Hospital Readmission Analysis
Overview
This project analyzes hospital Excess Readmission Ratios (ERR) using linear mixed-effects models to evaluate how hospital characteristics and county-level socioeconomic factors influence readmission performance.
The analysis focuses on identifying structural and socioeconomic drivers of readmissions rather than hospital-level volume effects alone.
Data
- CMS Hospital Readmission Measures
- Area Health Resources Files (AHRF)
Raw data are not publicly shared due to usage and privacy restrictions.
All analysis code is fully reproducible given access to the original data sources.
Methods
- Data cleaning and multi-source joins (ZIP → county FIPS)
- Feature engineering and standardization of socioeconomic variables
- Linear mixed-effects models with hospital-level random intercepts
- Marginal effect visualization using standardized predictors (
ggpredict)
Key Findings
- Higher county-level poverty and unemployment rates are associated with higher adjusted ERR, even after controlling for hospital characteristics.
- Physician-owned hospitals exhibit lower adjusted ERR compared to other ownership types.
- Socioeconomic context explains a meaningful portion of between-hospital variation in readmission performance.
Outputs
🌐 Live HTML Report (Recommended)
👉 https://ada-nguyen-ds.github.io/hospital-readmission-analysis/
📥 PDF Report
👉 Download PDF
R, tidyverse, lme4, emmeans, ggeffects