Bridging the Knowledge–Behavior Gap: A Cross-Sectional Study on Food Safety and Pesticide Exposure Among Migrant Workers in Chiang Mai

Highlights

Public health relevance—How does this work relate to a public health issue?

• Pesticide-contaminated food and unsafe food-handling practices remain important public health concerns in Northern Thailand because they increase the risk of pesticide exposure.
• Myanmar migrant workers are especially vulnerable due to limited access to food safety information and preventive resources, underscoring the need for targeted public health interventions.

Public health significance—Why is this work of significance to public health?

• Our findings suggest that improving food safety knowledge and attitudes alone may be insufficient to promote safe food-handling practices, highlighting the need for interventions that also target behavior change.
• The study shows that food safety knowledge, attitudes, and practices vary by sociodemographic characteristics among Myanmar migrant workers, supporting the development of targeted public health interventions.

Public health implications—What are the key implications or messages for practitioners, policy makers and/or researchers in public health?

• Food safety interventions for Myanmar migrant workers should go beyond education alone by addressing the barriers that limit the adoption of safe food-handling practices.
• Multilingual, culturally appropriate, and occupation-specific programs, supported by employers and migrant health services, may help improve food safety behavior and reduce pesticide exposure.

Abstract

Pesticide contamination in fruits and vegetables remains a critical food safety concern in Northern Thailand, posing risks to environmental and human health. Improving knowledge, attitude, and behavior (KAB) in food safety is essential for reducing pesticide exposure, particularly among migrant workers. This cross-sectional study evaluates food safety knowledge, attitude, behavior, and pesticide exposure among Myanmar migrant workers in Chiang Mai Province. Face-to-face interviews were conducted with 137 participants between August and September 2025 in Mueang, Mae Rim, and San Sai districts. Mann–Whitney U test and linear regression were used to explore demographic differences and associated factors of knowledge, attitude, and safety behavior. Most participants demonstrated adequate knowledge (65.7%) and positive attitude (80.3); however, 78.1% reported only moderate safety behavior. Higher knowledge and more favorable attitudes were observed among non-Shan ethnic groups and participants with higher education (p < 0.001). Women and non-farming workers exhibited significantly safer behavior (p < 0.05). Knowledge and attitude were strongly correlated (ρ = 0.556, p < 0.01), but neither was associated with safety behavior. Most participants had cholinesterase activity within the normal reference range (73.0% cellular AChE; 81.0% plasma BChE). Low enzyme activity was observed in 13.9% and 6.6% of participants, respectively. Enzyme activity did not differ significantly by occupation, sex, or ethnicity and was not significantly associated with KAB scores. Linear regression analyses identified ethnicity as significantly associated with knowledge and attitude; whereas, sex and occupation were significantly associated with safety behavior. These findings suggest that individual-level knowledge and attitude may not be sufficient to drive protective behavior among migrant workers, though the cross-sectional study design limits causal inference and generalizability.

1. Introduction

Food safety represents a critical public health priority and a fundamental element of global development and human security. Safe food is essential not only for human nutrition but also for economic stability and population well-being [1]. Annually, foodborne diseases affect an estimated 600 million people worldwide, leading to approximately 420,000 preventable deaths and substantial economic losses [2]. This burden disproportionately impacts low- and middle-income countries, where vulnerable populations have limited access to healthcare, preventive resources, and regulatory protection [3].

Beyond microbial contamination, the presence of chemical hazards, particularly pesticide residues, poses an increasing environmental health risk within the global food supply chain. In Northern Thailand, where agriculture remains a major economic sector, pesticide use has expanded dramatically to control pests and increase crop productivity [4]. According to the Ministry of Agriculture and Cooperatives, organophosphate and carbamate pesticides are among the most extensively utilized agrochemicals in agricultural practices [5]. Numerous monitoring studies have reported frequent contamination of fruits and vegetables with diverse pesticide residues, at levels above European Union maximum residue limits [6,7,8]. Prolonged reliance on chemical pesticides not only contaminates food and the environment but also elevates human exposure, contributing to detrimental public health effects [5].

Human exposure to pesticides occurs primarily through food consumption, inhalation, and dermal absorption. Such exposure can be linked to various adverse health outcomes, such as acute poisoning, cancer, developmental abnormalities, respiratory diseases, neurological disorders, and reproductive health problems [9,10,11]. Studies in Thailand have documented high rates of acute symptoms among farmers with unsafe pesticide handling practices [12,13]. Evidence from Northern Thailand also revealed widespread exposure to organophosphate and carbamate pesticides among adolescents and adults in agricultural communities, which was associated with chronic exposure and cognitive decline [14,15]. These health risks are exacerbated by unsafe pesticide use and inadequate food safety behavior among farmers, laborers, and consumers [4,16,17].

Migrant workers in Thailand experience elevated exposure risks within the agricultural sector. In 2023, Thailand registered over 2.3 million migrant workers from neighboring countries, with approximately 73% originating from Myanmar. Many are employed in agriculture, manufacturing, construction, and service industries, where they frequently encounter hazardous working conditions, inadequate occupational safety measures, and limited healthcare access. Structural disadvantages, including low educational attainment, lack of protective equipment, insufficient safety training, frequent misinformation, and absence of regular health monitoring, further increase their vulnerability to pesticide exposure and related health risks [18,19].

Several studies in Southeast Asia have assessed food safety knowledge, attitude and behavior concerning pesticide exposure among farmers and migrant workers. In Thailand, nearly one-third of migrant agricultural workers reported direct pesticide exposure, with common symptoms such as headaches, dizziness, and abdominal pain, reflecting unsafe behavior and knowledge gaps [18]. A recent study in Chiang Mai showed that higher education and safety training were associated with improved pesticide safety awareness, safer practices, and normal blood cholinesterase levels [4]. Similarly, a study in Malaysia reported that enhanced food safety literacy can promote positive attitudes and safe behavioral changes among migrant workers [3]. However, evidence from Bhutan and Nigeria demonstrated that improved knowledge and attitude do not always lead to safer behavior, highlighting the need for comprehensive behavioral interventions beyond education alone [20,21].

