Improved disease tracking with AI

Epidemiology Powerful machine-learning tools have the potential to improve real-time estimates of disease activity, such as for COVID-19 or influenza. Aiken et al. developed such a method that produces dynamic weekly and city-level forecasts of disease cases and is more accurate than existing techniques, including those that use real-time data from web searches. The researchers' neural network integrates disease activity data across cities and time to produce highly accurate forecasts of activity up to 8 weeks in the future. These findings illustrate the potential of modern artificial intelligence (AI) to better track and predict local epidemic dynamics up to several weeks ahead of current health care–based surveillance systems, which would support better public health policy and individual decisions. Sci. Adv. 10.1126/sciadv.abb1237 (2021).