Abstract 12010: Penalizing Hospitals for Excess Heart Failure Readmissions: Probability of Hospital Misclassification Error
Introduction: The Centers for Medicare & Medicaid Services (CMS) adjust hospital payments for heart failure (HF) admissions if the estimate of risk-standardized excess readmission ratio (ERR) is above 1. The 2012 Inpatient Prospective Payment Systems Final Rule does not take into account the uncertainty of ERR measurement raising concerns about possible consequences.
Hypothesis: Ignoring the uncertainty of ERR measurement results in significant probability of hospital misclassification error (HME) in HF, the most costly of applicable conditions.
Methods: For a cohort of 1,085,568 Medicare fee-for-service patients undergoing 1,494,993 HF admissions in 4,930 U. S. hospitals between July 1, 2005 and June 30, 2008, we reproduced ERR distributions for individual hospitals using hierarchical logistic regression model and bootstrap method employed by CMS to obtain confidence intervals of ERR. Based on ERR distributions, we calculated probability of HME for each hospital and stratified results by hospital volume.
Results: Mean probability of HME due to ignoring the uncertainty of ERR measurement is 28.2% (95% CI: 27.7%, 28.6%). Probability of HME varies with hospital volume: 41.5, 33.9, 26.9, 21.5, 16.7% in the 1st-5th volume quintiles (p<.001). In the subset of 2,381 hospitals with ERR above 1 (subject to payment adjustment per existing regulations), the falsely penalized rate was 27.5% (95% CI: 26.7%, 28.3%). In the subset of 2,549 hospitals with point estimate ≤ 1 (not subject to payment adjustment), the probability of not penalizing a poorly performing hospital was 28.8% (95% CI: 28.2%, 29.4%). (Possible maximum of probability of HME is 50% equivalent to no discrimination).
Conclusions: Ignoring uncertainty of ERR measurement leads to errors in imposing financial penalties on hospitals associated with heart failure readmissions: good performers are penalized and poor performers escape penalties with estimated probability ~30%. Furthermore, disparity exists among hospitals with different volume: probability of erroneous adjustment of payment is inversely related to hospital volume.
Author Disclosures: E.A. Sosunov: None. N.N. Egorova: None. K. McCardle: None. A.C. Gelijns: None. A.J. Moskowitz: None. H. Lin: None.
- © 2015 by American Heart Association, Inc.