DATTA MEGHE INSTITUTE OF HIGHER EDUCATION AND RESEARCH

(Deemed to be University)
FACULTY OF ENGINEERING AND TECHNOLOGY

Module 5

Subject: Management of Information in Healthcare

Class: B.Tech (TE)
Semester: VI

By
Dr. Sarang Pande
Professor
(AIDS/AIML/CSD/CSME/CSE)

1
 Course Contents

Module 5: Health Information Analytics [2]

Health information analytics and its applications in healthcare organizations, Data

mining and predictive analytics in healthcare, Healthcare business intelligence and
performance management.

Applications:
1.Predictive analytics
2.Business Intelligence
3.Comparison of performance
 Health Information Analytics

Types of Health Data

•Clinical data (EHRs, imaging, etc.).
•Administrative data (billing, operations).
•Patient-generated data (wearables, surveys).
Analytical Techniques
•Descriptive analytics: Summarizing historical data.
•Predictive analytics: Forecasting future trends using machine learning.
•Prescriptive analytics: Recommending actions.
 Health Information Analytics

Tools and Technologies

•Examples: Tableau, Python, R, SAS, etc.
•Big data technologies: Hadoop, Spark.
Applications
•Early disease detection.
•Optimizing hospital operations.
•Personalized medicine.
•Population health management.
Health Information Analytics

Descriptive Analytics
 Hospital Performance Reporting
Goal: Summarize and analyze the performance of a hospital's services over a specified period.

Average Length Patient Bed Occupancy

Month ER Visits
of Stay (days) Satisfaction (%) Rate (%)

Jan 1200 4.5 88 85

Feb 1100 4.2 90 82

Mar 1250 4.7 87 88

Outcome:
•Descriptive analytics could generate monthly reports showing trends in ER visits, average stay
durations, and patient satisfaction levels.
•This helps identify if patient volumes are increasing and if satisfaction or length of stay requires
improvement.
 Emergency Room (ER) Utilization

Goal: Summarize patient flow and wait times in the ER to identify inefficiencies.

Total ER Average Wait Average Treatment ER Admissions

Month
Visits Time (mins) Time (mins) (%)

Jan 5000 30 90 60

Feb 4800 25 80 58

Mar 5200 35 95 62
Outcome:
•Descriptive analytics can identify trends in ER visits, wait times, and treatment durations.
•This data helps optimize staffing and operational procedures in the ER.
 Patient Satisfaction Analysis
Goal: Analyze patient feedback surveys to identify common issues and overall satisfaction.

Wait Time Staff Rating Cleanliness Overall

Survey ID
(mins) (1-5) Rating (1-5) Satisfaction (1-5)

001 30 4 5 4

002 45 3 4 3

003 20 5 5 5
Outcome:
•Descriptive analytics can generate a summary of the average wait times, staff ratings, cleanliness
ratings, and overall satisfaction scores.
•This analysis helps identify the key drivers of patient satisfaction and areas for improvement.
Health Information Analytics

Predictive Analytics
 Chronic Disease Management
Goal: Predict disease progression to personalize care.

Patient Glucose Physical Activity Predicted Risk

Age HbA1c (%)
ID Variability (mins/day) Progression

201 58 8.5 High 20 Rapid

202 45 7.2 Medium 45 Moderate

203 70 6.8 Low 30 Slow

•Models Used: Regression models, support vector machines, or clustering.

•Outcome: Proactive adjustments in medication or therapy.
 Disease Prediction
Goal: Identify patients at risk for chronic conditions such as diabetes or heart disease.

Patient Cholesterol Blood Pressure Glucose Smoker Family Risk

Age BMI
ID (mg/dL) (mmHg) (mg/dL) (0/1) History (0/1) Score

1 55 28 240 140/90 150 1 1 0.85

2 45 22 180 120/80 95 0 0 0.15

3 60 30 250 150/95 160 1 1 0.92

•Models Used: Logistic regression, decision trees, or neural networks.

