Quality Measures, decoded
Most SNF underwriters look at the 1-5 star QM rating and stop there. The data underneath that rating is publicly available from CMS — and it answers the questions a star number can't. This is a short, no-jargon walkthrough of what we surface, what each number means, and what the gaps tell you.
What the QM Star Rating actually measures
The CMS overall 5-star rating for a nursing home is a blend of three sub-ratings: Health Inspections, Staffing, and Quality Measures (QM). The QM sub-rating is itself a roll-up of about 17 individual clinical measures — things like the percentage of long-stay residents who lost too much weight, who experienced a fall with major injury, or who were rehospitalized within 30 days of admission.
Each measure is published with its raw number (e.g. "6.5% of long-stay residents had a fall with major injury this year"). CMS turns those numbers into the 1-5 star QM rating using percentile thresholds within each measure. The star rating is the headline; the underlying measures are where the actual diagnostic signal lives.
Long-stay and short-stay are rated separately
CMS publishes three QM star ratings, not one: the overall rating plus a long-stay and a short-stay rating. They describe different businesses inside the same building. The long-stay rating scores custodial residents over months — decline in daily activities, decline in walking, pressure ulcers, falls with major injury, urinary tract infection, catheter use, and antipsychotic use — plus hospitalizations and outpatient emergency-department visits per 1,000 long-stay resident days. The short-stay rating scores post-acute rehab stays over days to weeks — rehospitalization within 30 days of admission, outpatient emergency-department visits, newly-received antipsychotics, and the SNF Quality Reporting Program function measures. Note that both stay types carry hospital-utilization measures, and that they are built differently: short-stay as a percentage of stays, long-stay as a rate per 1,000 resident days. A building can run a strong rehab operation and a weak custodial one, or the reverse, and the overall rating hides that distinction.
Not every measure we publish feeds a rating. CMS flags which ones count ("used in quality measure five star rating"), and several widely-cited measures are excluded — long-stay weight loss, for instance, is reported but does not affect any star rating.
The overall rating is not the average of the two. CMS scores every measure into points, sums the points, and cuts the total into stars — so a facility can sit above both of its own component ratings, or below both, depending on which measures carry weight. As of the June 2026 publication the national means were 3.83 long-stay, 3.26 short-stay and 3.66 overall — each computed over the facilities CMS rated on that component (14,149, 11,848 and 14,501 respectively), so the short-stay mean systematically excludes the low-volume facilities described below. Those cohorts do not subtract cleanly: 11 facilities carry a short-stay rating without an overall QM rating, so 14,501 minus 11,848 is not the number of rated facilities missing a short-stay score.
Why the short-stay rating is often missing
Of the 14,501 facilities CMS gave an overall QM rating in the June 2026 publication, 2,664 have no short-stay rating. CMS gave a reason for each: 2,603 were marked "not enough data available to calculate a star rating" and the other 61 "the accuracy of the data for this rating could not be validated by CMS" — two different problems, and we keep them distinct rather than collapsing both into "no data". A further 352 facilities are missing the long-stay rating, all of them for insufficient data. Where CMS withholds a rating we show the reason it gave rather than a bare blank, and any average we publish across a portfolio or chain covers only the facilities CMS actually rated — a smaller group than the overall QM average, which is why we show that count alongside it.
One practical consequence for a facility with no short-stay rating: its overall QM rating rests on long-stay measures. Read it as a custodial score rather than a whole-building one — particularly on a building whose rehab census you are underwriting.
The two datasets behind it
CMS publishes the QM detail in two separate files:
~17 clinical measures sourced from the MDS (Minimum Data Set) — the standardized resident assessments every nursing home submits to CMS. These measure clinical processes and resident-level outcomes: ADL decline, falls, pressure ulcers, depressive symptoms, restraint use, vaccinations. Reported as quarterly scores plus a 4-quarter average.
4 outcome measures sourced from Medicare claims — what actually happened to residents (rehospitalization, ER visits, etc.) regardless of what the SNF reported. These are risk-adjusted, meaning each facility's rate is compared against what would be expected given its specific patient mix.
Observed vs Expected vs Risk-Adjusted
The claims-based outcome measures are reported as three numbers per facility. Most platforms only show one. The other two are where the underwriting signal actually lives.
The raw rate at this facility. Just numerator / denominator. "X of Y short-stay residents went back to the hospital within 30 days." Doesn't account for who the patients are.
