The record is the dataset — versioned methods and inspectable evidence
Every published figure is traceable to a named public source on a named date. Registered edition status and hashes are checkable at /verify, and the methods behind published results are available for inspection. The public record requires no access request. The live record is the demo: use it, cite it, reproduce it, and try to break it.
Current state: 0 live Network Integrity findings. All 189 prior findings were withdrawn in two separate correction events for different methodological reasons. On 2026-08-04, TrustNav withdrew 182 licensing-screen findings after an audit showed that the flagged class did not discriminate from the broader candidate universe (72.2% of the flagged matched an active Colorado licensee, against 68.2% across the 146,310-provider comparison population). Seven remaining findings had already been withdrawn on 2026-07-31 after a separate reconciliation control showed that an apparent concentration could be explained by differences between the compared source regimes.
The two withdrawals remain separate in the correction record because they were different failures with different causes. The superseded method is historical only: it is not rerun as a current finding method — a successor must pass a base-rate control against the full provider universe as a release gate — but the preserved record remains available for studying methodological failure, correction provenance, and release-gate design.
- The regime-reconciliation set — where state-submitted and federal-published records differ, with cohorts retained separately rather than collapsed (state-only vs present-in-both).
- The schema-drift census — how published-file structures change across carriers and vintages.
- The comparability ladder — including UNRESOLVED as a first-class state when the evidence does not support a benchmark.
- The identity-resolution discard taxonomy — including a documented pass in which ~99% of mechanically generated candidates were excluded on identity grounds before any finding count was stated.
- The correction trail — the record documents its own supersessions and withdrawals as part of the dataset.
The public record is free to inspect, reproduce, and cite. Independent replication and critical method review are welcome; a method that survives external scrutiny is more useful than one that is merely described as rigorous.
Deeper collaboration can be scoped through a grant, contract, or institutional partnership. Where nonpublic or regulated data are involved, applicable data-use, privacy, and IRB requirements govern access. TrustNav does not condition access on authorship or a favorable interpretation.
Public-record access is free. Commissioned or grant-funded research can be scoped separately; funding does not determine the method or result.
The record establishes what published sources said and where designated reference sources agree or conflict under the published method. It does not determine network adequacy, compliance, or wrongdoing. Public-facing records withhold individual clinician names; NPIs and license identifiers remain the verification keys.
The record is still developing. Independent replication, methodological criticism, and research collaboration are welcome.