Researchers say their machine learning model using Google data was much more accurate at identifying sources of foodborne illnesses than traditional investigational methods.
 |
| A study found machine learning using Google data was significantly more accurate in identifying potentially unsafe restaurants when compared with consumer complaints and routine inspections |
By Allen Cone, UPI
Machine learning using Google data was significantly more accurate in identifying potentially unsafe restaurants compared with consumer complaints and routine inspections, according to a study.
Researchers at the Harvard T.H. Chan School of Public Health worked with Google to develop search and location data in a system called FINDER.
In findings published Tuesday in the journal npj Digital Medicine, the researchers concluded that their method can spot problems more quickly than the slow, cumbersome process of determining outbreaks based on consumer complaints or routine inspections.
"Foodborne illnesses are common, costly and land thousands of Americans in emergency rooms every year," corresponding author Dr. Ashish Jha, a professor of global health at Harvard Chan School and director of the Harvard Global Health Institute, said in a press release. "This new technique, developed by Google, can help restaurants and local health departments find problems more quickly, before they become bigger public health problems."
Google researchers, who developed the machine-learned model, worked with Harvard to test it in Chicago and Las Vegas.
The model first classifies search queries that can indicate foodborne illness, such as "stomach cramps" or "diarrhea." Then it uses de-identified and aggregated location history data from the smartphones of people searching those terms and of places recently visited.
In both cities, health departments were given a list of restaurants that were identified by the model as being potential sources of foodborne illness. Health inspectors, who didn't know whether their inspection was prompted by this new model or traditional methods, visited these restaurants.
The odds ratio that a FINDER restaurant was unsafe was 1.68 times higher than complaint inspections.
In Chicago, the model prompted 71 inspections between November 2016 and March 2017. The unsafe rate was 52.1 percent compared with 39.4 percent among inspections triggered by a complaint-based system. The researchers noted that Chicago already employs social media mining techniques, but this new model proved more precise in finding food safety violations.
In Las Vegas, there were 61 model-prompted inspections between May and August 2016. Complaint-based inspections from Las Vegas were not examined because the city complaints are handled differently -- the investigation is more focused on the nature of the complaint, rather than the entire establishment -- and the frequency of Las Vegas restaurant patrons, many of whom are tourists, leads fewer complaints to be filed.
The researchers noted that in 38 percent of all machine model cases, the restaurant potentially causing foodborne illness was not the most recent one visited by the person searching keywords related to symptoms. Foodborne illnesses can take 48 hours or even longer to become symptomatic after exposure, the authors said.
"In this study, we have just scratched the surface of what is possible in the realm of machine-learned epidemiology," said co-author Evgeniy Gabrilovich, senior staff research scientist at Google. "I liken the analogy to the work of Dr. John Snow, the father of modern epidemiology, who in 1854 had to go door to door in Central London, asking people where they took their water from to find the source of a cholera outbreak."
"Today, we can use online data to make epidemiological observations in near real-time, with the potential for significantly improving public health in a timely and cost-efficient manner," Gabrilovich said.
COMMENTS