Food Control 115 (2020) 107268

Contents lists available at ScienceDirect

Food Control
journal homepage: www.elsevier.com/locate/foodcont

Microbial awareness and risk perceptions are key to thermometer ownership T

and use
Eun Sol Hera, Barbara A. Almanzaa, Jing Mab,∗, Li Gec, Yiran Liub,1, Amy Landod, Fanfan Wud,
Linda Verrilld
a
School of Hospitality and Tourism Management, Purdue University, 900 West State Street, West Lafayette, IN, 47907, USA
b
Department of Hospitality Business Management, Alfred Lerner College of Business & Economics, 208 Raub Hall, 14 W. Main Street, University of Delaware, Newark, DE,
19716, USA
c
The Collins College of Hospitality Management, California State Polytechnic University, Pomona, 3801 West Temple Avenue 79B, Pomona, CA, 91768, USA
d
US Food and Drug Administration, Center for Food Safety and Applied Nutrition, 5001 Campus Drive, College Park, MD, 20740, USA

ARTICLE INFO ABSTRACT

Keywords: While government agencies recommend using a food thermometer to ensure the doneness of food, most con-
Food safety sumers still rely on unsafe methods including visual cues or cooking time. To enhance the understanding of
Thermometer consumer characteristics affecting thermometer ownership and use, this study examined whether consumers'
Food safety survey demographics, microbial awareness, and food safety perceptions and self-reported behaviors are associated with
Doneness
thermometer ownership and usage frequency for roasts, chicken parts, baked egg dishes, and hamburgers. Using
Pathogen awareness
Food safety practice
the data collected from the 2016 FDA Food Safety Survey, 4169 responses from the U.S. population were
analyzed using probit and ordered probit sample-selection models. Findings showed that 67% consumers owned
a thermometer, but rarely used it when cooking hamburgers (66% never) or egg dishes (58% never). The
findings also showed that consumers with more awareness of foodborne pathogens and better overall food safety
practices were more likely to own a thermometer and use it frequently for cooking all types of food. Certain
demographic factors, risk perceptions about raw and undercooked food, and risky food consumption were also
associated with thermometer ownership and usage frequency. Suggestions to increase consumers’ thermometer
ownership and use based on demographic categories and their awareness and motivation are discussed.

1. Introduction recommendations, however, most consumers still rely on unsafe

methods including visual cues or cooking time. In fact, it has been re-
Each year, approximately 48 million cases (one in six Americans) of ported that visual inspection was the primary method for testing food
foodborne illness occur in the United States with an estimated 128,000 doneness long before the introduction of food thermometers, especially
hospitalizations and 3000 deaths Centers for Disease Control and for certain types of food such as ground beef/hamburger (Lyon, Berry,
Prevention [CDC], 2016; Maughan et al., 2016). Eight known patho- Soderberg, & Clinch, 2000; U.S. Department of Agriculture [USDA,
gens - Escherichia coli, Salmonella, Clostridium perfringens, Staphylococcus 2018]). For example, an observational study of consumers’ food pre-
aureus, Listeria monocytogenes, Norovirus, Toxoplasma gondii and paration practices found that 22% of participants declared hamburgers
Campylobacter - account for the vast majority of illnesses, hospitaliza- done before they had reached the recommended internal cooking
tions, and deaths (CDC, 2016; FDA, 2017). Research has shown that temperature; the primary reason being that these consumers did not use
such pathogens can be contracted and spread by humans through raw a food thermometer (Phang & Bruhn, 2011).
or undercooked foods (CDC, 2018; Painter et al., 2013; Scallan et al., The use of a food thermometer to ensure foods reach an internal
2011). The Food and Drug Administration (FDA) recommends the use temperature sufficient to destroy foodborne pathogens is backed by
of food thermometers as the only dependable way of ensuring the ap- numerous research studies. One experimental study found that beef
propriate internal temperature of cooked food (FDA, 2018). Despite the patties can have considerable variability in the degree of doneness

∗
Corresponding author.
E-mail addresses: [email protected] (E.S. Her), [email protected] (B.A. Almanza), [email protected] (J. Ma), [email protected] (L. Ge),
[email protected] (Y. Liu), [email protected] (A. Lando), [email protected] (F. Wu), [email protected] (L. Verrill).
1
Present address: School of Hospitality and Tourism Management, Purdue University, 900 West State Street, West Lafayette, IN 47907, USA.

https://doi.org/10.1016/j.foodcont.2020.107268
Received 26 November 2019; Received in revised form 18 March 2020; Accepted 20 March 2020
Available online 26 March 2020
0956-7135/ © 2020 Elsevier Ltd. All rights reserved.
E.S. Her, et al. Food Control 115 (2020) 107268

