In 2017, a total of 5,911,567 individuals had registered as agricultural-related workers in Thailand ³⁰ . Only workers who had worked for the past 3 months were eligible to participate in the survey. The prevalence of substances used from the latest national household survey in 2016 was used as a study parameter to calculate sample sizes ³ . The sample size calculation method for prevalence survey ³¹ was employed with adjustment of design effect from multistage sampling. With an expected 10% non-response and ±1%% error, a sample size of 3140 individuals was targeted. Stratified multistage cluster sampling was employed. Thailand was stratified into 10 geographical zones on the basis of governmental narcotic operation areas; each zone comprised 7–12 provinces. Each zone was a systematically sampled province (one zone per province). The 10 sample provinces were Chiangmai, Phitsanulok, Khon Kaen, Nakornrachasima, Samitprakarn, Chonburi, Samutsakorn, Krabi, Satun, and Bangkok. Then, the agricultural clusters of each province were mapped. Only one cluster was randomly selected from each province, which gave 10 clusters altogether. Households were systematically selected from the updated map of individual clusters. The number of households was determined on the basis of the probability proportional to size. Simple random sampling using a table was employed to obtain a sample from each household (one household per sample). Of 3140 intended samples, 2936 (93.5%) individuals agreed to participate in the survey (ages ranging from 16 to 69 years old, with an average of 36.6 years old; 61.3% (1799) were males).

After the baseline survey, 124 current methamphetamine users agreed to participate in structured observation regarding their drug use (age ranging from 18 to 58 years old, with an average of 27.4 years old; 88.7% (110) were males).

Outcomes of the study were identified as drug use prevalence. In this study, drug use was defined as the use of illicit drugs (IDs) (heroin, cannabis, methamphetamine, etc.); legal substances, such as non-prescription drugs alcohol, tobacco and caffeine, were not included. Of all IDs, methamphetamine is hypothesized to be widely used among agricultural rated workers to accelerate productivities resulting in higher prevalence. Annual prevalence as well as past month prevalence was obtained.

The interview questionnaire and structured observation guideline were developed by 12 experts ³² . The interview questionnaire comprised individual-level, interpersonal-level, and cluster-level variable sections, as well as substance use. The structured observation guideline emphasized pattern of methamphetamine use and its effects.

A total of 20 research assistants (two for each province, one male, one female) were trained to conduct field work (updating a cluster map, devising a household sampling frame, randomly selected samples, requesting consent, and interviewing the samples). The research assistants asked the intended participants to provide written informed consent, waited a couple of days to allow the participants to make their own decision, and, then, made an appointment for them to come back to be interviewed/observed. Data from the survey were gathered from September to November 2017 via private, face-to-face interviews. Interview guides are available as Extended data. The participants chose a convenient place and time for their interviews (e.g., the participant’s household). An average of 1–1.5 hours was spent for each interview. After the interview, the research assistants requested written consent to participate in structured observation of the drug users. Only those who consented (n=124) were observed regarding their drug-use behavior—such as when and how IDs were used, and the amount and frequency used—and its effects. The research team conducted participant observations from November 2017 to July 2018 (3–5 times for each sample, 1–2 days for each observation).

The outcome variable was substance use. The current use in this study was defined as having taken drugs at least once in the past month from the interview date.

The individual-level variables included sex, age, education, monthly income, and marital status. All variables were categorized as dichotomous variables.

The interpersonal-level variables were assessed to reflect the extent to which the drug user’s peers used methamphetamine or other IDs. Potential covariates were accounted for by adjusting for individual and interpersonal-level variables. These variables were selected on the basis of evidence regarding factors that affect substance abuse and the neighborhood environment ^{20, 21, 23, 25, 28} .

The one cluster-level variable was the level of drug use in each of the provinces sampled in 2017. The level of drug use was defined as the proportion of the population that used drugs and was partly assessed by the number of drug users who accessed the treatment system. In this study, the treatment rate was the proportion of registered clients of governmental treatment centers in the population in a given area. The estimated rate (per 1000 people) of ID users/methamphetamine users was determined from the survey in a given area in 2017.

Data double-entry was employed. SPSS for windows version 16.0 was used. Descriptive statistics were used to analyze all socioecological factors. Multilevel binary logistic regression analysis via generalized linear mixed models was used to analyze the associations among cluster-level variables, each covariate, and substance use. Level 2 comprised individuals at level 1 (individual-level variables and interpersonal-level variables) nested within clusters at level 2 (consisting of two variables). The model building process started with a null model consisting of no predictors, and a series of two-level models was developed. Model 1 fit only individual-level variables into the model. Then, in model 2, all interpersonal-level variables were entered into model 1. Finally, in model 3, all cluster-level variables were entered into model 2. The median odds ratio (mOR) and interval odds ratio (IOR) were applied to measure the variation of substance use in different clusters and the effects of cluster-level variables, respectively. The level of statistical significance was set to p-values < 0.05.

