How Twitter Tracks the Flu

harmacist Jason To gives a flu shot to Diane Marie Dash.

harmacist Jason To gives a flu shot to Diane Marie Dash. Jack Dempsey/AP

The social-media site can raise flags about the spread of the flu weeks before CDC data becomes available.

New research suggests Twitter—the social-media site best known for spot news and cat GIFs—can also be used to pinpoint flu outbreaks.

A team of researchers at Johns Hopkins University and George Washington University found that tweets are an accurate predictor of where cases of the flu are highly concentrated.

Because Twitter produces information in real time, local public health officials can use the aggregate tweet data to act early to make resources available to combat the spread of the flu. Flu reports from the Centers for Disease Control and Prevention—which are currently used to make public health decisions—have about a two-week delay from the time the information is collected to the time it is released, said lead author David Broniatowski.

"I don't see this as replacing what [the CDC] does," Broniatowski said, "but it might be used as an early warning system, in parallel with the hospital data as it's collected."

Armed with the early information, public health officials can increase vaccinations and ready hospitals for an influx of patients.

Tweets are tied to a location, and the researchers were able to accurately gauge the spread of the flu at both a national and local level by checking their data against reports from the CDC and the New York City Department of Health and Mental Hygiene.

"This has major public health implications, because if you are trying to figure out what the likelihood is that you need to prepare for a flu outbreak and you're looking at national data, it doesn't help you if they're concentrated in Oregon and you're in Atlanta," Broniatowski said.

In addition to preparation, public health officials can use the Twitter data to understand the social aspect of how the flu spreads, Broniatowski said, because they are able to separate tweets that reveal awareness of the flu in the community from the tweets that indicate the user is infected with the flu.

Future research could explore Twitter's effectiveness in tracking other disease outbreaks. One of the limitations of Twitter is its demographics, however, which tend to under-represent children and older populations. In replicating the results with communities smaller than New York City, a smaller number of Twitter users could also decrease the reliability of the tweets in tracking the flu trends, the researchers said.

The team evaluated Twitter's accuracy during the last full flu season, Sept. 30, 2012, through May 31, 2013. One of the researchers on the team was supported in part by an award from the National Institutes of Health and another received a grant from the National Science Foundation. They conducted their study independent of any outside organizations, and the findings were published in the peer-reviewed journal PLOS ONE.