Partly Cloudy With a Chance of Influenza: Forecasting the Flu
A new computer model can predict the timing and severity of flu outbreaks the way meteorologists predict weather patterns. The new tool follows a similar association effort.
How would you feel about waking up to a flu forecast along with your daily weather report?
That day may be coming soon: Researchers at Columbia University’s Mailman School of Public Health and the National Center for Atmospheric Research announced last week that they have developed a new computer model that can predict the timing and severity of influenza outbreaks.
Using near real-time, web-based data, the researchers analyzed estimates of flu-related illness from 2003 to 2008 and found that they were able to predict the peak timing of an outbreak seven weeks in advance.
This new model provides “a window into what can happen week to week as flu prevalence rises and falls,” said Jeffrey Shaman, assistant professor of environmental health sciences at Columbia University.
A similar idea and predictive tool was implemented last year by the American Public Health Association. In collaboration with the Skoll Global Threats Fund and Children’s Hospital of Boston, APHA released Flu Near You, a user-contributed log of nationwide flu data in the U.S.
Every week, registered participants receive an email from the website, asking them to check off any flu symptoms they exhibited in the preceding seven days. Flu Near You then catalogs the information on a map, where viewers can see what areas of the country are experiencing high instances of the illness.
“The idea is that people can take a look at Flu Near You, look at the map, and then make decisions about their health,” said Michelle Holshue, a Flu Near You fellow at APHA.
Holshue explained that public health departments could use predictive tools like the Columbia University computer model and Flu Near You to estimate how much flu vaccine to have in stock, and hospitals could adjust staffing levels when high flu levels are estimated in an area.
Additionally, if computer models could predict outbreaks several weeks in advance, public health departments could advise people when to get a flu shot, which takes a couple of weeks to take effect, and businesses could make a plan to coordinate teleworking schedules and encourage employees to stay home if they are exhibiting symptoms.
Holshue said she’d be happy to see televised flu predictions.
“It would be great if people could take a look at the flu forecast—just like they would take a look at the weather every morning, and if it’s raining bring an umbrella—if they look at the flu forecast and say, ‘I’ve to get my flu shot.’”