Analyzing veterinary surveillance data: Approaches to model the relationship between disease incidence and cattle trade

Abstract

Two approaches to the analysis of registry data for bovine diseases with regard to the relationship between disease incidence and cattle trade are proposed. Firstly, a parameter-driven spatio-temporal disease mapping model
formulated in a hierarchical Bayesian framework is used. Various cattle movement parameters, e.g. the number and proportion of in-movements from infected regions, can be included as potential covariates. Within this context problems of such an endogenous covariate are discussed. Since a purely parameter-driven approach is often not adequate to depict local epidemics, a so-called observationdriven infectious disease model is proposed as a second possibility. It includes an autoregressive part for counts in the region of interest in the past. Additionally,
the sum of previous cases in other regions weighted by cattle movements is added to assess the spread of the disease by trading. Both models are applied to cases
of Coxiellosis in Switzerland, 2005 to 2009.