2.1 Making preliminary questionnaire

1. The literatures on chemical or pesticides prevention behaviors, farmers’ working characteristics, risk of chemical hazards and social cognitive theory were reviewed to examine the contents and psychometric properties of existing measures to prevent chemical hazard. We developed the original questionnaire using 63 items consisting four scales; knowledge, self-efficacy, outcome expectations and behaviors to prevent chemical hazard of pesticide. Each scale was made to represent a unidimensional construct derived from Bandura [2].
2. The panel of five experts in the area of occupational health and scale development reviewed the original questionnaire sheet. The panel of five experts was asked to rate each item on how strong it was relevant to the underlying scale. Based on Waltz and Bausell [6] advise, a 4-point ordinal scale was used to avoid a neutral and ambivalent midpoint. They were: 1= not relevant, 2 = somewhat relevant, 3 = quite relevant, 4 = highly relevant. Then, content validity index (CVI) was computed as proportion of items of which rating score was either 3 or 4 among total items of each underlying scale. A minimum of CVI was recommended to be 0.80 as an acceptable standard [7]. When any experts rated 1 for a item, the item was excluded from the questionnaire. When any experts rated 2 for a item, its expression was revised to become more suitable for the main concept of the underlying scale or to understand easier.
3. A focus group was made with 10 rice farmers to verify the expression and content of each item of the original questionnaire sheet reviewed by the panel of five experts. The subjects were recruited from Thai rice farmer who seemed to have same characteristics as the subjects of preliminary examination.
4. After revised by a focus group discussion, the preliminary questionnaire sheet was reviewed by the panel of five experts again. Rating each item was performed using method same as rating at the first review.

2.2 Evaluating questionnaire items to make scales

1. Three-hundred-fifty-three Thai rice farmers were recruited to complete the questionnaire including demographic data and items possibly relevant to knowledge, self-efficacy, outcome expectation and behaviors to prevent chemical hazard of pesticide. The one of researchers explained the aims and methods of the research, protection of the personal information, that participation was voluntary, that there was no disadvantage in not participating and absolute confidentiality of the individual both in oral and written forms. The participants who agreed with our survey gave the informed consent.
2. This survey was approved by Ethical Committee for Human Research of the Faculty of Public Health, Mahidol University (No. MUPH 2014-149 and Protocol No. 86/2557).
3. The Standards for Educational and Psychological Testing Guide [8] was used in this study to assess the validity and reliability of the scales.
4. Good-Poor (G-P) analysis was used to evaluate discriminative power of items. The participants were divided into four groups using interquartile range: highest, higher, lower, and lowest groups. Average score in each item was compared between the highest group and lowest group. The items showing no significant difference between them was excluded.
5. Item-Total (I-T) analysis was used to evaluate the internal consistency. The correlation coefficient (Spearman Rank correlation test) between the item and total score without subjective item was calculated. The item of which correlation coefficient was less than 0.4 was excluded.

6. Factor analysis was used using maximum-likelihood and Promax rotation to evaluate structural validity of the four scales. The item of which standardized factor loading was 0.4 and more was selected.

   To evaluate the discriminative validity, the subjects were divided into two groups: those who used PPE and those who did not from the demographic data. The mean values of total scores of items on knowledge, self-efficacy, outcome expectation and behaviors to prevent the chemical hazards of pesticide were compared between the two groups.

7. The international consistency of the final version of items was evaluated using Cronbach’s alpha coefficient.

8. The steps of developing questionnaire items is shown in Figure 1.

   

   Figure 1: Step of questionnaire development.

3.1 Making preliminary questionnaire

1. Original questionnaire contained 63 items: knowledge 25 items, self-efficacy 10 items, outcome expectation 10 items and behaviors to prevent the chemical hazards of pesticide 18 items.
2. The CVI at the first review of the panel of five experts was shown in Table 1. The CVI of knowledge was smallest but exceeded an acceptable standard (0.80). Corresponding to the protocol, the researchers excluded 4 items of 25 ones on knowledge.
   

