Spatially-structured human mortality modelling using air pollutants with a compositional approach

Abstract

The human mortality models with a demographic approach are performed in function of time. The addition of information (social, economic, and environmental) in the structure of demographic models allows fitting observed values better. Air pollution influences human mortality and could be used as an environmental covariate in the demographic models. The levels of air pollutants describe quantitatively the parts of a whole (air), called composition, and their statistical treatment should consider this characteristic in the modelling process. This article evaluated the association between human mortality data with levels of air pollutants as a composition using a spatially-structured model. The spatially-structured modelling approach in the human mortality data captures the spatial heterogeneity of air pollutant concentrations (local environmental conditions). Human mortality data is defined as the number of deaths, and in this work, it was analyzed with both total and disaggregated presentation. The disaggregation was by (i) sex and (ii) sex and age-group. A likelihood ratio test suggested the model with air pollutants as covariates treated under a compositional approach (proposed model) is more appropriate than the model based only on time explanatory variable in yearly basis. The proposed model was evaluated in 48 counties in Spain, each with its mortality and air pollution dataset. The modelling approach in this work presented adequate quality model indexes and could be applied to make short-term predictions with different air pollution scenarios.