Cost vs quality

Do higher-rated hospitals cost Medicare more or less? This page joins fct_hospital_utilization (one row per hospital ccn, Medicare fee-for-service inpatient payments) to the Care Compare overall star rating in dim_hospital. All dollar comparisons use payment per discharge, never raw totals, so hospital volume doesn't confound the star groups.

Payment per discharge by star rating

Median Medicare payment per inpatient discharge, by overall star rating. The raw medians are nearly flat — 1-star hospitals collect about a thousand dollars more per discharge than 4-star hospitals, with 5-star hospitals ticking back up.

But star groups don't treat equally sick patients: the median average HCC risk score falls from 1-star through 4-star hospitals (ticking up slightly at 5), so low-star hospitals see sicker-than-expected inpatients. Crudely scaling each hospital's payment per discharge by its average HCC risk score flips the endpoints — per unit of expected patient risk, 5-star hospitals are paid the most and 1-star hospitals the least.

Case mix, not stars, tracks payment

Each point is one star-rated hospital. Payment per discharge rises with the hospital's average HCC risk score (correlation ≈ 0.29 among star-rated hospitals), while its correlation with the star rating itself is ≈ 0 — case mix, not measured quality, is what tracks Medicare's per-discharge cost.

Most expensive high-volume hospitals

Star-rated hospitals with at least 1,000 inpatient discharges, ranked by Medicare payment per discharge. High HCC risk scores in this list are the case-mix confounder made concrete: these are largely academic referral centers treating the sickest patients.

Caveats

• Single snapshot, no trend. The utilization mart is a single snapshot with no year column (its vintage is the mart's as_of date), so nothing here is a time series and figures can't be labeled with a claims year.
• Inpatient dollars only. The mart's outpatient dollar columns are estimates (per-service average × service count) that exclude outlier payments, while inpatient totals include them — the two aren't comparable, so this page shows inpatient payments only.
• Join coverage. 129 of 3,216 utilization CCNs have no dim_hospital row and drop out of every view above; 3,087 hospitals survive the join.
• Unrated hospitals excluded. The star-rated views inner-join on a non-null rating, excluding 434 matched-but-unrated hospitals.
• Stars aren't risk-adjusted here. The overall rating is a hospital-level summary and the payment comparison is unadjusted; differences across star groups reflect case mix (HCC risk), teaching status, and payment-policy add-ons — not quality alone. The per-risk-unit scaling above is a crude ratio, not CMS risk adjustment.