- Click here for explanation of classification scheme.

To determine baseline food safety practices as part of the development of a questionnaire to evaluate a food safety education program.

All female adult Women, Infant and Children (WIC) Program recipients and female guardians of pediatric clients (usually the mother or grandmother) who were able to read and speak English.

Those who could not read and speak English.

Recruitment

• All subjects were recruited form one of the Miami-Dade County Health Department WIC clinics
• Recruitment occurred once during every time period that the clinic served clients
• Consenting clients were enrolled until the quota for the given time period was reached
• The quota for the time period was weighted on the basis of the typical census for that time period
• After clients completed the survey, they were given a thank- you gift of $10 for their time. 

Design

• Cross-sectional study using a survey designed to capture five constructs of food safety behaviors with the first four from the Partnership for Food Safety Education's Fight BAC! campaign. These behaviors include clean, separate, cook, chill and avoid unsafe foods during pregnancy.
• Two studies were conducted to determine the psychometric characteristics of the adapted survey so that it could be used in the evaluation of the food safety education program.

Statistical Analysis

• A univariate analysis of variance and a multivariate analysis of variance were performed on each of nine participant characteristics
• The univariate analysis was performed on a total score across all questionnaire items, with each of the nine participant characteristics of interest as the fixed effect
• This was followed by a multivariate analysis of variance on each of the same nine participant characteristic items as a fixed effect with the family of dependent measures being each of the four construct scores (clean, separate, cook and chill, for a total of four dependent measures in the multivariate analysis)
• The multivariate analysis provided a view of the participant character tics on the joint and four individual food safety constructs that was contained in the total score.
• A type one error rate was set at P≤0.05 for each participant characteristic, assuming that there was an independent hypothesis for each participant characteristic
• To describe the size effects found in the analysis, a partial square (η²) was obtained for each statistical test of a fixed effect
• Following the recommendation of Cohen, a value of η²≥0.03 was considered non-trivial, representing an effect equivalent to 3% or more of the variance attributed to the statistical test represented by the F₁-statistic. 

Timing of Measurements

Consenting women completed the survey in April 2005.

Dependent Variables

Food practice safety score based on 23-item self-administered survey:

• Clean construct
• Separate construct
• Cook construct
• Chill construct
• Food eaten during the past month by pregnant participants.

Independent Variables

Pregnant women and mothers in the Women, Infants and Children (WIC) Program in Miami, Florida.

Control Variables

• Age group
• Education
• Race/ethnicity
• Country of birth
• Employment status
• Number of children
• Diarrhea among household members in the past month
• Household member at risk for food-borne illnesses, excluding young age or pregnancy.

• Initial N: 342 eligible clients
• Attrition (final N):  299 consented to participate
• Age: 

Age 18 to 24	39.8%
Age 25 to 34	45.2%
Age 35 or older	15%

• Ethnicity:

Non-Hispanic, non-Haitian black	64.4%
Hispanic	27.1%
Haitian	5.8%
Non-Hispanic white/Other	2.7%

• Other relevant demographics:
  • 64.8% were born in the United States
  • 34% were full-time homemakers or unemployed
  • 5.9% had no children and were WIC clients on the basis of their pregnancy
• Location: Miami-Dade County, Florida, US.

 

Key Findings

Food Safety Behaviors

• In general, a high percentage of participants reported "almost always" or "always" following good practices in the clean and separate constructs, although the frequency of "always" or "almost always" washing hands after changing diapers was significantly lower (83.6%) than the frequency of "always" or "almost always" washing hands after using the toilet (93%) (P<0.001)
• Also, a small percentage of participants (12.6%) reported not properly cleaning cutting boards after contact with raw meat
• In the cook construct, however, safe practices were less common, particularly regarding thermometer use
• Only one-fourth of the participants reported using a cooking thermometer "almost always" or "always" for cooking whole chicken or turkeys (23.4%) or other large pieces of meat (22.3%). 24.4% reported owning a thermometer.
• 24.7% reported usually eating undercooked eggs. Usually eating pink or undercooked meat was, however, rare (3.5%). 
• Safe practices in the chill construct were not practiced among a sizable proportion of participants
• About one third (32.2%) of the participants reported usually leaving food out for more than two hours.
• Few (17.3%) reported refrigerating large amounts of leftovers in shallow containers
•  Importantly, 10.8% reported leaving formula or bottled breast milk outside of the refrigerator for more than two hours "most of the time", "almost always" or "always" and 61.8% reported thawing foods on the countertop or in the sink in standing water
• 62 pregnant women participated in the survey. More than one half (51.6%) reported eating hot dogs or deli meats without first reheating sometimes or more frequently since becoming pregnant and 35.5% reported eating soft cheeses and blue-veined cheeses sometimes or more frequently since becoming pregnant. Both of these practices carry a risk of acquiring listeriosis. 

