Every death counts: investigating the role of “avoidable” and “unavoidable” deaths to compare hospital quality. The case of circulatory diseases in England.

Since the 19th century, standardised mortality rates have frequently been used as an indicator of quality for the purposes of rating hospital quality. Despite being a measurable and clearly meaningful metric, there has been a sustained opposition to this practice. Lilford and Pronovost, for example, describes them as ‘a bad idea that won’t go away’, arguing that there is simply too much noise to avoid spurious conclusions. In response to these objections, we explore the economic and econometric issues of using avoidable deaths as signals of hospital quality. We illustrate this with an econometric estimation of mortality of inpatients with circulatory diseases treated in hospitals in England. Using data for 2003/04 to 2008/09 period from the Hospital Episode Statistics (HES), we obtain Empirical Bayesian Estimates of hospital specific effects on the probability of in-hospital deaths by exploiting the panel data nature of the HES dataset and using a random effects specification and adjusting for case-mix, improvements over time and other ‘noise’. We find statistically significant differences between hospitals. It is not clear however that this is a valid reason for ignoring exploration of variations in death rates in conditions that are considered ‘non-amenable’.