Assessing human risk of illness with West Nile virus mosquito surveillance data to improve public health preparedness

Surveillance for West Nile virus (WNV) and other mosquito-borne pathogens involves costly and time-consuming collection and testing of mosquito samples. One difficulty faced by public health personnel is how to interpret mosquito data relative to human risk, thus leading to a failure to fully exploit the information from mosquito testing. The objective of our study was to use the information gained from historic West Nile virus mosquito testing to determine human risk relative to mosquito infection and to assess the usefulness of our mosquito infection forecasting models to give advance warning. We compared weekly mosquito infection rates from 2004 to 2013 to WNV case numbers in Illinois. We then developed a weather-based forecasting model to estimate the WNV mosquito infection rate one to 3 weeks ahead of mosquito testing both statewide and for nine regions of Illinois. We further evaluated human illness risk relative to both the measured and the model-estimated infection rates to provide guidelines for public health messages. We determined that across 10 years, over half of human WNV cases occurred following the 29 (of 210) weeks with the highest mosquito infection rates. The values forecasted by the models can identify those time periods, but model results and data availability varied by region with much stronger results obtained from regions with more mosquito data. The differences among the regions may be related to the amount of surveillance or may be due to diverse landscape characteristics across Illinois. We set the stage for better use of all surveillance options available for WNV and described an approach to modelling that can be expanded to other mosquito-borne illnesses.