Here's an angle Mark Zuckerberg has probably not yet mined: restaurant reviews and bookings as surveillance tools for public health departments.
Speaking at a mobile health conference, John Brownstein, PhD, an associate professor of pediatrics at Harvard Medical School's Center for BioInformatics, said 10% of the restaurant reviews on Yelp are related to food-borne illness, and consumer's guesses about which ingredients provoked their distress are surprisingly accurate.
A study he co-authored and published in Preventive Medicine in October, demonstrated that Yelp could be used as a real-time surveillance tool.
"That informal data, that data that's not necessarily tied to the healthcare setting, is actually helping paint an important picture of what's happening," said Brownstein.
Cancellations on Open Table, an online service for booking restaurant reservations, are another surprisingly useful data source as they appear to peak during flu season. "By tapping into what tables are available at restaurants in different cities, you can actually map out flu," he told MedPage Today. That study was published in the Journal of Medical Internet Research.
A third report geocoded Facebook "likes" of sedentary activities, such as TV and movie watching, and compared them with "likes" for physical activity. The study conducted in urban areas across the country including New York City, showed that the couch potato "likes" mirrored the obesity prevalence in certain New York neighborhoods.
Facebook, Twitter, online chat rooms, even the search terms typed into Wikipedia, all of these digital channels can be used to help trace diseases and public health threats.
"What we're really interested in is what specific pieces of that data are related to disease and how we tap into that particular conversation set," he said.
Brownstein is the director of the Computational Epidemiology Group at Boston's Children's Hospital, who, along with Clark Freifeld, developed Health Map, a public health surveillance tool used to chart public health threats and disease outbreaks including Ebola.
He gave the keynote speech at the mHealth Summit, an international conference focused on mobile and connected health at the Gaylord National Resort and Convention Center in National Harbor, Maryland on Wednesday.
Traditionally, Brownstein said, knowledge of an outbreak happens through a ladder of communication: from the sick patient to the health worker, to the lab that verifies the disease on to the ministry of health who announces the public health threat.
"The ultimate time for detection of an event can take weeks, if not months," he said. His goal, right now, is to unwind that "hierarchical flow of information."
Health Map uses data from hundreds of thousands of sources -- social media channels, online chat rooms, blogs, news sites and government sources -- in 15 languages to map thousands of disease events that go unreported in traditional public health channels, Brownstein said.
The site uses natural language processing to deconstruct data and organize it into readable content. There's plenty of noise and there are "false positives" he admits, but using machine learning -- where algorithms train themselves through experience -- the site now boasts a 95% accuracy rate.
Brownstein and his colleagues are also currently building a platform that would allow state and local departments to see what communities are saying about certain health threats.
So, how does raising red flags help prevent a threat or stop an outbreak? Using the Yelp example, researchers could alert public health officials to the name and location of the restaurant where the food poisoning occurred and provide contact information for the individual with the complaint. At the consumer level, surveillance teams might set up automatic messages that pop-up when a person posts about a certain illness.
For a long time data mining has been one-directional, said Brownstein, but it doesn't have to be. "If you give a little bit back, people are more willing to be a part of it." One new app, he helped developed, Flu Near You, asks consumers to to comment on how they're feeling each week. The high resolution zip code data is available to the consumer in real-time, just as it would be to any other researcher.
For certain conditions, creating a sense of partnership with consumers around disease surveillance is less practical.
"For some reason people love to disclose that they have diarrhea on Twitter, but if we build Diarrhea Near Me, nobody wants to come and tell us they have diarrhea," Brownstein said.
Still, he and his colleagues are always looking for ways to give back to the individual. "Is it advice? Is it information that can funnel back to their electronic medical health record? Is it a health card that they can bring to their physician?" he asked.
"It's very difficult to target one sort of tool that has one sort of incentive structure."
He told MedPage Today there may be ways to, for example, harness the Fit Bit craze for public health purposes as well as clinical uses. Consumers could be given the choice to opt-in or not.
"If we have that kind of incredible Apple watch, Fit Bit concept plus your clinical record, the opportunities to mine, to look for potential risk factors or potential opportunities to deal with disease in different ways, I mean that's really exciting."