What CMS predicted for this facility's specific patient mix, using a risk-adjustment model that weights age, prior hospitalizations, comorbidities (diabetes, COPD, heart failure, dementia, etc.), and acuity on admission. A facility taking sicker post-stroke patients gets a higher Expected than one taking healthier post-knee-replacement patients — because that patient population is statistically more likely to bounce back to the hospital regardless of facility quality.
The standardized rate that lets you compare facilities apples-to-apples. Roughly: (Observed / Expected) × national-average-rate. It answers "what would this facility's rate look like if it took the national-average patient mix?" Use this number for cross-facility comparison.
Five patterns underwriters should recognize
The signal is in the gap, not any single number. SNF Shark auto-classifies each measure into one of these patterns and surfaces a one-line takeaway under the chart so you don't have to do the math.
Why some scores are blank (footnote codes)
When CMS suppresses a score (you'll see no number, just an explanatory note), it's almost always because the facility doesn't have enough residents in the measure's denominator for a reliable estimate. The most common codes:
| Code | Meaning |
|---|---|
| 9 | Insufficient claims data for a reliable score (typically: facility has too few short-stay admissions to compute the rehospitalization or ER-visit rate) |
| 10 | Too few resident assessments to report |
| 11 | Too few stays to report |
| 21 | Score not reported — insufficient data on file |
| 5 | Newly certified facility — not yet enough data |
Suppressed measures aren't errors or red flags — they just mean the facility's case volume on that specific measure is too low to publish a reliable estimate. SNF Shark hides suppressed rows from the main table and surfaces a count at the bottom ("+ 4 measures suppressed by CMS") so you know the data exists for some facilities but not this one.
Trending a measure over time
The Quality Measures Trend chart plots a single measure quarter by quarter against its state, county, and national averages — the same idea as the Star Ratings trend. The same series is also published as a table: on a facility page as Quality Measures by Quarter (every measure down, quarters across), and on a portfolio's Clinical tab as the same grid with one row per facility and a dropdown to choose the measure. Chart and tables read the same underlying per-quarter rows, so they cannot disagree. Three things are worth knowing about how the series is built:
- MDS process measures are plotted as single-quarter rates. Each quarterly CMS file reports four individual quarters, and we chart each one on its own calendar quarter — so a single release already yields a four-quarter trend. These single-quarter rates are higher-resolution but noisier than the four-quarter average shown in the QM detail card further down the page, so the most recent trend point won't exactly match the card's headline number. Claims-based outcomes are a single rolling rate per release (the risk-adjusted score), so they add one point per quarter going forward.
- An empty cell is labelled with its reason, not just blanked. Three different things produce a missing score, and the tables keep them apart. S — CMS withheld the value because the facility had too few cases in the window to report reliably; that is a fact about the facility's census, often a small population for that measure, not a gap in our coverage. A — the measure is calculated annually and not by quarter at all (this is the flu vaccination measures, CMS footnote 28), so those quarterly cells are structurally empty and say nothing about the facility. “–” — nothing published and no reason given. The CMS reason shows on hover. None of the three is ever treated as zero, and none enters any state, county, or national average.
- Claims and MDS rows are not co-timed. In the facility table, which stacks every measure on one quarter axis, a claims row's cell is a risk-adjusted rate for the twelve months ending that quarter, while an MDS row's cell is the unadjusted rate for that quarter alone. A claims move and an MDS move in the same column are therefore not the same event, and the two shouldn't be read as simultaneous. Each row's tooltip states which basis it is on.
- Benchmarks need at least three facilities. A state, county, or national average for a given quarter is only drawn when at least three facilities reported that measure that quarter — below that, an "average" is one or two facilities' own values and isn't a meaningful peer line. Where a comparator can't be drawn its toggle stays visible but disabled, and says how many facilities were available, so a thin basis is stated rather than silently hidden. The averages include the facility being viewed, consistent with the other benchmarks in the product, so in a small county the facility is a visible share of its own comparator — the facility count in the tooltip is what makes that judgeable. Benchmarks are computed across currently-certified facilities only (decertified excluded).
Sources & cadence
- NH_QualityMsr_MDS & NH_QualityMsr_Claims — published quarterly by CMS at data.cms.gov. SNF Shark refreshes both within a week of each CMS publish.
- State and national averages — computed in our database across every facility in the most recent CMS snapshot, so the comparison you see is always against the same vintage as the facility's own score.
- For the technical methodology, see CMS's Five-Star Quality Rating System Technical Users' Guide — the official document covering risk-adjustment models, scoring thresholds, and footnote definitions.