Food Protection Trends - January/February 2020 - 58

Figure 1. Example of CART output (8).

effect in that if sanitarians know which restaurants will get
along without an inspection for longer than recommended,
they are likely to skip them.
This model was never deployed, with no reason ever
offered. However, it had never been requested by the City of
Columbus, and no city time was used for its development.
The Ohio State University provided the statistical software
for initial processing of the data as well as the CART software. At that time (1992), the City of Columbus may not
have had the necessary data processing resources; it had not
yet even computerized its complaint system (8). The model
is mentioned here because it seemed to work, has potential
as a predictive model, and is in general agreement with the
first Chicago model, described later; as Table 1 shows, its
input was remarkably similar. The CART model also had
some of the predictor variables in the earlier models previously cited, but with some important differences: it could
be run separately for different kinds of food operation (e.g.,
markets and food service operations); it could predict various outcomes (such as inspection failures or time-temperature violations, specifically); and it was self-updating, in the
sense that after the results of inspections it had scheduled
were entered, CART could be re-run to predict the next
batch of inspection results.
Surveillance system using a weight-loss website
The Columbus Health Department (now Columbus Public
Health) applied in 1995 and 1997 for grants from the U.S.

Food Protection Trends January/February

Department of Agriculture Cooperative State Research,
Education, and Extension Service National Research
Initiative Competitive Grants Program (NRICGP) to fund
a surveillance system for sporadic foodborne illness, using a
weight-loss website. After health department management
priorities changed, the Scioto Valley Health Systems Agency
(SVHSA) applied for the same grant for the same purpose in
2000. SVHSA was a consortium of 15 county and city health
departments, including the Columbus Health Department.
The original proposal (9) was modified after each grant
application and has also been described in a short article in
Food Protection Trends (10).
The basic idea was to identify potential sporadic cases of
foodborne disease originating with restaurants among users
of a weight-loss website that would allow people with weight
problems or other conditions requiring dietary monitoring
to log on and enter a daily food history. It was expected that
a small number would experience symptoms of a foodborne
disease later and that they would be able to report them on
the site. This would minimize problems of recall bias, etc. The
Health Department would investigate commercial sources of
food with methods usually used in outbreak investigations.
The proposal included use of ATP bioluminometers to focus
environmental testing and immunomagnetic separation,
followed by PCR, to expedite microbiological testing.
Offering a bounty for stool samples was proposed to obtain
clinical specimens from squeamish clients. By using a
prospective rather than a retrospective design, the system

Food Protection Trends - January/February 2020

Table of Contents for the Digital Edition of Food Protection Trends - January/February 2020