This research project was approved by the Human Research Ethics Committee of Khon Kaen University and was conducted in accordance with the principles of the Declaration of Helsinki and International Conference on Harmonization Good Clinical Practice standards. Written informed consent was obtained from the participants and from the parents/guardians of minors.

The median age of the agriculture-related workers was 36 years; 15.9% of the workers were <25 years old and 47.7% had obtained elementary education and lower ( Table 1). One-third of the workers were poor (earned < per annum), 29.4% were current smokers and 19.1% were habitual smokers (>20 days in the past month), and 17.9% had smoked >1 pack daily. Regarding alcohol, 48.7% were current alcohol drinkers, 30.4% were habitual drinkers, and 24.1% had drunk >330 mL of beer or the equivalent amount of alcohol every time they drank (which resulted in a blood alcohol level >50 mg/dL). Regarding substance abuse, 57.7% (n=1695) of the participants currently used non-medical pharmaceutical drugs, mostly analgesics, codeine mixture cough syrup, sedatives, or opioids. Among the agriculture-related workers, 73.5% (2158) had used IDs at some point in their life, and 13.9% (408) were current ID users. This number is much higher than 1.99 in general population (Kanato et al. 2017). Current ID users were mostly regular and contracted workers also. There are a few reported histories of drug treatment for these categories of workers in the government system. All information gathered during this study is available as Underlying data ³² .

Among 408 illicit drug current users, 350 currently used methamphetamine. The rest used kratom, cannabis, inhalants, opiates/opioid, ecstasy, ketamine, and cocaine.

Users used methamphetamine 1–3 times daily (1–6 pills each time), around 30 minutes before work for first intake and every 5–6 hours booster. Mostly, users smoked methamphetamine (put 1 pill on foil then used lighter to burn and applied straw to take it). A few put 1–2 pills into drinking water. They recognized the effect of the drug (energetic, painless etc.) 1–2 minutes after smoking, or around 10 minutes after drinking. There are few reports of drug treatment for these categories of workers in the government system.

Concerning both the ID use ( Table 2) and methamphetamine use ( Table 3), the treatment rate showed a statistically significant association with decreased substance use. Regarding ID estimation, only ID use ( Table 2) was significantly associated with increased substance use; however, the effect was extremely small.

The respondents who were single, had low incomes, and currently used non-medical pharmaceutical drugs were more likely to use IDs and methamphetamine ( Table 2 and Table 3).

The mOR in all models was >1, which indicated that the between-cluster variation in substance use was greater than the within cluster-level variation; all of the IOR-80% intervals contained 1, which confirmed this finding further ( Table 2 and Table 3).

Table 2 presents the results of the multilevel binary logistic regression models for ID use. In model 1, the analysis of the association of the individual-level variables with substance use showed that a higher likelihood of ID use was associated with males, older age, and non-prescription drug use. In model 2, drug use among peers was not associated. In the final model (model 3), 2 cluster-level variables (i.e., treatment rate and ID estimation variables) were added to the model, and each 1-unit score increase in treatment rate was associated with a significant 12-fold increase in the likelihood of ID use.

We hypothesized that methamphetamines were used to accelerate the productivity of agriculture-related workers. A series of similar models was used to estimate the odds ratio of methamphetamine use, as shown in Table 3. In models 1 and 2 of this category, older respondents were more likely to use methamphetamines than were younger respondents. Males and non-prescription drug use were significantly associated with an increase in the likelihood of methamphetamine use. Finally, in model 3, the results showed similar associations between individual and interpersonal-level variables with substance use, as in model 2. In the final model (model 3), 2 cluster-level variables, i.e., treatment rate and ID estimation variables, were again added to the model, and each 1-unit increase in treatment rate was associated with a significant 15-fold increase in the likelihood of methamphetamine use.

This study used data from a cross-sectional survey; therefore, it cannot establish a temporal relationship or causality. However, the study strengths are the use of treatment rate as a cluster-level factor (enables an objective assessment), providing more reliability than that of previous subjective assessments ^{38, 39} . Another study strength is it controlled for a wide range of covariates and used a large sample size and a nationally representative sample, meaning that the findings can be generally applied to the larger Thai population. Not only did this study provide further insights into cluster-level affects on substance use, but also describes associated factors adjusted for cluster environment among agricultural at national scale. However, prospective studies to further investigate may be required.

This study was successful in collecting data with a high response rate. The findings of the present study can also be applied to other similar contexts. The authors suggest further qualitative study on the beliefs of among agricultural regarding drug use for working. However, patterns of drug use may be better measured using open-ended questions, where among agricultural are asked to provide information freely without being influenced.

Open Science Framework: Drug use among agriculture-related workers in Thailand.

https://doi.org/10.17605/OSF.IO/DWV9G ³² .

This project contains the following underlying data:

Open Science Framework: Drug use among agriculture-related workers in Thailand. https://doi.org/10.17605/OSF.IO/DWV9G ³² .

This project contains the following extended data:

Data are available under the terms of the Creative Commons Attribution CC 1.0 International license (CC0 1.0 Universal).