An end-to-end healthcare analytics project that asks: which U.S. hospitals readmit Medicare patients more than expected within 30 days, and what patterns explain the gap?
Live dashboard
Data source: CMS Provider Data Catalog, Hospital Readmissions Reduction Program, FY 2026 release
ReadmitScope is built as a complete analyst workflow, not just a visualization. It covers problem framing, live data acquisition, cleaning, quality checks, exploratory analysis, statistical testing, enrichment with hospital attributes, and a deployed interactive dashboard.
The core metric is Excess Readmission Ratio (ERR):
ERR > 1.0: hospital readmits more patients than expected after CMS risk adjustment.ERR < 1.0: hospital performs better than expected.- ERR is useful because it adjusts for patient risk, making comparisons fairer across hospitals.
- Excess readmissions are systemic: 48.1% of reported condition-measures are worse than expected.
- 77.2% of hospitals exceed expected readmissions on at least one tracked condition.
- Volume matters: small hospitals show the highest risk, with a statistically significant negative association between discharge volume and ERR.
- Surgical and cardiac measures are slightly worse, but every tracked service line has elevated readmission risk.
- For-profit hospitals readmit more than non-profit hospitals after risk adjustment.
- CMS star rating is strongly ordered: 1-star hospitals perform worse far more often than 5-star hospitals.
See docs/05_findings.md for the full write-up.
| Path | Purpose |
|---|---|
| src/fetch_data.py | Pulls the latest CMS HRRP data and logs provenance. |
| src/build_aggregates.py | Cleans, enriches, and exports dashboard-ready aggregates. |
| notebooks/01_cleaning.ipynb | Cleaning workflow and suppression handling. |
| notebooks/02_eda.ipynb | Exploratory data analysis. |
| notebooks/03_analysis.ipynb | Statistical analysis and hypothesis tests. |
| notebooks/04_enrichment.ipynb | Ownership and CMS star-rating enrichment. |
| docs/00_problem_statement.md | Project framing and analytical goal. |
| docs/01_data_sources.md | Source data notes and API details. |
| docs/02_data_dictionary.md | Field definitions and metric meanings. |
| docs/03_data_quality_log.md | Data quality decisions and exclusions. |
| docs/04_decisions.md | Modeling and analysis decisions. |
| docs/05_findings.md | Executive findings. |
| docs/assets/ | Dashboard screenshots used in this README. |
| data/processed/ | Reproducible processed outputs. |
| dashboard/ | React, TypeScript, Vite dashboard. |
| requirements.txt | Python dependencies for the pipeline and notebooks. |
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
python src/fetch_data.py
python src/build_aggregates.py
jupyter nbconvert --to notebook --execute --inplace notebooks/0*.ipynbcd dashboard
npm install
npm run devLocal dashboard: http://localhost:5181
The dashboard is deployed on Vercel:
https://readmitscope.vercel.app
To redeploy from this workspace:
cd dashboard
vercel deploy --prodCenters for Medicare & Medicaid Services, Hospital Readmissions Reduction Program dataset 9n3s-kdb3.
This is public U.S. Government data. Analysis and interpretation are the author's own and do not represent CMS.
Built and maintained by Pavan Venkata Manjunath Mallipudi.
This repository is intended to show Pavan as the sole project contributor.