In Northern Thailand, food safety behavior and pesticide exposure have been extensively studied among adolescents, consumers, farmers and their families [4,12,14,16]. In contrast, evidence remains limited for Myanmar migrant workers in Chiang Mai province. This gap is particularly important, as migrant agricultural workers, food handlers, and residents living near pesticide application areas are likely to experience repeated and prolonged exposure to pesticide residues [17,22]. A comprehensive understanding of their knowledge, attitudes, and behavior (KAB) is therefore crucial for reducing exposure risks across occupations and improving overall community well-being.

Blood cholinesterase (ChE) activity is an established biomarker of exposure-related effects of organophosphate and carbamate pesticides, which inhibit acetylcholinesterase (AChE) and butyrylcholinesterase (BChE) enzyme activities. ChE biomonitoring has been widely used among agricultural workers in Thailand, including migrant farm workers, to complement questionnaire-based assessments of pesticide exposure and safety behaviors. Measuring AChE and BChE alongside KAB assessment provides complementary biomarker evidence of pesticide-related biological effects among migrant workers with potential occupational and non-occupational exposure.

The study addresses this knowledge–behavior gap by examining the factors associated with safety behavior among Myanmar migrant workers in Chiang Mai. It evaluates pesticide exposure using blood cholinesterase (ChE) biomarker. Specifically, the study aims to (i) examine the association between knowledge, attitudes and behavior (KAB); (ii) identify demographic factors associated with knowledge, attitude, and behavior (KAB); (iii) assess blood cholinesterase activity among the participants and examine its association with KAB.

2. Materials and Methods

2.1. Study Design and Participants

This cross-sectional study was performed during August and September 2025 in three districts of Chiang Mai Province, northern Thailand: Mueang, Mae Rim, and San Sai (Figure 1). Participants were recruited using a combination of purposive and convenience sampling. Purposive sampling was first employed to identify and access Myanmar migrant workers through community and group leaders, who had established connections with the target population. This approach was necessary because migrant workers are often dispersed and difficult to reach through formal sampling frames. Convenience sampling was then applied to recruit eligible participants who were readily available and willing to participate during the data-collection period.

The sample size was calculated a priori using G*Power (version 3.1.9.4), assuming a two-tailed test with a significance level of 0.05, a medium effect size (Cohen’s d = 0.30), and a statistical power of 0.95. The required total sample size was 134 participants, which was increased to 147 to account for a 10% attrition rate; however, only 137 participants were ultimately recruited and included in the final analysis due to time constraints and non-response. This calculation served as a general planning assumption for overall between-group comparisons rather than being powered for a single pre-specified primary regression outcome. While the achieved sample size provides adequate power for the main group comparisons, the regression and subgroup analyses were exploratory and should be interpreted with appropriate consideration of statistical power.

Eligible participants were active migrant workers aged 20–60 years who had lived and worked in Chiang Mai Province for a minimum of six months. Individuals with conditions that could affect cholinesterase activity, such as anemia, pregnancy, liver disease, and neurological disorders, were excluded from the study.

2.2. Data Collection Instrument

2.2.1. Questionnaire Survey

Data was collected through structured interview questionnaire administered by trained interviewers and research assistants affiliated with the Research Institute for Health Sciences, Chiang Mai University. The questionnaire was developed based on the theoretical framework and previously validated studies [3,4,16,21,23,24]. It was prepared in English, Thai, and Burmese languages to ensure comprehensibility among participants. The Burmese version was translated by the Language Institute of Chiang Mai University and pilot-tested with 10 Myanmar migrant workers. Item wording was revised based on pretest feedback. A formal independent back-translation procedure was not performed, which we acknowledge as a methodological limitation. Its internal consistency was assessed through a pilot test. It took about 15–20 min to finish the questionnaire. The full questionnaire is provided in Supplementary Materials (File S1). This questionnaire comprised four parts, including:

• Socio-demographic characteristics: Data were collected on participants’ age, sex, marital status, ethnicity, residence in Myanmar, education, occupation, monthly income, smoking and alcohol consumption habits, co-morbidity, working experience, participation in training, and exposure to relevant information.
• Knowledge in food safety and pesticide exposure: This section had 11 items assessing food contamination, poisoning, prevention, and exposure to pesticide residues and protection. Participants were presented with “yes”, “no”, or “don’t know” options. Each correct response was assigned 1 point, while incorrect or do not know responses received 0 points, resulting in a maximum total score of 11.

The total knowledge score was classified as a high level (≥80%, ≥9 points), moderate level (60–79%, 7–8 points), and low level (<60%, <7 points).

3.

Attitude towards food safety and pesticide exposure: This section comprised 7 items evaluating attitude towards food safety behavior and pesticide exposure protection.

Responses were rated on a 3-point scale (“Agree”, “Not sure”, and “Disagree”), with negatively worded items reverse-coded. The maximum total score was 14.

The total attitude score was classified into positive attitude (≥80%, ≥12 points), neutral attitude (60–79%, 9–11 points), and negative attitude (<60%, <9 points).

4.

Behavior in food safety and pesticide exposure reduction: This section included ten items (items 35–44) rated on a 3-point scale (“Always”, “Sometimes”, and “Never”), with reverse scoring for negatively phrased items, which cover general food-handling hygiene (e.g., handwashing and cooking practices) and practices intended to reduce dietary pesticide-residue intake (e.g., washing/soaking produce and selecting unblemished produce). In addition, six items (items 45.1–45.3 and 46.1–46.3) assessed food-sourcing and exposure context (e.g., the proportion of home-grown versus market-purchased produce before and after migration) rather than protective behaviors, presented with “yes”, “no” options. Therefore, the behavior score reflects dietary exposure-reduction behavior and general food hygiene. The maximum score was 28.

