What a Nursing Home's Five-Star Rating Is Made Of

Two nursing homes in Texas, both showing three stars on Care Compare in the July 2026 refresh.

Home A was inspected on 12 June 2025 and cited for nothing at all. It also turned over 95.6 percent of its nursing staff and 90.9 percent of its RNs in a year. Home B is a 44-bed nonprofit that picked up five deficiencies on its September 2025 survey, two of them arising from complaints, and staffs at nearly five nursing hours per resident per day with sixteen-percent RN turnover.

Same three stars. Almost nothing else in common.

That is not a glitch. It is arithmetic, and the arithmetic is published.

The overall rating is addition, and no domain has a weight

CMS assigns the overall star in three steps, set out on page 24 of the Five-Star Quality Rating System Technical Users' Guide, July 2026 edition, which I downloaded and read on 22 August 2026:

1. Start with the health inspection rating.
2. Add one star if the staffing rating is five; subtract one if it is one.
3. Add one star if the quality measure rating is five; subtract one if it is one.

Nothing else happens. The result cannot go above five or below one, and there is one extra brake: a home whose inspection rating is one star cannot be lifted by more than a single star no matter how good the other two domains look. The guide says why in plain terms — staffing and quality measures "should not significantly outweigh the rating from actual onsite visits from trained surveyors."

So Home A ran 5, minus one for staffing, minus one for quality measures. Home B ran 2, plus one, plus nothing. Both land on three.

Count the possibilities and you get the size of the problem. In the July 2026 Provider Information file — 14,693 homes, processing date 1 July 2026 — 34 distinct combinations of the three domain ratings produce an overall rating of three stars. Nine homes reach three stars from a 5-1-1 split. Sixty-five reach it from 2-5-4. The overall star tells you the sum. It cannot tell you the terms.

The inspection domain still dominates. Overall and health inspection are identical for 8,623 of the 14,561 rated homes, 59.2 percent. But 163 homes were lifted two full stars by staffing and quality measures, and 105 hold a five-star overall on a three-star inspection record.

Health inspections: a curve drawn inside your state

This domain produces a score first and a star second, and the two are worth separating.

The score covers the two most recent recertification surveys, three years of complaint investigations, three years of focused infection control surveys, and revisits. Each deficiency earns points by scope and severity — a D is 4 points, a G is 20, an L is 150 — which is the scope-severity grid printed on the inspection report itself. The most recent survey cycle carries a weight of three quarters; the one before it, one quarter.

Both Texas homes show their working in the downloadable file:

• Home A: cycle 1 score 0, cycle 2 score 32. (0 × 0.75) + (32 × 0.25) = 8.0
• Home B: cycle 1 score 166, cycle 2 score 76. (166 × 0.75) + (76 × 0.25) = 143.5

Then the star is assigned by rank within the state. Top 10 percent by lowest score, five stars. Bottom 20 percent, one star. The middle 70 percent split three ways. CMS recalibrates the cut points monthly, and states the reason openly: survey management, state licensure law and Medicaid policy vary enough that a national comparison would be measuring the state as much as the home.

Which means the same score buys different stars depending on the border you are on. From the July 2026 file, the worst weighted score that still earned five stars:

Home A's 8.0 is a five-star score in Texas. In Alabama, South Carolina, Iowa, Kentucky or North Carolina it is not. Put two homes from different states side by side and the health inspection stars are not one measurement taken twice. They are two separate rankings that happen to use the same symbol.

Three more things the inspection star quietly excludes. Findings from the third-oldest survey cycle are displayed but not scored. Citations under informal dispute resolution are displayed but held out of the score until the dispute closes. And homes cited for harm-level abuse get an icon, an inspection rating capped at two stars, and therefore an overall capped at four — 1,435 homes carry that icon in this file, 21 of them with four stars overall.

