Clinical Data Management

(Process and practical guide)

Nguyen Thi My Huong, MD. PhD

WHO/RHR/SIS

Training Course in Sexual and Reproductive Health Research

Geneva 2013
 OUTLINE

Overview of Clinical Data

Management (CDM)
CDM: processes, practical guide and
challenges
WHO online data management (DM)
system (OpenClinica)
 Clinical Data Management (CDM)

Data - important products of the scientific

research
CDM - a critical task in clinical research that
involves in all aspects from data collection to
data extraction for analysis
End result for CDM:
– Provide a study database that is accurate,
secure, reliable and ready for analysis.
– Accelerate timeline from data collection to
analysis
Clinical Data Management (CDM)

Good CDM - foundation for good clinical

trial (CT) that ensures the delivery of the
quality data on-time and within the trial
budget
Good DM practices will enable you to
effectively create, organize, manage and
store data that makes your data easier to
use, analyze, share and REUSE in the
future
Clinical Data Management process
 CDM process
DM plan development
Study setup
Training
Data collection
Data processing
Quality Assurance & Quality Control
Audit trail
Monitoring data quality and data safety
Security and confidentiality
Database Closure, Data storage and archive
 WHY - CDM?

WHY do we need to process the research

data?
WHY do we implement edit checks and
query inconsistencies?
WHY do we need GCP-compliant CDM?
Data Management Plan (DMP) development

• DMP describes all the components of the DM

process to ensure consistent and effective DM
practices
• Good DMP - successful DM implementing
• Each component in the DM process should
specify:
• Work to be performed
• Responsible staff for the work
• Guidelines and/or SOPs will be complied with
• Output will be produced
 DMP development (cont'd)

• DMP should be developed for each study and

early during the setup of the study
• Provide budget information for DM
• Responsible staff should review and agree with
the DMP to make sure a consistent approach
to the process and guidelines
• DMP - a living document throughout the study
life cycle, to address any updates/changes
made during the conduct of the study
 Study setup

Includes:
• Case report form (CRF) design
• CRF completion guidelines
• Trial database (DB) setup
• Validation checks
 Case report form (CRF) design

• Quality of the data relies on the quality of

data collection instruments (CRFs)
• CRFs design:

• during the protocol development

• cover all the data specified by the
protocol
• Collection of extraneous data adversely

affects data quality

 CRF design (cont'd)
• Avoid redundant data
• Unnecessary work for the site staff
• Unnecessary need for checking data
consistencies
• Data based on the same measurement should
not be collected more than one.
• Raw data are generally preferable to the
calculation based on raw data (example: DOB
is preferable to the age)
 CRF design (cont'd)
• Flow of data from perspective of the person
completing the CRF
• Flow of study procedures and organization of
data in medical records define the flow of CRFs
• Logically related data should be grouped together
• Separate CRF for each visit: SCR, ADM, FUP…
• Questions and instructions - clear and concise
- Use consistent codes, appropriate date and time

formats and units of measurements

 CRF design (cont'd)
• Data in coded form:
• Minimize errors
• Reduce processing time
• Coded formats: drop-down list
• 1=yes
• 2=no
• 3=not sure
• Consistency in the order of similar response options
1=yes, 2=no throughout the CRF
• Minimize free text
• Pilot-testing
 CRF completion guidelines
• Full, accurate completion of CRFs is critical to:
• Quality of data captured
• Fewer queries
• Quicker validation of data
• Complete, concise and logical guidelines for CRF
completion ensure:
• All required fields are completed
• Data recorded in the CRFs are logical
• Free text entries are clinically appropriate
 CRF completion guidelines (cont'd)
• Definitions
for data items that are not directly
measurable (hypertension)
• Procedures for making corrections to data

• Handling completed CRFs

• Shipping the CRFs from sites to the DM center

• Update the guidelines

 Trial database (DB) setup

• Clinicaltrial DB contains clinical data and

metadata that is structured using CDM
software (rows, columns)
• Readable format of the clinical trial DB:
SAS, SPSS, Excel spreadsheet…
• Captured clinical data must be entered
and stored in a computer system
 Trial database setup (cont'd)

• Success of a clinical trial depends on quality

and integrity of its DB
• A poor DB design adversely impact DE, data
cleaning, extraction and data storage
• Key goal for the DB setup:
• high quality DB
• meet both clinical and regulatory
requirements
• store data accurately
 Trial database setup (cont'd)
• DB structure considers:
• ease and speed of DE
• prevention of errors in data creation and
modification
• efficient creation of data sets for analysis
• formats of data files requirements
• GCP: "Ensure that the systems are designed to
permit data changes in such a way that the
data changes are documented and that there is
no deletion of entered data (i.e., maintain an
audit trail)".
 Validation checks

