How we read the data

Every number on this site comes from a public federal file. This page says which file, how we calculate what we calculate, and where the data is weaker than it looks — so you can check us.

Last reviewed 9 August 2026.

Where the data comes from

All facility information is published by the Centers for Medicare & Medicaid Services (CMS), the federal agency that certifies and inspects nursing homes. We download the current files directly from CMS rather than copying another website, and we resolve the file locations through CMS's own catalogue each time, because the download URLs rotate with each release.

We use these CMS datasets:

• Provider Information — address, bed count, ownership type, star ratings, staffing hours, and staff turnover.
• Health Deficiencies — every inspection citation, its severity code, whether a complaint prompted it, and its correction status.
• Fire Safety Deficiencies — life-safety-code citations.
• Penalties — fines and payment denials.
• Ownership — disclosed owners, officers, managers and other parties CMS requires facilities to declare.
• Survey Summary and Inspection Dates — when inspections happened.
• Quality Measures — both the resident-assessment (MDS) and Medicare-claims-based measures.
• State and National Averages — the comparison baselines.

Distances use the 2025 U.S. Census ZIP Code Tabulation Area Gazetteer, a separate federal file that gives each ZIP area a representative point.

CMS republishes most of these files monthly. We rebuild the whole site from a fresh download rather than patching individual pages, so nothing on the site can drift out of step with its source.

How each figure is worked out

Weekend nurse coverage

This is the measure you will not find on most comparison sites, and it is the reason this one exists. CMS publishes registered-nurse hours per resident per day twice: an overall figure and a weekend figure, both derived from payroll data rather than from what a facility says about itself.

We express the difference as a percentage of the weekday level. A facility at 0.80 hours on weekdays and 0.40 at weekends has a 50% weekend drop.

We originally intended to flag any drop over 30%. Then we measured the real distribution: 56% of facilities drop more than 30%, and the median drop is 33%. A warning that fires on more than half of all facilities tells you nothing about any of them. So the callout threshold is a drop greater than 50%, which is roughly the worst sixth. Smaller drops are still shown in the numbers and described in the text — they are just not elevated into a warning.

We do not compute this where CMS reports no weekend figure, or where the weekday figure is so low that a percentage would be unstable. Just over 3% of facilities have no usable comparison, and their pages say so rather than guessing.

Inspection findings

We count citations from the last three years and group them by the severity inspectors assigned, because a raw total is close to meaningless — a records-keeping slip and a resident being harmed each count as one citation. CMS uses a letter scale that combines how serious the problem was with how many residents it affected:

• J, K, L — immediate jeopardy. Inspectors judged that residents were placed in danger of serious injury or death.
• G, H, I — actual harm, without immediate danger to life.
• D, E, F — potential for harm. No one harmed yet, but conditions could allow it. This is the most common band by a wide margin.
• A, B, C — no harm. Rule and paperwork findings.

We separately count citations that came from someone filing a complaint rather than from a routine inspection, because a complaint means a person picked up the phone about something.

Whether findings were corrected

Most citations are corrected, and a page that implies otherwise is unfair. CMS records a correction status in words rather than as a yes/no, and we map each one explicitly: a recorded correction date, past non-compliance, no revisit needed, or a granted waiver all count as resolved. We describe a finding as unresolved only where CMS records a correction plan that has not yet been confirmed, or no plan at all. Across the whole country that is about 1% of citations.

If CMS publishes a status we have not seen before, our build stops rather than guessing. We would rather ship late than publish a false statement about a named business.

Staffing and turnover

Staffing hours come from payroll-based journal data, not self-reporting. Turnover is the share of nursing staff who left over the measurement period. We show the state average beside both, because the raw number means little without it — what counts as high turnover in one state is ordinary in another.

Ownership and affiliations

CMS's ownership file is broader than its name suggests. Of roughly 248,000 disclosed relationships nationally, only about a tenth are actual ownership interests. The rest are corporate officers, directors, managing employees, trustees, lenders holding a mortgage or security interest, and entities CMS calls "additional disclosable parties."

So we do not say anyone "owns" a facility unless CMS's own role field says so. Each page lists disclosed parties with the exact role CMS recorded, and shows an ownership percentage only where CMS published one — for most relationships it publishes none.

Distance and nearby facilities

Distances are straight-line, calculated from the Census representative point of a ZIP area to a facility's published coordinates. They are not driving distances and will understate a real journey, sometimes badly where a river, a bay, or a motorway junction is in the way. Treat them as a way to find candidates, then check the actual travel time.

What we deliberately don't do

We report what the record says and name the comparison whenever a figure is relative. We do not rank facilities against each other, score them, or describe any facility as good, bad, safe or unsafe. Those are conclusions, and a conclusion about a named business drawn from government data is both the least defensible thing we could publish and the least useful thing for you to read. What helps you is the underlying fact and the question it suggests you ask.

We also don't accept payment from facilities, take referral fees, or let anyone pay to change, improve or remove a page.

Where this data is weak

Being straight about the limits is part of the method:

• Different measures cover different periods. Some CMS measures run close to a year behind others, which is why we date each measure separately instead of putting one "last updated" stamp on a page.
• Inspection frequency varies. A facility inspected recently may show more citations than a similar one not visited for two years. More findings does not always mean more problems.
• Star ratings are partly self-referential. The overall rating is built from the health inspection, staffing and quality-measure ratings, so it is not independent evidence on top of them.
• Quality measures are risk-adjusted by CMS, and the adjustment is imperfect for facilities with unusual resident populations.
• Records lag reality. A facility that changed hands or replaced its management last month looks exactly like the one that didn't.

None of this makes the data useless. It makes it a starting point for questions rather than an answer, which is how every facility page is written.

Corrections

If a figure here is wrong, we want to know — including from people who work at the facility. See how we handle corrections, or email corrections@wheretheycare.com.