How we counted it
Every figure in the assessment comes from one public file. None of it is a sample, an estimate or a model. The exact queries are below, they work as printed, and you do not need to ask us for anything to check them.
The source
| Publisher | Centers for Medicare and Medicaid Services |
|---|---|
| Dataset | CMS Nursing Home Health Deficiencies (r5ix-sfxw) |
| File | NH_HealthCitations_Aug2026.csv |
| Published at | https://data.cms.gov/provider-data/dataset/r5ix-sfxw |
| As at | 2026-08-01 |
| Surveys through | 2026-07-23 |
| Refresh | monthly |
| What we count | census of every citation in the file from 2023-08-01 onwards |
The window, and why it starts in August 2023
CMS publishes the last three inspection cycles for standard surveys, but complaint deficiencies appear only for the last three years. In this file the first complaint-flagged citation is dated 2023-08, and the file was processed 2026-08-01, which is exactly three years. So this piece counts citations from 2023-08-01 onwards, the window where a complaint can appear at all. Before that date 88,492 citations are present and every one is marked standard, not complaint. They are not unflagged; they are outside the retention window, and counting them as 'not a complaint' understates the finding.
88,492 citations before 2023-08-01, all marked standard. Including them gives 24.3% rather than 30.8%, so the exclusion moves the headline UP and is stated here for that reason.
This explanation is falsifiable and we would rather it were checked than believed. If it is a retention window, next month's file should start its complaint flags at 2023-09 and the earliest month should drop out. If it is instead a change in what CMS records, the 2023-08 boundary will stay put. We will say which happened.
| Survey year | Citations | On a complaint | Share |
|---|---|---|---|
| 2023 part year | 45,687 | 15,073 | 33.0% |
| 2024 | 121,547 | 35,582 | 29.3% |
| 2025 | 108,049 | 33,563 | 31.1% |
| 2026 part year | 55,704 | 17,773 | 31.9% |
The headline figure
Of 330,987 citations issued from 2023-08-01 onwards, 101,991 (30.8%) were cited on a complaint investigation rather than a standard survey.
Reproduce it with two requests. Read count from each
response. The denominator:
https://data.cms.gov/provider-data/api/1/datastore/query/r5ix-sfxw/0?limit=1&conditions[0][property]=survey_date&conditions[0][value]=2023-08-01&conditions[0][operator]=>=and the numerator, which is the same request with the two flags added:
https://data.cms.gov/provider-data/api/1/datastore/query/r5ix-sfxw/0?limit=1&conditions[0][property]=survey_date&conditions[0][value]=2023-08-01&conditions[0][operator]=>=&conditions[1][property]=standard_deficiency&conditions[1][value]=N&conditions[1][operator]==&conditions[2][property]=complaint_deficiency&conditions[2][value]=Y&conditions[2][operator]==No key, no account, no registration. If you are using
curl, add -g so it does not try to interpret
the square brackets, and quote the URL.
The six measures
Each question in the assessment maps to one federal citation tag. Prevalence is the share of the 14,339 facilities cited at least once in this window. On a complaint is the share of that tag's citations found on a complaint investigation rather than a standard survey.
| Tag | What it covers | Citations | Facilities | Prevalence | On a complaint |
|---|---|---|---|---|---|
F0689 | accident hazards and supervision | 18,123 | 9,079 | 63.3% | 55.5% |
F0684 | care delivered as the doctor ordered | 13,675 | 7,365 | 51.4% | 48.8% |
F0609 | reporting allegations on time | 7,831 | 5,277 | 36.8% | 70.0% |
F0607 | abuse prevention policies and training | 2,390 | 2,022 | 14.1% | 53.6% |
F0580 | telling the resident, their doctor and their family | 5,045 | 3,763 | 26.2% | 61.6% |
F0610 | responding to every allegation | 4,839 | 3,566 | 24.9% | 64.9% |
Between them these six cover 84.9% of the facilities in this window, meaning that share has been cited for at least one of them.
How the score is weighted
Points are the citation prevalence, normalised proportionally to 10 and rounded: 63.3, 51.4, 36.8, 14.1, 26.2, 24.9 becomes 3, 2, 2, 1, 1, 1. Nothing in the scoring is a judgement. All six are measured on the basis stated above, the same basis as every other figure here.
Where the assessment says a failure is the most or second most cited, that is measured across every tag in this window. Where it compares shares between tags, the comparison is restricted to tags cited more than 2,000 times, because a tag with a few dozen citations can show a very high share on almost no evidence.
The state comparison
The assessment ranks states on the same measure as the headline: the share of a state's citations that were found on a complaint investigation rather than a standard survey, over the same window. It is counted in one pass over the same file, so the state figures and the national figure are one read rather than two, and they reconcile exactly: the state totals sum to 330,987, which is the national figure above.
A small state can post an extreme share on very little evidence, so anything with fewer than 500 citations in the window is left out of the ranking. Held back this time: AK (421 citations), GU (51 citations), PR (377 citations). That is 50 of 53 regions ranked. The ones held back are named on the page rather than quietly dropped, because a reader who lives in one would otherwise be told nothing at all.
These are counts of citations, not of homes, and they are state totals only. Nothing here is attributed to an individual facility.
What this does not tell you
- Severity is not counted. Every citation weighs the same here, whether it was a paperwork finding or immediate jeopardy. CMS records a scope and severity code and this piece does not use it.
- Complaint-only does not prove the annual survey missed it. A complaint citation often covers something that had not happened at the last standard survey. That is timing, not a detection failure. The gap is real; the reason for it is an inference and is labelled as one.
- 29,324 citations are flagged both standard and complaint and are excluded from the numerator. Counting every complaint-flagged citation instead gives 36.4%.
- Prevalence is the share of facilities cited at least once since August 2023, not the share of all American nursing homes. Homes never cited do not appear.
- It counts what inspectors cited, not what happened. A failure nobody was cited for is not in the file.
If a figure moves
The headline was 30.8% when this was built. We watch it. If it leaves the range 24.0% to 38.0% we will say so rather than quietly changing the page.
Computed 2026-09-02 on the 2023-08-01 basis. Only two complete years are available: 2024 at 29.3% and 2025 at 31.1%. 2023 is partial because the retention window opens in August, and 2026 is partial because the year is not over. A band from two points is thin, so it is set well outside them at 24 to 38 rather than tuned tightly to a short history. It is deliberately wide, and it is a property of this source only. Never carry it to another deal.
Questions about any of this go to support@precisionleadsmedia.com.