The total behavior score was classified into good behavior (≥80%, ≥23 points), moderate behavior (60–79%, 17–22 points), and poor behavior (<60%, <17 points).

The number of items, scoring weights, and maximum scores for each domain were developed based on the theoretical framework and previously validated instruments. The percentage-based cut-off points for categorizing total scores as high, moderate, or low were based on Bloom’s cut-off criteria, which are commonly used in KAB survey research [12,25]. Content validity was supported by an index of item-objective congruence (IOC) above 0.5 for all items. The reliability coefficients (Cronbach’s α) were 0.81 for knowledge, 0.89 for attitude, and 0.82 for behavior in food safety and pesticide exposure reduction, indicating good internal consistency. While these indices support content validity and internal consistency, they do not by themselves establish full construct or cross-cultural validity of the instrument.

2.2.2. Measurement of Blood Cholinesterase Activity

Blood cholinesterase activity was measured using the original Ellman kinetic method to assess enzyme inhibition specific to organophosphate and carbamate pesticides. Both erythrocyte acetylcholinesterase (AChE) and plasma butyrylcholinesterase (BChE) were measured using acetylthiocholine iodide (ATChI) and butyrylthiocholine iodide (BTChI) as substrates, respectively. Reference values were calculated as the mean ± 2 standard deviations (SD) of enzyme activity levels measured in the study population sampled in the same area and period (erythrocyte AChE: 4.4–6.4 U/mL; plasma BChE: 2.7–4.6 U/mL), rather than derived from externally validated clinical or population references. Values below the range were classified as low enzyme activity, within the range as normal enzyme activity, and above the range as high enzyme activity.

Chemical reagents

Acetylthiocholine iodide (ATCh), butyrylthiocholine iodide (BTCh), and 5,5′-dithiobis-2-nitrobenzoic acid (DTNB), were purchased from Sigma-Aldrich (St. Louis, MO, USA).

Blood sample collection

Capillary finger-prick blood samples were collected by trained personnel prior to questionnaire administration.

Assay procedure

The assay was based on the reaction between thiocholine, produced from the enzymatic hydrolysis of ATChI or BTChI, and DTNB (Ellman’s reagent), resulting in the formation of a yellow-colored 5-thio-2-nitrobenzoate anion. The intensity of the yellow color, measured at 405 nm, is directly proportional to cholinesterase (ChE) activity [26].

In brief, the samples were diluted with phosphate-buffered saline (PBS) at pH7.4. Washed red blood cells (RBCs) were resuspended and diluted 1:600, while plasma was diluted 1:100 in PBS. A 1:1 mixture of ATChI or BTChI and 5,5′ dithiobis (2-nitrobenzoic acid) was prepared in 3.0 mM saline phosphate buffer (DTNB) solution. Then each 100 μL of diluted blood samples was mixed with 200 μL of the premixed solution in a 96-well microplate. The final concentrations in the assay mixture were 1.0 mM ATChI for the RBC assay or 3.0 mM BTChI for the plasma assay, with 0.30 mM DTNB. Immediately, the absorbance was measured at 405 nm, recording absorbance values at 0 and 5 min using UV-vis spectrophotometry microplate reader (CLARIOstar Plus, BMG LABTECH, Ortenberg, Germany). All measurements were conducted in duplicate, and enzyme activity values were used to evaluate the level of pesticide exposure risk among participants.

Calculation of enzyme activity

The reaction rate was expressed as the change in absorbance over 5 min (ΔA₄₀₅). The ChE activity was calculated in units per milliliter (U/mL), where one unit represents the amount of enzyme that hydrolyzes 1 μmol of the substrate in 5 min at 25 °C. The calculation was performed using the following equation [27]:

ChE activity (units/mL) = ∆Absorbance × factor

Factor = [1000/5(1.35 × 10⁴)] × [Total vol./Sample vol.] × dilution factor

where 1.35 × 10⁴/M/cm = extinction coefficient of yellow anion; 1000 = conversion from mM/mL to µM/mL; units/mL= U/mL [26]. For the RBC assay (1:600 dilution), the factor was 26.64. For the plasma assay (1:100 dilution), the factor was 4.44.

2.3. Statistical Analysis

Statistical analyses were conducted using SPSS software (version 25). Descriptive statistics were summarized as frequency (n), percentage (%), mean and standard deviation (SD), median (interquartile range; IQR). The normality of the data was assessed before conducting inferential analyses. The Mann–Whitney U test was applied to examine differences in knowledge, attitude, behavior, and enzyme activities across demographic factors. Categorical variables that originally contained more than two groups were dichotomized before analysis due to small individual subgroup sizes. All Mann–Whitney U test comparisons were performed between two groups per variable. Spearman’s correlation analysis was used to evaluate the associations among knowledge, attitude, and behavior. Independent variables associated with each outcome at p < 0.2 in bivariate analysis were entered into hierarchical linear regression models, with covariates added sequentially according to the predefined model structure rather than through automated stepwise, backward, or forward selection. Unstandardized regression coefficients (B) with 95% confidence intervals were reported, and model fit was assessed using R², adjusted R², and the overall F-test.

2.4. Ethical Consideration

Ethical approval for this study was obtained from the Human Experimentation Committee, Research Institute for Health Sciences, Chiang Mai University (Approval no. 9/68). Written informed consent was secured from all participants prior to data collection. All methods were performed in accordance with the Declaration of Helsinki and relevant guidelines and regulations.

3. Results

3.1. Demographic Information

In this study, a total of 137 Myanmar migrant workers participated. Most of the participants were aged <40 years (64.2%), with a mean age of 36.7 years. The majority were women (58.4%), originated from non-Shan ethnic groups, including Burmese, Karen, and Karenni ethnic groups (55.5%), and had secondary or higher education (58.4%). Farming was reported by 43.1% of the participants. However, the proportion of the participants with smoking (18.2%), alcohol consumption (26.3%), and co-morbidity (29.2%) were relatively low. Regarding working experience, 53.3% had worked in Chiang Mai for 5 years or longer (53.3%), while 46.7% had worked there for less than five years (

Table 1. Demographic characteristics of the study participants (n = 137).