•Outcome: High risk patients are flagged for lifestyle intervention or preventive care.
 Health Information Analytics

Predictive Analytics -
Example
Logistic Regression
 Define Relevant Risk Factors (Variables)

Identify the key risk factors associated with heart disease. Common risk factors include:
•Age (Years)
•Gender (Male/Female)
•Blood Pressure (Systolic/Diastolic)
•Cholesterol Level (Total cholesterol, LDL, HDL)
•Smoking (Yes/No)
•Diabetes (Yes/No)
•Family History (Yes/No)
•Physical Activity (Yes/No)
•Body Mass Index (BMI)
 Risk Factors and Points
Framingham Heart Study Risk Score
1. Age: 3. HDL Cholesterol (mg/dL):
•For males and females, the points are assigned •≥60: -1 point (negative points, as high HDL is protective)
differently. Here’s a simplified version for men: •50–59: 0 points
• 20–34 years: 0 points •40–49: 1 point
• 35–39 years: 2 points •<40: 2 points
• 40–44 years: 4 points 4. Systolic Blood Pressure (mmHg):
• 45–49 years: 6 points •<120: 0 points (for people not on medication)
• 50–54 years: 8 points •120–129: 1 point
• 55–59 years: 10 points •130–139: 2 points
• 60–64 years: 12 points •140–159: 3 points
• 65–69 years: 14 points •160–179: 4 points
• 70+ years: 16 points •≥180: 5 points
•For women, the points are slightly different. 5. Diabetes:
2. Total Cholesterol (mg/dL): •No diabetes: 0 points
•<160: 0 points •Diabetes: 2 points (For both men and women)
•160–199: 1 point 6. Smoking Status:
•200–239: 2 points •Non-smoker: 0 points
•240–279: 3 points •Smoker: 4 points (For both men and women)
•280+: 4 points
Apply Logistic Regression Model (or Scoring System)
 Compute Risk Score
An example using sample values for risk factors and assuming we have the following
coefficients based on prior studies (these are for illustrative purposes):

Coefficient (β\
Risk Factor Value Contribution to Score
betaβ)

Age 55 0.02 0.02×55=1.10.02 \times 55 = 1.10.02×55=1.1

0.015×240=3.60.015 \times 240 =

Cholesterol 240 0.015
3.60.015×240=3.6

Blood Pressure 140 0.01 0.01×140=1.40.01 \times 140 = 1.40.01×140=1.4

Smoking (Yes = 1) 1 2.5 2.5×1=2.52.5 \times 1 = 2.52.5×1=2.5

Family History (Yes = 1) 1 1.5 1.5×1=1.51.5 \times 1 = 1.51.5×1=1.5

The Risk Score formula:
Risk Score = 1.1 + 3.6 + 1.4 + 2.5 + 1.5 = 10.1
 Transform the Risk Score into a Probability &
Interpretation

•Risk Score Interpretation: If we assume a threshold (e.g., 0.3 or 30%), we can

classify patients into different risk categories:
•Low risk (e.g., < 0.1 or 10% probability)
•Moderate risk (e.g., between 10% and 30% probability)
•High risk (e.g., > 30% probability)
 Data Mining Techniques in Healthcare
•Classification models assign labels (e.g., "High Risk" or "Low Risk").
• Classification (Predicting Disease •Example algorithms: Decision Trees, Naïve Bayes, Support Vector
Machines (SVM), Neural Networks
Outcomes) •Use case: Predicting whether a patient has heart disease (Yes/No).
•Clustering algorithms find hidden patterns.
•Example algorithms: K-Means, DBSCAN, Hierarchical Clustering
• Clustering (Grouping Similar Patients) •Use case: Identifying groups of patients with similar disease
progression.
• Association Rule Mining (Finding •This technique finds relationships between different data attributes.
•Example algorithm: Apriori Algorithm
Relationships) •Use case: Finding common disease co-occurrences like "Patients with
diabetes have a 50% chance of developing hypertension."

