Prediction of equine risk of West Nile Virus infection based on dead bird surveillance

Since the introduction of West Nile Virus (WNV) to the United States in 1999, the efficacy of dead bird surveillance for the prediction of human and veterinary WNV infection has been an issue of debate. We utilized South Carolina's Department of Health and Environmental Control surveillance data from 2003 to determine whether dead bird surveillance accurately predicts equine WNV infection on a county level. We adjusted for human population density as a potential confounder of an association between WNV-positive dead bird counts and mammalian WNV risk. We found a strong positive association between avian risk of WNV death and subsequent equine mortality due to WNV in South Carolina even after adjusting for human population density. Sensitivity of dead bird surveillance as a predictor of future equine WNV risk was far superior to mosquito surveillance (95% vs. 9.5%, respectively). A Poisson regression model of the equine WNV rate as a function of WNV-positive dead bird rate, adjusting for population density and taking into account effect modification by population density shows a good fit with the data. Unlike most previous studies, we control for potential confounding of the dead, WNV-positive bird-equine WNV infection association by human population density. Yet, the positive association between dead bird surveillance and equine WNV risk remains strong and statistically significant, indicating that dead bird surveillance remains a valuable tool of WNV surveillance.