Demographic Characteristics		Frequency (n)	Percentage (%)
Age (years)	<40	88	64.2
	≥40	49	35.8
	Mean ± SD = 36.7 ± 10.17
Gender	Man	57	41.6
	Woman	80	58.4
Ethnicity	Shan	61	44.5
	Other ethnic groups £	76	55.5
Education levels	No formal and primary	57	41.6
	Secondary and higher	80	58.4
Occupation	Farming	59	43.1
	Non-farming	78	56.9
Smoking	No	112	81.8
	Yes	25	18.2
Alcohol consumption	No	101	73.7
	Yes	36	26.3
Co-morbidity	No	97	70.8
	Yes	40	29.2
Years working in Chiang Mai	<5 years	64	46.7
	≥5	73	53.3

^£ Burmese, Karen, Karenni ethnic groups.

).

3.2. Knowledge, Attitude, and Behavior Regarding Food Safety and Exposure to Pesticide Residues

3.2.1. Levels of Knowledge, Attitude, and Behavior

The overall distribution of knowledge, attitude, and safety behavior levels is presented in

Table 2. Distribution of scores for knowledge, attitude, and practice among the study participants (n= 137).

Variables	Level (Scores)	Frequency (n)	Percentage (%)
Knowledge	Low (<7 scores)	9	6.6
	Moderate (7–8 scores)	38	27.7
	High (≥9 scores)	90	65.7
Attitude	Negative (<9 scores)	5	3.6
	Neutral (9–11 scores)	22	16.1
	Positive (≥12 scores)	110	80.3
Behavior	Poor (<17 scores)	23	16.8
	Moderate (17–22 scores)	107	78.1
	Good (≥23 scores)	7	5.1

. The results showed that most migrant workers demonstrated a high level of knowledge regarding food safety and pesticide exposure (65.7%), followed by those with moderate (27.7%) and low knowledge levels (6.6%). For attitude, the majority exhibited a positive attitude (80.3%), while 16.1% demonstrated a neutral attitude, and 3.6% reported a negative attitude. Regarding behavior, 78.1% of the participants had a moderate level of behavior, with small proportions reporting good (5.1%) or poor behavior (16.8%).

3.2.2. Distribution of Responses to Knowledge, Attitude, and Behavior Among the Study Participants

Figure 2 presents the proportion of correct responses to knowledge concerning food safety and pesticide exposure, with accuracy levels ranging from 49.6% to 100%. Most participants demonstrated strong awareness of fundamental food safety principles and pesticide-related risks. All respondents (100%) correctly identified the risk of consumer harm from the improper use of pesticides. More than 90% recognized key preventive practices, including handwashing during food handling, washing vegetables and hands before eating, and using gloves to reduce dermal contact with pesticide residues. However, substantial gaps remained in understanding exposure pathways. Only 49.6% correctly rejected the misconception that ingestion is the only route of pesticide exposure, and 67.2% recognized dermal absorption as an exposure pathway. These findings indicate adequate overall knowledge but persistent misunderstandings regarding the routes of pesticide exposure among the study participants.

In terms of attitude regarding food safety and pesticide exposure, the responses demonstrated a general trend toward high awareness and cautious preference among participants, as illustrated in Figure 3. More than 80% of participants agreed that food safety knowledge is essential for preventing food poisoning, preferred consuming fresh and pesticide-free vegetables, and recognized that pesticide residues can enter the body through both ingestion and dermal contact. High levels of disagreement were also observed for negatively worded items. Specifically, most participants disagreed that pesticide residues do not affect their health (84.7%) and that vegetables can be consumed without washing if one is healthy (90.5%). Additionally, 73.7% disagreed with the statement that pesticides provide benefits without any associated risks. Across all attitude items, the proportion of “Not sure” responses was low, ranging from 3.6% to 16.8%, further indicating a generally clear and consistent attitude profile among the participants.

Figure 4a,b illustrates the distribution of food safety behavior among participants. Overall, adherence to basic hygiene practices was high. Most respondents reported consistently engaging in key safe behaviors, with 81.0% always washing hands before cooking or serving food, 75.2% always washing hands with soap, and 89.8% routinely rinsing fruits and vegetables with clean water. More than half also reported always cooking their own food and peeling fruits before consumption. Unsafe practices were generally uncommon. Only 10.2% reported always consuming unwashed produce, and 15.3% reported always eating food without handwashing, while over 70% indicated that they never engaged in these unsafe behaviors. Enhanced cleaning practices showed moderate uptake among participants. Approximately 40.9% always cleaned vegetables by soaking in salt, vinegar, limewater, bicarbonate, or rice-wash water, and 47.4% did so sometimes. However, only 16.1% consistently soaked produce for ≥15 min, indicating limited adherence to extended decontamination practices (Figure 4a).

Changes in food-sourcing behavior following migration were also observed, as presented in Figure 4b. Before moving to Thailand, most participants (70.8%) obtained produce from both planting and purchasing, while 12.4% relied exclusively on homegrown vegetables. After migration, reliance on market-purchased produce increased substantially, with 64.2% buying all their produce and only 4.4% continuing to grow their own. This shift may reflect reduced access to land and an increased dependence on commercial food sources after relocation.

3.2.3. Comparison of Knowledge, Attitude, and Behavior Scores by Demographic Characteristics

Table 3. Differences in knowledge, attitude, and behavior of the study participants by demographic characteristics (n = 137).