• Anomaly Detection (Identifying •Detects unusual patterns that may indicate fraud or rare diseases.
•Example algorithms: Isolation Forest, One-Class SVM
•Use case: Detecting health insurance fraud based on unusual billing
Fraudulent Claims) patterns.

•Predicts numerical values like hospital costs or patient readmission rates.

• Regression Analysis (Predicting •Example algorithms: Linear Regression, Ridge Regression
•Use case: Predicting ICU patient stay length based on symptoms and
Future Values) past data.
 Predictive Analytics Techniques in Healthcare
•Logistic Regression → Predicts binary outcomes (e.g., "Will a
patient develop diabetes?" Yes/No).
• Machine Learning Models •Random Forest & Decision Trees → Used for predicting
disease progression and hospital readmissions.
•Neural Networks & Deep Learning → Analyzes complex
patterns in medical imaging (MRI, X-ray, CT scans).

•Used for patient monitoring (e.g., blood sugar level trends in

diabetics).
• Time-Series Forecasting •Example algorithms: ARIMA, LSTM (Long Short-Term
Memory).

•Extracts insights from clinical notes, radiology reports, and doctor

prescriptions.
• Natural Language Processing (NLP) •Example: Predicting mental health disorders based on patient medical
records.
 Benefits of Predictive Analytics in Healthcare

• Early Detection of Diseases → Enables proactive treatment.

• Reduced Hospital Readmissions → Saves healthcare costs.

• Personalized Patient Care → Improves treatment effectiveness.

• Optimized Hospital Resources → Better staff and bed management.

• Fraud Detection → Identifies false insurance claims.

 Challenges in Predictive Analytics

• Data Privacy Issues → Need for HIPAA compliance in patient data.

• Data Quality & Integration → EHRs have inconsistent formats.

• Interpretability of AI Models → Complex models (deep learning) lack

explainability.

• Bias in Healthcare Data → AI models may be biased due to

underrepresentation of certain demographics.

 Healthcare Business Intelligence

• Healthcare Business Intelligence (HBI) uses data analytics, reporting,

and visualization to improve decision-making in healthcare organizations.

It helps hospitals, clinics, and policymakers optimize operations, improve

patient care, reduce costs, and enhance overall efficiency.

 Importance of Business Intelligence in Healthcare

• Data-Driven Decision Making → Helps administrators make informed choices.

• Operational Efficiency → Reduces wait times, optimizes staff schedules.

• Financial Management → Tracks revenue, expenses, and insurance claims.

• Patient Care Improvement → Identifies trends in disease outbreaks and

treatment effectiveness.

• Regulatory Compliance → Ensures HIPAA, GDPR, and other healthcare

standards compliance.
 Components of Healthcare Business Intelligence
•Electronic Health Records (EHRs): Integrates data from multiple
hospitals, labs, and insurance providers.
• Data Warehousing & Integration •ETL (Extract, Transform, Load) Processes: Aggregates data from
different sources for analysis.
•Big Data in Healthcare: Uses AI and cloud-based storage for real-time
analytics.

•Descriptive Analytics: Summarizes historical data (e.g., patient visits,

readmission rates).
• Data Analytics & Reporting •Predictive Analytics: Forecasts disease outbreaks, hospital admissions,
and patient deterioration.
•Prescriptive Analytics: Recommends the best course of action (e.g., AI-
driven treatment plans).

•Hospital Performance Metrics (e.g., Bed Occupancy Rate, Average

Length of Stay).
• Performance Dashboards & KPIs •Clinical KPIs (e.g., Infection Rate, Patient Satisfaction Score).
•Financial KPIs (e.g., Revenue Cycle, Insurance Claim Processing Time).

•AI-driven clinical decision support systems (CDSS) help doctors diagnose

diseases faster.
• Decision Support Systems (DSS) •Machine Learning Models predict patient deterioration in ICUs.