Data-driven restaurant inspections

How can open data be useful to our department? How can it benefit the public?

Some variation of these questions arose frequently during our interviews with Hamilton County employees. It’s challenging to prioritize open data when budgets are tight and its usefulness seems remote.

Metro Ideas Project took a look at how one county department would benefit from open data.

The Hamilton County Health Department handles a wide range of data and records. Much of it consists of sensitive private health information. But the department also collects one type of record in which there is much public interest.

Converting restaurant inspection reports to open data would increase public accountability and lead to more efficient government.

Benefits

This data would allow easier comparisons between restaurants, neighborhoods and regions. It could be integrated into existing apps and services like Yelp to give consumers reliable information about sanitation and safety.

Data-driven solutions would also increase department efficiency and lower costs. Inspection reports are filled out by hand and later transcribed to a digital format before being sent to a statewide database.

The scores published by the state of Tennessee are only accessible through a search form, which makes app integration and comparisons difficult. State data is available through a records request, but the information is frozen at the time it was sent.

Challenges

We spoke with Hamilton County and Tennessee officials and identified several challenges converting restaurant inspection to open data.

The Health Department has six environmental inspectors responsible archives written inspection reports. It does not maintain a digital database of records. The reports are sent to the state.

The state’s restaurant inspection search portal is hard to use and error-prone. Some restaurants are shown with scores of 0. There are no fine-tuned search fields (e.g. search by city or ZIP code).

The underlying data is only available with a records request. Over the summer, the state sent us inspection data in a 200-megabyte XML file, which is only useful to people with programming experience. Additionally, that dataset is outdated.

There are some simple steps the Health Department could take to help overcome these challenges.

• Request and maintain a local subset of the state’s larger dataset
• Publish the subset on a data portal or the department’s own website
• Ensure the information is updated monthly

Limited personnel

Hamilton County’s Health Department has a limited number of inspectors, who follow guidelines from the Food and Drug Administration. The FDA recommends that every food establishment be inspected twice a year. The federal agency also recommends there be one full-time inspector for every 280 to 320 inspections.

Hamilton County has six environmental inspectors who are responsible for checking more than 1,400 food establishiments. Each inspector has to make 630 visits in order to ensure each restaurant is inspected twice a year.

These inspectors are also responsible for permitting and inspecting every public swimming pool, hotel, motel, organized campground, tattoo parlor and piercing studio in the county.

Data-driven solutions

Health departments in Chicago and Montgomery County, Maryland, faced similar pressure to meet inspection requirements with limited personnel.

Chicago’s Public Health Department partnered with the Innovation and Technology Department in 2014 to prioritize restaurant inspections with open data.

They combined datasets to create a forecasting model that helps prioritize inspections by establishments where critical food violations are most likely to occur. Inspectors found 14 percent more critical violations in the first two months of implementation. Chicago still uses this model and has open-sourced the code behind it.

Montgomery County, Maryland, adopted the Chicago model in 2015. The county worked with a private vendor to tailor the software to its own needs. After two months, county inspectors identified 27 percent more violations. Officials there expected to save $2 million in personnel costs in the first year as a result.

Chicago and Montgomery County started with one advantage Hamilton County doesn’t have. They both had existing open data portals. But their work in predictive restaurant inspections demonstrates a practical benefit of open data.

The Hamilton County Health Department and county government would have to take several steps to begin moving in this direction:

• Make restaurant inspection data open
• Work with other departments and agencies to flesh out the data requirements
• Consult with a vendor or app developers to adapt the Chicago model

The Health Department has the potential to lead an open data initiative in Hamilton County. A forecasting model would help the department provide better service for citizens, improve quality of life and save taxpayers money. And it would position other county departments to adopt their own data-driven solutions.