Demographic Characteristics	Median (IQR)
	Knowledge	Attitude	Behavior
Age (years)
<40	9.00 (8.00–10.00)	13.00 (12.00–14.00)	19.00 (17.25–21.00)
≥40	9.00 (8.00–10.00)	13.00 (11.00–14.00)	19.00 (17.50–20.00)
p-value	0.521	0.120	0.339
Sex
Man	9.00 (8.00–10.00)	13.00 (12.00–14.00)	19.00 (17.00–20.00)
Woman	9.00 (8.00–10.00)	13.00 (12.00–14.00)	19.00 (18.00–21.00)
p-value	0.431	0.649	0.023 *
Ethnicity
Shan	8.00 (8.00–9.00)	12.00 (10.00–13.00)	19.00 (16.00–20.00)
Other ethnic groups	9.50 (9.00–10.00)	14.00 (13.00–14.00)	19.00 (18.00–20.00)
p-value	<0.001 **	<0.001 **	0.081
Education levels
No formal and primary	8.00 (7.00–10.00)	12.00 (10.00–13.00)	19.00 (16.00–20.00)
Secondary and higher	9.00 (9.00–10.00)	13.00 (12.00–14.00)	19.00 (18.00–20.00)
p-value	0.001 **	<0.001 **	0.233
Occupation
Farming	8.00 (7.00–10.00)	12.00 (10.00–13.00)	18.00 (16.00–20.00)
Non-farming	9.00 (9.00–10.00)	14.00 (13.00–14.00)	19.00 (18.00–20.00)
p-value	0.001 **	<0.001 **	0.010 *
Smoking
No	9.00 (8.00–10.00)	13.00 (12.00–14.00)	19.00 (18.00–20.00)
Yes	9.00 (8.00–9.00)	13.00 (11.50–14.00)	19.00 (15.50–19.00)
p-value	0.071	0.709	0.018 *
Alcohol consumption
No	9.00 (8.50–10.00)	13.00 (12.00–14.00)	19.00 (18.00–20.00)
Yes	8.00 (7.00–9.00)	12.00 (10.25–14.00)	19.00 (17.00–20.00)
p-value	<0.001 **	0.019 *	0.350
Co-morbidity
No	9.00 (8.00–10.00)	13.00 (12.00–14.00)	19.00 (17.00–20.00)
Yes	9.00 (8.00–10.00)	13.00 (11.00–14.00)	19.00 (18.00–20.00)
p-value	0.495	0.862	0.512
Years working in Chiang Mai
<5	9.00 (9.00–10.00)	13.00 (12.00–14.00)	19.00 (18.00–20.00)
≥5	9.00 (8.00–10.00)	12.00 (11.00–14.00)	19.00 (17.00–20.00)
p-value	0.004 **	0.006 **	0.604

Abbreviations: IQR—interquartile range; significant difference at * p value < 0.05; ** p value < 0.01; data analyzed with Mann–Whitney U test.

shows that multiple demographic variables demonstrated significant differences in the knowledge, attitude, and behavior scores across the study participants. Age did not significantly influence participants’ knowledge (p = 0.521), attitudes (p = 0.120), and behaviors (p = 0.339) when comparing individuals under 40 years of age with those aged 40 years and older. In contrast, women demonstrated significantly better food safety and pesticide-protective behaviors than men (p = 0.023), although knowledge and attitude did not differ by sex.

Ethnicity showed a significant effect on both knowledge and attitudes. Participants from non-Shan ethnic groups demonstrated higher knowledge levels (p < 0.001) and more positive attitudes (p < 0.001) than Shan participants, although behavioral differences were not statistically significant (p = 0.081). Similarly, higher educational attainment was associated with greater knowledge (p = 0.001) and more favorable attitudes (p < 0.001), while education levels did not significantly influence behavior. Significant differences were also observed by occupation. Non-farming workers exhibited higher knowledge (p = 0.001) and attitude scores (p < 0.001), as well as better food-handling and pesticide-protective behaviors (p = 0.010), compared with farming workers.

Smoking status influenced behavior but not knowledge or attitude. Non-smokers had significantly higher behavior scores (p = 0.018); whereas, no differences were noted in knowledge (p = 0.071) and attitude (p = 0.709). However, alcohol consumption was associated with lower scores in knowledge and attitude. Participants who consumed alcohol had significantly lower knowledge (p < 0.001) and less favorable attitudes (p = 0.019) than non-drinkers, while behavior scores did not differ between the two groups (p = 0.350). For comorbidity status, there were no significant variations in any score. Finally, years of work experience in Chiang Mai demonstrated an inverse relationship with knowledge and attitude scores. Participants with less than five years of work experience in the province showed higher knowledge (p = 0.004) and attitude scores (p = 0.006), though their behavior scores did not differ significantly (p = 0.604).

3.2.4. Correlation Between Participants’ Knowledge, Attitude, and Behavior

Table 4. Correlation between knowledge, attitude, and safety behavior scores.

Variables	Median (IQR)	Correlation Coefficient
		Knowledge		Attitude		Behavior
		Not Controlled	Controlled	Not Controlled	Controlled	Not Controlled	Controlled
Knowledge	9.00 (8.00–10.00)	-	-	0.556 **	0.533 **¥	0.031	−0.019 †
Attitude	13.00 (12.00–14.00)	0.556 **	0.533 **¥	-	-	0.132	0.058 ‡
Behavior	19.00 (17.50–20.00)	0.031	−0.019 †	0.132	0.058 ‡	-	-

Abbreviations: IQR, interquartile range; ** significant difference at p-value < 0.01, by Spearman’s rho and partial correlation. ^¥ controlled for behavior; ^† controlled for attitude; ^‡ controlled for knowledge.

presents the correlations among knowledge, attitude, and safety behavior scores. A strong positive correlation was observed between knowledge and attitude (ρ = 0.556, p < 0.01). This association remained significant after controlling for behavior (ρ = 0.533), indicating that higher knowledge levels were consistently associated with more positive attitudes.

