A model framework for mortality and health data classified by age, area, and time

This article sets out a modeling framework for modeling health outcomes over area, age, and time dimensions that takes account of spatial correlation, interactions between dimensions, and cohort as well as age effects. The goals of the framework include parsimony and parameter interpretability. Multivariate extensions may be made allowing interdependent or shared effects between different outcomes (e.g., ill health and mortality). A particular focus is on assessing the proportionality assumption whereby separate age and area effects multiply to produce age-area mortality or illness rates, and age-area interactions are assumed not to exist. A trivariate (mortality-health) application of the framework involves cross-sectional data in the 33 London boroughs, while a longitudinal univariate application involves deaths for the same areas over four 5-year periods starting in 1979.