How we rank providers

For assisted living, we rank a facility against other facilities in the same state and of a similar size, using the enforcement record that state publishes. Band 1 holds the facilities with the fewest findings; band 5 holds those with the most.

Every listing states the direction in words as well as the number, and names the group it was compared against, because "band 4 of 5" on its own is a stronger claim than the data supports.

Not every finding is equal, so they are weighted. A serious citation counts ten times a minor one, and a substantiated complaint counts five times.

"Serious" means the state's own most serious tier: California's Type A, which is an immediate risk to health, safety or personal rights, and Florida's Class 1 and 2, which are imminent danger and direct threat. A substantiated complaint is one an inspector independently verified, which is a different thing from a citation and is counted separately.

This is a deliberate simplification of the federal scope-and-severity grid, which runs from 4 to 150. We would rather use a scale every state can actually support than a finer one we would have to invent numbers for.

States publish counts covering a facility's entire life, so without adjustment a well-run thirty-year-old home looks worse than a poor two-year-old one purely for having existed longer. Findings are therefore counted per year licensed, capped at twenty-five years because published history further back is not reliably complete.

Size is handled by only ever comparing facilities of similar size. This matters more than it sounds: in Florida, substantiated complaints average 0.23 at homes with six beds or fewer and 5.64 at communities with over 120 beds. That is a twenty-four-fold spread driven by size rather than by quality. A six-bed residential home and a 200-bed community are different businesses, inspected differently, and we never rank them against each other.

States do not publish comparable things. Florida grades deficiencies in four classes; California uses Type A and Type B. California publishes how many times an inspector visited; Florida does not. A national ranking built on top of those would be comparing numbers that do not mean the same thing.

Medicare cuts its own nursing home inspection ratings within each state for the same reason. So a band tells you where a facility sits among its neighbours, and it will not tell you whether one state is better than another.

Between 4% and 40% of a peer group has a spotless record, and therefore an identical score. Splitting that block across two bands would give two facilities with identical clean records different published rankings for no reason at all, so identical records always receive identical bands. This means the bands are not five equal fifths, which is the correct outcome rather than a compromise.

Most providers in the directory carry no ranking, and it never counts against them. A listing that says the state publishes no inspection data means exactly that.

• Skilled nursing and home health are not ranked by us at all. Medicare already publishes its own risk-adjusted star ratings, and we show those instead of inventing a second opinion.
• Florida home care is not ranked, because the state publishes no enforcement record for it.
• California home care is not ranked, because there is no published measure of size to compare agencies fairly against.
• Subsidised senior housing is not ranked, because HUD publishes a building condition score, and the condition of a building is not a rating of the care inside it.
• Any group with fewer than 25 comparable facilities is left unranked, because a fifth of a group that small is meaningless.

It is not a safety rating. It is not a prediction of what will happen to your parent. It is not comparable across states. It does not measure kindness, food, staff turnover, or whether anyone is happy there, none of which any regulator publishes.

It is one thing only: where a facility sits among its own state's similar-sized peers on that state's published enforcement record. A facility can be fully licensed and in good standing with its regulator and still sit in band 4 because its neighbours happened to have cleaner records.

We do not use Google or Yelp ratings. Those track marketing spend and provider size, which is the exact bias this method exists to remove, and their terms forbid using their content to build a directory like this one.

Where a patient experience measure exists, we use Medicare's own survey and always show how many people answered it. Medicare publishes percentages backed by as few as one response, and a percentage without its sample size is not a fact a family should act on.

Tell us and we will look. Every number here comes from your regulator's published file rather than from us, so the usual answer is that the state's record needs correcting at the source, and we will point you to which file and which field it came from. Where we have made the mistake, we will fix it.

Get in touch.

Every set of results says when we collected each source and, where the publisher stamps its own date, when they last published it. Those are different questions and we report both: our copy can be behind, or the state itself can be. New York's adult care file dates from March 2022, and no amount of re-downloading on our part makes it newer.

Most sources refresh monthly and automatically. Florida is collected by hand, because its state search tool cannot be read any other way, so it is the one most likely to lag. The results page says so rather than leaving you to guess.