Hospital quality

Quality of Medicare-certified hospitals from CMS Care Compare: the overall star rating from dim_hospital, and healthcare-associated infection (HAI) results from fct_hospital_quality. Quality scores are only comparable within a measure, so each view below is filtered to specific measure_ids and never aggregates across measures.

Overall star ratings

Healthcare-associated infections

Standardized infection ratios (SIR): below 1.0 means fewer infections than the national baseline predicts, above 1.0 means more. Each SIR measure is summarized separately. Averages are unweighted means across reporting hospitals — small and large facilities count equally.

Hospital Readmissions Reduction Program (HRRP)

CMS's Hospital Readmissions Reduction Program docks Medicare inpatient payments for hospitals whose 30-day risk-adjusted readmissions exceed the national expected rate for six clinical conditions: acute myocardial infarction (AMI), coronary artery bypass grafting (CABG), chronic obstructive pulmonary disease (COPD), heart failure (HF), elective primary hip or knee replacement (HIP-KNEE), and pneumonia (PN). The core signal is the excess readmission ratio — a hospital-condition-level ratio of predicted to expected readmissions where ERR > 1 triggers a penalty.

Every hospital appears in all six conditions (18,330 rows = 3,055 hospitals × 6 conditions), but CMS suppresses low-volume cells, so only 11,720 of the 18,330 rows carry an ERR. Roughly half of those measured rows (48%) sit above 1.0 — that's expected by construction, since ERR is calibrated so the national aggregate lands at 1.0.

ERR distribution by condition

Reporting depth varies by condition: pneumonia (2,715 hospitals) and heart failure (2,621) are close to universal, while coronary artery bypass (878) is limited to hospitals that perform enough cases. Penalized shares hover near 46–50% for every condition — the tails are what differ.

HRRP by state

Joining to dim_hospital for the state attaches state labels; 20 CCNs in the HRRP file don't reconcile with the Care Compare hospital roster and are dropped by the inner join.

Hospital Value-Based Purchasing (VBP)

The Value-Based Purchasing program redistributes 2% of Medicare inpatient payments based on a total performance score built from four equally-weighted domains: clinical outcomes, person and community engagement, safety, and efficiency and cost reduction. The domain scores below are the weighted contribution to the 100-point total (each domain nominally worth 25 points before reweighting).

FY2026 covers 2,455 hospitals; Maryland participates under an all-payer waiver and does not appear in the file. 156 hospitals were reweighted on at least one domain — CMS drops a domain when it can't score enough measures — so the domain averages below exclude those NULLs.

Total performance score distribution

TPS is right-skewed: the median hospital lands at 29.5 out of 100, but a long tail runs into the 70s and higher — those hospitals capture the positive side of the 2% redistribution.

Domain decomposition

Safety and person-and-community-engagement contribute the most on average; clinical outcomes and efficiency contribute the least. The domains are equally weighted before reweighting, so these gaps reflect how hard each domain is to score highly on rather than a policy preference.

TPS vs overall star rating

Both signals summarize hospital performance but from very different inputs: the overall star rating rolls up dozens of Care Compare measures into a 1–5 rating, while TPS blends the four VBP domains. Plotting average TPS by star bucket confirms they move together — five-star hospitals average roughly twice the TPS of one-star hospitals. The chart covers 2,417 of the 2,455 VBP hospitals: 9 CCNs don't reconcile with dim_hospital (the same vintage gap as the HRRP state view) and another 29 hospitals have no published star rating.

Caveats

• Scores are comparable within a measure only. score semantics in fct_hospital_quality vary by measure_domain and measure_id — a HCAHPS star rating, a mortality rate per 100, a standardized infection ratio, and a readmission rate all live in the same column. Every view on this page filters to a specific measure (or a small family of related SIR measures) and never aggregates across them. The HAI averages are additionally unweighted means across reporting hospitals, so small and large facilities count equally.
• Suppressed and "Not Available" cells are kept as NULL, not zero. fct_hospital_quality, fct_hospital_readmissions, and fct_hospital_vbp all preserve CMS's small-count suppression and 'Not Available' sentinels as nulls. Coverage counts on this page are always non-null observations, not row counts — e.g. 11,720 of 18,330 HRRP rows carry an excess_readmission_ratio; the other 6,610 (36%) are ineligible or suppressed.
• HRRP is calibrated to a national average of 1.0. The excess readmission ratio is a predicted-to-expected ratio; roughly half the measured hospitals sit above 1.0 by construction, so a ~48% penalized share is the baseline outcome of the program, not a quality signal on its own. Penalized shares below in the by-state and by-condition tables should be read against that ~48% baseline.
• HRRP and VBP snapshots trail the Care Compare provider file. Both mart tests join to dim_hospital at warn severity: HRRP (2026-01-26 vintage) has 20 CCNs / 120 rows that don't reconcile with the 2026-07-22 provider snapshot; VBP has 9. This is a structural annual-vs-quarterly cadence gap — the state-level HRRP view and the TPS-vs-stars chart drop those rows via inner join, so their coverage (e.g. 2,417 of 2,455 VBP hospitals in the star-rating view) trails the mart totals.
• VBP reweighting excludes some domain scores. 156 of the 2,455 FY2026 hospitals are missing at least one weighted domain score because CMS reweights the remaining domains when a hospital lacks enough measure data. Those rows are kept in the mart; the domain-average table on this page necessarily excludes each domain's null hospitals, so scored-hospital counts differ across domains. Maryland is absent entirely — its hospitals are exempt from VBP under the state's all-payer model.
• HRRP measurement window is fixed. Every row in the current HRRP vintage covers the same 2021-07-01 → 2024-06-30 window (CMS publishes annually over a three-year rolling period), so nothing in the HRRP sections is a time series.