This is a community-based cross-sectional study on butchers of Bangladesh from January 2023 to December 2024. This study used a cross-sectional research design with a quantitative approach to evaluate the factors affecting food safety knowledge of butchers and beef handlers of selected districts. A two-stage stratified random sampling method was applied to select the district and the respondent. From each division of Bangladesh, two districts were randomly selected for the study. Our sampling frame was butchers and beef handlers of the randomly selected districts. A total of 16 districts included Bagerhat, Barisal, Bhola, Chattogram, Dhaka, Feni, Gaibandha, Gazipur, Khulna, Moulvibazar, Mymensingh, Rajshahi, Rangpur, Sherpur, Sirajganj, and Sylhet. A total of 160 butchers and beef handlers from eight divisions were randomly surveyed. Butchers and beef handlers were selected randomly from the market, that ensures each of them had an equal chance of being selected for the survey. Although it is understood that a big sample size increases the strength or statistical power of cross-sectional studies, the study was restricted to a small number of participants. Due to resource limitations, we only sampled 160 butchers and beef handlers. In spite of this restriction, the study made sure to adhere to appropriate sampling techniques in order to improve representativeness across the study area. As a result, this study offers valuable insights into the variables influencing food safety knowledge, but more research with a larger sample size may be done to further evaluate the facts.

A structured KAP questionnaire was developed to assess the food safety awareness, hygiene behavior, and meat handling practices of butchers and beef handlers in Bangladesh. The questionnaire was designed based on existing literature, food safety guidelines, and expert consultations to ensure its relevance and validity. It was divided into three key sections: (a) sociodemographic information—collected data on age, education, work experience, and type of meat-related occupation (butcher or meat seller); (b) knowledge assessment—included questions on awareness of Good Agricultural Practices (GAP), Good Manufacturing Practices (GMP), Hazard Analysis and Critical Control Points (HACCP), hygiene standards, and food safety laws in Bangladesh; and (c) attitude evaluation—assessed respondents’ perspectives on the importance of food safety training, hygiene regulations, preservation facilities, and their willingness to adopt improved practices.

The study emphasized the current status of food safety knowledge of the butchers and beef handlers of Bangladesh and evaluated whether education, experience, legal knowledge, cleanliness practice, etc., are affecting the knowledge of food safety.

Studies involving analysis of a categorical dependent variable often use a multinomial logit model or an ordered logistic model [17–19]. A multinomial logistic regression model was not considered in this study because of its inability to account for the inherent ordering of the dependent variable. Therefore, the ordered logistic regression model was found to be the most appropriate in this study because of the ordinal nature of the dependent variable. According to Fullerton (2009) [20], the ordered logistic model is commonly employed when the dependent variable has an ordinal outcome.

The ordered logistic regression model makes the assumption of proportional odds, which means that the link between predictors and the dependent variable remains the same throughout the ordered outcome’s thresholds. The Brant test (omodel) was used to confirm this hypothesis. According to the test results, the chi-square (χ²) value was statistically significant (p < 0.001), indicating that the proportional odds assumption was violated for several predictors, such as age, education, legal knowledge, and profit. The use of a conventional ordered logit model is not appropriate when the proportionate odds hypothesis is broken [21, 22].

The generalized ordered logit model was taken into account due to this infraction. In order to strike a balance between flexibility and frugality, the PPOM was estimated in Stata using the gologit2, pl command. Only variables that violated the proportional odds assumption are relaxed by the PPOM, while those that obeyed the assumption are kept, which enhances its accuracy, and the models are noted for their effectiveness and interpretability [15, 21, 22]. For clarity of understanding, odds ratios (ORs) were presented.

The generalized ordered logit model effectively builds the relationship between an ordinal outcome and a combination of ordinal and continuous explanatory variables, preserving the inherent order of the dependent variable and providing more accurate and interpretable estimates. As McCullagh notes, the ordered logit model is specifically designed for situations where the response variable is ordinal, making it ideal for this analysis [23]. Several studies have employed ordered logit regression to assess hygiene practices among beef handlers. For instance, Akpan et al. [24] utilized this method to evaluate the relationship between hygiene practices and various factors among beef handlers in Ibadan, Nigeria. Similarly, Viator et al. [25] applied ordered logit regression to examine food safety practices among meat processing workers in the United States. These studies demonstrate the applicability of ordered logit regression in analyzing ordinal dependent variables related to hygiene practices in the meat industry.