• Validation checks:
• Crucial tool for each study DB
• Created for all study endpoints and safety data
• Identify data inconsistencies and potential
errors
• Increases data quality
• Provide greater efficiency for data cleaning
• Validation check document - a living document
throughout the study life cycle, is updated to
CRF changes or errors need correcting
 Validation checks (cont'd)

• Missing values
• Valid range
• Logically inconsistent checks across
fields or across CRFs
• Protocol violations
• Checks for duplicates
 Training
• Effective training ensures:
• Regulatory compliance
• Performance effectiveness
• Job satisfaction of CDM staff
• Staff involve in the DM process must be trained
• GCP: "Each individual involved in conducting a
trial should be qualified by education, training,
and experience to perform his/her respective
task(s).”
 Training (cont'd)

• Training documentation (SOPs,

guidelines)
• Training content:
• Consistent across all training materials
• Consistently conveyed by instructors or
mentors
• Types of training is defined by the roles
 Data collection
• Clinical data capture can be done using:
• Paper CRFs (pCRFs)
• EDC system (online, offline, combination of both)

• GCP requirements:
• All clinical trial information should be recorded,
handled, and stored in a way that allows its
accurate reporting, interpretation, and verification
• Data reported on the CRF, that are derived from
source documents, should be consistent with the
source documents or the discrepancies should be
explained.
 Data processing
Data processing workflow at the DM center:
• Data receipt
• CRFs tracking
• Data review and coding
• Data entry (DE)
• Data validation
• Query management
 Data receipt
• Data receipt vary across the clinical research
that may be received through:
• Fax transmissions
• Regular mail
• Web entry or transferred through other
electronic means
• …
• The processes by which data are received,
confirmed as received, and made available
for DE should be documented
 CRF tracking

• All CRFs should be tracked

• CRF logging can be:
• Manual Subject Form Register
(SFR)
• e-SFR
• Missing CRF should be specified
 Data review and coding
• Manual review of CRFs
• For all CRFs before DE
• Detect data errors, frequently encountered
problems
• Medical coding - properly classify the medical
terminologies
• Manual coding using common coding
dictionaries (ICD-10, ICPM, WHO Drug
dictionary, Medical Dictionary for Regulatory
Activities "MedDRA")
• Automated coding
 Data Entry systems
Local DE system:
• Data entered onsite by local staff
• Quick data resolutions for omissions, errors,
inconsistencies
Central DE system:
• Completed CRFs sent to DM center
• Data entered by experienced DE operators
• Forms stored centrally
Web-based DE system:
• Require only web browser and internet connection
• Secure link provided
• Data transmission is not necessary
 Data Entry methods

• Double DE - independent verification: Two people

enter data and a third person resolves discrepancies
between both entries
• Double DE - blind verification: Two people enter
data (unaware of what values the other entered) and
the 2nd DE operator verifies data, determines the
appropriate entry and saves data (overwrite the prior
value)
 Data Entry methods (cont'd)
• Double DE - interactive verification: Two people
enter data and the 2nd DE operator resolves
discrepancies between 1st and 2nd entry while being
aware of the previous values
• Single data entry – review: One person enters data
and 2nd person reviews the entered data against the
source data
• Optical character recognition (OCR): Software is
used to recognize characters from pCRFs or faxed
images then these data are placed directly into the
database. Data obtained through OCR should always
be reviewed for accuracy
 Data Entry guidelines

• Standard conventions for DE ensures

consistency in the entry of data throughout the
study
• DE timelines: timing expectations between data
collection and DE
• Instructions for handling error messages
triggered from edit checks
 Data validation

• DB automatically checks data against the pre-

defined validation rules to detect:
• Missing values
• Outliers
• Inconsistencies
• Protocol violations
• Validation checks:
• At the time of DE
• Run on batches of data
 Data validation (cont'd)

• Data validation using descriptive statistics

• Manual review for data validation vs

Programmatic validation

• Data validation focus:

• Primary and other endpoints
• Key safety fields
 Query management

Ensure rapidity of query generation and

problem resolutions
• Review validation outputs
• Confirm queries and create query
sheets
• Resolve returned queries
• Update pCRFs and DB
 Quality Assurance & Quality Control

