Yelp reviews and Twitter can track food poisoning sources, Harvard study finds


On New Year’s Day in 2010, a group of three friends in Worcester visited a popular pizzeria for lunch. But by the next morning, all three regretted that trip.

“Today, I woke up deathly ill, and proceeded to kick off 2010 by vomiting,” one member of the group wrote on Yelp on Jan. 2. “The three of us had different items—not sure what took us all down—but we suspect Zorba’s as we all went our separate ways and are all deathly ill today.”

Posts like that, it turns out, can prove quite valuable to health officials.

Not only can such a review warn off diners, but it could also tip off public health departments to restaurants that are flouting health codes. Researchers at Harvard Medical School and Boston Children’s Hospital argue in a recent report that Yelp can serve as an indicator of  food poisoning outbreaks, and point to the food types that were the source of the bug.

“On Yelp, you can write [reviews] as long as you want,” Elaine Nsoesie, a post-doctoral researcher at Harvard Medical School and Boston Children’s Hospital, said. “They will tell you about when they got sick, what symptoms they have. All this information is pretty useful.”

Nsoesie worked with Boston Children’s researcher Sheryl Kluberg and John S. Brownstein, founder of social media disease tracker HealthMap, on the report published in Preventive Medicine. They read Yelp reviews posted to the websites of 5,800 restaurants around the country, written between 2006 and 2011. They searched for tell-tale words—vomiting, diarrhea, puking—to flag potential reviews and found that beef, dairy, and poultry were the most often involved. These matched the results reported by the Centers for Disease Control and Prevention.

Meanwhile, health officials are already putting Yelp reviews to work. Last year, the New York City health department collaborated with Yelp and Columbia University to analyze some 294,000 reviews written between July 2012 and March 2013. They uncovered three outbreaks that had slipped under the radar, and the foods that were the source of the illness, they reported in May this year.

Soon, the NYC department hopes to tap Twitter as a source of these tip-offs, a method that’s showed promise in yet another US city.

In Chicago, the city’s health department partnered with Smart Chicago, a Civic Technology group, to help citizens quickly report food poisoning incidents.

In March 2013, the FoodBorne Chicago project was launched. An algorithm read and flags tweets that carry tell-tale keywords, and real people read through the selected tweets, inviting a selection of tweeters to fill out an official report (they send a link to them).

In the first 10 months, the staff replied to 270 tweets, replying to complaining tweeters with a link to the reporting form. 193 reports were submitted, and officials inspected 133 restaurants. Twenty-one restaurants failed this inspection, and 33 were operating in conditions that the researchers described as “critical or serious violations.”

Back in Boston, Elaine Nsoesie is also developing a food poisoning tracker for Twitter. The projects in Chicago and NYC worked with historical data, but she hopes to build a system that can call for help in real time.

Image of suspicious tomatoes via Shutterstock

Nidhi Subbaraman writes about science and research. Email her at [email protected]
Follow Nidhi on Twitter - Facebook