Knowledge and behavior were not significantly correlated (ρ = 0.031), and the relationship remained non-significant after adjusting to attitude (ρ = −0.019). Similarly, attitude showed no significant correlation with safety behavior in both the unadjusted (ρ = 0.132) and adjusted analyses (ρ = 0.058, controlling for knowledge).

Overall, while knowledge demonstrated a strong association with attitude, neither knowledge nor attitude showed a meaningful correlation with safety behavior, suggesting that behavior may be influenced by factors beyond knowledge and attitude domains.

3.2.5. Comparison of Enzyme Activity by Participants’ Demographic Characteristics (n = 137)

Overall, plasma (BChE) activity ranged from 1.4 to 8.6 U/mL (median 3.5, IQR 3.0–4.1) and cellular (AChE) activity ranged from 2.1 to 7.8 U/mL (median 5.4, IQR 4.8–6.1). Applying the study-derived reference ranges (erythrocyte AChE: 4.4–6.4 U/mL; plasma BChE: 2.7–4.6 U/mL), 100 participants (73.0%) had cellular enzyme activity within the normal range (4.4–6.4 U/mL), 19 (13.9%) had low cellular enzyme activity (<4.4 U/mL), and 18 (13.1%) had high cellular enzyme activity (>6.4 U/mL); for plasma enzyme activity, 111 (81.0%) were within normal range (2.7–4.6 U/mL), nine (6.6%) were low (<2.7 U/mL), and 17 (12.4%) were high (>4.6 U/mL).

Table 5. Differences in plasma and cellular enzyme activity by demographic characteristics (n = 137).

Demographic Characteristics	Median (IQR)
	Plasma Enzyme Activity	Cell Enzyme Activity
Age (years)
<40	3.31 (2.88–4.05)	5.45 (4.71–6.12)
≥40	3.74 (3.29–4.18)	5.49 (4.88–6.19)
p-value	0.01 *	0.900
Sex
Man	3.73 (3.08–4.13)	5.59 (4.75–6.19)
Woman	3.41 (2.95–4.12)	5.43 (4.81–6.09)
p-value	0.314	0.748
Ethnicity
Shan	3.42 (3.00–3.97)	5.31 (4.68–5.96)
Other ethnic groups	3.58 (3.03–4.26)	5.54 (4.86–6.28)
p-value	0.313	0.256
Education levels
No formal and primary	3.32 (2.90–3.98)	5.31 (4.67–5.94)
Secondary and higher	3.53 (3.08–4.22)	5.57 (4.85–6.30)
p-value	0.247	0.103
Occupation
Farming	3.42 (2.85–4.00)	5.45 (4.68–6.13)
Non-farming	3.51 (3.10–4.14)	5.46 (4.82–6.21)
p-value	0.321	0.764

Abbreviations: IQR—interquartile range; significant difference at * p-value < 0.05; data analyzed with Mann–Whitney U test; Reference ranges: plasma (BChE) activity 2.7–4.6 U/mL; cellular (AChE) activity 4.4–6.4 U/mL.

presents plasma and cellular cholinesterase activities across demographic characteristics. Plasma enzyme activity was significantly higher in participants aged ≥40 years compared with those <40 years (3.74 vs. 3.31 U/mL; p = 0.01); whereas, cellular enzyme activity did not differ significantly between age groups (5.49 vs. 5.45 U/mL; p = 0.900). Plasma and cellular enzyme activities were slightly higher in men than women (Plasma: 3.73 vs. 3.41 U/mL, p = 0.314; Cell: 5.59 vs. 5.43 U/mL, p = 0.748), and in participants from other ethnic groups compared with Shan participants (Plasma: 3.58 vs. 3.42 U/mL, p = 0.313; Cell: 5.54 vs. 5.31 U/mL; p = 0.256), though these differences were not statistically significant. Similarly, plasma and cellular enzyme activities tended to be higher in participants with secondary or higher education (Plasma: 3.53 vs. 3.32 U/mL, p = 0.247; Cell: 5.57 vs. 5.31 U/mL, p = 0.103), and in non-farming compared with farming participants (Plasma: 3.51 vs. 3.42 U/mL, p = 0.321; Cell: 5.46 vs. 5.45 U/mL, p = 0.764); although, these differences did not reach statistical significance.

3.2.6. Correlations Between Cholinesterase Activity and KAB Scores

Spearman’s correlation analysis showed no statistically significant association between cholinesterase activity and KAB scores (

Table 6. Spearman’s correlations between plasma and cellular cholinesterase activity and knowledge, attitude, and behavior scores.

Enzyme Activity	Correlation Coefficient (ρ)
	Knowledge	Attitude	Behavior
Plasma enzyme activity	0.048, p = 0.574	0.021, p = 0.804	−0.066, p = 0.440
Cell enzyme activity	−0.008, p = 0.923	0.066, p = 0.444	−0.042, p = 0.623

Abbreviations: ρ, Spearman’s correlation coefficient; p, p-value.

). Plasma enzyme activity was weakly correlated with knowledge (ρ = 0.048, p = 0.574), attitude (ρ = 0.021, p = 0.804), and behavior (ρ = −0.066, p = 0.440). Similarly, cellular enzyme activity showed very weak and non-significant correlations with knowledge (ρ = −0.008, p = 0.923), attitude (ρ = 0.066, p = 0.444), and behavior (ρ = −0.042, p = 0.623).

3.2.7. Linear Regression Analysis for Knowledge, Attitude, and Behavior Scores

The linear regression analysis identified significant associations between selected sociodemographic characteristics and KAB scores among migrant workers as shown in

Table 7. Linear regression analysis for predictors of knowledge, attitude, and behavior scores by demographic characteristics among Myanmar migrant workers.

