About this atlas

Purpose and current estimate

The Rural Cardiology Desert Atlas presents area-level public data for the 3,144 counties and county equivalents in the 50 states and District of Columbia. It is a companion to the HEARTLAND Protocol v3.3. It contains no patient-level data.

Under the preregistered address-geocoding and fallback rules, 1,869 of 3,144 counties and county equivalents (59.4%) had no qualifying provider assigned in the June 8, 2026 NPPES snapshot. The Atlas counts NPPES-listed individual providers with an active NPI record and a primary adult-cardiology taxonomy.

Provider selection and location assignment

Provider selection is restricted to Entity Type 1 records with exactly one primary taxonomy flag and one of five adult-cardiology taxonomy codes. Deactivation and reactivation dates are evaluated as of June 8, 2026. The counted unit is a distinct NPI after strict duplicate checks. The Atlas uses the main-file Business Practice Location address; it does not use the separate Practice Location Reference File.

Street-capable addresses are submitted to the Census Batch Geocoder using fixed benchmark 8 (Public_AR_ACS2025) and vintage 825 (ACS2025_ACS2025). Valid matches supply both county and coordinates. Direct-ineligible, tied, unmatched, or invalid matches with a syntactically valid ZIP receive one HUD-USPS fallback attempt using the unique highest same-state BUS_RATIO from the pinned 2025 Q4 release. HUD is a ZIP-level allocation proxy, not address-level geocoding; fallback coordinates are Census county internal points. No method equates postal ZIP5 with a Census ZCTA.

Provider counts and distances consume the same content-hashed canonical provider-location table. Great-circle distance from the Census county internal point to the nearest canonical qualifying-provider location; it is not road distance or travel time.

Hospitals and Critical Access Hospitals

Hospitals are independently normalized and assigned from the pinned CMS Hospital General Information release. Facility ID is treated as an opaque identifier: numeric six-digit identifiers are CCNs, while five-digit identifiers ending in F identify federal facilities. CAH status requires the exact normalized hospital type and a numeric CCN. The national ≥20-mile statistic counts distinct evaluable CAH facilities, not counties containing a CAH. In this build, 909 of 1,372 evaluable in-scope CAHs (66.3%) meet the unrounded ≥20.0-mile threshold; 6 of 1,378 in-scope CAHs lack an evaluable canonical facility location.

Contextual data

• CDC PLACES: CDC PLACES 2025 release model-based age-adjusted prevalence of coronary heart disease among adults, estimate year 2023 (%). Source-suppressed and unavailable area values remain null and are never imputed.
• ACS 2024 five-year estimates: B01003 county population and B19013 median household income.
• USDA RUCC 2023: the Atlas explicitly classifies RUCC values 4–9 as rural and 1–3 as urban.
• HRSA primary-care HPSA: Official daily HRSA primary-care HPSA warehouse snapshot captured July 10, 2026. Included normalized statuses are DESIGNATED and PROPOSED FOR WITHDRAWAL. HRSA status reflects primary-care HPSA evidence, not a cardiology shortage designation.
• Census 2025 geography: the county Gazetteer defines the analytical universe and county internal points. Locally vendored 1:5,000,000 Census cartographic boundaries render the map only and do not assign entities or calculate distance.

Versioning and reproducibility

Every analytical input, canonical table, county asset, and national summary is version-pinned and hash-validated. The browser loads the precomputed raw-first national summary and verifies that its counties_json_sha256 value matches the exact county JSON bytes. It does not reconstruct the rural median from rounded public distances. Automated release checks are versioned; there is no unattended “latest” or automatic monthly publication path.

The April 2026 v0.1.0 Atlas result is superseded because its taxonomy, NPI-status, address-geocoding, ZIP-to-county fallback, and metric-label rules differed from the documented v0.2.0 method.

Limitations

• NPPES is provider-reported administrative data. An active NPI is not proof of current practice, employment, availability, licensure, credentialing, full-time status, or acceptance of patients.
• The main-file Business Practice Location address may be administrative, stale, or different from service locations. It is not described as a verified primary service site. Additional locations, telehealth, locum tenens, schedules, and cross-county service patterns are not measured.
• Census matching can misassign a county. HUD-USPS fallback can misassign an individual provider or facility, omits PO-box-only and some other ZIPs, and may contain rounded ratios. Direct-only, residential-ratio, and state-unrestricted sensitivity analyses accompany the analytical review.
• Fallback coordinates are county internal points and are coarser than direct matched addresses. Great-circle county-internal-point distances suppress within-county variation and are not road distance or travel time. They do not measure appointment access or establish which facility is closest.
• CMS registration and hospital type do not establish current cardiology capability, operating status on the review date, or service availability.
• PLACES values are model-based adult CHD prevalence percentages, not mortality, heart-failure prevalence, incidence, or individual risk. ACS estimates have sampling uncertainty and use a 2024 five-year vintage.
• The 50-state/DC analysis excludes Puerto Rico and all other US territories.

Citation and license

Muller Ferreira V. Rural Cardiology Desert Atlas [software and dataset]. 2026. Available from atlas.heartlandprotocol.org. Version-specific citation metadata accompanies each published release.

Atlas source code is MIT licensed. Upstream data and derived assets retain their source-specific terms and attribution requirements.

Author

Vicky Muller Ferreira, MD. ORCID 0009-0009-1099-5690.