About the Healthcare Dearth Map

The Healthcare Dearth Map is an interactive tool that visualizes healthcare access gaps across all 3,109 counties in the contiguous United States. It covers 15 medical specialties and uses a composite Dearth Score (0–100) to quantify how underserved each county is — higher means worse access. The map is designed for health policy researchers, healthcare planners, and anyone interested in understanding where provider shortages are most severe.

How It Was Built

The application was built as a data pipeline that processes publicly available federal datasets, computes access metrics using spatial algorithms and road-network routing, and exports the results as static files served on GitHub Pages — no backend server required.

1. Provider data collection — We downloaded the full NPPES National Provider Registry (9.7 GB), filtered to 1,560,696 active individual providers, and mapped their 139 NPI taxonomy codes to 15 specialty categories.
2. Geographic data — County boundaries, centroids, and populations come from the US Census Bureau Gazetteer files. We used the Census ZCTA-to-county crosswalk to map 33,012 zip code tabulation areas to their primary counties.
3. Spatial analysis — Using PostGIS spatial queries, we computed provider counts, density (per 100k population), and nearest-provider distances for every county-specialty pair.
4. Drive time computation — We processed the full US road network from OpenStreetMap (Geofabrik) using OSRM (Open Source Routing Machine) to calculate actual road-network drive times from each county centroid to its nearest provider for every specialty.
5. Score calculation — Each metric was percentile-ranked across all counties within each specialty, then combined into a weighted composite Dearth Score. Scores were validated against HRSA Health Professional Shortage Area (HPSA) designations.

The Dearth Score

The Dearth Score is calculated per county-specialty pair using two components, each normalized via percentile ranking across all 3,109 counties:

1. Provider Density (60%) — Number of providers per 100,000 population. Higher density means better access. Score = 100 × (1 − percentile rank of density). A county in the bottom percentile for provider density scores close to 100.
2. Drive Time (40%) — Road-network drive time (in minutes) from the county centroid to the nearest provider, computed using OSRM with real OpenStreetMap road data. Score = 100 × percentile rank of drive time. A county with the longest drive time scores close to 100.

Density is weighted higher (60%) because a county can have a short drive to one provider who serves 200,000 people — the proximity alone doesn't mean the area is well served. Drive time (40%) captures geographic accessibility through real road networks, which matters most in rural areas where the nearest specialist may be hours away through mountain roads or across rivers.

Score Interpretation

Specialties Covered

The map tracks 15 medical specialties, covering 139 NPI taxonomy codes:

Data Sources

• Provider data: CMS NPPES National Provider Registry — 1,560,696 active individual providers with practice locations and taxonomy codes
• Population & geography: US Census Bureau Gazetteer files — County and ZCTA centroids, populations, and land areas
• Road network: OpenStreetMap via Geofabrik — US road network processed with OSRM for drive time routing
• Validation: HRSA Health Professional Shortage Area (HPSA) designations used to validate score accuracy

By the Numbers

Current Limitations

• NPPES data currency — Provider counts are based on NPPES registry data, which may not reflect whether a listed provider is actively seeing patients or has retired.
• Practice location only — Providers are counted at their registered practice address. The model does not account for providers who serve multiple locations or cross-county patient flows.
• No wait time data — The score does not include appointment wait times, as no reliable nationwide data source currently exists for this metric.
• No telehealth — The model does not account for telehealth availability, which may improve access in underserved areas without requiring physical proximity.
• County-level granularity — Scores are computed at the county level. Access can vary significantly within large counties, especially in urban areas.
• Drive time from centroid — Drive times are measured from the county geographic centroid, which may not represent where most residents live, particularly in large or irregularly shaped counties.

Planned Improvements

• Real-time drive times via Google Maps API — Replace OSRM static routing with Google Maps Distance Matrix API for real-time drive times that account for current traffic conditions, road closures, and construction.
• Appointment wait time integration — Incorporate actual appointment availability data through partnerships with scheduling platforms (e.g., Zocdoc, Healthgrades) or direct provider surveys. This would add a third component to the Dearth Score capturing how long patients wait for care.
• Provider acceptance & availability — Integrate data on whether providers are accepting new patients, their insurance network participation, and office hours to give a more accurate picture of accessible care.
• Sub-county visualization — Drill down to zip code level within counties for more granular analysis, particularly useful in urban areas where access can vary block-by-block.
• Telehealth overlay — Layer in broadband availability data to identify where telehealth could bridge physical access gaps, and estimate the impact on effective Dearth Scores.
• Trend analysis — Track changes over time using monthly NPPES updates to identify areas where provider shortages are worsening or improving.

Disclaimer