Timely interventions like vaccinations can prevent or control the adverse impacts of epidemics on human health. However, prediction of epidemics is extremely challenging. For example, the incidence of dengue – a vector-borne disease affecting approximately 100 million people per year – can increase 3-5 fold during an epidemic, yet no clear indicator of the intensity or timing of an epidemic exists until it is already underway. Influenza and other globally important diseases present similar challenges. Advances in forecasting for these diseases and others are continually occurring, yet research gaps limit forecasting model development, evaluation of forecasts, and adoption by decision-makers. The Epidemic Prediction Initiative (EPI) aims to improve the science and usability of forecasts by addressing these challenges.
Since January 2016, EPI has published influenza forecasts from participating teams in real-time on the EPI website. This was the first time that infectious disease forecasts from multiple groups were been published jointly in real-time, facilitating forecast comparison and evaluation by public health officials. EPI also initiated and maintains anopen online repositoryof code and data related to epidemics. This activity aims to reduce redundancy in data cleaning, standardize data formats, and support forecasting research. Finally, EPI has been engaging in outreach efforts within CDC, among other federal government agencies, with state and international public health officials, and in the academic community to better understand how to improve forecast accuracy and how forecasts can be used in public health decision making.