The degree of severity and trends in hospital standardized mortality ratios in Japan between 2008 and 2012: a retrospective observational study.

Abstract

Hospital standardized mortality ratio (HSMR), an indicator that adjusts hospital mortality for case-mix differences, is used as a hospital performance measure. The aim of this study is to build a new HSMR model in Japan and examine HSMR trends according to the degree of severity. Observational retrospective study. Data from the Japanese Administrative Database. A total of 3 813 492 admissions from 278 Japanese acute-care hospitals were extracted from the database (patients between 2008 and 2012, from July to December in each year). We estimated the probability of in-hospital death by fitting a logistic regression model, and assessed the performance of the models with the c-index. In each year, HSMRs were obtained by calculating the ratio of the number of observed deaths to the number of expected deaths. The HSMR trends, including trends in comorbidity subgroups defined by the Charlson comorbidity index, were analysed. The c-index value was 0.871 for the HSMR model. The HSMR followed a constant decreasing trend over time; it fell by 18.8% from 110.3 in 2008 to 91.5 in 2012. The reduction in HSMR was not present in the severe comorbidity group, while the reduction trend was observed in the mild comorbidity group. Our model demonstrated excellent discrimination without detailed clinical data. The Japanese HSMR followed a constant decreased trend, while the reduction trend was not present in the severe patients. Our study implies the need to consider severe patients for assessing hospital quality by HSMR.