Methodology

Sources

Three files from the CMS Provider Data Catalog, joined on CMS Certification Number and nothing else. No fuzzy matching on names or addresses is used anywhere: attributing one facility’s record to another would be defamatory in effect even though every underlying fact is public and true.

Refreshed monthly. CMS publishes on a rolling three-year window, so a citation older than three years leaves the file and therefore leaves this site.

Two tag families, one file

The Fire Safety Deficiencies file carries two namespaces: K-tags (the Life Safety Code, Form CMS-2786R) and E-tags(Emergency Preparedness, 42 CFR §483.73, Appendix Z). A facility’s citation total counts both, because a surveyor writes both on the same visit; its page and the Emergency Preparedness Index break them out. The two are never merged into one tag list, because they come from different rules.

A third family, from a second file

A nursing home is also surveyed by health surveyors under Appendix PP, and their findings are F-tags, published by CMS in the separate Health Deficiencies file. The Physical Environment Index takes only the thirteen F-tags that concern the building and its equipment — essential equipment in safe operating condition, the call system, bed frames and rails, the environment, ventilation, handrails, pest control, smoking policy, hazards, water supply, emergency power for lighting and alarms, and infection control where it is water management — and ignores the rest of the health survey. Those citations are shown on a facility’s page and counted on their own. They are never added to the Life Safety total: the medians, the position within a state, the survey window and the clean-record badge are the fire-safety record only, because they come from a different survey with a different cadence. Citation alerts likewise cover the Life Safety survey.

Population

Medians and percentiles are computed over every Medicare-certified nursing home in a state, including those with no citations on file. Restricting the comparison to cited facilities only would drop the best performers out of it and flatter everyone who remained — and a percentile over a filtered population is not a percentile.

Measures

Position within state

The share of certified facilities in the same state with strictly fewer citations over the three-year window. Ties are handled by rank, not by ordinal position.

Repeat citations

A tag counted once per facility when it appears at two or more separate surveys. Two citations of the same tag on the same day are one finding, not a pattern. Nationally 10,096 of 13,911 cited facilities — 73% — have at least one; the median such facility repeats three tags.

Trend

Citations at the most recent Life Safety survey against the mean of earlier surveys in the window, requiring at least two surveys. Reported with both underlying numbers because at two or three surveys this is a direction of travel and not a conclusion.

Severity — measured, and deliberately not weighted

95.2% of all life-safety citations sit in a single severity band (D–F), and findings of actual harm or immediate jeopardy together account for 0.10%. Severity differentiates hospital safety data; here it barely moves. Weighting by it would add a defensible-looking term that does almost no work, so the site reports severity where CMS records it and does not weight by it.

Why the comparison is within-state

This is the finding the whole design rests on, so the numbers are given in full. Across the lower 48 (n = 14,600), a one-way analysis of citation totals by state gives η² = 0.301 — which state a facility is in explains about 30% of the variance in how many citations it has. Certified bed count explains 0.6% (r²).

State medians run from 1 in Mississippi and South Carolina to 31 in Washington, against a national median of 11. A thirtyfold spread is not a thirtyfold difference in how those buildings are run — it is a difference in inspection cadence, surveyor headcount and how aggressively findings are written up. A national ranking would tell a well-run Washington facility it sits in the bottom decile.

Because bed count explains so little, no size adjustment is applied. Every additional adjustment is one more thing to defend, and the simplest defensible model is the strongest.

Ownership

Who operates a facility now comes from the CMS provider file (the chain name and ownership type). Changes of ownership come from CMS’s separate Change of Ownership dataset, which records sales, acquisitions and consolidations of certified nursing homes since January 2016 as dated events between organisations. A facility page lists its events, newest first, and calls one New ownership only while it is recent — within the last two years — because that is a fact with a shelf life. A facility with no event on record has had none since the records begin, and the page says exactly that. The companion CMS file that names individual owners is not used anywhere on this site.

Survey windows

Each facility page states a next-survey window: the range after its last Life Safety survey in which facilities in its state are typically surveyed again. It is the one thing about the future the public record supports, and it is measured, not assumed. CMS’s rule of thumb is nine to fifteen months; in the records the national median gap is closer to fifteen, and states differ from about eleven months to three years.

The interval is computed over consecutive Fire Safety Standard surveys at the same facility. Gaps under six months are excluded (a revisit recorded as a standard survey is not a cycle). Only surveys in the two years before the snapshot count: the pandemic backlog pushed the national median gap to 33 months in 2022 and it has since returned to about 14, so an all-time figure would overstate today’s cadence. A state with fewer than 50 such gaps is too thin to characterise and uses the national figure, which the page says.

The window opens at the state’s 25th-percentile gap after the last survey and closes at the 75th; by the 90th, nine in ten facilities in the state have been surveyed. A facility beyond that is described as past the typical window, which is what the record says, not “overdue”, which is a judgment. Nothing here predicts a date.

The clean-record badge

A facility with nothing on file can place a badge on its own website that links back to its page here. It is offered only where the record earns it: no citations— Life Safety K-tags or emergency-preparedness E-tags — across CMS’s rolling three-year window, at least two Life Safety surveys in that window, and the most recent survey within the last 24 months. About 690 of 14,690 facilities meet it today.

The image is drawn from the database on request and cached for a week, so it follows the monthly file. A facility that is later cited does not get a “cited” badge on its own site; the same image quietly becomes a neutral “Life Safety record” mark that claims nothing. The badge says what the CMS file says and cites it. It is not a TagProof endorsement, and whether a facility is a customer plays no part.

What this does not do

• No grades, letters or stars. A grade is a judgement that can be disputed; a percentile is arithmetic that can be checked.
• No rankings or league tables.A facility can see its own position. There is no “worst facilities” list and there will not be one.
• A facility’s position never depends on whether it is a TagProof customer. Every figure is computed from public CMS files by the methodology on this page. Nothing about a commercial relationship enters the calculation, and no customer data is used here.
• Nothing is shown to anyone that a facility cannot see about itself, free. There is no private tier and no score withheld from the operator it describes.
• No fines, penalties or named individuals. CMS publishes those; this site does not republish them.

Corrections

CMS’s own dispute flags are effectively unused — 36 citations of 200,327 carry one — so their process cannot be relied on to catch an error here. Our correction route is documented separately and answered.

Versioning

This is version 1.0. If a definition changes, the version changes with it and the previous definition stays published — a figure that moves because we changed the rules should be distinguishable from one that moved because a facility did.