What’s the best way to raise awareness of a disease? A macabre joke, well-known to those in public health, says that your best bet is for a celebrity to contract it.
Dark, but true. “When a celebrity has a disease, money and attention flow from that,” says Mark Dredze, an associate professor in computer science at Johns Hopkins University (and a member of the data science team at Bloomberg) who also presented research at the 2016 Data for Good Exchange about how people respond to the gun control debate on social media. Michael J. Fox, for example, has been hugely influential in helping the public better understand Parkinson’s disease.
But how well does that awareness translate into action? Dredze, along with a team of researchers, looked at Charlie Sheen’s November 2015 announcement that he was HIV-positive for evidence of just this sort of behavior. In this case, they wanted to understand whether Sheen’s announcement spurred consumers to actually test their own HIV status.
In a paper published in JAMA Internal Medicine in February 2016, Dredze and his collaborators showed that, in the weeks following Sheen’s announcement, Google searches for terms related to HIV, as well as for information about symptoms of the disease, spiked to all-time highs. Concurrently, there was a jump in the number of news items that mentioned the condition. Together, those were promising signs.
The trouble with using awareness as a metric, however, is that it still doesn’t tell us if people took action. So while Sheen raised awareness about AIDS, it wasn’t clear if his disclosure changed behavior and inspired more people to get tested. For that, Dredze and his colleagues worked with one of the only manufacturers of in-home HIV testing kits, which made its sales data available to them.
The results were stunning. “We found significant increases in the number of tests sold,” says Dredze. “Tests almost doubled the week of Sheen’s disclosure, and remained higher the rest of the month.” Their research was published in the July 2017 issue of Prevention Science. “This sort of celebrity effect on public health is well-known,” says Dredze, “but the magnitude of it was still surprising.”
“The behavior we observed online actually translated into the real-world behavior of being tested,” says Dredze. “That’s really good news in finding a way to measure the impact of public-health initiatives. In the future, if we see a spike in searches, we can now say that’s a potential sign that people are going out and buying test kits.” And it supports the idea that web searches and social media activity around, say, The Great American Smokeout, are really an indicator that more people are trying to quit smoking. In fact, public health advocates, Dredze says, are already trying to better understand trends in mental health in the population by looking at activity on different search terms.
Dredze’s research about the so-called ‘Charlie Sheen effect’ demonstrates how data collected for one use can have broader applicability. He notes that Bloomberg collects news stories from all over the world which are initially analyzed for their relevance to the financial markets. But that data also includes many other things that have broader applications, such as news stories about HIV. “Once we have all these data, public health researchers can use it as well,” says Dredze. “The same things we use to analyze the market, others can use to improve the public good.”
Such uses of big data for the public good will be showcased at Bloomberg’s 4th annual Data for Good Exchange (D4GX) on Sunday, September 24, 2017.