Keywords: Bayesian model; hierarchical model; Poisson regression
Although the health of the U.S. population has improved as a whole, geographic disparities in mortality rates remain large. We develop Bayesian approaches to estimate achievable national, regional or state-specific target mortality rates by age and sex and the corresponding county-specific excess mortality (deaths above the expected number of deaths under the target mortality rates) to facilitate the examination of disparities in health. These methods were then applied to a case study of premature mortality in the U.S. between 1999 and 2014 to determine the extent and spread of geographic disparities in excess death across the nation. Of the more than 1.7 million annual total deaths, nearly 1 million, or more than 1 out of every 2 (55%), deaths in the United States were considered in excess as compared to the targets. These numbers varied within age groups and by gender, state, and county. Estimation of target rates and excess death to measure health disparities is advantageous and straightforward in a Bayesian framework. Characterizing these targets and disparities can aid in tracking progress toward objectives as well as in guiding policymakers at all levels.