C. K. Zogg1,2,3, Z. G. Hashmi3, J. R. Thumma2, A. M. Ryan2, J. B. Dimick2 1Yale University School Of Medicine,New Haven, CT, USA 2University Of Michigan,Center For Healthcare Outcomes And Policy,Ann Arbor, MI, USA 3Brigham And Women’s Hospital,Center For Surgery And Public Health,Boston, MA, USA
Introduction: Since passage of the 2010 ACA, the Centers for Medicare & Medicaid Services have begun to tie surgical reimbursements to hospital performance on 30-day mortality and readmission rates. Under this system, there remain concerns that some high-performing hospitals with a lower risk of 30-day mortality may suffer from higher readmissions simply by saving lives. This creates the potential for reimbursement strategies to unfairly penalize such hospitals for providing superior care. The objective of this study was to determine whether benchmarking results are similar when hospitals are profiled based on 30-day mortality versus readmission rates.
Methods: Older adult (≥65y) patients presenting for 3 common operations (elective colectomy, CABG, AAA) were identified using 2013-2014 100% Medicare fee-for-service claims. Each hospital was benchmarked on each outcome using risk-adjusted observed-to-expected (O/E, current Medicare standard) and shrinkage-adjusted (SA) rates (multilevel-modeling that accounts for variability due to hospitals with small sample-size). These estimates were then used to generate hospital performance profiles which were compared using: 1) linear regression with weighted correlation coefficients, 2) concordance among high/average/low performers with thresholds set as ±1 SD above/below the mean, and 3) magnitude of difference in quintile rank.
Results: Little to no correlation was found between mortality and readmission (Figure)—colectomy r=0.110; κ=0.002, p-value=0.111. Only 26.4% (707/2673) of hospitals performing colectomies had identical rankings for both metrics (CABG 24.8%, AAA 26.2%). Four percent had completely different rates (CABG 12.9%, AAA 12.5%)—an inverse association which became significant, r=-0.241, and markedly more pronounced, 25.0%, among high-risk patients with LOS ≥30d. SA demonstrated similar results. Discrepancies between mortality/readmission ranks were most pronounced among large hospitals (4-quintile difference vs no difference, ≥400 beds: 21.5 vs 17.9%, p=0.014), with more surgical admissions (highest quartile: 32.3 vs 29.3%, p<0.001), lacking certifications from organizations such as the Joint Commission and Council of Teaching Hospitals but with a larger resident role, more complex case-mix, and lower number of RNs/bed (p≤0.013 for each).
Conclusion: Mortality and readmission benchmarking do not identify high-quality hospitals in the same way. This creates a dichotomy between standards used to determine Medicare reimbursement rates. Implementation of benchmarking that reflects multiple aspects of quality is needed in order to avoid inconsistent penalization of large, outlying, teaching hospitals providing high-quality mortality care.
