Abstract 167: Correlation of 30-day Hospital Readmission Rates by Payer and Penalty Status

Abstract

Background: Readmission rates for heart failure (HF), acute myocardial infarction (MI), and pneumonia (PNA) hospitalizations among Medicare beneficiaries are used to assess hospital quality and determine global reimbursement. However, whether these measures reflect readmission rates for other conditions or insurance groups is unknown. Methods: All hospitals with >30 index admissions for HF, MI or pneumonia in 2013 in the all-payer Nationwide Readmissions Database (NRD) were included. For each hospital, we estimated 30-day all-cause risk-standardized readmission rates (RSRRs) for 3 groups: (A) Medicare beneficiaries admitted with HF/MI/PNA, (B) Medicare beneficiaries admitted for other conditions, and (C) non-Medicare beneficiaries admitted for HF/MI/PNA. Pair-wise correlations of these measures were assessed using Spearman correlation coefficients. Hospitals were divided into quartiles based on their calculated RSRR for each group, and inter-rater agreement between groups was assessed using weighted kappa statistics. We then examined pair-wise Spearman correlations among subgroups based on hospital characteristics (size, ownership status, and teaching status). Results: Among 1,228 hospitals, wide variation in RSRRs was seen across conditions and insurance type. Groups A vs. B had moderate correlation and agreement (Spearman 0.631, quartile weighted kappa 0.440). However, A vs. C had only modest correlation and fair agreement (Spearman 0.399, quartile weighted kappa 0.243). Compared with their quartile rank for the publicly reported group A, 46% of hospitals were ranked in the same quartile for group B and 35% for group C (Figure). Correlations between groups A vs. B and A vs. C were strongest among metro teaching hospitals (Spearman 0.674 and 0.507, respectively). Conclusions: Risk-standardized readmission rates for HF/MI/PNA are generalizable measures of a hospital’s readmission rate for other conditions among Medicare-insured patients. However, publicly reported RSRRs for HF/MI/PNA among Medicare patients are less suitable measures of hospital performance for non-Medicare patients admitted for these conditions.