Food Protection Trends - July/August 2020 - 237

TABLE 2. Pearson correlation matrix for 2016 census-level demographic variables
in Toronto
Variable

POP

AGE

LIM

IMM

LAN

Population density per km2 (POP)
Median age (AGE)
Low-income measure (LIM)
Immigration status (IMM)
Non-official language spoken at home (LAN)

-0.377
0.378
0.012
0.074

-0.359
-0.018
-0.019

0.511
0.519

0.916

-

Labor population working in accommodation
and food service (LAB)

0.172

-0.155

0.502

0.309

0.415

LAB

-

TABLE 3. Bivariate negative binomial regression modeling results of the association
between census-level demographic variables and the rate of CP/C instances
per census tract (n = 469)
Variable

IRR

95% CI

P-value

Alphaa

Population density (per km2)
Median age
Low-income measure (LIM-AT)
Immigration status
Non-official language spoken at home
Labor population working in accommodation and food service

1.000
1.011
1.013
1.014
1.015
1.012

1.000, 1.000
0.996, 1.027
1.002, 1.023
1.008, 1.021
1.008, 1.022
0.983, 1.041

0.485
0.146
0.013
< 0.001
< 0.001
0.430

0.445
0.434
0.426
0.389
0.388
0.445

CI = confidence interval; IRR = incidence rate ratio
a
This value is the over-dispersion parameter, which would equal zero if over-dispersion were not present (i.e., a Poisson model).
In all models, the parameter was significantly different from zero (P < 0.001) with use of a likelihood ratio test.

households that are classified as immigrants (IRR = 1.014,
95% CI = 1.008, 1.021) and that speak non-official languages
at home (IRR = 1.015, 95% CI = 1.008, 1.022) were associated
with higher rates of CP/C outcomes across both study years
(Table 3). A similar association was found with regard to
the percentage of low-income households (LIM-AT) (Table
3). For a 10% increase in these variables in a census tract,
the rate of CP/C outcomes increased by a factor of 1.15
(immigration status, 95% CI = 1.08, 1.23), by a factor of 1.16
(non-official language, 95% CI = 1.09, 1.25), and by a factor of
1.14 (LIM-AT, 95% CI = 1.03, 1.26). Population density was
not related to the CP/C rate in a census tract, while median
age and the percentage of working households employed
in accommodation and food service were not consistently
associated, with 95% CIs crossing the null (Table 3).
DISCUSSION
Our results contrast with Lee et. al's spatial analysis of
Miami restaurants (21); we did not identify Toronto's

waterfront and downtown core areas as significant hot
spots, or the areas near the island airport as a significant
cold spot. In 2017, we found, with 99% confidence, that
two of four hot spot clusters were located along the eastern
shoreline. They overlapped with two of 11 public beaches
in the city. However, in 2018, the only hot spot identified
with 99% confidence along Toronto's waterfront did not
overlap with any public beaches. Additionally, no hot
spots of any significance were identified within the core
downtown of Toronto in either year, although a few hot
spots were identified around the edges in both years. These
findings suggest that the distribution of restaurant inspection
infractions may be highly localized.
Although the downtown region had the highest number
of CP/C outcomes, our results showed that this area did
not differ significantly from other areas when the total
percentage of restaurants inspected was considered, because
of the high concentration of restaurants in the downtown
area. Downtown Toronto is a major tourism destination

July/August Food Protection Trends

237
Food Protection Trends - July/August 2020

Table of Contents for the Digital Edition of Food Protection Trends - July/August 2020
