Assessing the Utility of Google Health Trends for Forecasting Dengue in Brazil

Dengue is a mosquito-borne disease spread primarily by Aedes aegypti and Aedes albopictus mosquito species. Almost half of the world's population lives in regions where dengue is present, and 400 million people acquire new infections every year. About 5% of dengue cases develop into severe dengue or hemorrhagic fever, which is a dangerous medical emergency. Dengue has recently been expanding to new areas in the Northern and Southern Hemispheres, pointing to a need for accurate forecasting of outbreaks. Using linear regression with feature selection and regularization (ElasticNet in R), we incorporated previous dengue case counts and Google Health Trends (GHT) search terms to forecast weekly dengue cases at the state level for Brazil. We tailored the predictions by selecting top-performing search terms and lag times for each state, and evaluated performance with Root Mean Squared Error (RMSE) and the Coefficient of Determination (R2). Fine-tuning statistical forecasting models based on regional performance has the potential to extend the forecasting horizon for dengue in Brazil, which can help communities prepare for outbreaks and reduce the burden of this disease.

JAS was supported by an appointment to the Intelligence Community Postdoctoral Research Fellowship Program at Los Alamos National Laboratory administered by Oak Ridge Institute for Science and Education (ORISE) through an interagency agreement between the U.S. Department of Energy and the Office of the Director of National Intelligence (ODNI).

Data calibration and coding automation were supported by the Laboratory Directed Research and Development program of Los Alamos National Laboratory under project number 20210062DR.

LA-UR-22-27854