Disease mapping with individual level information; a case study of acute myocardial infarction mortality

When mapping relative mortality risk under specific causes of death in time, one can use small areas and single year mortality data to explore the space time variation in detail. To reduce the variability of the initial mortality risk estimates and help explain their differences, hierarchical Poisson models are typically used. Here we deal with the situation where besides aggregated small-area level data necessary for that, one also has complete individual level data about the presence of certain risk factors in the population, which is now rare but it should become routine in places with universal health coverage using a medical record sharing system. In particular, we consider the convenience of including individual level covariates in the models, and mapping relative mortality risk adjusted for them. That is illustrated by exploring how mortality due to acute myocardial infarction varies in space and in time in Catalonia between 2014 and 2019 using individual data on obesity, diabetes, dyslipidemia and smoking habits.