The variables utilized in the regression model are listed in Table 1, along with the anticipated sign of the association between the predictors and the food safety knowledge of butchers and beef handlers.

Stata (version 15.1) was used for the data analysis. To describe the socioeconomic profile of survey participants as well as their food safety knowledge, beliefs, and behaviors, descriptive statistics were calculated. χ² tests were used to analyze the bivariate relationships between food safety knowledge levels and explanatory variables. At the 5% level, statistical significance was evaluated. We treated the five-point scale responses to the food safety knowledge questions (covering GAP, GMP, HACCP, meat safety law, workplace and personal hygiene, and withdrawal periods of antibiotics, steroids, and veterinary drugs) as the ordinal dependent variable. The categorization of food safety knowledge levels used in this analysis is presented in Table 2.

The three categories: low, medium, and high, of food safety knowledge, as specified in Table 2, were considered to be an ordinal dependent variable. An ordered regression approach was utilized in order to discover the variables that are linked to the understanding of food safety. The Brant test (omodel) was used to assess the proportional odds assumption. Because the hypothesis was broken for many covariates (p < 0.001), a PPOM was calculated using the gologit2, pl command. For ease of comprehension, ORs are presented with 95% confidence intervals. The estimated PPOM having a single latent continuous variable with multiple cut-points is given in this equation:

l n   P   (Y   >   j) P   (Y   ≤   j)   =   α j + β 1 j   A g e   +   β 2 j   E d u c a t i o n   +   β 3 j   E x p e r i e n c e   +   β 4 j   L e g a l   k n o w l e d g e   +   β 5 j   P r a c t i c e   +   β 6 j   R e s p o n d e n t   t y p e   +   β 7 j   C l e a n l i n e s s   +   β 8 j   P r o f i t ;   f o r   j   =   1 ,   2 , …

Figure 1 shows that the sample consisted predominantly of butchers (90.62%), with the remaining 9.38% being meat sellers. The age distribution indicates that most respondents were between 21 and 30 years old (40.63%), followed by those aged 31–40 years (34.38%), suggesting a largely young to middle-aged, energetic workforce.

Education levels were fairly low: 63.75% had only a secondary education, and 13.75% had only a primary education. Higher education was uncommon, with just 0.62% holding an undergraduate degree. Regarding work experience, 40.63% reported 11–20 years of experience, while only 7.50% had more than 30 years. These figures imply that the industry is dominated by mid-career professionals who possess practical experience but generally have limited formal education in food safety.

In spite of the limited knowledge, butchers showed a positive attitude toward improving food safety. While most respondents lacked awareness of the hazards associated with veterinary drugs, antibiotic residues, and contamination risks, more than 90% agreed that scientific training should be provided to butchers. Furthermore, 93.75% believed that traditional meat-handling practices should be improved, and 95.63% supported the need for better preservation facilities. On the contrary, their actual practices were far away from what they needed to do. Only 8.75% performed post-mortem inspections to detect abnormalities, and almost none recorded animal diseases, followed scalding or chilling procedures, or used grading techniques. A significant lack of adoption of technology was also observed, as 99.38% of butchers reported not using any modern processing tools. Results from Table 3 state the current status of knowledge, attitudes, and practices among butchers. While butchers acknowledge the need for improvement, they require resources and training to implement better meat-safety practices.

The findings from Figure 2 indicate that meat handling practices in Bangladesh are still mostly traditional. The carcass rejection rate was comparatively low, with 83.13% of butchers rejecting less than 5% of carcasses, which suggests that unfit animals might still be entering the food chain. Most butchers (86.87%) stored unsold meat in deep freezers, while the rest relied on other methods that could compromise food safety. Meat was generally sold within three hours of slaughter (81.25%), but some butchers (3.75%) kept meat for more than nine hours, increasing the risk of bacterial contamination. Meat cutting methods were predominantly traditional (88.75%), with very few butchers following wholesale or retail cutting practices.

Figure 3 illustrates the distribution of hygiene and cleanliness ratings, showing that the majority of butchers rated carcass surface hygiene as very poor, followed by poor workplace and process hygiene. Hygiene standards among butchers were generally poor, particularly in terms of carcass surface hygiene, which 78.13% of respondents rated as “very poor”. Process hygiene and workplace hygiene also scored poorly, indicating an overall lack of sanitary measures. Equipment hygiene fared slightly better, with 85.63% and 6.88% of butchers rating it as “fair” or “good”. However, general cleanliness remained inadequate, with only a few respondents maintaining satisfactory hygiene levels.

