Red meat handlers’ food safety knowledge, attitudes, and practices in the Dhaka megacity of Bangladesh

ABSTRACT Foodborne disease is a major health problem worldwide and unhealthy food handling practices are a cause of foodborne disease (FBD) transmission among – populations. Red meat handlers play an important role in ensuring food safety throughout production, processing, transportation, and preparation. This study investigates the food safety knowledge, attitudes, and practices (KAP) of red meat handlers in the Dhaka megacity of Bangladesh (DMCB). Four hundred cross-sectional samples were collected from the DMCB using a pretested survey-structured questionnaire. Descriptive, and inferential statistics, and the KAP model were applied using partial least square structural equation modeling (PLS-SEM) with the help of SmartPLS 0.3 software to achieve objectives. All factor loading scores (>0.5) and Cronbach’s alpha (α ≥ 0.69) were greater than the reference value, which means that indicators of reliability and internal consistency measure latent variables. The results found a significant positive association between red meat handling practice and knowledge and attitude. Red meat handlers’ knowledge has a moderating effect on the relationship between attitude and practice, where attitude positively affects practices toward food safety. The findings revealed that the KAP model met the goodness-of-fit criteria (HTMT<1, SRMR<0.08, NFI>0.90) and was acceptable (AVF≤0.50). The estimated model for food safety knowledge, attitude, and practices is well-suited and deemed acceptable. The findings also reveal that food safety knowledge has a significant and beneficial impact on food safety attitudes. Furthermore, attitudes regarding food safety and its measures are linked in a significant way. The findings could aid public health professionals and practitioners in developing focused initiatives to improve meat handlers’ food safety knowledge and practices and prevent FBD. The government should pay special attention to improving food safety knowledge and ensuring quality handling practices to protect FBD transmission in the DMCB.


Introduction
Foodborne diseases (FBD) occur in developing countries, and each year 420,000 people die, and 10% of the population gets sick from eating contaminated food. [1] In Bangladesh, FBDs are prevalent due to poor food handling practices, inadequate sanitation facilities, insufficient food safety legislation, weak regulatory systems, lack of financial resources to invest in food safety equipment, and lack of food handlers' education and knowledge. Microbiological agents cause infections, and biotoxins, The written close-ended questionnaire was developed in the following ways: First, a two-round Delphi method was conducted to identify the key statements regarding the KAPs of the meat handlers to prevent FBDs and ensure food safety. Second, the study adopted and modified the statements with using previously published studies [18][19][20][21][22][23] to develop a survey schedule as study's requirements. Later, the study designed a self-administered KAP questionnaire to collect data. It was divided into socioeconomic questions and food safety questions regarding knowledge (13 items), attitude (11 items), and practices (11 items).

Data analytical technique
Descriptive and inferential statistics were used to summarize variables of interest, while structured equations were used to summarize variables of interest (the multicollinearity among independent variables was checked using a variance inflation factor and tolerance; Shapiro-Wilk test for normality; Cronbach's alpha for data reliability and correction of the relationship between variables; Kaiser-Meyer-Olkin test for validity, etc.). The KAP model was applied by using partial least square structural equation modeling (PLS-SEM) with the help of SmartPLS 0.3 ( Figure 1).
SmartPLS Procedure: A partial least square (PLS) that is used in this research. According to, [24] PLS is simultaneous modeling (SEM) technique capable of analyzing latent variables, indicators, and measurement errors in real time. PLS can be used with few samples and applied to all data scales. [25] We have summarized the process for calculating PLS-SEM in SmartPLS in five stages, as shown in Figure 1.

Construction of latent variables
The red meat handlers' KAP question patterns are divided into several points: personal hygiene, prevention of cross-contamination and sanitation, food handling, and health problems. The following latent variables are constructed based on the meat handler KAP questionnaire's in Table 1.  Cleaning food contacting and after surfaces before processing 3.46 1.52 0.512 P 7 Proper cleaning and handling of instruments reduces the risk of food contamination.

Food safety knowledge
The food safety knowledge variables in Table 2 were generated based on thirteen observations (K1-K13). Most respondents were aware of handling procedures (personal hygiene and sanitary) to ensure food safety. So, we used a binary (yes/no) response for each indicator. All standard factor loadings were >0.40 and Cronbach's alpha was (α = 0.69). All observation reliability and internal consistency contract latent variable knowledge (K).

