Dos conciencias

Researchers are mining social media for insights into our well-being — or lack thereof.

Alice Mongkongllite / BuzzFeed

For years, Alicia McGarry's elementary school–age sons tended to get sick easily. But two years ago, McGarry downloaded Sickweather, an app that broadcasts infectious illness warnings based on location-specific chatter on social media. Now, when an alert goes around, the 35-year-old mother doubles down on cold and flu precautions, reminding her boys to wash their hands and eat leafy greens. "We've been able to minimize family illness," McGarry, of Kansas City, Missouri, told BuzzFeed News.

We yak endlessly on social media about our lives and the physical discomforts we sometimes encounter as we live them. We tweet about our headaches and chills, and seek solace on Facebook when we're feverish. In doing so, we're creating an unprecedented trove of anecdotal health data. And now researchers are digging through it to unearth some surprising things about our well-being — or lack thereof.

One of the first programs in this vein, Google Flu Trends, attempted to use aggregated search data to instantly estimate flu outbreaks worldwide, though its accuracy was questioned by some critics. That was 2008, though, and the flu was just the start. Researchers have since tapped other social media reservoirs — namely Twitter — to search for new insights into mental health, hospital admissions, heart disease mortality, and air pollution. Meanwhile, startups like Sickweather, which tracks and maps reports of illness on social media, and Iodine, which crowdsources medication reviews, are now translating digital health complaints into meaningful information for people who aren't doctors and data scientists.

Sickweather CEO Graham Dodge claims that his app has successfully warned people about the start of flu season before the Centers for Disease Control and Prevention did, and beat Chicago media in identifying whooping cough around the area. Dodge, whose background is in web development, started the company about four years ago, after he'd contracted a stomach virus and saw a friend complaining of similar symptoms on Facebook. "That's when it occurred to me that social media could be used as a valid source of health data — if it could be mined correctly [and] you could reduce the signal-to-noise ratio," Dodge told BuzzFeed News.

A snapshot of ailments in the San Francisco Bay Area on June 25, according to Sickweather.

Sickweather / Via sickweather.com

And there is a lot of noise. That's one reason some scientists look askance at efforts like the one Dodge is undertaking with Sickweather. Having "Bieber fever" and being "so hot right now" are clearly not clinical symptoms, though software — unaware of such colloquialisms — might flag them as such. Avid social media users also aren't necessarily representative of the population. For those reasons, public health agencies are taking a cautious approach to social media health analytics.

"Kids under 13 aren't allowed to sign up for [Twitter] accounts. Adults who are over 65 are less likely to be present on Twitter," said Dr. Matthew Biggerstaff, an epidemiologist at the CDC, which has experimented with flu tracking on social media. "Those are groups where we feel like we might not be getting as complete a picture on social media as we do in our more traditional surveillance systems, where we have good coverage of all age groups in the U.S."

---

View Entire List ›