Modelling a discrete spatial response using generalized linear mixed models: application to Lyme disease vectors

Predicting disease risk by identifying environmental factors responsible for the geographical distribution of disease vectors can help target control strategies and optimize preventive measures. In this study we present a hierarchical approach to model the distribution of Lyme disease ticks as a function of environmental factors. We use the Poisson framework natural for count data while allowing for spatial correlations. To help identify environmental factors that best explain tick abundance, we develop an intuitive procedure for covariate selection in the spatial context. These methods could be useful in analysing effects of environmental and climatological changes on the distribution of disease vectors, and the spatial extrapolation of vector abundance under such scenarios.