Pennsylvania rural hospitals: FHIR endpoint publication and EHR concentration
All 187 hospitals CMS lists in Pennsylvania, joined to the certified-EHR service-base-URL bundles their vendors publish under HTI-1. The join answers two questions at once: whether a hospital is reachable by FHIR, and which EHR it runs. Result: 137 of 187 (73%) appear in a vendor bundle. Rural hospitals are not behind. 81% of hospitals in nonmetro counties appear, and 65% of Critical Access hospitals. The EHR market is concentrated: Epic 86, athenahealth 24, MEDITECH 8, TruBridge 7, Oracle Health 5. The structural finding is about how vendors publish, not about coverage: all 137 matched hospitals resolve to an endpoint, but only 51 carry `Organization.endpoint` on their own record. The rest sit beneath a brand-level organization that holds the endpoint, reached through `partOf`. Epic publishes that hierarchical shape (every one of its 1,187 national brand records carries an endpoint, while its 83,678 facility records do not), and the flat-shape vendors put an endpoint on each organization. Both are valid FHIR. An integration that checks only the matched record reports no endpoint for an Epic hospital whose endpoint is live. Separately, none of Epic's 84,865 published organizations carries an NPI, so its footprint cannot be joined to NPPES or the NDH on identifier at all.
Headline
137 of 187 Pennsylvania hospitals resolve to a published FHIR endpoint (73.3%). 43 sit in nonmetro counties and 17 are Critical Access. 50 could not be matched to any vendor bundle.
137 / 187 = 73.26%
unit: count
What this means
Startups + integrators
Resolve `partOf` before deciding a hospital has no endpoint. Vendors publish in two shapes: an endpoint on each organization, or an endpoint on the brand with facilities hanging beneath it. Checking only the record you matched reports a false negative for every hospital published in the second shape, which is most Epic sites.
CMS publishing the data
Two publishing shapes coexist in the HTI-1 bundles and a consumer must handle both. The larger obstacle for NDH ingestion is identifiers: Epic publishes 84,865 organizations nationally with no NPI on any of them, so that footprint cannot be joined to the federal directory on identifier. Naming an expected identifier in the submission spec would do more than requiring an endpoint reference.
State Medicaid PI offices
The 50 hospitals with no vendor-bundle match are the list to check first, not a list of hospitals without endpoints. Some publish under a parent system name and federal facilities publish differently. Every row carries its match tier.
Researchers
Rural disadvantage does not show up in endpoint publication here, which is worth stating because the opposite is usually assumed. Nonmetro PA counties do carry lower median household income ($59,743 against $71,832) and an older population (23.5% aged 65+ against 20.9%), so the demand-side gap is real even where the connectivity gap is not.
Methodology readers
Hospital-to-vendor matching is by name and city because most bundles omit the NPI, so results are tiered: 114 exact, 23 by token overlap, 50 unmatched. Per-hospital Health Information Organization and TEFCA participation are deliberately absent because no machine-readable participant list exists.
Null hypothesis
Hospitals in nonmetro Pennsylvania counties and Critical Access hospitals publish FHIR endpoints at a materially lower rate than metro hospitals, and a published organization record carries its own endpoint reference. The first half is rejected: rural publication runs above the statewide rate. The second half is rejected in a different way: resolution succeeds for every matched hospital, but for most of them only after following `partOf` to a brand-level record.
Denominator
The 187 hospitals with a Pennsylvania address in CMS Hospital General Information. County rural status is the USDA ERS Rural-Urban Continuum Code 2023 (4-9 nonmetro), covering 34 of 67 counties. Critical Access is the CMS facility-level federal rural designation, 17 hospitals. The endpoint denominator is the set of Pennsylvania organizations published in certified-EHR bundles (10,177 at the pinned snapshot).
Data source
Five public sources, no BigQuery and no paid API: CMS Hospital General Information; USDA ERS Rural-Urban Continuum Codes 2023; USDA ERS county median household income; Census Population Estimates vintage 2023 for median age and share 65+; and ftrotter-gov/npd_slurp_cehrt_clientfhir_cache, the CMS directory team's cache of ONC Lantern certified-EHR bundles. Compute script: `analysis/pa_rural_health.py`.
Notes
Pennsylvania has five certified Health Information Organizations connected to the P3N hub, and QHIN designation under TEFCA continues on a rolling basis. Neither the HIO participant lists nor the QHIN participant rosters are published in machine-readable form, and the PA eHealth Partnership states that the P3N and its HIOs are not sub-participants in any QHIN. This dashboard therefore reports FHIR endpoint publication and EHR vendor, which are measurable, and does not assign HIO or TEFCA participation per hospital.
Get the next finding in your inbox. One email per release, no filler.