Dependent Variables	Model	Ethnicity	Sex	Age	Occupation	Education
Knowledge	Model 1	1.003 **	0.205	0.040	-	-
	Model 2	0.766	0.176	0.026	0.299	-
	Model 3	0.593	0.248	−0.014	-	0.533
	Model 4	0.578	0.225	0.035	0.095	0.545
Attitude	Model 1	1.626 **	−0.078	−0.279	-	-
	Model 2	0.940 *	−0.161	−0.319	0.867	-
	Model 3	1.352 **	−0.055	−0.280	-	0.427
	Model 4	0.855	−0.139	−0.315	0.775	0.246
Safety behavior	Model 1	0.565	1.118 **	−0.379	-	-
	Model 2	−0.423	0.998 *	−0.438	1.250	-
	Model 3	0.568	1.118 **	−0.379	-	−0.003
	Model 4	−0.311	0.969 *	−0.443	1.372 *	−0.325

Unstandardized regression coefficients (B) shown from linear regression models. Reference groups: Shan ethnicity; male sex; farming occupation; lower education. Statistically significant at * p < 0.05, ** p < 0.01; “-“ indicates that the variable was not included in that model.

. Four hierarchical models were constructed for each dependent variable, with covariates entered sequentially to examine associations after adjustment for different combination of sociodemographic factors. For the knowledge score, ethnicity was significantly associated with knowledge scores in Model 1 (B = 1.003, p < 0.01), indicating that participants from non-Shan ethnicities demonstrated higher knowledge scores compared to the Shan participants after adjusting for age and sex. However, this association became non-significant after further adjustment for occupation in Model 2 and/or education in Models 3 and 4 (B = 0.578–0.766, p > 0.05). Age, sex, occupation, and education were not significantly associated with knowledge scores in any of the models.

Similarly, ethnicity showed a robust, positive association with the attitude score in Models 1 (B = 1.626, p < 0.01), Model 2 (B = 0.940, p < 0.05), and Model 3 (B = 1.352, p < 0.01), but the association was attenuated in Model 4 (B = 0.855, p > 0.05). Other variables such as sex, age, occupation, and education did not exhibit significant associations with attitude scores across any model. For safety behavior, sex was significantly associated with higher scores across all models (B = 0.969–1.118, p < 0.05–0.01), indicating higher protective behaviors among women. Occupation was additionally associated with safety behavior scores in Model 4 (B = 1.372, p < 0.05); whereas, ethnicity, age, and education were not significantly associated with safety behaviors in any model. These results suggest that ethnicity was significantly associated with knowledge and attitude; whereas, sex and occupation were significantly associated with safety behavior. Overall, the models were statistically significant for knowledge (R² = 0.120–0.150; adjusted R² = 0.097–0.124; F-test, p ≤ 0.002), attitude (R² = 0.226–0.250; adjusted R² = 0.209–0.225; F-test, p < 0.001), and safety behavior (R² = 0.081–0.109; adjusted R² = 0.054–0.080; F-test, p ≤ 0.023). Variance inflation factors across all models and outcomes ranged from 1.00 to 3.04, indicating no serious multicollinearity among predictors.

4. Discussion

This study examined food safety knowledge, attitude, behavior, and pesticide exposure among Myanmar migrant workers in Chiang Mai, and identified the sociodemographic factors associated with KAB outcomes and cholinesterase enzyme activity. Overall, most participants demonstrated moderate (27.7%) to high (65.7%) levels of knowledge and positive attitudes toward food safety and pesticide risks, but 78.1% showed only moderate adherence to recommended safety behaviors, with just 5.1% classified as having good behavior. This knowledge–behavior gap is consistent with findings from Malaysia, Bhutan, and Nigeria, where adequate knowledge and positive attitude does not always translate into protective actions [3,20].

Despite a strong correlation between knowledge and attitude, neither was associated with protective behaviors. Similar findings among migrant and agricultural workers in Malaysia and Thailand indicate that knowledge alone has limited influence on behavioral change [3,12,28]. This persistent gap is attributed to structural and contextual constraints rather than individual awareness, including limited access to protective equipment, economic pressure, time constraints, and normalized risky practices in the workplace [18,22,29]. Evidence from northeastern Thailand showed that protective equipment availability, comprehension of pesticide labels, and workplace social support were the strongest predictors of safe pesticide behaviors, emphasizing the need for interventions beyond education-focused approaches [22].

In addition, the result shows significant differences in KAB outcomes across demographic profiles. Participants from non-Shan ethnic groups, such as Burmese, Karen, and Karenni, had better knowledge and more positive attitudes than Shan ethnic workers. This discrepancy may reflect variations in community networks, language proficiency, literacy, or occupational roles within migrant communities [3,19,30]. Similarly, participants with secondary or higher education reported significantly better knowledge and attitude, aligning with evidence showing that literacy facilitates access to safety information and comprehension of food safety and pesticide risks [4,31]. Non-farming workers demonstrated higher knowledge, attitude, and behavior scores than farming workers, reflecting persistent gaps in safety training and access to protective equipment among agricultural laborers [32]. In addition to occupational differences, gender also significantly influenced safety behavior, with women exhibiting safer behavior than men. This finding may be linked to women’s primary role in household food preparation and their greater attentiveness to hygienic and protective practices [33,34].

The marked shift from home-grown to market-purchased produce after migration may reflect migrant workers’ limited access to land and housing with growing space in their host communities, together with reliance on wage labor rather than subsistence farming. While this shift may reduce workers’ control over the specific pesticide handling practices used on the produce they consume, it may also increase reliance on the food safety practices captured in this study (e.g., washing, soaking) as their primary means of reducing dietary pesticide-residue intake.

In this study, cholinesterase enzyme activity, used as a biomarker of pesticide exposure, was largely unrelated to participant characteristics, except for the age groups. Workers aged 40 years and older demonstrated higher plasma cholinesterase activity than younger workers. Similar counterintuitive patterns have been reported in recent biomonitoring studies, in which age-related physiological variation, differences in baseline cholinesterase levels, or reduced involvement in high-intensity spraying among older workers my influence measured enzyme activity [35,36,37,38]. Furthermore, single-point biomonitoring may not reflect peak or seasonal exposure, potentially underestimating short-term cholinesterase inhibition [39,40]. Although other demographic variables did not show significant associations, male gender, individuals from non-Shan ethnic groups, participants with higher education, and non-farming workers exhibited slightly higher enzyme levels, suggesting marginal variations in exposure pathways or safety practices.

