Bayesian Testing of Relative Mortality, with Application to Occupational Cohort Studies

An important issue in epidemiology, often exploratory, is to compare the mortality of a certain cohort, exposed to some dangerous conditions, to that of a reference population. The main statistical tool to study relative mortality has long been the subject-years method. We propose a Bayesian version of the subject-years method, particularly suited to analyse the mortality of small cohorts. In particular, we employ the Bayes factor in order to test whether the mortality of a cohort exceeds that of a reference population. The methodology is applied to an occupational epidemiology study. The results are compared to those obtained from the classical one-sided tests on the same data.