Finding a healthcare practice on the web and knowing which registered organization it is are different problems. A website shows that a practice exists and is active. It does not, by itself, establish the practice's organizational NPI. Without that identity, the registry-backed facts that segmentation, market mapping and diligence depend on (legal name, practice location, specialty, and the clinicians linked to the organization) cannot be attached with confidence.
We measured that gap in GPH's proprietary observed practice graph. At the 2026-09-19 census, 4,973 of 16,751 active practices (29.69%) carried an organizational identity that met the graph's promotion rule. The other 11,778 (70.31%) were observed and active but not yet identity-resolved.
Unresolved does not mean a practice is missing, closed, misidentified or unusable. It means the graph has not established a qualifying organizational identity at the required confidence level, so identity-dependent registry attributes are withheld rather than guessed. It does not mean that no corresponding registry identity exists. This is a measurement of one observed graph, not an estimate of how many U.S. practices are unresolved, and the resolved subset should not be treated as representative of the whole for identity-dependent claims.
How the link breaks
Many practices are registered under one name and known by another. The organizational NPI is held by the legal entity, for example a professional limited liability company or corporation, while the website, signage and invoices carry a brand name that may never have been filed as an assumed name or added to the registry record. The registered practice location can also lag the practice's actual address, for example when the entity is formed before the practice settles on its premises. When both the name and the address differ, the obvious matching keys fail, and where a practice publishes little else about itself, public evidence alone may not be enough to connect it to its registration.
We publish these figures as a baseline, so later measurements against the same definitions can be compared with it.
| Measure | Counted population | Aggregate result | What it means |
|---|---|---|---|
| Active practice entities | All active practice entities in the observed graph | 16,751 | The denominator for this release |
| Identity-resolved | Active entities with an organizational-NPI identity meeting the graph promotion rule | 4,973 (29.69%) | Registry-linked attributes can be attached |
| Identity-unresolved | Active entities without that organizational-NPI identity | 11,778 (70.31%) | Identity-dependent dimensions remain unavailable |
| Governed resolution ceiling | A bounded, reviewable set based on evidence already present in the graph | 1,368 (11.61% of unresolved) | A reviewable ceiling, not an expected recovery rate |
Why this matters for anyone using practice data
A web domain is useful for discovering an observed entity. It is not enough to make that entity decision-ready. The identity-resolution layer determines whether a record can be connected to location, specialty, legal-name, and clinician-linkage evidence that make segmentation, market mapping, diligence, and account research more reliable.
At the 2026-09-19 census, every practice in the observed graph had a web domain by construction, yet 70.31% had not been resolved to a qualifying organizational identity. Treating domain presence as equivalent to a complete practice record would therefore have overstated usable coverage.
What can be improved without buying new data
1,368 of the 11,778 unresolved practices (11.61%) fell within a bounded, reviewable set based on evidence already present in the graph. These are reviewable records, not verified identities or an expected recovery rate.
The remaining 10,410 of 11,778 unresolved entities (88.39%) have no measured existing-data resolution lever in this graph. That is not an assertion that they cannot be resolved. It is a scope-bound finding that the current graph lacks a measured path for them without a new source or method.
Methodology and boundaries
The source is a verified, read-only census of the proprietary practice graph, with an independently verified entity-resolution clarification. An active practice is a practice entity with ACTIVE status. Identity-resolved means an organizational-NPI identity satisfying the graph's confidence promotion rule; identity-unresolved means that rule is not met.
The release reports only aggregate findings. It intentionally omits entity names, domains, addresses, NPIs, raw identifiers, row-level source evidence, state-by-state values, and combinations that could be used to recreate the underlying inventory. Reviewable records are reported as aggregate counts and are not represented as verified identities.
The observed graph spans 19 states according to its verified member-level state assignment. That figure describes graph coverage only. It is not a national coverage rate, a market-share estimate, or a complete census of practices in those states.
Citation
Cite this research using the stable path:
GPH, “The Practice Identity Resolution Problem: Why a Website Is Not an Organizational Identity,” September 2026,
/research/practice-identity-resolution-gap-2026-09/.