depending on certain properties (such as the fat content) when cooked ownership and use. In this regard, this study examines how consumers’
for the same amount of time using the same amount of heat (Liu & food safety-related awareness, perceptions, and self-reported behaviors
Berry, 1996). This study shows that reliance on cooking time alone does as well as demographics are associated with their thermometer own-
not guarantee the product is fully cooked. Another study found that the ership and use. Once these factors are better understood, it would be
brown color typically used to judge the doneness of meat could occur possible to determine what kinds of information are needed to educate
prematurely at internal temperatures that are not sufficient to kill or motivate consumers in the use of food thermometers. This study
foodborne pathogens (Lyon et al., 2000), confirming that the color does therefore addressed the following research questions using the 2016
not accurately reflect degree of doneness. Recent research has even FDA Food Safety Survey data.
demonstrated that beef can appear done (premature browning or PMB),
but still contain viable E. coli (Boqvist, Fernström, Alsanius, & Lindqvist, Q1 (demographics). Which demographic groups are more likely to
2015). Moreover, another study found that various endogenous (i.e., own and frequently use thermometers?
pH, muscle source, species, redox state) and exogenous (i.e., packaging, Q2 (awareness). Are consumers that have heard of foodborne pa-
ingredients, and storage) factors may influence the internal color of thogens more likely to own and frequently use thermometers?
cooked meat (Suman, Nair, Joseph, & Hunt, 2016). Q3 (perceptions). Are consumers who believe that food contamina-
The finding that some products are already contaminated when tion is a serious problem and who think raw and undercooked food
consumers buy them (Berrang, Ladely, & Buhr, 2001) further justifies contains pathogens more likely to own and frequently use thermo-
the need for thermometer use to confirm meat doneness. Poultry, for meters?
instance, is often linked to Campylobacter and Salmonella. One study Q4 (self-reported behaviors). Is thermometer ownership and use as-
found that Campylobacter jejuni and Campylobacter coli were detected on sociated with other food safety behaviors such as handwashing,
94.5% and 5.5% of the tested surfaces in a poultry processing plant in kitchen cleaning, food preparation, and risky food consumption?
Spain (García-Sánchez et al., 2017) and an additional study discovered
that “1.00% (5 of 500 broiler carcasses) of composite samples and 2. Materials and methods
0.47% (7 of 1500 samples) of individual samples were positive for
Salmonella” (Sexton, Geornaras, Belk, Bunning, & Martin, 2018). Sample and Data Collection. This study was approved by the FDA
Unfortunately, research has also shown that consumer use of food to use primary, raw data from the 2016 Food Safety Survey (FSS)
thermometers is generally low (Lando & Chen, 2012; Maughan et al., conducted by the U.S. Food and Drug Administration (FDA) in colla-
2016). Consumer surveys conducted by the U.S. Food and Drug Ad- boration with the U.S. Department of Agriculture (USDA). Started in
ministration found that only 49% of U.S. consumers who cook raw meat 1988, the FSS is a periodic national telephone survey that currently
and poultry reported owning a food thermometer in 1998, although this includes both English and Spanish speaking adults in the 50 states and
number increased to 70% in 2010 (Lando & Chen, 2012). When all re- the District of Columbia. Telephone interview data are collected by a
spondents were asked (including those who did not prepare food), 67% random digit dialing (RDD) process, yielding an equal probability of
reported owning a thermometer (FDA, 2016a; 2016b). Data from 1998 to selection and a single-stage sample of telephone numbers. The details of
2010 showed that males, non-Hispanic whites, people with higher in- sampling selection and data collection methods in the previous FSS's are
comes, people with some college education or higher, and people who reported in Fein, Lando, Levy, Teisl, and Noblet (2011). The 2016 FSS
were 65 years and older were more likely to own a food thermometer. was the seventh iteration of the survey and, for the first time, sampled
The percent of owners who use the thermometer also increased from not only landline telephone numbers but also cell phone numbers to
1998 to 2016, although usage varied with the type of food (FDA, 2016b) ensure responses from those who only have cell phones. A total of 4,169
and demographic characteristics of the respondents (Lando & Chen, interviews were completed between October 2015 and January 2016,
2012). For example, although the use of a thermometer is recommended consisting of 2021 landline and 2148 cell phone interviews. Since no
for all meat products, among those who owned a thermometer in 2016, meaningful differences were found between landline and cell phone
only 81% used it for roasts, 58% used it for chicken parts, and less than a respondents with the exception of an only marginal difference (p = .09)
third (33%) used it for hamburgers; and for baked egg dishes, merely in self-reported thermometer ownership between landline phones
18% of thermometer owners who cook the food used it (FDA, 2016a). (slightly higher) and cell phones, the phone type of respondents was not
Compared to people aged 65 or older, people aged 18 to 29 were more considered in the subsequent data analyses. The average length of the
likely to use a thermometer for roasts and chicken parts after controlling phone interviews was about 17 min. The survey data were then
for thermometer ownership (Lando & Chen, 2012). weighted so that no demographic groups were over- or under-
Usage variability was also shown in a study by (Kosa et al., 2014), represented—the details of the weighting procedure are reported in
which found that among the 62% of consumers who owned food Lando, Bazaco, and Chen (2018).
thermometer(s), 73% used it to test the temperature of whole turkeys, Measurements. The survey included food safety questions about
but only a very small percentage (12–26% “depending on the cut or home practices, attitudes, and risk perceptions. Responses to the
type of poultry”) used it when cooking smaller cuts of poultry. Similar questions specifically pertaining to the variables of interest in this
ownership and usage results were shown in a 2018 USDA study (USDA, study, i.e., food safety awareness, perceptions, and behaviors, were
2018), finding that the majority of participants (66%) did not use a food selected for data analysis. The responses were either used in their ori-
thermometer to check internal temperatures of food before they were ginal format or summed to create indices.
given training interventions. The low adoption of food thermometers is Dependent variables. Thermometer Ownership. Participants were
also reflected in more recent studies (Feng & Bruhn, 2019). A recent asked whether they have a food thermometer (“Do you have a food
USDA study found that “while preparing meals with meat, poultry, or thermometer, such as a meat thermometer?”). Responses were coded as
seafood, only 14 percent of at-home meal preparers in the United States “Yes” = 1 and “No” = 0 and treated as binary data.
used a food thermometer” (Rhodes, Kuchler, McClelland, & Hamrick, Thermometer Use. Thermometer owners were asked how often they
2019). In summary, research clearly supports the fact that many of used the thermometer for four foods (roasts or large pieces of meat,
today's consumers do not follow governmental agencies' recommenda- chicken parts, baked egg dishes, and hamburgers) over the past year.
tions for food thermometer use, exposing themselves to the risk of Responses ranged from “Never”, “Sometimes”, “Often”, to “Always”
contracting a foodborne illness. and were treated as ordinal data. If participants responded that they did
Food thermometer research, however, has not investigated how not cook that food, the responses were treated as missing.
consumer food safety-related variables such as microbial awareness, Independent variables. The descriptive statistics of all independent
risk perceptions, and food safety behaviors are related to thermometer variables except demographics are provided in Table 1.

2
E.S. Her, et al. Food Control 115 (2020) 107268

Table 1
Descriptive statistics of independent variables including food safety awareness, perceptions, and behaviors.
Variable Mean/Frequency Standard Deviation Range Valid N

Awareness
Awareness of Foodborne Pathogens Index 2.8 1.1 [0, 5] 4151

Perceptions
Microbial Food Contamination Risk Perceptiona 4085
Serious 1393 (34.1%)
Somewhat 2315 (56.7%)
Not at all 377 (9.2%)
Raw and Undercooked Foods Risk Perception Index 3.0 2.8 [-8, 8] 4104

Behaviors (self-reported)
Food Safety Practices Index 9.7 6.5 [-12, 29] 3831
Risky Food Consumption Index −2.0 1.9 [-14, 0] 4125

a
Different from the other continuous index variables, Microbial Food Contamination Risk Perception was a single categorical variable.

