Description of study area
This cross-sectional study was conducted in Khalkhal and Kosar counties, located in the southern part of Ardabil Province, northwestern Iran. The study area covers approximately 5,204 km², with geographical coordinates ranging from 37.6° N to 37.8° N and 48.3° E to 48.6° E. The study area features diverse climatic conditions, including warm, temperate, and cold regions at higher elevations. Agriculture and livestock farming are the region’s primary economic activities. Major crops include cucumber, tomato, sunflower seeds, rice, melon, and watermelon, while orchards produce apples, pomegranates, apricots, cherries, quinces, and figs. The target population of this study consisted of farmers engaged in cucumber cultivation.
Pesticides, including insecticides, herbicides, and fungicides, are widely used in the study area. The main reasons for their use are the high prevalence of pests and plant diseases, especially in cucumber and tomato farming, as well as the need to control weeds in agricultural fields18. Moreover, climate change and regional conditions have worsened the incidence of certain fungal diseases, leading to increased fungicide use. For many farmers, maintaining product quality and avoiding yield losses are vital economic goals that drive their greater reliance on chemical pesticides.
Sampling and data collection
Data were gathered through a structured questionnaire administered via face-to-face interviews conducted by trained interviewers. Before the interview, the study objectives were clearly explained to the farmers, and their informed consent was obtained. The target population included farmers involved in cucumber cultivation. A two-stage cluster sampling method was employed to select participants. In the first stage, a list of villages primarily engaged in cucumber farming was compiled from records of local agricultural offices. The interviewer visited these villages and compiled a list of farmers who had cultivated cucumbers in this area. During sampling, farmers who did not personally perform pesticide spraying or whose main income did not come from farming were excluded. In the second stage, all eligible farmers who agreed to participate were interviewed, resulting in a total of 352 participants.
This study was approved by the Ethics Committee of Khalkhal University of Medical Sciences (Approval No. IR.KHALUMS.REC.1402.015). Written informed consent was obtained from all participants prior to data collection. All methods were performed in accordance with the relevant guidelines and regulations.
Measures and statistical analysis
Data were collected using a structured interviewer-administered questionnaire11 The instrument’s validity and reliability were previously reported in the referenced study. In the present study, the internal consistency of the 41-item safety behavior scale was assessed using Cronbach’s alpha. The scale demonstrated acceptable reliability (Cronbach’s α = 0.79).
The questionnaire consisted of two main sections. The first section included demographic, managerial, health-related, and attitudinal variables. Health-related variables were based on self-reported responses. Participants were asked whether they had experienced health problems that they attributed to pesticide exposure, whether similar health problems had occurred among children or workers involved in farming activities, and whether access to medical care was available during pesticide poisoning incidents. These variables reflected respondents’ perceptions and experiences and were not based on medical diagnoses or clinical assessments. Demographic variables included age, education level, farming experience, household income, cultivated land area, and proportion of family labor. Managerial variables included household consumption of agricultural products and participation in training programs. Attitudinal variables included safety knowledge level and perceived barriers to safe pesticide use.
In the second section, comprising 41 items, we assessed farmers’ safety practices across four key dimensions: use of PPE (11 items), avoidance of health risks (12 items), hygiene practices after pesticide use (6 items), and appropriate pesticide use (12 items). Responses were recorded on a five-point Likert scale to evaluate perceived importance (1 = very low, 2 = low, 3 = moderate, 4 = high, and 5 = very high). Pesticide usage frequency was measured using a five-point Likert scale (1 = never, 2 = rarely, 3 = sometimes, 4 = often, and 5 = always).
To calculate the final safety behavior score, the total points from the 41 items were summed and normalized using the following formula (1) to yield a dimensionless index ranging from 0 to 1:
$$\:\text{Safety\:Level}=\frac{{\text{SB}}_{k}-{\text{SB}}_{\text{min}}}{{\text{SB}}_{\text{max}}-{\text{SB}}_{\text{min}}}\:\:$$
(1)
where SBk is the safety behavior score for the kth farmer; SBmin is the minimum score for safety behavior in the sample; and SBmax is the maximum score for safety behavior in the sample, enabling comparison of safety behavior across individuals. The safety behavior index was calculated exclusively from the reported frequency of actual safety practices (current use scores). Perceived importance ratings were analyzed separately and were used only to compare perceived importance with actual practices through the Wilcoxon signed-rank test. Importance scores were not included in the construction of the safety behavior index.
Subsequently, based on the five-stage model proposed by Ko (2005), safety behavior was classified into five levels: safe behavior (excellent; 0.81–1.00), potentially safe behavior (good; 0.61–0.80), intermediate behavior (0.41–0.60), potentially unsafe behavior (poor; 0.21–0.40), and unsafe behavior (very poor; 0.00–0.20).
Data analysis was conducted using SPSS software, version 26. Descriptive statistics, including mean, standard deviation, frequency, and percentage, were used to summarize the farmers’ socio-demographic, managerial, and agricultural characteristics. The Kolmogorov–Smirnov test was applied to assess the normality of continuous variables.
Prior to regression analysis, binary variables were coded as 1 (No) and 2 (Yes) in accordance with the questionnaire coding scheme. Continuous variables were entered in their original form, and ordinal variables were entered using their coded category values as defined in the questionnaire. Because some participants did not respond to all questionnaire items, sample sizes varied across variables. Missing values were not imputed, and analyses were conducted using the available data for each variable.
Since most variables did not meet the normality assumption and were measured on ordinal scales, nonparametric statistical methods were used for inferential analysis. Differences between farmers’ perceived importance and actual practice of pesticide safety behaviors were examined using the Wilcoxon signed-rank test for paired data across several dimensions, including the use of PPE, post-spraying hygiene, and safe handling and application of pesticides. For the large sample size, the Wilcoxon test statistic was standardized using a normal approximation and reported as a Z value. Negative Z values indicate that perceived importance scores were significantly higher than the corresponding practice scores.
To identify the determinants of farmers’ safety behavior, a multiple linear regression model was developed. The model included a wide range of predictors, such as personal attributes (age, educational level, farming experience, and income), managerial characteristics (farm size, proportion of family labor, and frequency of pesticide application), and attitudinal and educational factors (participation in training programs, safety knowledge level, and perceived barriers to safe pesticide use). These variables were selected to account for background characteristics, exposure-related conditions, and key components of the knowledge–attitude–practice (KAP) framework that are theoretically and empirically associated with safety behaviors in pesticide use.
The variable “source of guidance for pesticide use” consisted of nominal categories representing respondents’ primary source of information. For analytical purposes, these categories were coded numerically according to the questionnaire coding scheme and entered into the regression model. Therefore, the associated regression coefficient should be interpreted with caution as reflecting differences among coded categories rather than a true ordinal relationship. A p-value < 0.05 was considered statistically significant for all analyses.