Factors Associated With Behaviors

• There were no statistically significant multivariate effects for any of the major demographic variables except for educational level. However, the effect size for educational levels was in the trivial range (η²=0.024).
• The univariate effects on the sum of the chill items for educational level and for race and ethnicity characteristics were significant although the effects were quite small (η²=0.033 for education and η²=0.041 for race and ethnicity)
• Chill scores were directly related to the amount of education (ie, the lower the education level, the lower the chill score). The following were the lowest mean scores:

Black non-Hispanic, non-Haitian group	85.9%
Haitian group	88.1%
Hispanic group	89.4%
Others	93.6%

• In a post-hoc contrast between pairs of groups, the only pair to show statistically significant differences was between the non-Hispanic, non-Haitian black group and the Hispanic group, with a mean difference in chill scores of 3.5 (P<0.01)
• The statistical test for a multivariate effect across the dependent measures clearly indicated that there was a significant difference by pregnancy status, F(4, 283)=6.62, P<0.001, with a moderate-to-large value of η² (0.086)
• There was a statistically lower average total score for those who were pregnant, F(1, 286)=13.69, P<0.001, although the effect size was small, η²=0.0446
• The size of effect was stronger for the sum of the clean items, F(1,286)=24.02, with a moderate effect size of η²=(0.077)
• The multivariate test for the number of children was significant, F(12, 744)=3.47, P<0.001, with a small effect of η² (0.047), with all of the other scores, except for the chill items having effect sizes of η² ranging from small (0.028) for the cook items to moderate (0.073) for clean items
• There was an even stronger effect size (0.088) for the total score, with F(3, 284)=9.13, P<0.001. The women who were pregnant with their first child had the lowest scores on all dimensions.
• For the variable diarrhea among a household member, there were statistically significant but small effect sizes for the multivariate analysis and for the clean items [F(4, 277)=3.84, P=0.005 and η²=0.053 and F(1,280)=8.72, P=0.003) and η²=0.03, respectively]
• Those with a history of diarrhea among household members during the past month had lower clean scores.

Other Findings

Psychometric Characteristics of the Survey Instrument

The item distribution of the revised survey were less skewed and sufficiently spread out to increase the potential for detecting change in behavior as a function of training. Moreover, the confirmatory factor analysis fit indices improved with a chi-square per df=1.457, a confirmatory fit index=0.969, a Tucker-Lewis index=960, a root mean square error of approximation=.039 (90% CI; 0.025 to 0.052) and a probability that the root mean square error of approximation 90% CI covered the value of 0.05 to 0.92.

The measure of inter-item reliability within each factor was estimated by the standardized Cronbach's alpha:

• The standardized Cronbach's alpha coefficients for the six clean items were 0.828 and 0.753 for the four separate items
• The cook items consisted of two subsets that, taken together, had a standardized Cronbach coefficient of 0.751
• The lowest standardized Cronbach's co-efficient was for the seven chill items at 0.389
• Upon further analysis, it appeared that the chill item about how frequently food was left out for "no more than two hours" may not have been understood by the participants
• An additional confirmatory factor analysis was conducted without the question, but there was no improvement in the parameters.

• This study provided at least three important findings:
  • Client's food safety practices were most problematic in the cook and chill constructs. Using a cooking thermometer, refrigerating foods within two hours and thawing foods safely were the practices least commonly reported
  • Being pregnant for the first time was the factor most commonly associated with suboptimal practices
  • Among pregnant participants, the consumption of foods that could expose women to listeriosis was common
• The findings from this study should be considered in the development of food safety education programs in similar populations. However, it should be noted that good personal hygiene behaviors can prevent illnesses that have a particularly high incidence and cost and thus should not be excluded from food safety education curricula. The relatively high prevalence of some suboptimal food safety practices is of particular concern because these WIC clients are either highly susceptible to food-borne illnesses themselves, if pregnant or they prepare food for their highly susceptible infants and young children. Especially worrisome is the high percentage of pregnant women who reported the consumption of foods that put them at risk of listeriosis. Importantly, it was found that women who were pregnant with their first child had the poorest food safely practice scores. It would be important to attempt to duplicate this finding in other populations.
• The WIC program is in a good position to provide food safety education because it serves people who are highly susceptible to food-borne illnesses, including pregnant women, infants and young children and has regular contact with its clients. Although all women in the WIC program could benefit from food safety education, these findings indicate that women during their first pregnancy should be especially targeted.

Government:

National Integrated Food Safety Initiative of the Cooperative State Research, Education, and Extension Service of the U.S. Department of Agriculture

The authors note the following limitations:

• The food safety practices were self-reported and no actual practices were observed. Studies that compare reported with observed food-handling practices indicate that people tend to over-report their performance of good food safely.
• Although refusal rates were low, those who refused may have been particularly unconcerned with food safety and have had worse practices than those who participated
• Because of these first two limitations, it is likely that the prevalence of inadequate food safety practices found in this study underestimates the true prevalence
• There were inconsistencies in the responses between the two questions about cooking eggs and between the two questions about how promptly foods were chilled
• Although the participants represented the clinic population well regarding race and ethnicity, they did not necessarily represent those of other WIC clinics in the area, Florida or the United States. Thus, the results may not be applicable to other populations, particularly those serving a smaller number of minority clients.
• This study assessed only self-reported practices and did not assess knowledge or attitudes; thus, it is not possible to determine if specific unsafe practices were due to knowledge deficits, attitudinal issues or specific barriers.