Food safety attitude
The attitudes of food handlers were constructed based on eleven indicators (A 1 -A 11 ). Food safety attitudes and self-reported practices were evaluated using a 5-point Likert scale. Each indicator was scaled as follows: strongly agree (4), agree (3), neutral (2), disagree (1), and strongly disagree (0). In contrast, for negatively worded items, the lowest point (0) was given to "strongly agree," and the highest (4) was given to "strongly disagree." All standard loadings were >0.527, and Cronbach's alpha (α = 0.77) determines indicator reliability and internal consistency to measure the latent variable attitude (A).

Practices
The meat handlers' practices are constructed based on eleven indicators (P1-P11). This questionnaire comprises eleven items rated based on a 5-point Likert scale (1 = very low to 5 = very high). Each indicator was scaled according to the previous steps (b). The results showed that the loading factor scores were >0.484, and Cronbach's alpha (α = 0.83) designated that most of the indicators were consistent and reliable in measuring latent variable practices (P). A total of thirty-five questions (K1-K13, A1-A11, and P1-P11) were used to measure the following three latent variables: knowledge (K), attitude (A), and practices (P) to be accepted KAP, as shown in Table 1.
The present study's findings indicate that the questionnaire on food safety KAP and prevention was valid and reliable. The reliability study revealed that all items in the three latent constructs had strong internal consistency, with Cronbach's alpha coefficients of more than 0.50 for each construct. This finding was consistent with that of, [26] who found that all items in a food safety practice questionnaire met the acceptable criteria for internal consistency and that a Cronbach's alpha of 0.50 or greater was an acceptable criterion for internal consistency in reliability analysis. [27] Results and discussion

Sociodemographic profiles
The sociodemographic profiles of the respondents are presented in Table 2. Out of the 400 food handlers, 76.75% (n = 307) were male, and 23.25% (n = 93) were female. Most studies [20,28,29] have reported a higher proportion of males, involved in food handling. The mean age of the respondents was 39, ranging between the ages of 20 and 69, and 78% were married. Approximately 10.75% of the respondents had no formal education. More than half (70.75%) of respondents had worked for more than 5 (five) years, but very few respondents had food safety training (21.25%) or health certificate (25.75%). About 58.76% of the respondents reported having experienced symptoms of food poisoning diseases such as vomiting, diarrhea, headaches, stomach cramps, fever, and loss of appetite (Table 2). Figure 2 shows that 64.75% of respondents had adequate knowledge, 49.46% had adequate attitudes and.12% had adequate food safety and hygiene practices. Knowledge of food safety concepts does not necessarily corroborate self-reported practices for food safety behavior. Although 64.75% showed positive knowledge, only.12% implemented adequate food safety practices . [30,31] Table 3 shows an association between KAP and food handlers working in food courts. There was a statistically significant association between knowledge and attitudes (X 1 = 0.896, p = .014) and attitudes and practices (X 1 = 0.766, p = .049 and practices score (X 1 = 0.667, p-value = 0.501) ( Table 3). The strongest positive correlation was between attitude scores and practice scores. Knowledge is the key element influencing attitudes and practices among food handlers. This association shows that food handlers with excellent knowledge will have good attitudes and practices. These findings support a prior study that showed a strong positive relationship between knowledge, attitude, and behavior. [16,32] However, several studies have shown that simply learning about food safety does not instill a healthy sanitary attitudes and practices among farmers and food handlers. [33]

