Modeling west nile virus: Vector host preferences drive enzootic transmission in the northeastern United States

Background/Question/Methods   The transmission cycles of vector-borne pathogens with multiple hosts are ecologically complex and emergence is notoriously difficult to forecast. Adding to the complexity, recent studies suggest that vector host preferences may modify transmission because contact rates are no longer homogenous among susceptible hosts. We developed an empirically-informed model to investigate how local host preferences mediate enzootic transmission of West Nile virus at four sites in Connecticut. Using a standard epidemiological S-I-R framework, we modeled daily enzootic transmission in a community of multiple avian hosts with a single mosquito vector, Culex pipiens, for one year. Seasonal mosquito abundances were incorporated into the model using CDC-light trap collection data. Bird populations were incorporated using point counts from field surveys. Cx. pipiens preference for hosts was modeled using a field-derived feeding index, which is a measure of the proportion of Cx. pipiens blood meals from a certain host in relation to that host's abundance in the field.

Results/Conclusions   Our field data showed that Cx. pipiens exhibits preference for one particular host species, the American robin (Turdus migratorius), with the feeding index ranging from 6.70 to 31.90. The model accurately predicted the density of infected Cx. pipiens in three of the four sites when compared to the site-specific vector index (monthly mean abundance x monthly mean infection rate). Results show that the feeding index had a large impact on both transmission intensity and timing, with a consistent pattern across all sites. That is, enzootic transmission did not persist at any site when feeding index was less than 6. The feeding index had the greatest influence on transmission intensity and timing at moderate levels (7-15), but reached saturation at high levels (>20). Our findings highlight the importance of incorporating host preferences when modeling transmission, the challenges to forecasting when host preferences are not consistent across locations, and the benefits of an empirically-informed model.