bst2012 - Fotolia
BOSTON -- Big data has given rise to the "digital epidemiologist."
John Brownstein, a Harvard Medical School professor and director of the Computational Epidemiology Group at Children's Hospital in Boston, didn't coin the term, but he employed it enthusiastically in a rapid-fire presentation on putting text analytics at the service of public health officials performing disease surveillance.
For Brownstein, the keynote speaker at a conference sponsored by Fresh Data Boston and software companies Basis Technology and RapidMiner, parsing unstructured health-oriented text goes beyond making sense of physicians' clinical notes and aggregating EHRs, though it does that, too.
What Brownstein and his group of researchers does is scour textual information from such nontraditional sources as search engines and social media -- particularly Twitter and its hashtags, but also Instagram, Facebook and others -- to detect public health emergencies and gauge their true scope.
Norovirus, flu outbreaks first detected online
For example, a recent norovirus outbreak at a Vancouver hotel burst into social media via Twitter hashtags well before it was picked up in the traditional press. News of nasty new flu strain outbreaks around the world, especially in third-world countries, often is being disseminated first and widest through social outlets, Brownstein said. In China, health activists have snapped photos of EHRs and uploaded them to social media en masse to force the government to be more transparent, he noted.
John BrownsteinComputational Epidemiology Group director, Children's Hospital Boston
Text miners and epidemiologists then use their software tools to build medical taxonomies out of the information after trying to correct for duplication, rumor and inaccuracy with such techniques as cluster analysis.
"Social media is transforming the ways we think about disease outbreak," Brownstein said in his talk to about 200 attendees of the District Hall event. "The things people say and do online [are] different than what they report to physicians, and [it] doesn't show up on EHRs."
Yelp: Serious public health tool?
Brownstein also uses Yelp reviews as a serious source of data on food poisoning outbreaks. So does the New York's municipal health agency, which he said used Yelp to build part of its database.
"We're trying to think through various channels, and by using these kinds of data we can actually recognize these outbreaks much more quickly," Brownstein said. "The goal is to unlock data. It's digital detection."
Using unstructured text, Brownstein and his colleagues have already built a public, worldwide epidemic tracking site called HealthMap that uses geocoding and data mining to graphically represent disease surveillance and outbreaks. It is being used by the World Health Organization, CDC, U.S. Department of Defense, European Union and other organizations.
While Brownstein is an evangelist for nontraditional data, he also warned against hype, which he said can be controlled for by carefully curating taxonomies of data gleaned online. "One has to be very careful," he said.
In an interview with SearchHealthIT, Brownstein, who was wearing a Fitbit pedometer band, said such consumer mHealth devices have significant potential for patients to deliver data straight to doctors and researchers. "It's going to be huge," he said.
In particular, sleep research is a key area for using data mining to measure how people describe sleep quality in social media, he said.
Patients participate in their care through social sources
How social media and ACOs are linked
A healthcare-specific virtual assistant is on the radar