Demographics. Participants’ gender, race/ethnicity, education, age, safest level of risk, and the middle value—zero—represented neither an
U.S. Census geographical region, and income were included in the addition nor reduction of foodborne illness risk. If a question was
analysis as dummy variables. skipped because it was not applicable to the participant (e.g., question
Awareness of Foodborne Pathogens Index (awareness). Five questions about handwashing after cleaning a cat litter box when the participant
asked whether participants had ever heard of the following foodborne did not have cats) or the participant said he or she did not know, the
pathogens as a problem in food: Salmonella, Listeria, Campylobacter, E. response was also recoded as zero, so that the total score was not af-
coli, and Toxoplasma or T. gondii. For each question, participants re- fected and the sample size was retained. A response was treated as
sponded either “Yes” = 1 or “No” = 0. These responses were summed missing only when the participant refused to answer a question.
to create the awareness index ranging from 0 to 5. The response practices were first assessed independently by four
Microbial Food Contamination Risk Perception (perceptions). One researchers with expertise in food safety. The independent risk assess-
question asked whether participants think food contamination by mi- ments were then discussed with the other researchers until a consensus
croorganisms is a serious food safety problem. Responses ranged from was reached. Finally, the coding was confirmed by three FDA re-
“Serious”, “Somewhat”, to “Not at all” and were included in the ana- searchers responsible for the 2016 FDA Food Safety Survey. For ex-
lysis as dummy variables with “Not at all” as the reference category. ample, in the question of “After you have cracked open raw eggs, do
Raw and Undercooked Foods Risk Perception Index (perceptions). you usually continue cooking, or do you first rinse your hands with
Three questions asked the likelihood of pathogenic microbial con- water, or wipe them, or wash them with soap?”, the responses were
tamination in raw chicken, beef, and eggs (“How likely do you think it recoded as “Continue cooking” = −2, “Rinse or wipe hands” = −1,
is that the following foods contain germs or other micro-organisms that “Wash with soap” = +2, “Never handle raw eggs”, “Skipped” and
could make you sick?”), and one question asked the likelihood of get- “Don't know” = 0, and “Refused to answer” = missing. Although each
ting an illness from eating undercooked meat or chicken (“If you eat response was recoded using a 5-point scale (i.e., −2 to +2), the
meat or chicken that is not thoroughly cooked, how likely are you to get number of possible responses to each question varied. For some ques-
sick?”). Participants responded on a 5-point scale ranging from “Not at tions, two responses were offered (e.g., “yes/no” binary questions), for
all likely” to “Very likely”. Responses from these four questions were other questions, up to 12 responses were possible (e.g., how to clean a
coded into −2 (“Not at all likely”) to +2 (“Very likely”) and summed cutting board or other surfaces after cutting raw foods). Therefore, the
to create an index ranging from −8 to +8. sum of all responses created an index ranging from −12 to +29.
Food Safety Practices Index (self-reported behaviors). The index for Risky Food Consumption Index (self-reported behaviors). Participants
overall food safety practices was created using 28 questions about (1) were asked whether they had eaten any raw foods (n = 19), such as
handwashing (n = 7: handwashing frequencies before beginning food raw clams, raw oysters, steak tartare, sushi with raw fish, raw sprouts,
preparation, after yard work or gardening, and after cleaning a cat litter raw poultry, raw or unpasteurized milk, raw milk cheese, and eggs with
box and handwashing practices after cracking raw eggs, after handling runny yolks in the past 12 months. Responses were coded as
raw meat or chicken, after handling raw fish, and after handling elec- “Yes” = −1 or “No” = 0. The sum of the 19 responses resulted in an
tronic devices during food preparation); (2) kitchen and kitchenware index ranging from −14 to 0 (no participants scored between −19 and
cleaning (n = 7:cleaning practices of cutting board or other surfaces −15).
after cutting raw meat or chicken and after cutting raw fish or shellfish, Data Analysis. A probit model was used to test the associations
types of cloth or wipes used to clean kitchen counter tops after pre- between the independent variables (demographics, microbial aware-
paring raw meat or chicken, frequency of changing or laundering dish ness, food safety perceptions and practices) and the probability of
towels or cloths, type of serving ware used to carry food grilled outside, owning a food thermometer (binary). To test the associations between
and cleaning practice of the serving ware); and (3) food preparation and the independent variables and the probability of using the food ther-
storage practices (n = 14: cleaning frequencies of raw chicken pieces mometer frequently (ordinal) for the four foods, two-step ordered
and of whole poultry before cooking, cleaning practices of raw avo- probit models with sample selection were used. The two-step method
cados and cantaloupes before preparing, cutting; or eating, cooling and was selected because, while the use of thermometers in the sample was
storage practices of leftover soup, stew, or other food with meat or only possible for those who owned a thermometer, the decision to own
chicken, and with fish or shellfish, likelihood of following raw meat a thermometer was not random but determined by the individuals
cooking instructions on the package, and doneness of hamburgers themselves. As a result, the sample of thermometer owners inevitably
served at home). To assess the level of foodborne illness risk associated suffer from the self-selection problem, and conventional regression
with each response, the original responses were recoded using a 5-point models without techniques to address the sample selection processes
scale ranging from −2 to +2, where the lowest value represented the would be expected to produce biased coefficient estimates (Heckman,
highest level of foodborne illness risk, the highest value represented the 1979). In this regard, the two-step Heckman technique successfully

