Influenza forecasting system wins prize from CDC

A new system for predicting seasonal peaks of influenza in cities across the U.S., developed by a team of scientists including Marc Lipsitch of Harvard School of Public Health (HSPH), has won first place out of 11 teams in the Centers for Disease Control and Prevention’s “Predict the Influenza Season Challenge,” along with a prize of $75,000.

Participants in the CDC contest were asked to forecast the timing, peak, and intensity of the 2013-14 flu season using digital data from a variety of sources as well as innovative modeling approaches. From December 2013 through March 2014, teams were required to submit eight biweekly national and regional flu predictions, which the CDC subsequently matched against actual flu outbreak data from around the country.

The winning team, led by Jeffrey Shaman of the Columbia University Mailman School of Public Health, produced the most accurate and reliable forecasts for about 100 locations around the U.S. (see cpid.iri.columbia.edu). They based their predictions on data from Google Flu Trends (which estimates outbreaks based on the number of flu-related search queries), reports of flu-like illnesses, and verified cases of flu. They also incorporated earlier work by Shaman and Lipsitch that identified weather conditions, particularly absolute humidity levels, as important drivers of flu transmission. They then fed the data into a mathematical model that was calibrated to produce a more accurate and reliable forecast of national and regional flu activity.

Lipsitch, professor of epidemiology and director of the Center for Communicable Disease Dynamics at HSPH and an expert in the epidemiology and surveillance of flu, cited four important reasons for flu forecasting. For one, accurate forecasting provides scientific verification that the forecasters have a good understanding of how flu spreads. Second, if public health agencies know that a bad flu season is on the way, they can better prepare for higher-than-usual numbers of respiratory infections. Third, better forecasting can provide public health officials with a stronger scientific basis for urging people to get flu vaccines. And fourth, prediction during pandemics is particularly important for planning, and developing forecasting methods and experience during regular flu seasons will provide a basis for responding better in emergencies.

“This new forecasting method won’t give us a crystal ball with perfect predicting power, but it will help us refine our abilities to make predictions about the spread and intensity of flu each year,” said Lipsitch.

Read the CDC press release: CDC Announces Winner of the ‘Predict the Influenza Season Challenge’

Learn more

Dry winters linked to seasonal outbreaks of influenza (HSPH release)