CMS Hospital Quality Analytics

Project Overview

This project analyzes CMS hospital quality data to explore hospital characteristics, overall ratings, missing quality-measure reporting, and differences across states, hospital types, and ownership categories.

The goal is to build a clean, reproducible data analytics project using Python, pandas, and matplotlib while demonstrating a professional workflow for data collection, cleaning, exploratory analysis, and reporting.

Business Question

How do hospital quality ratings and reporting patterns vary across U.S. hospitals by geography, hospital type, and ownership?

Dataset

The project uses the CMS Hospital General Information dataset from Medicare.gov provider data.

The dataset includes hospital-level information such as:

• Facility ID
• Hospital name
• Address, city, state, and ZIP code
• Hospital type
• Hospital ownership
• Emergency services availability
• CMS overall hospital rating
• Quality-measure group counts

Tools Used

• Python
• pandas
• matplotlib
• Git and GitHub
• VS Code
• Git Bash

Project Structure

cms-hospital-quality-analytics/
├── data/
│   ├── raw/
│   │   └── hospital_general_info.csv
│   └── processed/
│       └── hospital_general_info_clean.csv
├── outputs/
│   └── figures/
│       ├── average_rating_by_hospital_type.png
│       ├── overall_rating_distribution.png
│       └── top_states_by_hospital_count.png
├── reports/
│   ├── average_rating_by_hospital_type.csv
│   ├── average_rating_by_ownership.csv
│   ├── average_rating_by_state.csv
│   ├── hospital_count_by_state.csv
│   ├── missing_values_report.csv
│   └── overall_rating_distribution.csv
├── src/
│   ├── download_data.py
│   ├── clean_data.py
│   └── eda_summary.py
├── .gitignore
├── requirements.txt
└── README.md

Workflow

1. Data Collection

The raw CMS hospital dataset is downloaded and saved to:

data/raw/hospital_general_info.csv

2. Data Cleaning

The cleaning script standardizes column names, trims text fields, preserves ZIP codes as strings, converts hospital overall ratings to numeric values, and creates a flag for whether a hospital has an available overall rating.

Run:

python src/clean_data.py

Cleaned data is saved to:

data/processed/hospital_general_info_clean.csv

3. Exploratory Data Analysis

The EDA script creates summary reports and charts for hospital counts, rating distributions, missing values, average ratings by state, average ratings by hospital type, and average ratings by ownership.

Run:

python src/eda_summary.py

Outputs are saved to:

reports/
outputs/figures/

Visualizations

Top 15 States by Number of Hospitals

Hospital Overall Rating Distribution

Average Overall Rating by Hospital Type

Key Findings

• The cleaned CMS hospital dataset contains 5,432 hospitals and 38 columns.

• Hospital counts were highest in larger states. Texas had the most hospitals in the dataset with 468, followed by California with 378 and Florida with 221.

• CMS overall ratings were missing for a large portion of the dataset. 2,250 hospitals, or 41.42%, did not have an available overall rating. Because of this, rating-based comparisons were interpreted only among hospitals with available ratings.

• Among hospitals with available overall ratings, ratings were most commonly concentrated around 3 and 4 stars. The dataset included 987 hospitals rated 3 stars and 950 hospitals rated 4 stars.

• State-level average ratings varied among rated hospitals. Utah had the highest average overall rating at 4.24 across 29 rated hospitals, followed by Colorado at 3.96 across 49 rated hospitals and South Dakota at 3.89 across 18 rated hospitals.

• Average ratings also differed by hospital type. Acute Care - Veterans Administration hospitals had the highest average overall rating at 4.16 across 112 rated hospitals. Critical Access Hospitals averaged 3.22 across 400 rated hospitals, while Acute Care Hospitals averaged 3.16 across 2,670 rated hospitals.

• Ownership categories showed meaningful variation, but sample size matters. Veterans Health Administration hospitals had the highest average overall rating at 4.16 across 112 rated hospitals. Tribal hospitals averaged 4.00, but this was based on only 2 rated hospitals, so that result should be interpreted cautiously.

• The missing-values report showed high missingness in several CMS quality-measure footnote and measure-count fields. These missing values were preserved because they may reflect measure applicability, hospital type, reporting eligibility, or CMS reporting requirements rather than simple data-entry errors.

Missing Data Handling

Missing values were reviewed and summarized in a dedicated missing-values report. Missing values were preserved rather than automatically removed because many missing fields relate to CMS quality-measure reporting categories that may not apply to every hospital.

This is especially important for smaller hospitals, specialty hospitals, psychiatric facilities, children’s hospitals, and critical access hospitals, which may not report the same measure groups as larger acute-care hospitals.

How to Run This Project

Clone the repository:

git clone https://github.com/dgraves4/cms-hospital-quality-analytics.git
cd cms-hospital-quality-analytics

Create and activate a virtual environment:

python -m venv .venv
source .venv/Scripts/activate

Install dependencies:

pip install -r requirements.txt

Run the scripts:

python src/download_data.py
python src/clean_data.py
python src/eda_summary.py

Next Steps

Potential future improvements include:

• Building an interactive dashboard in Power BI or Tableau
• Adding maps to show hospital distribution by state or region
• Comparing ratings across ownership types in more detail
• Adding additional CMS quality datasets
• Creating a final executive summary of findings
• Automating the full pipeline with a single command or workflow script

Project Status

Current status: Initial data collection, cleaning, exploratory analysis, reports, and visualizations are complete.