3
E.S. Her, et al. Food Control 115 (2020) 107268

Table 2 Thermometer ownership. The interpretation of probit coefficients is

Food thermometer ownership and usage frequency by food type. not as simple as conventional regression or logit models since the in-
Thermometer ownership (%) crease in the predicted probabilities of dependent variables due to the
increase in an independent variable was also influenced by other fac-
No 32 tors. Therefore, findings are described using the direction and statistical
Yes 67
significance of the relationships, similar to Lando and Chen (2012)'s
reporting of the probit model findings from the 1998–2010 FSS results
Food type Thermometer usage frequency among thermometer owners (%)
on thermometer ownership and usage.
Never Sometimes Often Always Never cook the Probit model results showed that most demographic variables,
food awareness of foodborne pathogens, overall food safety practices, and
risky food consumptions were significant predictors of thermometer
Roasts 15 25 18 38 3
Chicken Parts 40 26 13 19 1
ownership. Demographic variables of age, race/ethnicity, education,
Egg Dishes 58 12 4 6 20 region, and income were significantly associated with thermometer
Hamburgers 66 17 6 10 1 ownership. Consumers who were 1) 65 years old or older, 2) non-
Note. N = 4169. The percentages do not sum to 100 due to missing data; less Hispanic Whites, 3) with a high school degree or some college educa-
than 1% of the sample. tion, 4) living in the Midwest, and 5) in a higher income grouping were
more likely to own thermometers. Those who were less likely to own a
thermometer were 1) younger, 2) from other race/ethnicity groupings,
enabled estimation of non-biased regression coefficients and resolved 3) with less than a high school degree, 4) living in the Northeast or
the self-selection biases when estimating the probability of thermo- South, and 5) in an income level that was less than $25,000. No sig-
meter use, considering the probability of thermometer ownership in the nificant difference in thermometer ownership was found between men
first stage. Also, in reference to the literature (Lando & Chen, 2012), we and women.
assumed that the household income would be an important determi- Consumers were more likely to own thermometers if they were
nant for thermometer ownership, but not necessarily for the frequency more aware of foodborne pathogens, had better overall food safety
of thermometer use, thus income was included in the first stage only practices, and had less risky food consumption practices. However,
(i.e., exclusion restrictions, meaning “the selection equation should perceptions about risk of general food contamination by microorgan-
contain at least one variable that is not in the outcome equation”; isms and risk associated with consumption of raw and undercooked
StataCorp, 2019, p. 927; Bushway, Johnson, & Slocum, 2007). All foods were not significant predictors of thermometer ownership.
analyses were conducted using STATA version 15.1, following the Thermometer usage frequency. The results of the ordered probit
commands first introduced by De Luca and Perotti (2011). and ordered probit sample-selection models revealed that, even after
adjusting for the bias caused by the self-selection of thermometer
3. Results ownership and controlling for other variables in the model, awareness
of foodborne pathogens and overall food safety practices were the
Descriptive Statistics. Descriptive statistics for thermometer primary determinants of food thermometer usage frequency, regardless
ownership and use for the four foods are provided in Table 2 (also of food type (roasts, chicken parts, egg dishes, and hamburgers). Risk
available in FDA, 2016a). Most consumers (67%) reported that they perception about raw and undercooked foods was also a significant or
owned a food thermometer. Although overall use of thermometers was at least marginally significant (p < .1) factor for thermometer usage
not high (i.e., depending on food type, only 6%–38% consumers always frequency for all four foods, meaning that consumers who perceived
used it when cooking the food), among those who owned a thermo- more risk in consuming raw and undercooked foods were more likely to
meter, the thermometer tended to be most frequently used when use thermometers more frequently when cooking food.
cooking roasts or large pieces of meats (38% always and 18% often), Other associations among thermometer usage frequency, and de-
followed by chicken parts (19% always and 13% often). Consumers mographics and risky food consumption behaviors varied by food type.
were least likely to use it for hamburgers and egg dishes, with two- More frequent thermometer use for at least one food type was found for
thirds of thermometer owners never using it when cooking hamburgers men, younger ages (18–39), non-Hispanic Whites, and those who had
(66%) and more than half never using it for baked egg dishes (58%). less than a high school degree. There were also a number of demo-
graphic differences in thermometer usage frequency when cooking
3.1. Thermometer ownership and usage frequency - probit models roasts, chicken parts, and baked egg dishes, but only one (age) for
hamburgers. In general, thermometer use for hamburgers was con-
Sample-selection model assumption testing. Before running the four sistently low across all groups.
Heckman two-step correction models for thermometer usage frequency More specifically, men were more likely to use the thermometers
for four types of food, the likelihood ratio (LR) test was performed to more frequently for roasts, chicken parts, and baked egg dishes as
verify if the observed sample was indeed a non-random sample with a compared to women. Younger consumers (18–39 years old) were sig-
sample selection problem (Bushway et al., 2007). The LR test compared nificantly or marginally more likely to use thermometers more fre-
the log likelihood of the full two-step model with the sum of the log quently when cooking roasts, chicken parts, and hamburgers as com-
likelihoods for the separate ordered probit and sample-selection pared to older consumers (65 years old or older). Non-Hispanic Whites
models. The results showed that the use of sample-selection models was were more likely to use thermometers more frequently when cooking
appropriate for thermometer usage frequency for chicken parts chicken parts than the other ethnicity groups and were marginally more
(χ2(1) = 32.18, p < .001), egg dishes (χ2(1) = 24.32, p < .001), and likely to use them frequently for roasts and egg dishes than Hispanics.
hamburgers (χ2(1) = 12.79, p < .001). However, the sample of those Consumers with more education (high school graduate, Bachelor's and
who used thermometers for roasts did not have a selection bias due to a graduate degrees) were less likely to use thermometers very frequently
thermometer ownership (χ2(1) = 1.34, p > .1). Accordingly, a stan- when cooking egg dishes, as compared to those with less than a high
dard ordered probit technique was used for the usage frequency for school degree. Regarding geographic regions, consumers living in the
roasts and, as planned, two-step sample-selection techniques were used Midwest were less likely to use thermometers very frequently for roasts
for the other three models. The regression coefficients of probit, or- as compared to those living in the Northeast or West, but were more
dered probit, and ordered probit sample-selection models for owning likely to use them frequently when cooking chicken parts than those
and frequently using a thermometer are provided in Table 3. living in the Northeast and South. Lastly, consumers who consumed a

4
E.S. Her, et al. Food Control 115 (2020) 107268

Table 3
Probit model regression coefficients of owning and frequently using a food thermometer.
Variablea Thermometer Ownership Thermometer Use Ordered Probit Model Thermometer Use
Probit Model Ordered Probit
Sample-Selection Modelsb

