Sample Size for Estimating Disease Prevalence in Free-Ranging Wildlife Populations: A Bayesian Modeling Approach

A two-parameter model and a Bayesian statistical framework are proposed for estimating prevalence and determining sample size requirements for detecting disease in free-ranging wildlife. Current approaches tend to rely on random (ideal) sampling conditions or on highly specialized computer simulations. The model-based approach presented here can accommodate a range of different sampling schemes and allows for complications that arise in the free-ranging wildlife setting including the natural clustering of individuals on the landscape and correlation in disease status from transmission among individuals. Correlation between individuals and the sampling scheme have important consequences for the sample size requirements. Specifically, high within cluster correlations in disease status can reduce sample size requirements by reducing the effective population size. However, disproportionate sampling of small subsets of subjects from the greater target population, combined with high correlation of disease status, tends to inflate sample size requirements, because it increases the likelihood of sampling multiple animals within the same highly correlated clusters, resulting in little additional information gleaned from those samples. Our results are consistent with those generated using both previously established approaches and extend their ability to adapt to additional biological, epidemiological, or societal sampling complications specific to wildlife health.