Staffing: six measures, 380 points, three ways to hit the floor

The staffing star is a point total, not a ratio. Six measures, drawn from the Payroll-Based Journal system and a daily census computed from MDS assessments. Three are hours per resident per day, case-mix adjusted using PDPM nursing groups. Three are turnover, built from six consecutive quarters of payroll data by looking for 90-day gaps in days worked rather than trusting a termination date the home reports.

Points, per the guide's Appendix Table A2, run in deciles. Here is the tally for the two Texas homes:

Under 155 is one star. 320 and up is five. Both land exactly where Care Compare shows them.

Notice what the turnover measures did to Home A. Its hour figures are mediocre rather than alarming — the national median in this file is 3.770 total nursing hours and 0.586 RN hours per resident day. It is the 95.6 percent and the 90.9 percent that cost it 90 of the 100 available turnover points and pushed the total under the one-star line. Median nurse turnover nationally is 44.9 percent.

Three scoring exceptions override all of the above. A home that submits no staffing data by the deadline gets one star. A home reporting four or more days in the quarter with zero RN hours while residents were present gets one star. A home that fails a PBJ audit, or ignores the audit request, gets one star for three months. Those are not bad performance translated into a rating. They are a rating imposed in place of data.

Quality measures: ten the facility codes, five from claims

Fifteen measures — nine long-stay, six short-stay. Ten come from the Minimum Data Set, the resident assessment that the home's own clinical staff complete. Five come from Medicare claims: hospitalizations and outpatient ED visits among long-stay residents, rehospitalization and ED visits within 30 days for short-stay residents, and the rate of successful return home from a SNF. As of January 2026 the long-stay antipsychotic measure was respecified to draw on Medicare and Medicaid claims and encounter data as well as the MDS.

That split matters more than the measure names do. Whether a resident was readmitted to a hospital within 30 days is settled by a claim somebody else filed. Whether a resident has a stage II pressure ulcer is settled by the facility's own assessment.

CMS knows this. The guide contains a scoring exception for MDS schizophrenia coding audits: a home found to have coded inaccurately has its overall and long-stay quality measure ratings dropped to one star for six months, its short-stay rating suppressed for six, and its long-stay antipsychotic measure suppressed for twelve. A rating system does not build a penalty like that for a data source it considers reliable.

Two more mechanics are worth knowing before you read a quality measure star.

Small homes get state averages folded in. A measure needs 20 assessments across four quarters. A home with only 12 has the remaining eight imputed at the statewide average, and those imputed values feed the points even though the measure itself is then withheld from public display. Part of the star you can see is built from data you cannot.

The thresholds are absolute and rising, not a curve. Unlike the inspection domain, this one uses fixed point ranges — as of July 2026, 1,494 or more out of a possible 2,300 for five stars. CMS raises those thresholds twice a year by half the average improvement in scores. The result is a lopsided distribution: 4,233 homes hold a five-star quality measure rating and only 971 hold one star. A five-star quality measure rating is common. A five-star staffing rating, held by 1,888 homes, is not.

What none of the three domains look at

CMS prints its own warning at the top of the Five-Star Quality Rating System page: "No rating system can address all of the important considerations that go into a decision about which nursing home may be best for a particular person."

Specifics, from the guide and from the file:

• Enforcement. Fines, payment denials and penalties are published and never scored. Among the 3,006 homes rated five stars overall, 414 have at least one fine, together about $9.19 million, one of them $297,367 against a single home. Sixty-one have a payment denial.
• Fire and building safety. Life Safety Code deficiencies are excluded from the rating calculations. So are two health tags, F731 and F884.
• Complaints investigated and not substantiated. Only cited deficiencies carry points. The number of complaints filed against a home appears nowhere in the score.
• What residents and families think. There is no experience survey in any of the three domains. Not one of the fifteen quality measures asks a resident anything.
• Price. Nothing in the rating reflects the daily rate, the care-level surcharge structure, or the annual increase a contract permits.
• The homes with the worst records. The 88 Special Focus Facilities in this file receive no overall rating and no rating in any of the three domains; the guide says a yellow warning sign is displayed where the stars would be. A further 440 rows carry the value SFF Candidate in the file's Special Focus Status column. That is a flag in the data, which is not the same claim as a label on the page.
• Whether the inspection is recent. For 1,030 homes, the file flags the most recent health inspection as more than two years old. The star still sits there, built on it.
• Assisted living. Entirely. No federal survey, no Form 2567, no Care Compare page, no star, because assisted living is licensed state by state and the residency agreement carries the weight the federal regulation carries here.