The availability of adequate slaughterhouse facilities was another critical concern. Although 65.63% of respondents reported access to a slaughterhouse, a substantial proportion (34.38%) slaughtered animals in unsanitary locations such as roadsides, drains, or beside their shops. These unhygienic conditions greatly increase the risk of contamination and pose serious public health hazards.

Table 4 indicates that butchers sold an average of 236.25 kg of meat per day, with a mean profit of 22.38 BDT per kg. While these figures highlight the economic significance of the meat industry, the low profitability suggests that butchers may have limited financial resources to invest in better facilities, training, or food safety measures. This economic constraint could be a barrier to adopting improved meat-handling and hygiene practices.

Figure 4 illustrates the extent of knowledge across various meat-safety parameters. These parameters were measured using structured questions and recorded on a five-point scale: “Not Acquainted”, “Poorly Acquainted”, “Fairly Acquainted”, “Acquainted”, and “Highly Acquainted”.

Responses from butchers in different regions of Bangladesh reveal a substantial lack of understanding of critical food-safety concepts, as shown in Figure 4. Almost all butchers were unfamiliar with GAP, GMP, and HACCP, all of which are essential for ensuring meat quality and hygiene. Knowledge of Bangladesh’s food-safety laws and regulations was also very low, with only a minority of respondents demonstrating any awareness. Awareness of workplace and personal hygiene was poor: 69.38% of respondents were rated as “Poorly Acquainted”. Knowledge of withdrawal periods for veterinary drugs, antibiotics, and steroids was alarmingly low, with 80% of respondents showing little to no understanding, raising the risk of chemical residues contaminating meat. In summary, only a small number of beef handlers are aware of proper meat-safety practices.

The χ² test results (Tables S1–S7) showed that there was no statistically significant variation in food safety knowledge according to age, education, experience, respondent type, cleanliness, practice level, or legal understanding.

A number of diagnostic and validity tests were conducted to make sure the regression approach was appropriate. Spearman’s rank correlation (Table 5) was used to analyze multicollinearity among explanatory variables. There are moderate correlations between several predictors, like age and experience (ρ = 0.743) and age and education (ρ = –0.695), according to the results. Considering the nature of population trends, this was anticipated. Nevertheless, none of the correlations came close to unity, and other factors like cleanliness and practice (ρ = 0.161 with age; ρ = –0.129 with education) also played a role. Detailed marginal effects from the ordered logit model are summarized in Table S8, which supports the robustness of the PPOM results.

Second, the Brant test (omodel) was used to formally assess the ordered logit model’s parallel regression (proportional odds) assumption (Table S9). The PPOM was used in the study to deal with this problem since it maintains proportionality but only relaxes the proportionality assumption for the variables that violate it.

Third, the likelihood ratio (LR) χ² statistic was used to assess the model’s fit. The PPOM yielded a pseudo R² of 0.1916 and an LR χ²(18) = 33.64 with p = 0.0139 (Table 6), demonstrating that, taken as a whole, the predictors were able to predict well, which describes a significant amount of the variation in food safety understanding.

According to the findings, butchers between the ages of 21–30 (OR = 0.146, p = 0.032) and 31–40 (OR = 0.029, p = 0.003) had considerably lower odds of being in a higher knowledge category than the youngest group (20 years old or younger).

The PPOM findings, in general, highlight that gaps in food safety knowledge are widespread and more impacted by economic output and practices that are self-reported rather than based on demographic characteristics, emphasizing the need for comprehensive training and resource-oriented initiatives to raise the bar for meat safety in Bangladesh.

This multi-district survey demonstrates widespread deficiencies in the knowledge and practices of beef butchers in Bangladesh, particularly in post-slaughter inspection, cold storage, and awareness of international food-safety standards. Economic performance and reported practice levels were positively linked to food safety knowledge, while demographic characteristics were not. These findings suggest that financial capacity and practice-based experience may be more important determinants than background factors.

To improve beef safety, interventions should prioritize training programs tailored to butchers, investment in preservation and processing facilities, and stricter enforcement of existing regulations. Study limitations include the modest sample size, small numbers in higher knowledge categories, and reliance on self-reported practices. Future research should integrate microbiological assessments and larger, representative samples to better inform policy.