Accounting for the Simpson-Yule Effect in Standardized Mortality Ratios

Each year, medical errors are the cause of over 250,000 deaths in the United States alone, making them the United States’ third-leading cause of death. Society places a tremendous amount of trust in our medical professionals, and in return, we expect to be treated with the highest-quality medical care available. In this Preliminary Report, we use multivariate logistic regressions to predict the mortality rates of patients in hospitals based on any set of historical mortality data. We discuss a new method to directly compare hospitals to each other, protecting our model from a version of the Simpson-Yule effect, a critical flaw allowing objectively worse hospitals to score better standardized mortality ratios facing all current implementations of hospital standardized mortality models, including those used by the United States, the United Kingdom, Canada, and the Netherlands.