Roasts Chicken Parts Egg Dishes Hamburgers

Demographics
Gender (vs. Female)
Male .05 .12 * .10 * .18 ** -.03
Age (vs. ≥ 65)
† †
18-29 -.38 *** .15 .17 .14 .27 *
†
30-39 -.35 *** .21 * .06 .01 .18
40-49 -.43 *** -.02 -.12 -.03 -.07
50-64 -.21 ** -.01 .00 -.02 -.02
Race/Ethnicity (vs. Non-Hispanic White)
Non-Hispanic Black -.36 *** .02 -.30 *** .16 .06
† †
Hispanic -.68 *** -.17 -.30 *** -.19 -.04
Other -.30 ** -.12 -.27 * -.02 .18
Education (vs. < High school)
†
High school graduate .36 *** -.23 .09 -.30 * -.10
Some college .35 ** -.12 .10 -.22 .05
Bachelor's degree .14 .04 .05 -.53 *** -.06
†
Graduate degree .23 -.01 .00 -.70 *** -.08
Region (vs. Midwest)
Northeast -.18 * .27 *** -.18 * .11 -.05
South -.22 ** .07 -.19 ** .06 .07
West .03 .31 *** -.10 .03 .07
Income (vs. < $25,000)
$25,000-$50,000 .26 ***
$50,000-$75,000 .48 ***
$75,000-$100,000 .59 ***
> $100,000 .71 ***

Awareness
Awareness of Foodborne Pathogens Index .10 *** .07 ** .11 *** .12 *** .13 ***

Perceptions
Microbial Food Contamination Risk Perception (vs. Not all)
Serious -.00 -.03 .08 .06 -.02
Somewhat serious .04 -.10 .02 -.01 -.11
†
Raw and Undercooked Foods Risk Perception Index .00 .02 * .02 * .03 ** .02

Behaviors (self-reported)
Food Safety Practices Index .04 *** .02 *** .04 *** .03 *** .04 ***
Risky Food Consumption Index -.08 *** .00 -.00 .01 .05 **

Constant -.40 **
Intercept 1 -.46 1.01 *** 1.46*** 1.54***
Intercept 2 .36 1.63 *** 2.02*** 2.11***
Intercept 3 .86 2.01 *** 2.32*** 2.39***
rho .83 .74 .70

N (Ownership) 3035 2751 2550 2670

N (Usage) 1968 1752 1551 1671

*p < .05. **p < .01. ***p < .001. †p < .1.
a
Reference categories are shown in parentheses.
b
The Likelihood Ratio (LR) test statistics for the rhos were all significant at the α = .001 level. Therefore, the Heckman sample-selection models were used for
these analyses.

greater number of risky foods in the past 12 months were more likely to foodborne pathogen awareness and the respondent's food safety prac-
use thermometers more frequently when cooking hamburgers. All of tices score was evident as thermometer usage frequency in cooking any
the remaining demographic relationships and the relationships with of the four foods varied significantly with these two variables.
food contamination risk perception were not significant. Perception of risk associated with consumption of raw or undercooked
In summary, the greatest variability in thermometer usage fre- foods was significantly (roasts, chicken parts, egg dishes) or marginally
quency was associated with cooking roasts, where it varied significantly (hamburgers) associated with all of the four foods as well.
or marginally with gender, age, race/ethnicity, education, and geo-
graphic region. For cooking chicken parts, thermometer usage fre- 4. Discussion
quency varied significantly or marginally with gender, age, race/eth-
nicity, and region. For cooking egg dishes, thermometer usage This study examined the current use of food thermometers using
frequency was associated with gender, race/ethnicity, and education, data from a national phone survey to determine which demographic
and when cooking hamburgers, thermometer usage frequency varied characteristics, as well as food safety awareness, risk perceptions, and
only with age and risky food consumption behaviors. The influence of food safety practices were associated with thermometer ownership and

