Using millions of tweets, a study showed the efficacy of gathering public information via the microblogging service and how it can track illness behaviors, as well as what people are taking to alleviate symptoms.
Another use for the publicly available information on social networks has been published this week. Researchers studied how Twitter could be used as a crowdsourcing tool for ailments and health information last year, and were able to track things like allergy season in a new way from filtered tweets (full report PDF). While previous studies in this area have focused on influenza, the Johns Hopkins University report attempts a more general approach.
Researchers Michael J Paul and Mark Dredze collected two million tweets from May 2009 to October 2010, and used filters and keyphrases to identify health information and remove redundancies and links. This left over eleven million tweets which were then further classified to remove health words used in different context, "e.g., "'I'm sick of this," and justin beber yer so cool and i have beber fever.'" The remainders were categorized under "sick," or "health."
From the collected findings, after the irrelevant data was removed, three groups of analyzable material emerged: general ailments, specific ailments, and treatments that Twitter users were undertaking. These three types of data sets were applied to seven types: allergies, insomnia, obesity, injuries, respiratory, dental and aches/pains.
Web data can provide an effective alternative to published statistics from the Centers for Disease Control and Prevention (CDC), according to this report. Google Flu Trends tracks influenza rates daily, as well as up to seven to ten days faster than the CDC's FluView. The general approach attempted by the researchers on this Twitter project hope to extend this type of immediacy to the rest of health tracking. Their process tracked ailments and treatments, which were used to track diseases over time as well as geography available publicly through Twitter's location features. This method also could read how often people mis-medicate their own symptoms, as coverage from the BBC explains.