Methodology & Data
CareShed is built to be auditable: every county score traces back to public CDC data through a reproducible pipeline. This page documents the sources, the index, and its limits.
Data sources
| Source | Role | License |
|---|---|---|
| CDC PLACES — County (2025 release)swc5-untb | Chronic burden, access, prevention, and social-need measures (model-based small-area estimates) | CDC open data / U.S. Government work (public domain) |
| US county cartographic boundaries (FIPS)geojson-counties-fips | County geometry + FIPS join spine | Public domain (U.S. Census cartographic boundaries) |
Composite index construction
The Health-Equity Need Score aggregates four percentile-normalized domains (higher = greater need):
- Chronic burden (0.35) — diabetes, high blood pressure, obesity, coronary heart disease, COPD, stroke.
- Access gap (0.30) — uninsured rate, plus the shortfall in routine checkups and dental visits.
- Prevention gap (0.15) — the unscreened share for cholesterol, colorectal cancer, and mammography.
- Social vulnerability (0.20) — food and housing insecurity, transportation barriers, lack of social support.
Each measure is oriented so higher means worse, converted to a percentile rank, averaged within its domain, then combined with a weighted geometric mean and re-ranked to a 0–100 score. Counties are tiered (Critical / High / Moderate / Lower) by score quantiles.
Limitations (read before acting)
- PLACES values are model-based small-area estimates, not direct counts.
- Age-adjusted prevalence preferred for cross-county comparability where available.
- Model-based estimates. PLACES values are modeled small-area estimates from survey data, not direct counts; confidence intervals widen for small populations.
- Weighting is a choice. The domain weights are a documented v1 baseline; a sensitivity analysis and provider-shortage access data are the roadmap.
- Proposed enhancements not shown as results: HRSA provider-shortage (careshed access); Census ACS social vulnerability; CDC WONDER-validated explainable risk model with SHAP.
Reproducibility
The pipeline is one command — npm run data (python3 scripts/build_index.py) — pulling CDC PLACES, building the domains, computing the composite, and writing validated GeoJSON/JSON. No manual steps, no hand-edited numbers.