While scrolling through one's Twitter timeline, it might sometimes seem like everyone is tweeting about fear of contracting the flu, actually having the flu or just talking about their flu-ridden friends. From a research standpoint, such tweets can be hard to sift through when trying to determine how many people actually were infected with the flu in a given season.
To that end, Johns Hopkins University computer scientists in the School of Medicine recently developed a new Twitter screening method for delivering real-time data on flu cases that determines which publicly available tweets are linked to actual infection, according to a report in HUB, Johns Hopkins' medical research blog. The researchers said that their methods, based on analysis of 5,000 public tweets per minute, are more accurate than other tools and align more closely with government disease data.
Mark Dredze, an assistant research scientist at Johns Hopkins, said that people tweet about fear of getting the flu, whether to get a flu shot or mentioning someone else, such as a public figure, that has the flu. To improve the accuracy of using such tweets to determine how many people actually have the flu, he and his colleagues created "sophisticated statistical methods based on human language processing technologies" used to filter out the incessant tweet chatter. The system can separate between "I have the flu" and "I'm worried about getting the flu," according to JHU's report.
The JHU flu projection method also produces real-time results, whereas the U.S. Centers for Disease Control and Prevention take two weeks to publish flu data.
"In late December," Dredze said, "the news media picked up on the flu epidemic, causing a somewhat spurious rise in the rate produced by our Twitter system. But our new algorithm handles this effect much better than other systems, ignoring the spurious spike in tweets."
The JHU researchers certainly aren't the first to use social media as a predictor of illness. Last November, researchers from Columbia University and the National Center for Atmospheric Research used data from Google Flu Trends and weather forecasting techniques to predict the peak of flu outbreaks in specific areas more than seven weeks in advance, also using real-time data. And in January 2012, researchers from Harvard Medical School and Massachusetts General Hospital used Twitter to uncover a 2010 outbreak of cholera in Haiti; according to the researchers, the tweets would have predicted the outbreak two weeks earlier than health officials were able to report it.
To learn more:
- read the blog post
Twitter use forecasted cholera outbreak
Google Flu Trends a good warning system for EDs
CDC turns to social, mobile media for swine flu updates
Twitter helps to predict flu trends in New York