Spatiotemporal Model of West Nile Virus Detection in Northeastern Texas

Keywords: mosquito-borne disease, medical geography, logistic regression
Session Type: Virtual Poster Abstract
Day: Friday
Session Start / End Time: 2/25/2022 05:20 PM (Eastern Time (US & Canada)) - 2/25/2022 06:40 PM (Eastern Time (US & Canada))
Room: Virtual 21

Abstract

West Nile virus is the most common mosquito-borne disease in the United States. Drivers of West Nile virus activity tend to be place-specific and complex. While mosquito surveillance activities aim to anticipate West Nile virus risk, these measures are costly and time-consuming. In this study, we explored relationships between environmental factors and the probability of West Nile virus detection in mosquitoes from 555 locations across five northeastern Texas counties from 2010 through 2015. We considered quarterly measures of land surface temperature, the enhanced vegetation index, and precipitation in models with the binary outcome of annual West Nile virus detection at a one-kilometer resolution using a multilevel logistic regression model from the “lme4” R package. During initial data exploration, we found that first quarter enhanced vegetation index values across the sampled locations were dramatically higher in 2012, a year with epidemic West Nile virus transmission, compared to the other years. The results from the model fitting showed land surface temperature, enhanced vegetation index, and precipitation from the first quarter of the year were positively associated with the probability of West Nile virus detection, while precipitation during the third quarter had a negative association. This model provides a starting point for predicting where and when West Nile virus risk will be highest, allowing for more targeted interventions, such as surveillance, abatement, and public health messaging. Next steps will include models for two additional study areas to determine the extent to which environmental precursors of West Nile virus activity vary across Texas.