Twitter users send around 500 million tweets a day, an endless fire hose of information about how people feel, what they’re doing, what they know and where they are.
For epidemiologists and public health officials, it’s a potential gold mine of data, a possible way to track where disease is breaking out and how it spreads, as well as how best to help — but only if they can figure out how to find the useful signal amid all that noise.
“The question is: How do you take these billions of messages, find the useful information and get it to people who can respond?” says Mark Dredze, an assistant professor of computer sciences at Johns Hopkins University, who studies computational linguistics.
That’s a very big question, one whose difficulty has pushed many researchers away from the idea of using Twitter data, which they say is too messy and too uncontrolled compared with traditional methods of collecting health data, such as surveys and analyses of hospital visits. Others argue that, once we learn to effectively harness the data, Twitter’s very messiness (including the impulse to tweet what you had for breakfast or how annoying your runny noise is) will be what makes it an invaluable resource.
“It’s like a pulse on the world, because people will just tweet whatever, whenever,” explains Christophe Girraud-Carrier, an associate professor of computer science at Brigham Young University, who studies what he and his colleagues have dubbed “computational health science.” “Poll answers are filtered by perception or memory; on Twitter, we’re actually observing real behavior” in real time.
Using Twitter data has other advantages, Dredze says. For starters, it’s faster: It can take the Centers for Disease Control and Prevention about two weeks to publish findings, Dredze says. Those numbers can additionally be delayed by the fact that a sickness doesn’t show up in statistics until someone goes to the hospital or does something else that causes the ailment to be reported.
Twitter, on the other hand, might reflect it the first morning someone wakes up with a sore throat. Speed can be a big advantage when tracking epidemics and emerging diseases, says Taha Kass-Hout, director of the CDC’s Division of Informatics Solutions and Operations. “An emerging disease from Southeast Asia can be in your back yard in 12 to 14, maybe 24 hours. So you have to respect that.”
Twitter can also provide a more detailed picture of where disease is breaking out, since many tweets are tagged with their locations. That, coupled with faster data, could help keep hospitals and clinics from getting overwhelmed in the middle of an outbreak: Even a few days’ notice that disease occurrences are spiking can mean being prepared with extra beds, staff or medicine. Detailed, location-specific data can also identify clumps of noncommunicable diseases — cardiovascular disease or Type II diabetes, for example — allowing health officials to focus education efforts in the areas that need it most.
Twitter is also in increasingly wide use, including in countries that don’t have effective public health tracking agencies. “In that case, anything Twitter can provide — whether it’s fast, slow whatever — is really valuable,” Dredze says.
Those advantages, coupled with the fact that researchers are getting better at tracking and analyzing useful information, mean that “consensus is forming in the public health and health-care communities that we really need to pay attention to social media,” Kass-Hout says. However, he stresses that social media information is “a complementary tool, rather than a replacement” for more traditional methods of gathering information. It also depends on validation, the ability to prove that data collected through Twitter have real-world accuracy. That was one goal of Dredze’s research: to confirm the utility of Twitter data by studying if tweets about the flu could be filtered in such a way that they tracked with official flu rates.
Central to that effort is the signal-in-the-noise question, the effort to find and isolate useful information amid the barrage of tweets. In May 2011, Dredze and his colleagues were using a computer program to monitor mentions of the flu on Twitter. Suddenly, there was a massive spike in chatter. “It didn’t make any sense to us,” Dredze said. “The flu season was pretty much over.” They drilled down and discovered that people were discussing the fact that Kobe Bryant of the Los Angeles Lakers had played a game while sick.
That information may be interesting to basketball fans, but it’s not the kind of news that health researchers are looking for.
Dredze and his colleagues decided they needed a better algorithm, one that would allow the program to filter out tweets that aren't actually about people having the flu. Their system starts by searching for some key words (such as “flu,” “fever” and certain brands of medicine) and screening out others (including “Bieber” with “fever” is a good sign that someone’s not talking about having the flu; so is including a URL, since it probably means they’re simply sharing an article), then applying grammatical analysis to figure out whether someone actually has the flu or is just talking about it. (Is “flu” the subject or the object of the verb? Which verbs are used? Which pronouns?)
They tested the system when reports of the latest flu epidemic hit the media in January. The number of tweets mentioning the flu shot up, though most of them didn’t reflect actual cases. But when Dredze and his team filtered tweets through their algorithm, they matched the CDC’s findings about actual flu rates.
Meanwhile, another key problem — underrepresentation of certain demographic groups, including the very young and the elderly — is diminishing rapidly as Twitter use expands, Kass-Hout says. Likewise, research is beginning to show that location data is indeed accurate enough to be of statistical use.
That leaves researchers and public health officials pondering the possible applications of Twitter research — for example, using tweets to map urgent needs in the wake of natural disasters or to determine where vaccines are most needed following an outbreak.
Another possibility is using Twitter to better understand and respond to health-related behavior. For instance, Dredze says, the Johns Hopkins study turned up evidence that “a significant percentage of people who had the flu mentioned antibiotics”— a troubling finding since antibiotics don’t cure the flu, a virus, and their misuse can increase drug resistance. Knowing just what misinformation they’re combating can help officials better target educational efforts.
Giraud-Carrier’s work revealed details of prescription drug abuse; he and his colleagues also studied whether algorithms can be created that flag cases of potential suicide or domestic violence before they happen.
“I don’t want to just be listening and finding out all these bad things that are happening,” he says. “In the long run, our vision is to do something more than listen.”
Such ideas carry privacy concerns, which have yet to be resolved. Tweets are public information; still, Kass-Hout says, public health officials will have to be careful about respecting privacy as they figure out how to use information gleaned from them.
Another goal is to create ways for social media users to benefit directly from aggregated data — to be able to look up disease rates in their neighborhood, even to compare cases of food poisoning in order to trace them to a particular restaurant. Web sites are already popping up to provide this service; Greg St. Clair, whose day job is in software quality assurance for medical companies, created one of them, GermTrax.com, after a bad illness made him wish there were a way to figure out where he’d gotten sick. At first he simply asked users to input their own data; when there wasn’t enough, he started importing information from social media sites.
“We’re using social media to leverage the data,” says St. Clair, who never did figure out where he’d gotten sick. “But we’re also giving that information back to the public.”
That is the same reason why Dredze and his colleagues recently released their findings on improved algorithms for tracking the flu.
“Normally we wouldn’t publish material like this until we’d had a chance to really analyze what’s happened, the whole flu season. But we think this technology can be beneficial right now in the middle of the epidemic.”
Jarvis has written for Rolling Stone, the Atlantic and the American Prospect, among others. And, of course, you can follow her on Twitter.