5
E.S. Her, et al. Food Control 115 (2020) 107268

use. Overall, findings revealed that (1) age, race/ethnicity, education This idea might be supported by results from a nationwide survey
level, geographic region, and income level were significantly associated conducted in Australia in 2012 that clearly showed the growth in the
with food thermometer ownership, while their associations with its number of men becoming interested in learning how to cook different
usage frequency for different types of food varied; (2) gender was sig- types of meats, such that more men reported this interest as compared
nificantly associated with thermometer usage frequency for certain to women (Worsley, Wang, Ismail, & Ridley, 2014).
types of food, such that men were more likely to use a thermometer Third, although education levels were positively related to ther-
frequently than women; (3) microbial awareness, food safety practices, mometer ownership, they were negatively related to thermometer
and risky food consumption were significantly associated with ther- usage frequency for roasts (marginally) and egg dishes. That is, while
mometer ownership, while microbial awareness and food safety prac- consumers with higher education were more likely to own thermo-
tices were further associated with thermometer usage frequency across meters, they were less likely to use it frequently for roasts and egg
four types of food; (4) risk perception about raw and undercooked food dishes. The similar pattern (i.e., the reverse of the direction) also oc-
was significantly associated with thermometer usage frequency for curred for age in regard to cooking roasts, chicken parts (marginally),
certain types of food; and (5) general risk perception about microbial and hamburgers. While younger consumers (18–39 years old) were less
contamination of food was not associated with thermometer ownership likely to own thermometers than older consumers (65 years or older),
and use. younger consumers were more likely to use them frequently when
Theoretical Implications. First, findings regarding the low per- cooking roasts, chicken parts, and hamburgers as compared to older
centage of thermometer ownership and use were consistent with pre- consumers. A possible reason for the demographic differences in ther-
vious studies (e.g., Kosa et al., 2014; Lando & Chen, 2012). Specifically, mometer use may be that certain demographic groups consume certain
the findings regarding thermometer use for the four foods aligned with foods more frequently than other groups. For example, people with a
another study that described the introduction of cooking endpoint lower education may consume roasts or egg dishes more frequently
temperatures for these four foods into home cookbooks in the U.S. than those with a higher education; younger people may consume
(Almanza, Byrd, Behnke, Ma, & Ge, 2017). Almanza et al. (2017) found roasts or hamburgers more frequently than those who are 65 years of
that cooking endpoints for beef (roasts) were introduced into cook- age or older. Thus, if a demographic group is more likely to consume a
books in the 1930s, followed by endpoints for chicken and turkey in food, it might be that they are also more likely to use the thermometer
1965, and finally in 1996 - the cooking endpoints for eggs and ground for that food as compared to other foods. In support of this, studies
beef. A similar sequence was found in the current study in that con- found that men reported eating meat and certain types of poultry more
sumers most frequently used thermometers for roasts, followed by than women (Beletshachew et al., 2012) and that men (vs. women),
chicken parts, and then egg dishes and hamburgers. This possibly Whites (vs. Blacks and Hispanics), and consumers with less than a high
suggests the important role cookbooks play in consumers' food handling school degree (vs. more than a high school degree) reported a higher
and cooking. It also suggests the need for consumer education through meat consumption including red meat, poultry, and hamburgers (Wang,
additional media to 1) provide instructions on how to use a food Beydoun, Caballero, Gary, & Lawrence, 2010). Our findings also in-
thermometer with smaller cuts of meat and 2) encourage them to be dicated that these demographic groups were generally more likely to
more aware of the need to increase their use of thermometers using use thermometers frequently when cooking roasts, chicken, and ham-
today's cooking standards. burgers.
Second, regarding the findings on demographic relationships, re- Fourth, more importantly, this study contributes to the under-
sults showed that thermometer ownership was more common among standing of consumers’ thermometer ownership and use by showing
consumers who were older (65 years or older), non-Hispanic Whites, that other food safety variables such as foodborne pathogen awareness,
and those from higher income and education groups. This was con- risk perceptions regarding the consumption of raw and undercooked
sistent with the findings from the previous FDA FSS datasets collected food, food safety practices, and risky food consumption are significantly
from 1998 to 2010 (Lando & Chen, 2012). However, although Lando associated with thermometer ownership and usage frequency. Findings
and Chen (2012) found that men were slightly more likely to own showed that better pathogen awareness and overall food safety prac-
thermometers than women in the surveys from 1998 to 2010, no gender tices were most significantly correlated with both higher thermometer
difference in thermometer ownership was found in this study. The ownership and usage likelihoods for all four types of food. The strong
difference in findings on a gender effect might have come from the associations between microbial awareness and overall food safety
possible changes in gender roles and perceptions about education, use practices with both thermometer ownership and usage frequency con-
of technology and devices, or home cooking and meal preparation over firms the important roles of motivation and education. That is, while
time. For example, it is possible that men were more educated and several possible reasons exist as to why consumers may not have or use
interested in the mechanical devices than women in the past, which a thermometer, such as lack of awareness of its importance or limited
may have become attenuated nowadays. Other possible reasons include knowledge about how to use thermometers, the results showed that
the methodological changes with the introduction of cell phone num- awareness of foodborne pathogens, in fact, was associated with an in-
bers in the current dataset compared to prior datasets that used only crease in the likelihood of both owning a thermometer and using it
landline numbers in data collection, or the inclusion of different in- frequently—perhaps because an awareness of pathogens provided a
dependent variables in each study's models. Lando and Chen (2012) good motivation to own and use thermometers. In addition, consumers
included survey year and demographics in their models, while this who were better at handwashing, kitchen and kitchenware cleaning,
study included several other food safety variables in addition to de- and food preparation and storage, and thus who appeared to be more
mographics. Thus, when controlling for the other food safety variables, knowledgeable about safe food preparation and received more educa-
findings about gender may also have changed. tion in safe practices were more likely to own and frequently use
Interestingly, results of the current study found some unexpected thermometers. This viewpoint is in line with the finding that consumers
relationships as well. Gender differences in usage frequency of ther- who less frequently consumed risky foods were more likely to own
mometers were found with males being significantly or marginally thermometers, since food safety education would probably have con-
more likely to use thermometers more frequently than women for tributed to their knowledge about risky foods and the value of food
roasts, chicken parts, and egg dishes. The lack of a gender difference in thermometers.
thermometer ownership vs. a gender difference in thermometer use Fifth, in a similar vein, the findings showed that greater risk per-
might possibly be explained in that thermometers are “owned” by a ception about the consumption of raw and undercooked foods was
household and therefore may not reflect gender differences, whereas significantly or at least marginally associated with more frequent
thermometer use reflects individual differences in food preparation. thermometer use for all types of food. While it is possible that the

