Twitter Can Track Disease — Can It Predict Outbreaks?

 After analyzing 17 million mentions of illness in Facebook posts and tweets (separating "Bieber fever" from actual fever) and plotting genuine ailment mentions on a map, founder of Sickweather, Graham Dodge, has noticed trends in how disease spreads throughout the United States.  For one thing, disease spreads most quickly between Hartford, Conn. and Washington D.C., an area he has dubbed "contagion alley."  And it is impacted by big events like the Super Bowl. When the New York Giants played the New England Patriots in Indianapolis this February, Twitter mentions of illness in Indianapolis spiked, more than doubling those in demographically similar city Dayton, Ohio.  Several studies have already suggested that social media can accurately track disease. But observations like these, Dodge thinks, can eventually be used to anticipate outbreaks faster than ever before.  His company, Sickweather, is the first consumer-facing product to take a serious stab at tracking disease through social media chatter.  Its technology scans social networks for mentions of 24 different symptoms. Through semantic analysis, it decides which mentions actually refer to illness, and plots them at the location where they originate. The result is an interactive map searchable by location and illness. Users can also add their ailments directly to the map.Sickweather shows social media mentions of disease symptoms on a map Tracking disease through online activity has been done before. Researchers from Google and Yahoo found that search terms were good indicators of flu activity in 2008 and 2009. In 2010 Google launched Google Flu Trends, which provides public estimations for flu activity.  Social media, however, is an improved digital disease indicator.  "It inherently provides more context to the individual's situation for natural language processing to better qualify what the person means, e.g. a tweet of 'I have the flu' versus a search of the word 'flu,'" Dodge explains.  There are a variety of indicators that help track disease, including emergency room visits and over-the-counter medication purchases, but, says Mark Dredze — who has co-authored a study about Twitter's applicability to public health research — "social media is one of the fastest."  While it might take weeks for the CDC to report disease trends, social media is rife with real-time data that can help public health officials more quickly anticipate demand for medication and health services.  Therefore, it's not a surprise that the Department of Health and Human Services is interested in the approach. It recently launched a contest calling for apps that not only track how disease trends have developed in the past, but that also use social media as an advance signal of a public health emergency.  "When we looked back at the H1N1 pandemic, we saw that, in some cases, social media trends provided the first clues to flu outbreaks," said Dr. Nicole Lurie, the assistant secretary for preparedness and response, when the contest launched in March. Sickweather is among the entrants vying for the contest's $21,000 prize.  Here's where it differs from the methods that have been proven before: Dodge believes that adding data from travel patterns, event schedules, weather patterns or environmental roadside data into the mix could allow Sickweather to anticipate outbreaks even faster than social data alone.  "Think of it like the early days of weather forecasting," says Dodge. "We're basically transitioning from the Farmers' Almanac to Doppler radar."