Partial least square structural equation modeling (PLS-SEM) results
In contrast to previous investigations on the interrelationships of food safety KAPs, we examined the postulated links using PLS-SEM, a second-generation statistical approach. For this type of research design and causal link prediction, PLS-SEM is a good choice [34] as it is the most commonly used SEM approach when the sample size is small and the topic field is new. [35] In most cases, a PLS-SEM is analyzed and interpreted in two stages. The first step is to examine and refine the measurement model's adequacy, followed by examining and reviewing the structural model. This guarantees that the metrics are reliable and valid before attempting to draw conclusions from the structural model. We will test the following three hypotheses, H 1 : Red meat handlers' knowledge has a moderating effect on the relationship between attitude and practices toward food safety.    Red meat handlers' knowledge has a moderating effect on the relationship between attitude and practices, where attitude positively impacts practices toward food safety. The hypothesized model for food safety knowledge, attitude, and practices had a good fit and was acceptable. The structural model showing the relationship between the variables of food safety knowledge, attitude, and practice is shown. Hypothesis testing was performed using PLS-SEM. The PLS-SEM full model can be seen in Figure 3.
The PLS-SEM analysis showed that the path coefficient for food safety knowledge and practices was 0.218 (P < .05), suggesting that food safety knowledge affects food safety practices. The path coefficient is positive, signifying that attitudes toward food safety become more positive as knowledge of food safety increases. This result indicates the importance of developing programs that improve food safety knowledge of red meat handlers. Therefore, H 1 is supported. This result was similar to. [16,19,36] The path coefficient of food safety knowledge and attitude was 0.220 (P < .05), suggesting that food safety knowledge affects food safety attitude. This result is in accord with studies [16,19,36] and [37] that stated that good knowledge could lead to a positive attitude. Similarly, a study by [38] found a significant positive correlation between knowledge and attitude. This result indicates the need to improve food safety attitudes. The path coefficient of food safety attitude and practice was 0.322 (P < .01), suggesting a relationship between attitudes toward food safety practices, and the results showed that these two constructs are significantly and positively related. Thus, H 3 is supported, but the result is argued by. [39] Knowledge has a considerable impact on food safety attitudes and practices, and attitudes toward food safety and practices are also significant and positively related. These findings, comparable to, [36] imply the necessity to increase food safety knowledge to improve red meat handlers' food safety attitudes and practices.

Outer model test results
The outer model focuses on testing the validity and reliability of each indicator on its latent variable. This outer model (measurement model) is formed by testing convergent validity, discriminant validity, extracted average variance (AVE), and composite reliability.

Convergent validity
The average variance recovered should be greater than 0.5, which refers to how well individual items reflect concept coverage compared to items measuring various constructs [40] . The factor loadings in this investigation showed that the six constructs have convergent validity. All loadings are above 0.50, with a majority more than 0.60. The loadings of the factors ranged from 0.56 to 0.96The high factor loadings conclude that the measures have convergent validity. All construct factor loadings exceeded the 0.50 cutoff, except IS sophistication (AVE = 0.2). However, the IS sophistication dimensions were found to have adequate convergent validity based on their high composite reliability (>0.70) [41] .

Reliability of measures
The reliability of the construct items is the final stage in determining construct validity. A composite alpha value is a measure of internal consistency or composite dependability. This value was used to evaluate the ten constructs' dependability. Construct reliability coefficients should all exceed the 0.70 lower limits [42] The composite reliability and Cronbach's alpha values for the studied constructs were computed by SmartPLS and ranged from 0.690 to 0.830 and 0.255 to 0.5, respectively. From the Table 4 it is clear that the variables used in this research were reliable since they obtained composite reliability and Cronbach's alpha values greater than 0.5. All values fall within the acceptable range to conclude good reliability.

Discriminant validity
To measure discriminant validity, we used the Heterotrait-Monotrait Ratio (HTMT). [41] HTMT criterion. It examines the findings of the Heterotrait-Monotrait ratio of correlations (HTMT) criterion (Table 6), the cross-loadings of each item in the constructs (Table 2), and the estimated square root of average variance extracted (AVE) for all constructs ( Table 4). Table 5 represents the results of the Fornell-Larcker criterion. The square root of AVE of the latent variable should be more than other correlation values among the latent variables. [43,44] For example, the latent variable knowledge AVE is 0.116 (Table 4); hence, its square root becomes 0.3. This number is larger than the correlation values in the knowledge column (Table 5) and more significant than those in the knowledge row (0.220). A similar observation is also made for the latent variable of attitude and practices. The results indicate that discriminant validity is well established.

Heterotrait menotrait ratio (HTMT)
The correlation of indicators within the same notion refers to measuring distinct phenomena via indicators. [35] Smart PLS3 software was used to estimate the Heterotrait Menotrait ratio (HTMT). The threshold value of HTMT was found to be <1 [42,45] . The HTMT results revealed that all the values were substantially different from 1. The HTMT ratio of correlation in Table 6 shows that all values are below the threshold of 0.85, [43,44] indicating that the reflective constructs are discriminately valid.

