SNF Readmissions:
1 in 5 Bounces Back

1. Downloadable Raw Data

State-level SNFRM snapshot — includes all 51 jurisdictions, facility counts, ranks, and deviations from the national median. Filename encodes the snapshot date for reproducibility.

Files: placet_snfrm_state_2026-04-07.csv / placet_snfrm_state_2026-04-07.json — 51 rows (50 states + DC), facility-level granularity in the JSON metadata.

2. Data Dictionary

CMS technical specs for SNFRM: qualitynet.cms.gov/snf/measures/readmission · NQF endorsement #2510: qualityforum.org/QPS/2510

3. Reproducibility

Aggregation SQL (PostgreSQL syntax — adapt for your environment):

-- Reproduce state-level SNFRM aggregates from CMS facility-level VBP data
-- Source: CMS Provider Data Catalog, dataset ID: 284v-j9fz
-- API: https://data.cms.gov/provider-data/api/1/datastore/query/284v-j9fz/0

-- Step 1: Download the full dataset (13,900 rows) via Socrata API with pagination
--   GET https://data.cms.gov/provider-data/api/1/datastore/query/284v-j9fz/0
--   Key field: performance_period_fy_2024_riskstandardized_readmission_rate
--   Exclude rows where this field is NULL (facilities that did not meet ≥25-stay threshold)

-- Step 2: Aggregate to state level
SELECT
    state_code,
    COUNT(*) AS facility_count_reportable,
    ROUND(AVG(
        performance_period_fy_2024_riskstandardized_readmission_rate * 100
    )::numeric, 2) AS rate_pct
FROM cms_snf_vbp_fy2026
WHERE performance_period_fy_2024_riskstandardized_readmission_rate IS NOT NULL
GROUP BY state_code
ORDER BY rate_pct ASC;

-- Step 3: National benchmarks
SELECT
    PERCENTILE_CONT(0.5) WITHIN GROUP (
        ORDER BY performance_period_fy_2024_riskstandardized_readmission_rate
    ) * 100 AS national_median_pct,
    AVG(performance_period_fy_2024_riskstandardized_readmission_rate) * 100 AS national_mean_pct,
    COUNT(*) AS total_reportable_facilities
FROM cms_snf_vbp_fy2026
WHERE performance_period_fy_2024_riskstandardized_readmission_rate IS NOT NULL;

-- Results: median = 20.39%, mean = 20.53%, n = 10,305

4. Known Limitations

1. 1.
   Medicare FFS only — SNFRM captures only Medicare fee-for-service beneficiaries. Medicare Advantage (MA), now ≥50% of Medicare enrollment nationally, is excluded from this measure entirely. In states with MA penetration above 60% (FL, MN, PA among others), the measured population may not be representative of the facility's full patient mix.
2. 2.
   Risk adjustment is CMS's, not ours — We report CMS's own risk-standardized rates directly. CMS controls for patient demographics, comorbidities, and prior utilization, but not for socioeconomic factors like poverty rate or insurance mix. Our state aggregates inherit all assumptions embedded in CMS's risk model.
3. 3.
   Small-N suppression threshold: ≥25 stays — CMS suppresses rates for facilities with fewer than 25 qualifying stays in the performance period. 3,595 of 13,900 enrolled facilities (25.9%) are excluded. Rural states (AK, MT, WY, ND, VT) have the highest suppression fractions — their state averages rest on fewer reportable facilities.
4. 4.
   Simple vs. volume-weighted averages — State rates here are simple averages of facility-level rates (each facility weighted equally). Volume-weighting by Medicare days or stay count would shift results — likely toward larger urban facilities. We chose simple averaging for transparency; researchers should re-weight as appropriate for their analysis.
5. 5.
   Rolling window and fiscal year — Performance period: FY 2024. CMS defines the SNF VBP performance period as October 1, 2022 – September 30, 2024 (two-year rolling window). Baseline period is FY 2019. Dataset released 2026-01-21; next CMS update expected April 2026.

5. Open Questions — We Want Your Help

Things we haven't been able to answer well internally. If you have relevant work or data, we'd genuinely like to hear from you.

1. 1.

   Does coupling SNFRM with LTCFocus chain/ownership panels reveal differential readmission patterns by chain type (PE-owned vs. nonprofit vs. independent)? The ownership effect on readmission is understudied.

2. 2.

   How should we handle the MA blind spot for states where MA penetration exceeds 60%? Is there a defensible correction factor, or should those states carry an explicit asterisk?

3. 3.

   What's the right way to surface risk-adjusted vs. unadjusted rates side-by-side without misleading family readers who don't understand what risk adjustment means?

4. 4.

   Are there county-level or HSA-level patterns in SNFRM that state aggregation is masking? We'd love a sub-state view but can't do it at facility level without compromising facility privacy at small-N.

5. 5.

   What's the causal story for Alaska (#1) vs. Oklahoma (#51)? Case mix alone doesn't explain a 3.2pp gap. Is this primarily staffing, hospital capacity, or something else?

6. How to Cite This Page

Placet Intelligence (2026). Nursing Home 30-Day Readmission Rates by State (SNFRM). Snapshot dated 2026-04-07. Retrieved [date] from https://placet.org/transparency/hospital-readmission-report. Data source: CMS FY 2026 SNF VBP Facility-Level Dataset (ID: 284v-j9fz), CMS Provider Data Catalog. Reporting period: FY 2024 Performance Period.