As one of the inaugural recipients of a Twitter Data Grant, a team of researchers from Harvard Medical School and Boston Children's Hospital will study new methods of detecting outbreaks of foodborne illness.
Last week, Twitter awarded privileged access to its data to six research teams across the world, according to an announcement on its Engineering Blog. The Harvard project, led by Associate Professor of Pediatrics John S. Brownstein, is the only one focused on disease detection.
The project builds on a growing branch of public health that uses Internet data to help detect and prevent disease outbreaks. Researchers have combed a number of data sets—Google and Wikipedia searches, Facebook likes, and Yelp reviews—for indicators that can inform disease detection.
Twitter offers an advantage over other data sources by enabling public health officials to not only collect data but also follow up with people, according to Brownstein. He cited a project called Foodborne Chicago, which automatically sends the city public health complaint form to Twitter users who mention possible symptoms, as a model.
Until recently, research using search engines and social media had focused on the flu, according to Brownstein. A recent paper in the Journal of Medical Internet Research found that of 32 relevant studies, only four dealt with foodborne disease.
Brownstein said that gastrointestinal illness is an "area that's ripe for using these new technologies, because people tend to describe these illnesses [online]."
Moreover, the Centers for Disease Control recently released a report, which documented that foodborne illness remains a significant public health issue in the United States.
Some scientists have challenged the usefulness of aggregated social media data for addressing such problems. Two recent studies call into question the most well-known example, Google Flu Trends, which claimed to provide accurate detection of influenza outbreaks through search data. Those studies claimed that Google missed a significant outbreak in 2009, and often predicted much higher levels of illness than actually occurred.
Donald Olson, a research scientist in the New York City Department of Health and Mental Hygiene who authored one of those studies, said that secondary data "doesn't allow public health to do what public health needs to do."
To use the information, public health officials need to be able to follow up with a person to determine whether a tweet about vomiting after eating at a restaurant is an actual indicator of foodborne illness, according to Olson.
Olson said big data projects should be "evaluated in the context of actual public health practices, and that almost never happens."
Still, Brownstein said that the work of data aggregation researchers is meant to augment existing public health surveillance systems, rather than replace them.
"It's absolutely not a holy grail," he said. “And it's definitely not a replacement."