CEHRT-published FHIR endpoints missing from the NDH, by state
Certified EHR vendors must publish service-base-URL bundles under the HTI-1 rule: org name, address, often an NPI, and the FHIR endpoint URL. ONC's Lantern aggregates them, and the CMS directory team's public scrape (DSACMS/npd_ehr_fhir_npi_slurp, cached at ftrotter-gov/npd_slurp_cehrt_clientfhir_cache) flattens them into joinable records: 7,999 Organization entries and 31,255 Endpoint entries across 17 vendor hosts at the 2026-07-10 snapshot. H45 joins that scrape against the NDH to measure, per state, how many provider organizations with a publicly-published EHR FHIR endpoint either have no matching NDH Organization or match one that carries zero Endpoint references. The baseline from H5: 98.7% of NDH organizations carried no endpoint at all in the 2026-04-09 release.
What this means
CMS publishing the data
The directory team's own Lantern scrape is the ingestion candidate; H45 quantifies what it would close, per state, against the 98.7% zero-endpoint baseline (April release). That is the before-number to cite when the ingest lands.
Startups + integrators
If the NDH shows no endpoint for an organization, a publicly-published CEHRT endpoint may still exist. Until the gap closes, check the HTI-1 service-base-URL bundles (or Lantern) as a second source before concluding an org is unreachable by FHIR.
State Medicaid PI offices
The per-state gap count is a concrete, verifiable measure of how much of your state's provider-endpoint infrastructure is invisible in the federal directory today, with public data on both sides of the join.
FHIR implementers
Vendor-published Organization entries carry NPIs and addresses of varying quality; the match rate itself (NPI-join vs name+state fallback) is reported so you can judge how far self-published HTI-1 bundles can be trusted as a directory source.
Null hypothesis
The NDH already carries a FHIR endpoint for at least 90% of organizations that publish one through their certified EHR vendor's HTI-1 service-base-URL bundle, and the residual gap shows no meaningful state-level concentration.
Denominator
Organization entries in the CEHRT client-FHIR scrape that carry a US state and match an NDH organization. The registered join was NPI-primary with a normalized name-plus-state fallback. Measured 2026-08-04, that ordering does not hold: of 287,916 published organizations, 148,757 (51.7%) carry a valid NPI, but 139,159 of those belong to athenahealth alone. Excluding athenahealth the rate is 9,598 of 148,650 (6.5%), and 101 of 191 vendors publish no NPI on any organization. Name-plus-city is therefore the primary method for roughly half the population, not the fallback, and every matched row carries its tier (exact, token-overlap, or unmatched) so the join quality stays visible. Pennsylvania, run first, resolved 114 exact, 23 by token overlap, and 50 unmatched of 187 hospitals.
Data source
Public GitHub cache of the Lantern-derived CEHRT endpoint scrape (ftrotter-gov/npd_slurp_cehrt_clientfhir_cache), joined to `cms_npd.organization` and `cms_npd.endpoint` in BigQuery, one capped scan via `bq_job_config()`. Cross-references H5 (orgs without endpoints) and H28 (FHIR-REST vs HISP split). Per-state gap counts feed the /states/[state] slices.
Pre-registered: results not yet published.
This finding is listed here before results drop. That is the project's trust contract: the null hypothesis and the computation are public first, and numbers follow. Methodology: /methodology.
Get the next finding in your inbox. One email per release, no filler.