Is a low readmission rate indicative of a good hospital?

Abstract

Hospital readmissions are an increasing focus of health care policy. This study explores the association between 30-day readmissions and 30-day mortality for surgical procedures. California longitudinal statewide data from 1995 to 2009 were analyzed for 7 complex procedures: abdominal aortic aneurysm repair, aortic valve replacement, bariatric surgery, coronary artery bypass grafting, esophagectomy, pancreatectomy, and percutaneous coronary intervention. Hospitals were categorized based on observed-to-expected (O/E) ratios for 30-day mortality and 30-day readmissions. Hospitals were considered "high" or "low" outliers if the 95% confidence intervals of their O/E ratios excluded 1 and "expected" if they included 1. Hospitals that were outliers in at least 1 metric were classified as "discordant" if their readmission and mortality rates were not both "high" or both "low," and "poorly discordant" in the particular scenario of high mortality with "expected" or "low" readmission rates. A total of 1,090,071 patients and 299 hospitals were analyzed for 7 procedures, representing a total of 1,150 clinical encounters. The overall 30-day mortality was 3.79% and the 30-day readmission was 12.69%. Of the total, 729 (63.3%) had "expected" O/E ratios for both outcomes. Among outliers, 358 (85.0%) were "discordant" and 100 (23.8%) were "poorly discordant." Hospital readmission rate alone is a limited measure of quality given the poor correlation between hospital readmission and mortality rates. In this study, 85% of hospital outliers were "discordant" for readmission and mortality. Furthermore, almost a quarter of these discordant hospitals had "expected" or "low" readmission but "high" mortality rates. Quality metrics that focus exclusively on readmission rates overlook these discrepancies.