Consistent with the absence of an occupational difference in ChE activity, no statistically significant correlations were observed between cholinesterase activity and knowledge, attitude, or behavior scores. This finding may partly reflect the limited variation in ChE activity within the study population and the fact that the behavior scale primarily assessed dietary and food-hygiene behaviors rather than occupational pesticide-handling behaviors. As a result, self-reported knowledge and behaviors may not necessarily correspond to biomarker-based indicators of exposure to cholinesterase-inhibiting pesticides. These findings highlight that self-reported KAB measures, and biochemical biomarkers provide complementary but distinct information and should not be assumed to directly track one another.

Notably, although farming workers reported significantly poorer safety behavior than non-farming workers (Table 3), plasma and cellular cholinesterase activity did not differ significantly between the two groups (Plasma: 3.42 vs. 3.51 U/mL, p = 0.313; Cell: 5.45 vs. 5.46 U/mL, p = 0.764; Table 5). This finding suggests that poorer self-reported safety behavior among farming workers should be interpreted as an indicator of elevated behavioral risk rather than as evidence of greater biological exposure at the time of sampling. This discrepancy may reflect shared environmental or dietary exposure, differences in sampling timing, and heterogeneity in farming tasks. Therefore, poorer safety behavior among farming workers did not correspond to significantly different ChE activity in this study.

Furthermore, cholinesterase activity was not correlated with knowledge, attitude, or behavior scores, indicating that better self-reported knowledge or safety behavior did not necessarily correspond to less enzyme inhibition in this sample. These findings may reflect the influence of exposure pathways outside occupational farming, differences in exposure timing, and individual or task-specific variation in pesticide exposure. Therefore, the observed cholinesterase findings should be interpreted cautiously as evidence of possible enzyme inhibition rather than as a direct measure of current pesticide exposure. Future studies should combine cholinesterase measurements with exposure-timing information, pesticide-use history, and task-specific occupational classification to better characterize exposure.

Linear regression analyses showed that ethnicity was initially associated with knowledge and attitude, but its effect diminished after adjusting for education, occupation, and other demographic factors. This finding suggests that the observed ethnic differences in knowledge and attitude may be partly related to differences in other sociodemographic characteristics. Recent studies show that migrant ethnicities significantly shape food safety awareness and perceptions [3,22,41]. In contrast, sex and occupation remained significantly associated with safety behavior across all models, with women and non-farming workers demonstrating safer behavior. These results differ from a study in Thailand reporting safer pesticide-use behaviors among male farmers [22]. Furthermore, lack of an effect of education on safety behavior suggests that structural and contextual factors, such as work tasks, exposure patterns, and gender responsibilities, play a greater role in shaping safety behavior than knowledge alone [3,19].

This study uses a biomarker-based cholinesterase measurement and a multilingual, validated questionnaire, which enhances the accuracy and cultural appropriateness of exposure and KAB assessments. However, the use of purposive sampling and the restriction to three districts in Chiang Mai may constrain the generalizability of these findings to migrant people in other regions. The absence of formal records on approached and declining participants precludes assessment of selection and non-response bias, while the purposive and convenience sampling may limit generalizability to similarly recruited migrant communities. The cross-sectional design restricts causal inference, and self-reported behaviors may be influenced by recall or social desirability bias. Moreover, single-time biomonitoring cannot capture seasonal or task-specific fluctuations in pesticide exposure. As all participants in this study held valid work documents, the findings may not generalize to undocumented Myanmar migrant workers. Future studies employing longitudinal designs and broader geographic sampling are recommended to strengthen the external validity of these findings.

Understanding the factors affecting behavior in food safety and pesticide exposure among migrant workers will inform migrant-inclusive, targeted education and awareness-raising programs. The findings suggest that educational approaches may be strengthened by considering contextual and occupational factors, including work conditions, access to protective measures, and exposure-related behavior. Government and relevant organizations play a critical role in implementing effective policies for providing accurate information on food safety and pesticide residues and strengthening pesticide residue monitoring in local markets to address potential exposure pathways.

5. Conclusions

This study revealed that Myanmar migrant workers in Chiang Mai demonstrated a high level of knowledge and positive attitudes toward food safety and pesticide exposure; however, their safety behaviors remained predominantly moderate, with notable disparities across ethnicity, education, sex, and occupation. The study highlights the importance of considering these sociodemographic factors when planning and implementing effective interventions for migrant populations. Although knowledge and attitude were moderately correlated, neither was significantly associated with safety behavior after adjustment. This suggests that factors beyond individual awareness, such as work conditions, resource limitations, and economic pressures, may constrain the adoption of safe practices. However, these structural, occupational, and other individual-level factors were not comprehensively assessed in this study and therefore cannot be confirmed as associated factors of safety behavior.

Blood cholinesterase activity, used as a biomarker of organophosphate and carbamate exposure, indicated low activity in 13.9% (cellular) and 6.6% (plasma) of participants. However, it did not differ significantly across most sociodemographic and occupational groups, nor was it correlated with knowledge, attitude, or behavior scores. This finding, together with the behavioral results, suggests that self-reported practices and biological exposure markers should be interpreted separately rather than assumed to align.

The findings in this study support the potential value of targeted, culturally appropriate, and occupation-specific interventions that address both informational gaps and practical constraints faced by migrant workers. Additionally, multilingual community education, employer-supported safety training, and strengthened migrant health services are essential to reduce pesticide exposure risks and promote safer behaviors, although this study cannot directly demonstrate their effectiveness. By considering these diverse factors, future studies can help develop more comprehensive and effective approaches to promoting awareness and safer behaviors.