Spatio-Temporal Bayesian Regression of Pollutants on Reported Respiratory Disorders in Districts of Delhi, India

We propose Bayesian Poisson regression with log-linear link function to model prevalence of respiratory disorders (RD) as response to the concentration levels of five types of air contaminants. Intensity of individual pollutants is initially assessed through Heatmaps. Spatial random effects, global temporal effects and differential time trends are investigated across the nine districts of the union territory of Delhi by assuming intrinsic conditionally autoregressive priors. Impact of the pollutants on risk of RD admissions reveal evidence for spatial and spatio-temporal dependence indicating contribution of individual location on exposure to pollutants in addition to explaining dynamics across space in terms of relative risk associated with each pollutant.