How Big Data Helped Chicago Improve Its Food Safety

By Lisa Rabasca Roepe, Contributor

When it became clear that the city of Chicago’s three dozen food inspectors couldn’t possibly execute sanitation checks on the city’s more than 15,000 food establishments, the Chicago Department of Public Health turned to big data to help better allocate its resources.

The Chicago Department of Public Health teamed up with the Department of Innovation Technology and two external partners—Allstate and Civic Consulting Alliance—in the fall of 2014 to create and pilot test an algorithm that could forecast which food establishments were most likely to have critical violations.

Using data from Chicago’s open data portal—including the results of nearly 100,000 sanitation inspections and statistics from the city’s non-emergency 311 complaints—the Departments of Public Health and Innovation Technology created a predictive model that produced a risk score for every food establishment in the city. Establishments with higher risk scores were inspected first, allowing critical violations to be more quickly detected by the Department of Public Health’s inspectors before patrons became ill.

Using big data, the city of Chicago deploys its health inspectors to the highest risk establishments first, allowing inspectors to discover violations 25 percent faster, according to Danielle DuMerer, CIO and commissioner, Chicago Department of Innovation and Technology.

In 2016, the city began deploying its health inspectors based on the algorithm, allowing inspectors to discover critical violations 25 percent faster, on average, than if inspectors had used the traditional method, says Danielle DuMerer, CIO and commissioner for the city’s Department of Innovation and Technology. The algorithm is a prime example of how public health can use analytics to promote data-driven policies.

“Chicago Public Health is now able to discover critical violations and prioritize which inspections have to be done first with cutting-edge data analytics,” says Gerrin Cheek Butler, manager of food protection services at the Chicago Department of Public Health.

Reducing Food Borne Illness

The algorithm is aimed at prioritizing inspections to reduce occurrences of food borne illnesses, DuMerer says. According to the CDC, food borne illness is an under-reported problem that affects an estimated 48 million Americans annually, resulting in 128,000 hospitalizations and 3,000 deaths, exceeding $15 million in medical and industry costs. Food borne illnesses make up about 60 percent of outbreaks in dine-in restaurants, the CDC finds.

By reducing the number of food borne illnesses, Chicago could decrease the economic cost of food poisoning to its residents.

By reducing the number of these illnesses, Chicago could decrease the economic cost of food poisoning to its residents, including unanticipated healthcare costs and lost time at work, DuMerer says. The pilot project was so successful that, by spring 2015, Chicago was using the algorithm to prioritize all of its restaurant inspections.

Nine Key Pieces of Data

The algorithm looks at nine key pieces of data:

  • Previous critical or serious violations
  • The three-day average high temperature
  • Nearby garbage or sanitation complaints
  • Nearby burglaries
  • The type of eating establishment being inspected (dine-in versus takeout)
  • Tobacco or liquor licenses
  • Length of time since the last inspection
  • Length of time the establishment has been operating in Chicago
  • The inspector assigned to the establishment

Due to the complexity of the analysis, DuMerer says, it’s difficult to explain how each piece of data is used in the algorithm. However, restaurants with previous critical violations are more likely to have an incident of food borne illness, she says. For example, on the West Coast, Seattle’s King County has found a similar correlation. “Restaurants with poor inspection scores and violations of proper temperature controls of potentially hazardous foods were, respectively, five and 10 times more likely to have outbreaks than restaurants with better results,” according to a study in the American Journal of Public Health.

Although the Chicago Department of Public Health does inspect food trucks, their data collected from those inspections isn’t currently part of the algorithm. In addition to using the algorithm, the city of Chicago hired 10 additional inspectors, bringing the total to 46, says Butler. “Chicago Public Health is now able to discover critical violations and prioritize which inspections have to be done first with cutting-edge data analytics,” DuMerer adds.

Other Cities Can Adapt the Algorithm for Free

The city has made the algorithm available to other municipalities at no cost. The analytic code is written in R, an open source, widely known programming language for statisticians. “The data and software have been released as an open source project, so it can be reviewed, adopted, and improved by researchers and other cities,” Butler says.

“We have put the human resources in the right places to protect public health and food safety.”—Gerrin Cheek Butler, manager, food protection services, Chicago Department of Public Health

However, DuMerer admits it’s unlikely another city could simply pick up the code and run with it. Other municipalities would need to look at the data they collect and compare it to Chicago’s. Differences in city ordinances could lead to differences in how violation information is captured or assessed, DuMerer says. Before committing to the program, a municipality should run a pilot with its inspectors to identify whether the model actually reduces its time to discover critical violations, she suggests.

“We have put the human resources in the right places to protect public health and food safety,” Butler says. “By finding critical violations sooner, there is an opportunity for establishments to fix the problem sooner and prevent illness.”