Cross-loadings
All the constructs' elements were weighted more heavily on their respective constructs. As a result, concept discriminant validity was sufficient.

Average variance extracted (AVE)
The average commonality is used to assess the reflective construct's convergent validity (average variance extracted). It should be greater than 0.5 to be considered suitable. The AVE for each latent factor was more significant than the squared correlation between the factors, indicating discriminant validity. [46]

Goodness-of-fit indices
The SRMR is an index that measures the average of standardized residuals between observed and hypothesized covariance matrices and is an estimated measure of model fit. The estimated model has a good fit when SRMR = 0.08, with a smaller SRMR showing a better fit. Table 7 presents that value of SRMR was 0.073, indicating that it fit well, while the chi-square was 1101.74 and the NFI was 0.239.
The values for SRMR, dULS, and dG in the "Estimated Model" column for the assessment use the guidelines stated earlier. The SRMR value in our model is 0.086, which is slightly higher than the 0.08 criterion, showing a poor theoretical model fit.

Standardized root means square residual (SRMR)
We calculated and presented the results using the model evaluation findings as a guide The An SRMR of 0 implies a perfect fit, and an SRMR value of less than 0.05 suggests an adequate fit. [47] SRMR value of 0.08 or lower is acceptable. A value significantly greater than 0.08 suggests the absence of fit. Some studies use a more lenient cutoff of less than 0.10. The lower the SRMR, the more accurate the model. When SRMR is zero, the ideal fit is realized. An SRMR of 0.08 or less is considered acceptable. A score much more than 0.08 indicates a lack of fit. Others choose a softer cutoff of less than 0.10. [47,48]

Normed fit index (NFI)
The Bentler-Bonett index, known as the normed fit index (NFI) is an incremental fit metric that computes the proposed model's chi-square value and compares it to a meaningful benchmark, measuring how well something fits. [49] NFI values greater than 0.9 indicate a good match and are deemed acceptable for factor models. [47] The NFI levels for composite models have yet to be specified. The NFI is rarely used and should be implemented with caution when comparing models because it does not penalize increasing parameters.

D_ULS and D_G
According to, [48] dLS and dG are two distinct methods for calculating this discrepancy. The lower the dULS, the more accurate the model. The geodesic (dG) inconsistency was 0.516, indicating that the model in this investigation is better suited.

Conclusion
This study investigated red meat handlers' food safety KAPs in the DMCB. The results found that food safety knowledge has a significant and favorable impact on attitudes toward food safety. Additionally, attitudes regarding food safety and practices were strongly and positively interrelated. However, food handlers in Bangladesh have insufficient food safety knowledge and practices. They need to improve their expertise, increase secondary education attainment, gain more work experience, and more working hours per day, increase training, and raise income. These factors are linked to better food safety awareness among meat handlers in Bangladesh. Red meat handlers have enough food safety knowledge but do not convert it into tight hygienic standards of food processing and handling. Food handlers have an important role in preventing food contamination that can develop into foodborne disease outbreaks. Food handlers who work with red meat in the DMCB must handle meat properly to avoid food contamination. This study aimed to evaluate the knowledge, attitudes, and food handling practices regarding food safety and personal hygiene among food handlers at DMCB. Training programs must be institutionalized with specific guidelines that covering food safety and meat hygiene topics to educate meat handlers better. Finally, to reduce foodborne infections and diseases in Bangladesh, intervention and longitudinal studies including large, diverse samples of Bangladeshi meat handlers, are needed to investigate characteristics associated with their food safety knowledge and practices.

Limitations of the study
Our study has limitations due to the limits of food safety knowledge questionnaires, and only a few KAP items were included in this study. This study considers certain limitations when interpreting the findings because of the small sample size and a lack of KAP questions. The first was the intrinsic limitation of the cross-sectional sample, which prevents the formation of cause-effect correlations, and the second was the possibility of social desirability bias. Furthermore, because it was self-reported, this study did not provide a complete picture of the country, necessitating additional micro-level research that considers other aspects of food safety concerns.