6
E.S. Her, et al. Food Control 115 (2020) 107268

personal traits such as risk-taking avoidance affected the risk percep- Additional recommendations for future research can also be made
tion, the results clearly suggest that those who were more knowledge- based on these findings. That is, cooking endpoints for foods (such as
able or at least more conscious about the risks of consuming raw and hamburgers) that were introduced into home cookbooks more recently
undercooked food were actually more likely to pay attention to cooking (Almanza et al., 2017), were less frequently checked by consumers
endpoints. Finally, the finding that a general perception of microbial using a thermometer. This points out a possible problem in that con-
food contamination risk was not related to thermometer ownership or sumers may be relying on older cookbooks for their food handling and
use highlights that a general risk perception is not sufficient to drive the cooking information, perhaps even those that have been passed down
purchase or use of thermometers. This means that only a specific risk from a parent or grandparent. Further research is recommended on
perception about the consumption of raw and undercooked foods was publication or print dates of consumers’ cooking information because of
influential in frequent thermometer use. changes in knowledge regarding safe food handling. In regard to the
Implications for Policy Makers, Food Educators, and Consumers. gender differences in the findings, future studies may wish to consider
Altogether, the findings provide important educational implications re- gender differences in food preparation related to sources of cooking
garding thermometer ownership and use. The findings show certain information, recipes, and use of cookbooks. For example, the under-
consumer groups—perhaps related to cultural variations in food pre- standing about whether women are more likely to use cookbooks
paration, dietary patterns, or access to food safety information—need (perhaps even those inherited from mothers or grandmothers) in con-
more information about the importance of thermometers when cooking trast to men who might be more likely to find recipes on-line and
foods and/or more awareness of pathogens, safe food practices, and in- perhaps more current in their cooking guidelines would be useful to
formation on the risks associated with consuming raw and undercooked further explore and extend the current findings.
foods that precipitate the need to use food thermometers. These groups Finally, future research may wish to look more closely at the re-
for consideration can be categorized into the following three types: (1) lationships of other food preparation and consumption-related vari-
Low ownership and low use consumers who are both less likely to own ables, such as cultural differences in food preparation methods, with
thermometers and to use them when cooking certain types of food even thermometer ownership and use. This is important because cultural
though they own thermometers, and includes race/ethnicity groups variations may create differences in cooking times and food tempera-
other than non-Hispanic Whites (especially Hispanics) and residents of tures. For example, in certain cultures (e.g., Latin American and
the U.S. South; (2) Low ownership but high use consumers who as a Caribbean cuisines), the general cooking time of recipes may be longer
group are less likely to own thermometers, but includes individual and involve multiple stages, so that there may be less reliance on the
thermometer owners who are likely to use them more frequently, and is use of a food thermometer to determine doneness. Recipes may include
represented by younger consumers (age 18–39) and consumers with less fully cooking the ingredients followed by a final oven finish lasting
than high school degrees; and lastly (3) High ownership but low use several minutes to blend the flavors and complete dishes such as chili
consumers who are likely to own thermometers but are less likely to use con carne burritos, carnitas, and chiles rellenos (Sánchez, 2019). An
them when cooking certain types of food, and includes females, elderly example from Native American culture (bison pot roast with hominy)
consumers (age 65 or more), consumers with more than high school uses a long cooking time to tenderize and enhance the flavors of in-
degrees (except some college), and residents of the US Midwest. digenous foods (Sherman, 2020). Future research is strongly en-
The following recommendations are suggested for these groups couraged into cultural factors and cuisines that would provide stronger
based on the results of this study. First of all, the most comprehensive insights into relevant educational messages for different race/ethnic
educational approach will be needed for the (1) Low ownership and groups based on cooking practices.
low use consumers regarding the necessity of thermometer ownership
and use, such as why it is important to use them, where to buy them, 5. Conclusions
and how and when to use them. Our findings indicated that the increase
of foodborne pathogen awareness may be a good starting point for these The findings of this study suggested that food safety awareness and
people. On the other hand, the thermometer use of the (2) Low own- access to food safety information, reflected in consumers’ microbial
ership but high use consumers may improve by providing them with awareness, risk perceptions, and food safety practices, are key to in-
information about what type of thermometer to buy and low cost places crease thermometer ownership and use. While the best method of
to buy them. The thermometer use of the (3) High ownership but low communication is not yet determined, previous research based on the
users may improve with provision of recommendations about how to
Stages of Change (SOC) theory showed that most consumers are cur-
use thermometers, the types of foods for which thermometer use is now rently in the “precontemplation stage” when it comes to thermometer
recommended, and reminders to use them more frequently. This study's
use (i.e., never used a thermometer, no intention to use it soon, and
results also suggest that information about the risks of consuming raw little knowledge about the potential risks (Takeuchi, Edlefsen,
and undercooked foods may also help motive them to use thermometers
McCurdy, & Hillers, 2005);). Thus, campaigns focusing on raising
more frequently. awareness of food safety risks associated with consuming raw and un-
Limitations and Recommendations for Future Research.
dercooked foods and that of the benefits of using food thermometers are
Limitations of the study include the fact that no direct assessment of suggested to move consumers on to the contemplation, preparation,
food safety knowledge was used. Instead, food safety practices were
action, and maintenance stages of behavior change. Further, the use of
assessed because of their possible implications for food safety knowl- short educational messages such as a 3-min video or printed educa-
edge. Future research may wish to assess the direct relationships be-
tional material have been found to encourage consumers to follow re-
tween the food safety knowledge related to food thermometer use and commendations in the use of food thermometers during meal pre-
food safety behaviors to extend the findings of this study. In addition,
paration (McCurdy et al., 2006; USDA, 2018), and this strategy can be
this study did not assess where or how food safety practices, pathogen useful in the campaigns. In sum, it is anticipated that future researchers
awareness, or risk perceptions were acquired. Future studies may also
take advantage of the findings of this study and explore the most ef-
wish to explore where consumers learn such information to identify fective and engaging ways to stimulate consumer awareness and mo-
possible applications or approaches that could be used to educate
tivations to use thermometers for their safety.
consumers in the current recommendations for the use of thermo-
meters. Also, this study used a telephone survey, meaning that the
behaviors of the consumers were measured by participants’ self-re- Declaration of competing interest
porting. Thus, future studies may wish to compare these findings to
observations of actual practices. None.

7
E.S. Her, et al. Food Control 115 (2020) 107268

Acknowledgements a national survey. Journal of Food Protection, 77(Suppl A), 138.

