A new study published in the Proceedings of the National Academy of Sciences introduces a groundbreaking forecasting method called “epimodulation,” which could help health systems better predict and prepare for epidemic peaks. Developed by researchers at the University of Texas at Austin, the approach gives models a more intuitive understanding of how epidemics naturally evolve improving accuracy when predicting the most critical stage of an outbreak.
During an epidemic, decision-makers often struggle to determine when infections will peak, how many people will need care at once, and how long hospitals will remain under strain. Traditional models tend to falter at these turning points. The new method addresses this by embedding basic epidemiological principles such as how immunity builds and transmission slows directly into the forecasting process.
“It tells the model, in effect, ‘We expect the curve to bend as immunity builds,’” said Lauren Ancel Meyers, Cooley Centennial Professor in UT’s Department of Integrative Biology and director of epiENGAGE, a national center for excellence in outbreak forecasting. “The result is a better forecast that delivers real-time insight to hospitals and communities when it matters most.”
When tested on past influenza and COVID-19 data, the epimodulation technique improved hospital admission forecast accuracy by up to 55% during epidemic peaks without reducing accuracy during other periods. It also enhanced ensemble models that combine multiple forecasts, showing its potential as a powerful tool for managing rapidly changing outbreaks.
The research was funded by the U.S. Centers for Disease Control and Prevention the Council for State and Territorial Epidemiologists, and Tito’s Handmade Vodka.
Experts say the approach could be applied to a range of infectious diseases that spread in waves, including bird flu, Ebola, and Mpox. Such wave-like patterns typically occur as immunity builds, behaviors shift, or environmental factors change.
“Epidemics tend to follow recognizable patterns,” Meyers explained. “They grow rapidly at first, then slow as immunity or awareness increases. Our model helps capture those dynamics, providing more realistic forecasts for the moments that matter most.”