ADVANCEMENT OF THE SCIENCE
was published, “its readers were outraged— but not in the way Sinclair had hoped. Their primary concern was food quality.” Because of the public response, the Pure Food and Drug Act of 1906, which ensured that meatpacking plants processed their products in a sanitary manner, and the Fed- eral Meat Inspection Act of 1906, which required that the USDA inspect all live- stock before slaughter, were passed in 1906 (Lohnes, 2022). The findings by Mead et al. (1999), however, bring the overall eective- ness of these programs into question. Predictive Modeling for Foodborne Disease Hartman (2020a) described a variety of pre- dictive models for scheduling facility inspec- tions based on the facilities’ probability of food code violations; the violations can be thought of as both risk factors and risk mark- ers for foodborne illness. One early model (Hartman, 1992) used past inspection results to predict future ones by applying classifica- tion and regression trees (CART) software (Breiman et al., 1984) to food safety inspec- tion data provided by the city of Columbus, Ohio, to develop a new predictive model. This “tree-structured approach” is an example of artificial intelligence (AI): machine learning without specification of the variables to use. The class intervals of the continuous random variables were selected by the software. The results from one classification tree (Hartman, 1992) are shown in Figure 2. A learning sample of 1,000 full-menu restau- rants contained 528 restaurants that failed a standard inspection at least once in 5 years. CART made its first split with the question: “Was the standard deviation of scores in the previous year above 1.95?” Of the 573 estab- lishments included, 398 indicated they had failed at least once. The bar at the right end of the box indicates that this was a “termi- nal node” and was not split further. Of the 427 for which the first answer was no, 159 establishments that did not have an extra inspection included 86 failures and could not be split further. Another split of the 268 establishments that had an extra inspection identified 41 failures by asking if the average interval between inspections was ≥241 days. Of those establishments, 36 had 2 or 3 extra inspections in the previous year. This procedure was self-updating, in the sense that after the results from inspections
TABLE 1
Example of a Restart Inspection Form
Requirement
Yes
No
No self-service, including straws, stir sticks, buffets, and condiments. Self-service beverage stations are allowable. Employees must perform daily symptom assessments. Require employees to stay home if symptomatic of COVID-19. Ensure handwashing by employees when changing tasks, when hands become contaminated, and at a minimum once every 2 hr. Ensure social distancing between employees. Provide COVID-19-compliant floor plan for the kitchen. Post COVID-19-compliant seating plan that ensures social distancing and states maximum dining capacity. Number of employees permitted in break room limited to 10. Compliance with OAC training requirements for person in charge (PIC) and manager certification. Provide training for all staff on COVID-19 prevention. Have a plan to immediately isolate and seek medical care for any person who develops symptoms of COVID-19 while at the facility. Shut down areas affected for deep sanitation if possible. Perform enhanced cleaning of commonly touched surfaces such as doorknobs, railings, and countertops at least every 2 hr. Ensure high-volume shared surfaces in congregate areas, including playing cards, arcade controllers, and pool cues, are cleaned between customers. Clean and sanitize tables, chairs, and menus between customers. Ensure online and remote access reflects changes in response to COVID-19. Provide access to hand sanitizer for customers. Reservations for no more than 10 people per party. Ensure distancing of 6 ft between parties, otherwise use barriers. Barriers must meet all applicable building and fire code requirements. Maintain compliance with ODH sanitation and food safety regulations.
Note. OAC = Ohio Administrative Code; ODH = Ohio Department of Health. Source: Recreated from the Delaware General Health District, 2020.
were entered, CART could be rerun to predict the next batch of inspection results. Another predictive model for scheduling facility inspections is FINDER (Hartman, 2020a; Sadilek et al., 2018), mentioned pre- viously (see Misconceptions section). In this strategy, anonymous visitors to restaurants had their smart phones set to share their location data. The entire sequence of loca- tions each one visited within 3 days prior to the user performing a Google search of web pages about foodborne illness were included—pages such as Wikipedia articles and the CDC website about foodborne ill- ness. FINDER classified the web searchers as “sick” or “not sick.” The result was that
52.3% of restaurants identified by FINDER had serious health code violations, compared with 22.7% for other restaurants.
Risk Characterization and Predictive Modeling for COVID-19
Risk Characterization for COVID-19 After the first three suspected cases of 2019 novel coronavirus (initially known as 2019- nCoV or severe acute respiratory syndrome coronavirus-2 [SARS-CoV-2] and renamed by the World Health Organization as coronavi- rus disease 2019 [COVID-19]) in Ohio were confirmed on March 9, 2020, the State of Ohio closed restaurants and bars except for carryout
24
Volume 88 • Number 2
Powered by FlippingBook