If a close friend has a cold, chances are you might catch it. A startup called Sickweather hopes to tap into the social side of sickness with a social networking service that tracks illnesses within a user’s circle of friends, and to forecast outbreaks.
The startup mines publicly available data from social networks such as Twitter and Facebook, as well as from its users, to provide information on illness trends. Sickweather recently launched an early version of its site for closed beta testing, and plans to open to a broader audience in July.
While some people complain about the vast quantities of often mundane data uploaded to social networks every day, companies are increasingly interested in mining that information for commercial ends. Bluefin Labs, for example, uses social network data to determine users’ reactions to television shows and advertisements. Specialized social networks are also springing up to collect more precise data.
PatientsLikeMe, for example, is a social networking site on which patients can share information and experiences related to their condition.
Sickweather’s founders hope the data they collect can help users avoid catching their friends’ bugs. Users would be able to log in to the site and view a map showing illnesses in their area. They would also see updates from friend connections mentioning current illnesses. The company plans to release mobile apps for the site as well, so that users can view this information while on the go.
Graham Dodge, the company’s CEO, says he thinks Sickweather will be particularly useful to young families that want to be alerted to illnesses going around their social circle. “They can decide, ‘Maybe I won’t take my kids to that birthday party,’ ” Dodge says. Users might also be able to use the information to decide whether their symptoms match up to current common illnesses and merit more attention, or to prepare a health regimen before taking a trip.
Like any social network, much of the value of Sickweather will depend on getting enough people to sign up, but the founders hope to provide a useful service based on publicly available data on social networks as well.
Sickweather uses Twitter’s API to find information about sicknesses tied to a particular location. Michael Belt, Sickweather’s chief technology officer, says the company searches for keywords related to sickness. To tune the algorithm, the system uses a database of words that exclude certain posts. For example, the system would want to record a post that says, “I’m feeling so sick—my nose is running.” But it would exclude a post that says, “There were some sick beats last night at the club.”
Sickweather hopes to eventually use posts that simply mention the symptoms of an illness. But the initial version of the site is focused on words such as “bronchitis,” “pneumonia,” and “pertussis.” “It was eye-opening to see how much data we got just from specific technical terms,” says Dodge.
Dodge says the company is exploring advertising partnerships to gain revenue for the site.
Others are exploring mining social networks for public health data, and their experiences show how difficult it can be. For example, HealthMap, cofounded by Clark Freifeld and John Brownstein, brings together data from a variety of sources to show outbreaks as they happen around the world. HealthMap began by crawling news reports and blogs, but also has a partnership with Google to mine data from search terms. The company offers apps for smart phones, Twitter, and Facebook, through which users can report outbreaks.
For HealthMap’s purposes, Brownstein says, Twitter is a difficult source to mine. “You need a sizeable [health] event to get enough data,” he says. For example, flu outbreaks are usually large enough to get sufficient posts from social networks, but gastric illnesses are too isolated, he says.
Gunther Eysenbach, a senior scientist in the division of clinical decision-making and health care at the Toronto General Research Institute, has extensively studied tracking diseases through the Internet. He says that data from social networks is full of false positives. For example, people may post about articles they’ve read about a disease, and it’s hard to filter that out, he says.
Eysenbach also worries about the privacy implications. He says mining social connections could have a chilling effect on people’s willingness to post about illness, since it could create uncomfortable social situations, such as when no one shows up to that kid’s birthday party. Such an outcome, Eysenbach says, would be a shame, because such data has public-health value.