Preventable Deaths at Acute Care Hospitals

Abstract

Background Previous measurements of preventable death in hospital do not account for the uncertainty of preventability ratings. Objective To determine the proportion of deaths in hospital that a have high probability of being prevented with high quality care. Methods We created summaries for every death at a tertiary care hospital over 4-months. Four reviewers assigned preventability ratings to each death and latent class analysis was used to classify deaths into high and low preventability categories. Results There were 480 decedents with mean age of 73.9. Inter-rater reliability was poor with an intra-class correlation of 0.14. The best latent class model found that 6.2% (95% CI 0.00 – 15.2%) of deaths had a 31.0% probability of being rated more likely preventable than not by each reviewer. In contrast, 93.8% (95% CI 84.8 - 100.0%) of deaths had a 0.8% probability of being rated more likely preventable than not by each reviewer. The incidence of truly preventable deaths is less than the 6.2% that are deemed possibly preventable. xi Conclusion Very few deaths in hospital are preventable. The low incidence of preventable deaths and low inter-rater reliability means that peer review methodology is only sensitive to large differences in preventable death rate.