About this site

Nursing Home Records republishes the ownership and inspection data CMS already makes public, exactly as CMS publishes it, in a form built to be searched, read, and cited.

Why it exists

Most people assume nursing homes are government institutions. The federal data says otherwise: roughly three in four are operated for profit, a split the Facilities page re-counts from every published batch, and every facility in the file holds Medicare or Medicaid certification. Federal law requires them to disclose their owners, officers, directors, and managing employees.

Those disclosures are public records. But public is only meaningful if people can find and read them, and today they arrive as a zip file of wide CSVs. This site exists to close that gap and nothing more.

Who built this

Joseph Keegan built and maintains this site.

This site is an independent project, not affiliated with CMS or with any facility, owner, or operator named in the data. All of it is open: the source repository holds the pages, the transform that builds them, the checks that run on every deploy, and the decision log that records every choice and every reversal.

What the data is

These are disclosures filed by the facilities and their parties themselves. CMS publishes them as filed and does not independently verify them. Gaps and layered structures in the filings appear here as gaps and layered structures. That is worth knowing before quoting any number on this site.

The checksums on the Data page prove this site matches what CMS published; they say nothing about whether a filing matches reality. The batch history goes one step further. It is an append-only ledger of every build this mirror has served, with per-file checksums, so the site's own history is provable even though CMS's is not.

Who this serves

Families are welcome, and it is worth being plain about what they will find. This site publishes the same CMS star ratings and inspection records a family would see on Care Compare, shown as CMS files them. What it does not do is compare, rank, or recommend. Care Compare puts facilities side by side, which is what choosing between them requires; this site answers a narrower question, who stands behind a given building, in what disclosed role, since when, and it answers that without interpreting. That is a different question rather than a smaller version of the same one, and it stays different however polished this site gets.

The same records make this a professional tool that happens to be free and public: discharge planners, state agencies vetting owners, reporters, researchers, and workers deciding where to work.

The rules this site follows

• Display, never calculate. Every value is CMS's own, shown as published. Counts of rows and groupings by published fields are allowed; derived scores, rankings by judgment, and edited values are not.
• Exact strings. Names are grouped by the exact published spelling. Related companies often file under several names, and the site says so instead of guessing: it never matches, merges, or claims two names are the same person or company.
• Roles first. The disclosure file includes lenders, accountants, consultants, and managers as well as owners. A name is never shown without its disclosed role.
• Uniform rules, no targets. Every name gets the identical layout under the identical rules, and the sort keys are neutral, published columns. The site never singles anyone out; whatever pattern exists reveals itself to anyone who looks.
• Radical provenance. Every table names its source file, CMS dataset, and modified date. The untouched zip, manifest, and data dictionary are republished with checksums on the Data page, and processing warnings are public there too.
• The full record, one click away. Summaries never replace access to the complete published record.
• A claim ships with the check that earns it. Anything this site says about what it does at run time is checked by software that actually runs it, every build. Where that check cannot be built, the claim comes down until it can.

When one of these rules catches a feature this site already shipped, the feature is reversed and the reversal is written down. The public decision log records every correction at full length, next to the rule that forced it, including the corrections that were embarrassing.

Researching a name further

Researchers who need more than these filings typically consult other public records: court dockets, corporate and securities filings, and government enforcement and exclusion databases, each maintained by the court or agency that creates it. This site links to none of them and attaches none of them to any name, because a matching name string is not a matching person.

Many names in the disclosure file are common. Some of those databases exist only to record wrongdoing, so a coincidental hit reads as a verdict. And the official verification steps for the most consequential of them require identifiers the public does not hold. The copy button under each name does the one part a website can do honestly: it preserves the exact filed spelling for wherever the reader's own judgment takes it.

What this site does not do

No money flows, cost reports, or margins. No entity resolution or name merging of our own. No single-name campaigns: no one's personal experience is evidence on this site, the author's included, and nothing here rests on anything but the published record. No editorial layer: no adjectives, no verdicts, no "worst of" lists. If a fact here seems significant, that significance comes from the public record itself.

Limitations

The per-facility record tables and the live owner search run a database engine inside your browser. On a network or browser that blocks it, those views show an error message rather than failing silently, and every one of them links to the untouched source files on the Data page. The engine and everything it needs come from this domain, so no third party's availability decides whether the record can be read.

The pre-rendered owner pages and every facility, state, and download page work without the engine at all. That is the designed floor: when the interactive layer cannot run, the record is still there, just less convenient.

Continuity

This site is built to outlive attention. The code is MIT-licensed, the data is public and re-fetchable from CMS, the pipeline is documented for adaptation, and the hosting has no bill that can lapse. Every page carries its data batch's date, so if this mirror ever goes unmaintained it degrades into a labeled archive rather than a silent lie. Anyone can fork the repository, run the monthly workflow, and have a current copy serving again.

Corrections

If anything here does not match the CMS source files, that is a bug: report it on GitHub or email NursingHomeRecords.org@gmail.com and it will be fixed. Corrections to the underlying records belong with CMS, which publishes its own correction process on each dataset page.