• QA: Strategies used before and during data

collection to ensures the best possible data will
be collected (CRF design, training staff, testing
data collection tool and supporting technologies
"DE system, validation tool)
• QC: processes applied after data collection to
evaluate the quality of the collected data (data
cleaning, SDV, making decisions for data
issues)
Quality Assurance & Quality Control
Implementing QA and QC procedures
enhances the quality of data, minimizes
errors and identifies potential problems
and techniques to address them.
QA-QC processes vary substantially in
their cost and effectiveness (QC - more
difficult and resource-intensive than QA.
It's easier to prevent than repair problems
and much cheaper in the long run).
 Audit trail
• GCP requirement: Any change or correction to a
CRF should be dated, initialed, and explained and
should not obscure the original entry. That is, an
audit trail should be maintained. This applies to both
written and electronic changes or corrections
• Audit trail:
• Documents all modifications to a DB
• Is stored in a secure system
• All documentation of data changes
• Essential study documentation
• Is subject to audit
 Monitoring data quality and safety
• Routine progress reports:
- Recruitment report (actual vs target number of subject
recruited)
- FU report (overdue visits)
- Data monitoring reports (number of forms received,
entered, list of form errors & data errors)
- Adverse Event Reporting
• Interim analysis: follow predefined frequency and timing
• Site monitoring visits
• Auditing
• Data Safety Monitoring Board (DSMB)
 Security and confidentiality
• Keep identifying data (subject name, social security
number, medical record number) in a separate place and
restrict access to this data
• Make sure only subject ID links to the DB
• Data can be accessed by authorised staff (Password
protect)
• Qualified personnel for management and modifications
• Copy of data cannot be distributed without investigator’s
consent
 Database closure

• Proper closing a study DB:

• Preventing inadvertent or unauthorized
changes to data
• Ensuring the integrity for the generation
of results, analysis and submissions
• Process for closing the DB and conditions
for re-opening the DB must be followed
 Data storage
• Secure, efficient and accessible storage of clinical data
is very important
• Potential of unauthorized access and data corruption
during data storage and transfer are significant and must
be prevented to ensure consistency of results and data
quality
• Original data collected (e.g., CRFs, lab data, medical
notes and e-documents) must be protected and stored in
secure areas with controlled access (e.g., locks)
 Data storage (cont'd)
• Store clinical data in a way that backup copies can be easily and
frequently made
• (Paper documents should be scanned soon after receiving and
archived electronically, whenever possible, as the backup. E-
documents are regularly backup)
• Access permission control, especially important for a the EDC trial
that has no paper backups
• Minimize opportunity for data corruption via accidental or intentional
manipulation
• Use open formats for archival, storage, and transport of data (e.g.,
ASCII, SAS Transport, PDF, CDISC ODM Model)
 Data archive
• Maintain all documents and electronic records to ensure
their raw formats
• Archive clinical data and documents in a secure and
stable areas (no flood, fire protected, pest control)
Components must be archived:
• Original study documents: The original and/or scanned
images of all CRFs, clinical notes, lab data… DMP, data
handling guidelines
• Raw data files: The final raw data preserved in the study
DB format and all original data transfers in their raw
format
 Data archive (cont'd)
• Final data files: Preserved in a standard file format (e.g.,
ASCII, SAS transport, CDISC Operational Data Model)
• Audit trail
• Discrepancy management logs
• Database design specifications (metadata, validation
checks)
• DB closure documentation: of each DB-lock and unlock,
describing the time and conditions surrounding those
procedures
• Procedural variation documentation: Memos and
relevant information about any variations from SOPs or
working practices
 WHY - CDM?

WHY do we need to process the research

data?
WHY do we implement edit checks and
query inconsistencies?
WHY do we need GCP-compliant CDM?
 CDM - Conclusion

CDM is really a challenge to many

researchers who rarely have formal
training in DM
Implementing DM practices should take
the full study life cycle into account.
Good DM requires adopting best practices
 OpenClinica
WHO online DM system
05_XXX_MM52
100% web-based system that is built on a
modern, web-based technology architecture
The system is always accessible to those who
need to access it
Easy to generate and extract data in numerous
formats: SPSS, SAS, Excel
Users need a simple PC and internet
connection to use the system
Introduction
Open source clinical trials software for capturing
and managing clinical trial data
Software is free and available in source code
form
Free to run studies for any purpose
It is used for private and non-commercial
application
In addition to private and non-commercial
applications, AKAZA Research sells commercial
services where customers can get technical
supports
 Fully complies with GCP and regulatory
guidelines
– Keep trial data secure
– Monitor access and data changes
– Comprehensive auditing to any data changes
– Control access to study information via different
user roles and privileges
– Prevent unauthorized access to data via user
password
– Ability of electronic signature
– Daily data back-up
 WHO/RHR OC system
OC platform is hosted and technically
supported by Akaza Research
WHO/RHR/SIS is responsible for:
– Design and configuration of the online study
as well as guidelines for using the system
(SOP)
– User training (onsite and remote training) on
how to employ OC most productively
– Monitoring and managing database of the
project to ensure data is of highest quality
OpenClinica - Key functions

Submit Data, Notes & Discrepancies

Extract Data
Monitor and Manage Data
Setup and Manage Study
Administration