Bayesian Spatiotemporal Model for Life Expectancy Mapping; Changes in Barcelona From 2007 to 2018

When mapping life expectancy, and investigating its local variation in time, there is a conflict between using large areas and/or mortality data from long periods of time to have low variance life expectancy estimates, and using small areas and single-year mortality data to explore the space–time variation of life expectancy in detail, without bias. Here a Bayesian model is proposed to smooth annual small-area life expectancy estimates and help deal with that trade-off. The specific area effect on life expectancy, together with its spatial and temporal dependencies are modeled through random effects, while the effect of covariates is modeled through a fixed effect component. By smoothing life expectancy estimates directly, instead of smoothing age-specific mortality rates first the way done in the literature, the model used is easier to implement and interpret. The approach is illustrated, by using it to explore how life expectancy at birth of males and of females, and their gap, varied in space and in time in the city of Barcelona between 2007 and 2018, and their relationship with covariates. It is found that, on average, life expectancy has been growing by 0.23 years per year for males and 0.15 years per year for females. The female life expectancy is becoming more spatially homogeneous than the male one, while the rate of life expectancy growth for males turns out to be more homogeneous than for females.