34 - Spatially refined estimates of the risk of West Nile virus

Abstract: Since its introduction in 1999, West Nile virus (WNV) has established itself as the leading domestically acquired arbovirus in the United States. Transmission is driven by Culex spp. mosquitoes which predominantly feed on birds but also mammals in the late summer, resulting in West Nile virus spillover events in humans. Environmental factors such as temperature, precipitation, hydrology, and humidity influence WNV transmission dynamics and can inform more spatially refined forecasts. Identifying the spatial-temporal variability of vector-borne diseases is challenging because 1) outbreaks are heterogeneous and 2) only small portions of a given area monitored are typically positive for disease activity. Consequently, effective allocation of public health resources can be compromised. There is, thus, a need for accurate, spatially refined observations and forecasts of the burden of disease. We used spatially-refined mosquito data throughout the United States in conjunction with a mathematical model representing WNV transmission dynamics among mosquitoes and birds, as well as spillover to humans, to generate forecasts at sub-city spatial scales. The mathematical model is optimized using a data assimilation method and the two observation data streams: mosquito infection rates and reported human WNV cases. We present the forecasting framework, evaluate retrospective forecasts at different spatial scales, and discuss the limitations of the real-time monitoring network. Forecast skill decreases when predicting infectious mosquitoes at the sub-city scale; however, human case forecasting accuracy is maintained. This model-inference forecasting system can be used to understand better the relationship between monitored zoonotic amplification and human outbreaks. This work represents an initial step in standardizing observational data for optimal use observing and forecasting seasonal outbreaks of West Nile virus.