One of the most interesting presentations I attended at the 2012 Health 2.0 Conference in San Francisco was presented by Adam Sadilek from Fount.in. What Adam and his co-developer from the University of Rochester have done is develop a mechanism to analyze Twitter data in real time in order to identify disease prevalence based upon analysis of natural language within the tweets.
How it works: using an analytical formula, Tweets from a specific geographic area are monitored for key words. For example, “I feel sick”, “I have a sore throat”, or “My body aches”. This may indicate an outbreak of influenza. Similarly, tweets of symptoms for other diseases can be tracked. Using geo-mapping in conjunction with Twitter, Fount.in identifies the location of Tweets and colour codes them according to their match to key words. Closely matched Tweets are coded red while others that are less of a match are coded orange or yellow.
By tracking the geo-tagged tweets over time, it could be possible to identify outbreaks of a disease or disease spread in real time, providing new tools for surveillance during epidemics.
Click on the individual dots to read Tweets or view the chronological timeline.
We are just beginning to identify uses for large data sets and social media. This particular solution is fascinating in terms of how the technology could be applied.