Google Flu Trends: Take with grain of salt

A study out of the University of Washington finds that Google Flu Trends is least accurate during times when people use it most, as compared to CDC national surveillance programs.

Elizabeth Armstrong Moore
Elizabeth Armstrong Moore is based in Portland, Oregon, and has written for Wired, The Christian Science Monitor, and public radio. Her semi-obscure hobbies include climbing, billiards, board games that take up a lot of space, and piano.
Elizabeth Armstrong Moore
2 min read
Google Flu Trends is not as accurate as more traditional influenza surveillance programs, new research finds. Sir Sabbhat/Flickr

When compared to the Centers for Disease Control's national surveillance of influenza lab tests, data from Google Flu Trends is 25 percent less accurate at estimating rates of lab-confirmed influenza infections, according to a new study out of the University of Washington.

The findings are being reported at the American Thoracic Society 2010 International Conference in New Orleans this week.

Google Flu Trends estimates national rates of flu-like activity by monitoring the popularity of certain Google search queries in real time. This method is precisely the problem, because the influenza virus does not always cause influenza-like illnesses. (The researchers said studies vary widely on how often it does, with figures ranging from 20 percent to 70 percent of cases resulting in illness during the influenza season).

The researchers say that Google Flu Trends is still a highly valuable tool, especially since Google search queries can be counted and analyzed almost instantaneously, where traditional flu surveillance systems can take days or even weeks to collect, analyze, and release data.

"Google Flu Trends influenza surveillance provides an excellent public health service because it provides nationwide influenza activity data in a cheap and timely manner," said Justin Ortiz, clinical fellow at the University of Washington who led the study, and who is a known Google user. "Nevertheless, our study demonstrates that its data should be interpreted with caution and that other surveillance systems more accurately reflect influenza activity in the United States."

"Flu Trend is by no means a replacement of traditional surveillance systems," said Google spokesman Jamie Yood. "What we're showing is modeling influenza-like illness. We've released all our raw data and we welcome people studying our system so we can try to understand the space that much more."

In light of these findings, perhaps Google should consider renaming its tool Google Flu-like Trends?

After looking at the incidence of flu outbreaks in the U.S. between 2003 to 2008, researchers found that Google Flu Trends deviated greatest from CDC surveillance during the 2003-04 influenza season, a year notable for early and intense influenza activity with higher infant mortality and media attention.

These findings coincide with a new Harris Poll survey of 2,755 adults, which found that fewer people thought they had the flu in the winter of 2009-2010 than in recent years, in spite of (or is it because of?) all the hype around H1N1.