96
profile matching 199, 200
O protected health information (PHI) 120
obstetrics 162, 173, 174 public healthcare initiative 86
obstructive sleep apnea (OSA) 235 public healthcare services 85, 86, 90
online patient communities 232 PubMed 15, 17, 21, 23, 34
open source movement 162
Q
open source software (OSS) , 162, 176, 163,
164, 165, 167, 168, 169, 170, 175, 176, questionnaire for user interaction satisfaction
177, 178, 179, xii, 180, 181, 183, 184, (QUIS) 56, 57, 62, 64
185, 186, 187, 188, 189, 190, 191, 192,
193 R
ordinary line square (OLS) 214, 221 radio-frequency identification (RFID) , 69,
organization for economic cooperation and 80, 81, 195, 196, 197, 199, xii, 200, 201,
development (OECD) 171 202, 203, 204, 209, 210, 211
out patient card (OPCard) 46 random blood glucose (RBG) 250
outpatient clinic 276, 277, 278, 279, 280, 281 receiver operating characteristic (ROC) curves
306
P
reimbursable event 238
paper-based system 62 risk status , 102, 104, x, 105, 106, 107, 108,
paperless hospitals 209, 210 109
Parkinson’s disease 120 Royal Government of Bhutan (RGOB) 168,
patient profiling 198, 200 172
Patient Safety and Quality Improvement Act
76 S
Patients’ Rights Act 132 secure socket layer (SSL) 38
payment by results (PbR) 289, 301 Seinajoki Central Hospital 94, 95, 96, 97, 99
perceived ease of use (PEOU) 217 self-care 309, 310
perceived usefulness (PU) 217 self organising mapping (SOM) 248, 249
perinatal medicine 162 semi-automatically linking 237
personal digital assistant (PDA) 23, 24, 202, snapshot data analyses 96
203, 204, 206, 308 software 197, 198, 200, 201, 202, 204, 205,
personal ECG monitor (PEM) , 302, 303, 304, 206, 210, 211
305, 306, 307, 308, 309, 310, xv spirometry (SPR) 250
personal identifiable information (PII) , 112, state diagram 278, 280, 281, 285
113, 114, x, 115, 116, 117, 118, 119, 120, student inspired model for effective clinical
121, 122, 123, 124 teaching (SIMECT) 154, 155, 158
picture archival and communication systems SWOT , 65, 66, ix, 67, 77, 78, 80, 83
(PACS) 69 sympathetic behaviour 258
pimping 153, 154, 156, 160

351
 Index

system crashes 60 unified modelling language (UML) , 275, 276,

system dynamics 256, 257, 258, 262, 271, 273, 277, 278, 279, 280, 281, xiv, 282, 283,
274 284, 285, 286, 287
system speed 56, 61, 62, 63 U.S. Department of Health and Human Servic-
system thinking 257, 258 es (HHS) 70, 73, 113, 118, 121, 123, 124
user resistance 71
T
V
taxonomy 74, 79
technology adoption 214, 216, 217 vectors of alternatives 5
technology adoption model (TAM) 217, 218, very small aperture (VSAT) satellite 36, 41,
220 42, 43, 44, 51, 168
technology gap 256 Veterans Health Administration 67
telecare 2 victorian emergency medical data (VEMD)
telecommunication technologies 2 248, 253
telemedicine , 1, 2, 8, 9, 10, 163, 164, 165, virtual reality 215
172, 173, 174, 284, 286, 302, xv, 311
tele-nurse 309, 310 W
theory of constraints 94, 96 walk-in clinic 133, 134
theory of reasoned action (TRA) 220 web 2.0 232
time-motion study 61, 64 WebMD 15, 22
total decision weight 8 weight of attributes 5
total quality management 94, 95, 96 wide area network (WAN) 36, 37, 38, 40, 41,
treatment status , 102, 104, x, 105, 106, 107, 42, 43, 44, 50, 51
108, 109 Wiki , vii, 14, 17
triability 2, 3, 6, 7, 8, 9 Wikipedia 16, 17, 18, 25, 34
wired local area network (WLAN) 203, 204,
U
206
ubiquitous healthcare system 214 women, infants and children program (WIC)
u-health system 214, 215, 216, 217, 218, 220, 86, 87, 91
221, 222, 223, 224, 226, 227, 228 World Health Organisation 167, 170
UK National Health Service (NHS) 289, 294, write once, read many (WORM) 203
295, 297, 300
unified medical language system (UMLS) 21, Y
26, 27, 32 Y2K investments 96

352