Article Text
Abstract
Introduction The hospital standardised mortality ratio (HSMR) has been used internationally to help support improvements in hospital quality of care. Most commonly, HSMRs are computed using only in-hospital deaths, influenced by length of stay. Following a major review, the National Health Service in England recommended that HSMR calculations include both in- and 30-day post-discharge mortality. Our study investigated the relationship between these two measures.
Methods Retrospective analysis of probabilistically linked mortality and hospital data for patients admitted to 162 hospitals in New South Wales, Australia between July 2000 and June 2008. HSMRs for in-hospital and 30-day post discharge (30-day HSMR) were calculated by dividing an observed number of deaths by the logistic-regression derived expected number and compared using correlation coefficients and outlier status.
Results There were 270 456 hospital admissions ending in death either in-hospital (147 926; 55%) or up to 30 days post discharge (122 530; 45%). HSMRs ranged from 12.5 to 251.7 for in-hospital HSMR and 14.3 to 195.4 for 30-day HSMR. Correlation between the two measures was high (Spearman ρ=0.868, p<0.001) but there was moderate agreement on outlier status (κ=0.58). Five hospitals swapped status between significantly higher/lower HSMR of 100.
Conclusions In-hospital and 30-day HSMRs were similar across hospitals, even though there were differences in agreement on outlier status and whether HSMRs 'signalled' as significantly high or low. 30-day HSMRs are more difficult to calculate, and may present few advantages over conventional HSMRs where the aim is to initiate more detailed investigation of possible hospital performance issues.