Lando, A. M., Bazaco, M. C., & Chen, Y. (2018). Consumers' use of personal electronic
devices in the kitchen. Journal of Food Protection, 81(3), 437–443.
The authors wish to thank the U.S. Food and Drug Administration Lando, A., & Chen, C. (2012). Trends in ownership and usage of food thermometers in the
for the use of the dataset. This research did not receive any specific United States, 1998 through 2010. Journal of Food Protection, 75(3), 556–562.
grant from funding agencies in the public, commercial, or not-for-profit Liu, M. N., & Berry, B. W. (1996). Variability in color, cooking times, and internal tem-
perature of beef patties under controlled cooking conditions. Journal of Food
sectors. Protection, 59(9), 969–975.
Lyon, B. G., Berry, B. W., Soderberg, D., & Clinch, N. (2000). Visual color and doneness
References indicators and the incidence of premature brown color in beef patties cooked to four
end point temperatures. Journal of Food Protection, 63(10), 1389–1398.
Maughan, C., Chambers, E., Godwin, S., Chambers, D., Cates, S., & Koppel, K. (2016).
Almanza, B. A., Byrd, K. S., Behnke, C., Ma, J., & Ge, L. (2017). Cookbooks in US history: Food handling behaviors observed in consumers when cooking poultry and eggs.
How do they reflect food safety from 1896 to 2014? Appetite, 116, 599–609. Journal of Food Protection, 79(6), 970–977.
Beletshachew, S., Verrill, L., Booth, H., Zansky, S., Norton, D., Crim, S., et al. (2012). Sex- McCurdy, S. M., Takeuchi, M. T., Edwards, Z. M., Edlefsen, M., Kang, D.-H., Elaine Mayes,
based differences in food consumption: Foodborne diseases active surveillance net- V., et al. (2006). Food safety education initiative to increase consumer use of food
work (FoodNet) population survey, 2006-2007. Clinical Infectious Diseases, 54(Suppl thermometers in the United States. British Food Journal, 108(9), 775–794.
5), S453–S457. Painter, J. A., Hoekstra, R. M., Ayers, T., Tauxe, R. V., Braden, C. R., Angulo, F. J., et al.
Berrang, M. E., Ladely, S. R., & Buhr, R. J. (2001). Presence and level of Campylobacter, (2013). Attribution of foodborne illnesses, hospitalizations, and deaths to food
Coliforms, Escherichia coli, and total aerobic bacteria recovered from broiler parts commodities by using outbreak data, United States, 1998–2008. Emerging Infectious
with and without skin. Journal of Food Protection, 64(2), 184–188. Diseases, 19(3), 407.
Boqvist, S., Fernström, L.-L., Alsanius, B. W., & Lindqvist, R. (2015). Escherichia coli Phang, H. S., & Bruhn, C. M. (2011). Burger preparation: What consumers say and do in
O157:H7 reduction in hamburgers with regard to premature browning of minced the home. Journal of Food Protection, 74(10), 1708–1716.
beef, colour score and method for determining doneness. International Journal of Food Rhodes, M. T., Kuchler, F., McClelland, K., & Hamrick, K. S. (2019). Consumer food safety
Microbiology, 215, 109–116. practices: Raw milk consumption and food thermometer use, EIB-205. U.S. Department of
Bushway, S., Johnson, B. D., & Slocum, L. A. (2007). Is the magic still there? The use of Agriculture, Economic Research Servicehttps://www.ers.usda.gov/webdocs/
the heckman two-step correction for selection bias in criminology. Journal of publications/91110/eib-205.pdf?v=675.4.
Quantitative Criminology, 23(2), 151–178. Sánchez, A. (2019). Chef Aarón Sánchez, recipes – blog. https://www.chefaaronsanchez.
Centers for Disease Control and Prevention (2016). Estimates of foodborne illness in the com/recipes-blog.
United States. http://www.cdc.gov/foodborneburden/. Scallan, E., Hoekstra, R. M., Angulo, F. J., Tauxe, R. V., Widdowson, M.-A., Roy, S. L., &
Centers for Disease Control and Prevention (2018). Food poisoning symptoms. https:// Griffin, P. M. (2011). Foodborne illness acquired in the United States—major pa-
www.cdc.gov/foodsafety/symptoms.html#symptoms. thogens. Emerging Infectious Diseases, 17(1), 7.
De Luca, G., & Perotti, V. (2011). Estimation of ordered response models with sample Sexton, T. Y., Geornaras, I., Belk, K. E., Bunning, M., & Martin, J. N. (2018). Salmonella
selection. STATA Journal, 11(2), 213–239. contamination in broiler synovial fluid: Are we missing a potential reservoir? Journal
Fein, S. B., Lando, A. M., Levy, A. S., Teisl, M. F., & Noblet, C. (2011). Trends in US of Food Protection, 81(9), 1425–1431.
consumers' safe handling and consumption of food and their risk perceptions, 1988 Sherman, S. (2020). Sean Sherman's 10 essential Native American Recipes, the founder of
through 2010. Journal of Food Protection, 74(9), 1513–1523. the Sioux Chef, a company devoted to indigenous foods, created recipes to showcase
Feng, Y., & Bruhn, C. M. (2019). Motivators and barriers to cooking and refrigerator tribal diversity across the lower 48 states. https://www.nytimes.com/2019/11/04/
thermometer use among consumers and food workers: A review. Journal of Food dining/native-american-recipes-sioux-chef.html.
Protection, 82(1), 128–150. StataCorp (2019). Stata 16 base reference manual. College Station, TX: Stata Press. https://
Food and Drug Administration. FDA food safety survey. (2016). https://www.fda.gov/ www.stata.com/manuals/r/index2.html.
media/101366/download 2016. Suman, S. P., Nair, M. N., Joseph, P., & Hunt, M. C. (2016). Factors influencing internal
Food and Drug Administration (2016b). 2016 FDA food safety survey. https://www.fda. color of cooked meats. Meat Science, 120, 133–144.
gov/media/101375/download. Takeuchi, M. T., Edlefsen, M., McCurdy, S. M., & Hillers, V. N. (2005). Educational in-
Food and Drug Administration (2017). Bad bug book: Handbook of foodborne pathogenic tervention enhances consumers' readiness to adopt food thermometer use when
microorganisms and natural toxins. https://www.fda.gov/downloads/Food/ cooking small cuts of meat: An application of the Transtheoretical Model. Journal of
FoodborneIllnessContaminants/UCM297627.pdf. Food Protection, 68(9), 1874–1883.
Food and Drug Administration (2018). Kitchen thermometers. https://www.fsis.usda. U.S. Department of Agriculture. (2018). https://www.fsis.usda.gov/wps/wcm/connect/
gov/wps/portal/fsis/topics/food-safety-education/get-answers/food-safety-fact- cb222383-1e02-471a-8657-c205eda92acf/Observational-Study.pdf?MOD=
sheets/appliances-and-thermometers/kitchen-thermometers/ct_index. AJPERES.
García-Sánchez, L., Melero, B., Jaime, I., Hänninen, M.-L., Rossi, M., & Rovira, J. (2017). Wang, Y., Beydoun, M. A., Caballero, B., Gary, T. L., & Lawrence, R. (2010). Trends and
Campylobacter jejuni survival in a poultry processing plant environment. Food correlates in meat consumption patterns in the US adult population. Public Health
Microbiology, 65, 185–192. Nutrition, 13(9), 1333–1345.
Heckman, J. J. (1979). Sample selection bias as a specification error. Econometrica, 47, Worsley, A., Wang, W., Ismail, S., & Ridley, S. (2014). Consumers' interest in learning
153–161. about cooking: The influence of age, gender and education. International Journal of
Kosa, K., Cates, S., Bradley, S., Godwin, S., Chambers Iv, E., Ricketts, J., & Chambers, D. Consumer Studies, 38(3), 258–264.
(2014). Food thermometer usage among consumers who cook raw poultry: Results of