How HealthAtlas works
Geographic variation analysis modeled on the Dartmouth Atlas approach, extended with multivariable regression, clustering, international benchmarking, and live public data. All sources are public domain or openly licensed — no restricted claims, no EHR, no DUA.
What these numbers are, and are not
Everything here is a county-level or coarser public aggregate. Regression coefficients, cluster labels, variation indices and the synthesised county signals on the atlas are statistical associations across places — not causal findings, not clinical judgements, and not statements about any individual. Reading something about a person off a county average is the ecological fallacy; this kind of map invites it.
Two limits carry through everything below. Medicare rates are price-standardised but not age/sex/race standardised — that demographic standardisation cannot be reproduced from the public CMS files, so covariate case-mix adjustment is offered instead and labelled as such. And HRR/HSA views are a derived county rollup, not native Dartmouth region rates: about a third of counties span more than one HRR, and eleven sub-metro HRRs carry no value at all.
Questions about a specific figure, or about using this in your own work: about & contact.