Do CMS star ratings affect nursing home occupancy?

Key Findings

Across the June 2026 CMS data, average occupancy at one-star US skilled nursing facilities is 77.0%, and at five-star facilities it is 81.8%.

If the CMS Five-Star Quality Rating System were the primary driver of where post-acute patients go, occupancy would climb steadily as you move up the scale. It does not. Fewer than five percentage points separate the bottom of the scale from the top, and the correlation across all 14,396 rated facilities is r = 0.109, which squares to roughly 1% of explained variance.

For a 100-bed skilled nursing facility, moving from one star to five stars is associated with about five additional residents on an average day, and that association assumes the rating caused the difference rather than simply traveling alongside it.

The first reasonable objection to a national correlation this weak is that it might be an artifact of geography. States differ enormously in bed supply, payer mix and rating distribution, so a national figure could be washing out a relationship that shows up clearly inside any single market.

Measured within individual states, the average correlation between star rating and occupancy is 0.085, slightly weaker than the national figure of 0.109.

We re-ran the analysis state by state, limited to states with more than thirty rated facilities. The relationship did not strengthen. Whatever is filling skilled nursing beds is not showing up in the star rating at the market level either.

A note on the size of the claim. This is not zero. The correlation is real, it is positive, and a higher CMS rating is better than a lower one for reasons that extend well beyond census. The finding is that star rating is a very small part of the occupancy picture, not that it is irrelevant.

Figure 2

Each point is one skilled nursing facility, with horizontal jitter added within each rating for legibility. n = 14,396. The fitted line is nearly flat, and facilities at every rating operate at nearly every level of occupancy.

A study that only reports what does not matter is of limited use to anyone running a building, so we tested the other variables available in the CMS files. The strongest single relationship we found does not appear on the report card at all.

Nursing staff turnover correlates with occupancy at r = −0.188, nearly twice the strength of the star rating relationship.

US skilled nursing facilities keeping annual nursing turnover under 30% average 83.3% occupancy. Facilities above 60% turnover average 75.6%. That is a spread of nearly eight percentage points, against fewer than five across the entire five-star scale.

The inverse is just as telling. In the same release, 739 one-star facilities are operating at 90% occupancy or better, which is difficult to reconcile with the idea that referral sources screen strictly by rating.

This is the finding that tends to land hardest with operators, because most people in this industry can name one of these buildings without looking anything up. They are well run, they are clinically strong, they have the survey history to prove it, and they still have empty beds. Whatever is keeping those beds empty is not a quality problem, and no amount of further clinical improvement is going to solve it.

The CMS rating matters. It matters to families comparing options, it matters to the hospital case manager scanning a discharge list, and it matters for reasons that have nothing to do with census. But it is a small input into a decision made using a great deal of other information, most of it gathered before anyone visits the building.

For an administrator sitting at two or three stars and working hard to move up, that should come as some relief, because it means the census problem is not gated behind a number that takes years to change. And for the 762 facilities sitting at five stars with empty beds, it means the answer is somewhere else entirely, which is worth knowing before another quarter goes by.

Our own view, formed over a decade of doing this work inside a skilled nursing operating company, is that most of the gap sits in the space between the referral and the bed. How quickly the phone gets answered. What a family finds when they search the facility at eleven at night. Whether anyone is walking into the referring hospital often enough to be remembered. None of that appears on Care Compare, and all of it is fixable in months rather than years.

CMS Provider Data Catalog, Nursing Home Provider Information file, June 2026 release, merged with the payroll-based journal staffing file from the same release.

All 14,695 Medicare and Medicaid certified nursing facilities in the file. Analyses involving star rating use the 14,396 facilities carrying an overall star rating and a usable derived occupancy figure. Analyses involving turnover use the 13,249 facilities that additionally report nursing staff turnover.

CMS does not publish an occupancy figure directly. We derived it as average number of residents per day divided by number of certified beds. Facilities missing either field, or with a derived occupancy above 100%, were excluded.

Pearson correlation coefficients throughout, with Spearman run as a check against rank effects. The within-state analysis was limited to states with more than thirty rated facilities.

This is observational data and none of these relationships establish causation. Certified bed counts do not always reflect beds actually in service, which will overstate vacancy at some facilities. Occupancy and turnover are reported on different cadences. The analysis covers only variables present in the CMS files, which exclude payer mix, local bed supply, referral volume and every other market condition that plausibly matters a great deal.

Every figure here comes from public files anyone can download from the CMS Provider Data Catalog. If you want to check the work, or want the numbers for your own state, get in touch and we will send them over.