The spatial and temporal patterns of infectious diseases can provide information on the underlying drivers of infection as well as the design of surveillance programs and control strategies. However, such characterization of wildlife diseases is difficult due to uncertainty in the diagnostic process and clumping of surveillance efforts in time and space. Thus, apparent prevalence, calculated strictly from observed values, is a biased representation of the true prevalence of the disease. Bayesian statistical frameworks serve as solution to biased apparent prevalence calculations, by allowing uncertainty in diagnostics and sampling to be incorporated into estimates of true prevalence. Avian Influenza virus (AIV) is one such wildlife disease that also poses a threat to domestic poultry populations. Here, we used AIV sampling data from migratory waterfowl and their estimated migratory patterns in a Bayesian statistical framework. We estimated true prevalence, in areas and times with and without sampling data, to identify drivers of the spatial and temporal distribution of AIV in the continental United States.
Results/Conclusions
We used a model selection approach to compare several different spatial relationships that test the importance of the local transmission versus migration-driven transmission in determining the spatial distribution of AIV. Our results support migration-driven transmission as a stronger driver of the observed spatial distribution than localized transmission. Migration patterns representing the volume of movement resulted in lower DIC values compared to migration patterns representing only the presence or absence of migration. These results suggest that the amount of AIV introduced into an area during migration is important to the AIV spatial distribution and may have strong impacts on AIV surveillance efforts.