Hospital Readmission Risk Analysis

Predicting hospital readmission penalties using logistic regression and CMS quality data.

🎯 Project Overview

This 4-week project analyzes CMS hospital readmission data to:

• Identify patterns in hospital readmission rates
• Build logistic regression models to predict penalty risk
• Create risk scoring system for hospital quality improvement

📊 Key Skills Demonstrated

• Logistic Regression: Binary outcome prediction
• ROC Analysis: Model performance evaluation
• Risk Stratification: Creating actionable risk categories
• Statistical Analysis: ANOVA, correlation analysis
• Data Visualization: R plots and Tableau dashboards

📁 Project Structure

🚀 Getting Started

1. Download CMS data files (see Data Sources section)
2. Run scripts in order: week1_explore_data.R → week4_risk_scores.R
3. View results in outputs/ folder

🔍 Key Findings

Model Performance

• Predictive Accuracy: AUC = 0.629 (Fair performance)
• Dataset: 2,493 hospitals with complete data across 6 medical conditions
• Sensitivity: 69.9% (correctly identifies hospitals that receive penalties)
• Specificity: 48.1% (correctly identifies hospitals that avoid penalties)

Primary Risk Factors

1. Hospital Quality Rating (Strongest Predictor)

• Impact: Each 1-star increase reduces penalty odds by 29.8%
• Business Insight: 5-star hospitals have ~76% lower penalty risk than 1-star hospitals
• Recommendation: Quality improvement initiatives targeting star rating advancement provide highest ROI

2. Geographic Location

• Highest Risk Region: Northeast (+21.6% higher penalty odds vs Mid-Atlantic)
• Lowest Risk Region: Mountain West (-50.8% lower penalty odds vs Mid-Atlantic)
• Business Insight: Regional performance variations suggest opportunities for best practice sharing

3. Hospital Ownership

• Highest Risk: For-profit hospitals (baseline)
• Lower Risk: Government hospitals (-30.1% lower odds) and Non-profit hospitals (-23.6% lower odds)
• Business Insight: For-profit hospitals may benefit from quality practices adopted by government/non-profit sectors

Actionable Recommendations

Immediate Actions (0-3 months)

1. Focus quality improvement on 1-2 star hospitals - highest penalty reduction potential
2. Implement Northeast region support program - address highest-risk geographic area
3. Deploy risk scoring dashboard - enable proactive penalty prevention

Medium-term Initiatives (3-12 months)

1. Best practice sharing program - Mountain West hospitals mentor Northeast facilities
2. For-profit hospital quality enhancement - targeted support for highest-risk ownership type
3. Predictive model integration - incorporate into existing quality management systems

Statistical Significance

All primary predictors showed high statistical significance (p < 0.001), indicating robust and reliable relationships for business decision-making.

📊 Dashboard Visualization

Hospital Penalty Risk Assessment Dashboard

Key Features:

• Interactive geographic risk mapping
• Real-time KPI monitoring
• Risk category distribution analysis
• Top 10 highest risk hospitals identification

Built with: Tableau Desktop Data Source: CMS Hospital Readmissions Reduction Program (2020-2023)

🛠️ Tools Used

• R: Data analysis and modeling
• Packages: tidyverse, pROC, broom, corrplot
• Tableau: Interactive dashboard creation
• GitHub: Version control and portfolio showcase

📊 Data Sources

• CMS Hospital Compare: Hospital readmission rates and penalties
• CMS Hospital General Information: Hospital characteristics and ownership

📝 Methodology

1. Week 1: Exploratory data analysis
2. Week 2: Statistical pattern analysis (ANOVA, correlations)
3. Week 3: Logistic regression modeling and ROC analysis
4. Week 4: Risk scoring system and dashboard preparation

---

Part of healthcare analytics portfolio demonstrating progression from descriptive statistics to predictive modeling.