Regression

—

How well do 40 county-level health drivers collectively explain geographic variation in this layer? An OLS+ridge regression isolates each driver's independent contribution — going beyond a single pairwise correlation.

Live signals

Healthcare watchtower

Computed directly from the live data — no model involved: the weakest-reach specialties by county coverage, the strongest health-driver correlations by Pearson r, and the counties where disease burden most outruns clinical supply.

These are statistical associations over county-level public aggregates — not causal findings, not clinical judgements, and not a statement about any individual. A county's appearance on a watchlist reflects its data relative to other counties, subject to the limits set out in the methods page and the standardization caveats noted on the Medicare layer.

Computing signals…

Coverage gaps

Correlation watch

County watchlists

Operations

Freshness & roadmap

Monitors source currency, schedules automatic update checks, and prioritizes the next public datasets to ingest based on analytic value and coverage gaps.

Source freshness

Next sources

Methods

How this atlas works

Geographic variation analysis modeled on the Dartmouth Atlas approach, extended with multivariable regression, clustering, international benchmarking, and live data through 2024.

Full methods & sources →

About

Who builds this

HealthAtlas is built and operated by Astronexus LLC. CSV and PNG export are free and ungated, and stay that way.

About & contact → [email protected]

Distribution

Geographic distribution

Quantile-binned choropleth: darker = higher values. Hover for area-level detail; click any region to open its full data profile.

—

Bivariate lens

Relationship

Each dot is a county: the mapped layer going up, the variable you pick going across. Hover a dot to find the county on the map, or a county to find its dot. Drag across dots to outline those counties.

Compare against

Pearson r across county figures: an association between places, not a causal effect and not a statement about any individual. The line is a least-squares fit, drawn for reading only. For many drivers at once, see the regression panel.

Choose a specialty

—

Loading…

—