Andrew A. Gonzalez, Micah Girotti, Thomas W. Wakefield, Justin B. Dimick
University of Michigan, Ann Arbor, MI.
OBJECTIVES: To determine the extent to which historical readmissions rates accurately predict future readmissions rates for vascular surgery.
METHODS: We examined four years of national Medicare claims data for patients undergoing open repair of abdominal aortic aneurysm (n=36,830) and lower extremity arterial bypass (n=87,278). Using logistic regression, we calculated risk-adjusted readmission rates for each hospital and then stratified hospitals into quintiles of readmission during two periods of time. Using 2005-2006 as an initial time period, we applied Spearman's rank correlation and weighted kappa tests to determine if hospitals tended to remain in the same quintile or migrate to a different quintile by 2007-2008.
RESULTS: The Spearman's rank correlation coefficient between 2005-2006 rankings and 2007-2008 rankings was 0.07 (p<0.004) and weighted kappa was 0.068 (p<0.001) indicating a weak correlation between year groups. There were major re-classifications between years with 11% of the worst hospitals in 2005-2006 being reclassified into the best group by 2007-2008. Of hospitals ranked in the top quintile in 2005-2006, 88% were redistributed to a different quintile by 2007-2008, with 21% of these later ranked in the lowest quintile.
CONCLUSIONS: Risk-adjusted 30-day readmission rates for open repair of abdominal aortic aneurysms and for lower extremity arterial bypasses have poor reliability. Payers and policymakers should consider abandoning the practice of using historical rates of readmission as targets for current hospital performance.
AUTHOR DISCLOSURES: J. B. Dimick: ArborMetrix, Ownership or partnership; M. Girotti: Nothing to disclose; A. A. Gonzalez: Nothing to disclose; T. W. Wakefield: Nothing to disclose.
Posted April 2013