Four numbers per home, and the subtraction that shows which one moved

Open Care Compare and do this for each home on your list.

Write down four numbers, not one: overall, health inspection, staffing, quality measures. Then run the subtraction backwards. If overall equals health inspection, neither of the other two domains was at an extreme. If overall is higher, a five-star staffing or quality measure rating did it, and you can see which. If it is lower, a one-star domain did.

Compare inspection stars only within one state. Across state lines, use the raw score instead. It is not on the Care Compare page; it is in the Provider Information file at data.cms.gov/provider-data, column "Total Weighted Health Survey Score," alongside each cycle's date, deficiency count and score. Home A at 8.0 against Home B at 143.5 is a comparison the stars cannot express.

Read the staffing numbers, not the staffing star. Adjusted RN hours, adjusted weekend total hours and total nurse turnover are all on the staffing tab. National medians in the July 2026 file are 0.586 RN hours, 3.770 total hours and 44.9 percent turnover. A home at 0.45 RN hours and 95.6 percent turnover has told you something that the phrase "three stars" did not.

Check the survey date before anything else. If the most recent standard survey is more than two years old, every inspection figure on the page is describing a building that may have changed hands, administrators and staff since.

Open the penalties tab separately. It is not in the star.

Then pull the actual inspection reports for the two most recent cycles and read the F-tags and scope-severity letters yourself. Points are how CMS compresses a survey into a number. The narrative in the middle column is what the surveyor actually saw, and it is where you find out whether the transfer and discharge protections at 42 CFR 483.15 have ever been cited at this address — the same rules behind a fast appeal when a Medicare-covered stay is ending.

Sources

Retrieved and read on 22 August 2026:

• CMS, Design for Care Compare Nursing Home Five-Star Quality Rating System: Technical Users' Guide, July 2026 edition, 34 pages. The three-step overall algorithm and the one-star cap are on page 24; health inspection scoring and the three-quarter and one-quarter cycle weights on pages 4 to 7; the abuse icon rules on page 8; the six staffing measures, scoring exceptions and Table 3 on pages 8 to 16; the quality measure list, imputation rules, schizophrenia audit exception and Table 5 on pages 17 to 23; staffing decile cut points in Appendix Table A2. Downloaded with a browser user agent, since cms.gov returns 403 to plain automated requests.
• CMS, Five-Star Quality Rating System, source of the quoted "Caution" paragraph and of the state-level cut point tables that accompany the guide.
• CMS provider data catalog, Provider Information, file NH_ProviderInfo_Jul2026.csv, dataset last modified 29 July 2026, released 6 August 2026, processing date 1 July 2026. Every count and facility figure above is my own tabulation from that file: 14,693 homes, 14,561 with an overall rating, 34 domain combinations producing three stars, 8,623 homes where overall equals health inspection, 1,435 abuse icons, 88 Special Focus Facilities and 440 candidates, 1,030 inspections older than two years. The two Texas homes are rows in that file. They are not named here because the point is the arithmetic rather than the two buildings, but both rows can be pulled back out of the file: Home A is the only Texas row with an overall rating of 3, a health inspection rating of 5 and Collin in the county column, and Home B is the only Texas row with an overall rating of 3, a health inspection rating of 2 and a staffing rating of 5.
• CMS, Payroll-Based Journal staffing data submission, for the quarterly submission deadline of 45 days after the end of each reporting period.
• Medicare, Care Compare, the consumer-facing display of all of the above.

This page is general information, not legal, medical, or financial advice. See the terms.