The importance of estimating selection bias on prevalence estimates shortly after a disaster.