CMS Hospital Quality Command Center

An end-to-end Machine Learning pipeline and Tableau dashboard predicting CMS Hospital Star Ratings (1-5) based on clinical and patient experience metrics.

Architecture

• Ingestion: Automated retrieval of 4,800+ hospital records from CMS Provider Data.
• Processing: Handled schema drift, unified disparate facility IDs, and isolated Hybrid_HWR readmission metrics.
• Modeling: XGBoost Multi-class classifier (43% accuracy vs 20% baseline) to predict Star Ratings.
• Interpretability: SHAP feature importance revealed Patient Experience (Nurse Communication) as a primary driver alongside Clinical Mortality.

Tech Stack

• Language: Python 3
• Libraries: pandas, xgboost, shap, scikit-learn
• Visualization: Tableau Public

Execution

1. Run python src/ingestion.py to load raw CMS data.
2. Run python src/processing.py to build the master analytics table.
3. Run python src/modeling.py to train the model and generate SHAP insights.