How we count

This page exists so that anyone — a family, a facility administrator, a reporter, a regulator — can reproduce every number we publish and check our arithmetic. Where a figure is our own calculation rather than something a government agency published, we say so on this page and we show the working. That distinction matters and most sites blur it.

The two files everything comes from

Both are free, both are published by the Centers for Medicare & Medicaid Services, and both can be downloaded by anyone at the CMS Provider Data Catalog.

• Provider Information (dataset 4pq5-n9py) — one row per certified nursing home. Star ratings, certified beds, county, ownership type, chain name, staffing hours, whether ownership changed in the last twelve months, and the abuse icon.
• Health Deficiencies (dataset r5ix-sfxw) — one row per citation. The facility, the survey date, the deficiency tag, and the scope-and-severity grade.

Assisted living has no federal equivalent — no star rating, no national inspection file — so those counts come from each state’s own licensing data: the Department of Social Services in California, the Department of Health Services in Arizona, the Division of Public and Behavioral Health in Nevada, and the Health Care Authority in New Mexico.

The severity letters, which are the whole game

Federal inspectors grade every finding from A to L. Almost nobody explains this, and without it a violation count is close to meaningless.

Grade	What it means
A–C	No harm, minimal potential for harm
D–F	No actual harm, but more than minimal potential. About 93 of every 100 findings nationally sit here.
G–I	A resident was actually harmed
J–L	Immediate jeopardy — residents were in danger of serious injury or death

When we say a home was cited for something that actually hurt a resident, we mean grade G or above. When we also give a total, that total includes every grade. This is why we always publish both numbers, and why a home with 200 findings and none at G can be a safer bet than one with 40 findings and twelve.

What CMS published, and what we calculated

This is the section that matters most.

Reported directly from the government record — we are repeating what CMS published, and you can verify each one on that facility’s own Medicare page, which we link from every row: star ratings, certified bed counts, county, ownership type, chain membership, the abuse icon, and individual citation counts and their severity grades.

Our own calculation — CMS did not publish these; we computed them from the files above, and they are our analysis, not the government’s:

• State and national shares. “37 of Arizona’s 140 nursing homes carry the abuse warning — 26.4%” is us counting rows where abuse_icon = Y and dividing by the state’s total certified homes.
• State rankings. “Second-highest in the country” is us running that same calculation for all fifty states and sorting. We excluded territories. We did not exclude small states.
• Operator-level rates. “9 of Haven Health’s 20 Arizona homes” is us grouping by the chain_name field exactly as CMS records it.
• Star distributions. The bar charts on each state page are counts of the overall_rating field.
• Harm counts per facility. We filtered that facility’s citations to grades G through L and counted them.

The limits of this data — stated plainly

We would rather tell you these than have you find them out.

The star ratings are graded on a curve within each state. The bottom fifth of homes in California get one star regardless of how they would rank in Nevada. A three-star home in one state is not equivalent to a three-star home in another, and our charts compare a state against itself.

Citation rates vary by state survey practice, not only by care quality. Each state’s survey agency conducts federal inspections, and they do not all cite at the same rate or on the same schedule. A state that inspects more often and writes more down will look worse on paper than one that does not. Where we make a state-level comparison we check inspection recency alongside it and publish what we find, including when it cuts against our own finding.

Raw citation counts favour small facilities. A 300-bed home has more residents, more shifts and more opportunities to be cited than a 60-bed home. We publish bed counts beside every figure for this reason.

The abuse icon is a citation history, not a finding of abuse. CMS applies it where inspectors cited a home for abuse at harm level on its most recent survey or a complaint survey in the past twelve months, or at a lower severity across two consecutive survey cycles. It is an agency record, it can be appealed, and it comes off after a clean inspection. It is not a court ruling and it does not establish that any particular resident was harmed.

A home with the icon cannot be rated five stars. CMS caps a flagged home’s health inspection rating at two stars and its overall rating at four. So the absence of five-star homes from any abuse-flag list is arithmetic, not a discovery — and we say so wherever it could be misread.

The deficiencies file only holds about the last three inspection cycles. Older citations age out as facilities are re-surveyed. This is why we publish no multi-year trend lines built from it — a five-year series drawn from a single snapshot of this file is an artefact of how the file is maintained, not a finding. We removed three such charts from this site in August 2026 for exactly that reason.

Everything is a snapshot. CMS refreshes monthly. Every figure we publish carries the date we pulled it. A rating that has changed since is not an error in our reporting; it is the record moving, and we re-pull quarterly.

What we will not publish

A number we cannot point you to the source of. A trend the underlying file cannot support. A characterisation stronger than the record’s own words — we use CMS’s vocabulary, “citation,” “deficiency,” “scope and severity,” and we do not say a home “abused residents” where the record says it was cited. And a statistic whose origin we cannot trace: in August 2026 we removed a widely repeated figure about unreported elder abuse after tracing it to a 2011 study of a single state, about people living at home rather than in care facilities, that had been misattributed to the U.S. Department of Justice across much of the internet.

If you think we have something wrong

Tell us and we will check it against the source file and fix it publicly, with a date, rather than editing quietly. That applies to families, facility operators and regulators alike. The route is on our corrections page.