How we score
Every verdict on this site is computed by one published, deterministic formula, run identically on all 14,695 Medicare/Medicaid-certified nursing homes, from public CMS Care Compare data (currently the June 1, 2026 quarterly file). No facility, law firm, or advertiser can influence a score — the verdict exists before any money moves anywhere on this site.
The Safety Score (0–100)
Six weighted parts, each scored 0–100:
| Part | Weight | What goes in |
|---|---|---|
| Inspection record | 30% | CMS's Total Weighted Health Survey Score — every deficiency, weighted by CMS itself for severity, scope, and re-visits. We map it on a curve calibrated to the national distribution (the national median maps to ≈61; the worst 1% falls below 17). |
| Serious harm & abuse | 20% | Immediate-jeopardy citations (current inspection cycle counts hardest), actual-harm citations (severity G–I), the CMS abuse icon, and abuse-cluster citations (F600–F610) at the harm level. |
| Staffing | 20% | Payroll-reported nurse hours per resident-day vs research benchmarks (3.48 total is the floor CMS's own landmark staffing study recommends; 0.55–0.75 RN hours), weekend coverage vs weekday, and annual nursing turnover. When a facility's payroll data is suppressed, we fall back to CMS staffing stars and label the score "limited data." |
| Federal penalties | 15% | Fines (count and dollars) and Medicare/Medicaid payment denials in CMS's 3-year window. Half of all facilities have zero penalties and keep full marks here. |
| Trend | 10% | The most recent inspection cycle's score vs the average of the two prior cycles — is this facility getting better or worse? |
| Context | 5% | Special Focus Facility status, chain-average quality, ownership changes, and stale-data warnings. |
The verdict
- Avoid — score below 48.
- Recommended — score 70 or above and every gate passes (below).
- Caution — everything else.
Hard caps
Some findings must dominate, not average away:
- CMS Special Focus Facility → capped at 30 (always Avoid). SFF candidate → capped at 55.
- CMS abuse icon → capped at 55; abuse icon plus a current immediate-jeopardy or harm-level abuse citation → capped at 38 (Avoid).
- Two or more current-cycle immediate-jeopardy citations → capped at 45.
Recommended is gated
A facility cannot be Recommended — no matter its composite — with any of: a current-cycle immediate-jeopardy citation, the CMS abuse icon, a bottom-tier ("much below average") CMS inspection rating, an inspection sub-score below 55, a staffing sub-score below 45, or three-plus actual-harm citations in the current cycle. When a gate fires, the facility page says exactly which one.
What we deliberately leave out
- CMS quality-measure (QM) stars. They're largely self-reported by facilities and widely criticized as gameable. We show them for context; they never move the score.
- Family reviews. Shown as their own signal on each page, never blended into the CMS-derived score. When families and inspectors disagree, that divergence is information — we show both.
Honest limitations
- CMS covers the ~14,695 Medicare/Medicaid-certified nursing homes. Assisted-living-only communities are state-regulated, not federally inspected, and are not scored here — what to do instead.
- Ratings are lagging indicators, updated quarterly. A facility can change fast, especially after an ownership change. Every page shows its as-of date.
- The score is a screen, not a guarantee — it can't see your parent's specific needs or the aide who goes above and beyond. Always pair it with an in-person visit and the tour checklist.
Calibration & reproducibility
Curves are calibrated against the full national distribution each time CMS's data model changes materially, and every quarterly ingest stores a snapshot manifest so any published score can be recomputed from the source file. Current distribution: 6,375 Recommended · 6,376 Caution · 1,944 Avoid.
The integrity rule
The verdict is computed from public data by this published algorithm before any report is sold, any lead is routed, or any ad is shown. No paying party — facility, operator, law firm, insurer — can move a score. If we ever add operator-facing tools, a paying operator gets analytics, never verdict influence. That rule is the product.