Methodology

Two federal sources and every state that publishes a Medicaid exclusion list. The full ledger – including jurisdictions that are broken or absent – is on the coverage page.

Source URLs are re-checked against the live sites rather than taken from directories. That matters: when this was built, eight of the state URLs republished by commercial screening vendors returned 404, including Ohio and Pennsylvania, and several others were behind bot protection or broken TLS chains. Those are recorded as gaps rather than quietly dropped.

The bulk files published by HHS OIG and the states contain no Social Security Numbers – OIG strips them under the Privacy Act. Nothing in this data can therefore establish that a listed party is the person you have in mind. OIG maintains a verification tool that confirms identity by SSN or EIN; that is the only step that closes the gap.

Search widens recall using a phonetic key, so spelling variants like Smith and Smyth are found. That key is intentionally coarse and collides across distinct names, so every candidate is then scored on its actual spelling and anything below a similarity floor is discarded before it is shown. The score is displayed on each result. It measures how closely two strings resemble each other – never how likely it is that they are the same person.

This is the part most likely to harm someone if it is done carelessly. When a provider is reinstated, jurisdictions handle it in one of two ways: some publish a reinstatement date alongside the record, and others simply remove the row from the next file. New York does the latter; Texas does the former, and keeps every provider ever excluded in its file.

Records are never deleted here. Each refresh is compared against the previous snapshot: a record that has disappeared from its source is marked as removed rather than dropped, and a record carrying a past reinstatement date is marked reinstated. Both remain searchable and both are labelled, so someone who saw an earlier result finds the correction instead of an unexplained absence.

If a source suddenly parses to zero records, the refresh is rejected outright rather than deactivating an entire jurisdiction – that pattern almost always means the source changed shape, not that everyone was reinstated.

Sources are refreshed on a weekly pass. The federal LEIE publishes monthly, by the 10th; state cadences vary and several are unstated. Each record shows the date its source was last successfully fetched, and the coverage page flags any source whose fetch has been failing so staleness is visible rather than assumed.

It is not a consumer reporting agency, a background check, or a screening service of record. It does not store rosters, monitor individuals over time, or send alerts. Employment, contracting, and credentialing decisions carry obligations this tool does not discharge – use it to find and read the primary record, then act on the primary record.

If you believe something here is wrong about you or your organization, the corrections page explains what we can and cannot change.