   Table 1: Content Validity Index (CVI) of the knowledge, self-efficacy, outcome expectation and behavior on pesticide exposure prevention.
3. Following suggestions from the focus group, the researchers excluded four items on knowledge and modified some items to easier understand.
4. At the second review of the panel of five experts, the rating score showed that 5 items of knowledge were rated less than 3. Thus, the researchers excluded such items. Among the items on self-efficacy, outcome expectation and behaviors to prevent the chemical hazards of pesticide, all items were rated 3 or 4. All items were used for next analysis.
5. After these procedures, the preliminary questionnaire was consisted of following items in each scale.
   1. Knowledge consisted of 12 items including: characteristics of chemical substances (1 item), entry route to the body (4 items), chemical toxicities (2 items), chemical intoxication prevention (2 items) and fitting test for PPE use (3 items). The answers had 2 choices; correct and wrong.
   2. Self-efficacy contained 10 items, including the questions measuring farmers’ perception of their own ability to find and to use appropriate personal protective equipment (PPE) for pesticide exposure prevention. The answer had Likert 3 scales: disagree, neither agree nor disagree and agree.
   3. Outcome expectation consisted of 10 items, including the questions of expectation of good outcome for health if they do some good behaviors for pesticide exposure prevention. The answer had Likert 3 scales: disagree, neither agree nor disagree and agree.

   4. Behavior to prevent chemical hazard of pesticide consisted of 16 items including: reading the chemical instruction (1 item), following the chemical instruction (1 item), checking the equipment for chemical substances (1 item), wearing thick plastic hand gloves, chemical mask, long sleeve shirt/trousers, long boots, goggles against chemical substances (8 items), doing the fitting test (1 item), washing hands (1 item), taking a shower (1 item), washing the clothes separately from ones used for chemical substances (1 item), no entry to the area where chemical substances were used (1 item), storage chemical substances (1 item) and checking the finger chemical blood stripe for cholinesterase level after continuing use chemicals for 7 days. The answers had 3 Likert scales; never, sometimes, and always.

      Table 2 shows the characteristics of participants. Two-third of participants are males (66.6%). More than a half of them were at forties and fifties (58.6%) and an average age was 46.0 (Standard deviation: SD 11.4). A half of them had educational level at primary school (46.7%). Most of them were married (89.8%). More than a half of them had not enough family income (67.4%). Most of them use PPE (89.5%) when dealing with chemical substances. They were exposed to chemical substances by mixing (28.3%), spraying (51.8%) and both mixing and spraying (19.9%).

      

      Table 2: Characteristics of the participants (n = 353).

3.2 Good-poor (G-P) analysis

There were significant differences in the average between in the highest group and lowest group in all items. Thus, all items; knowledge 12 items, self-efficacy 10 items, outcome expectation 10 items and behavior 18 items were remained. We used all items for the next analysis procedure.

3.3 Item-Total (I-T) analysis

Item-total analysis showed that the knowledge 12 items and selfefficacy 10 items had the correlation coefficient exceeding 0.40. However, 2 items of the outcome expectation and 4 items of the behavior had the correlation coefficient less than 0.40. These items were excluded. Therefore, the items of knowledge 12 items, selfefficacy 10 items, outcome expectation 8 items and behavior 14 items, 44 items, were used for the next analysis procedure.

3.4 Structural validity

Table 3 presents the results of factor analysis using 44 items. There were 4 components. The first component was comprised of 10 items (No. 13-22). They were the items on self-efficacy of which standardized factor loading value was 0.4 and more in the first component. The scale on self-efficacy was then made using 10 items included in the first component.

Table 3: Result of factor analysis of knowledge, self-efficacy, outcome expectation and behavior on pesticide exposure prevention (n = 353).

The second component was comprised of 14 items (No. 31-44). There were the items on behavior of which standardized factor loading was 0.4 and more. The scale on behavior was then made using 14 items included in the second component.

The third component was comprised of 12 items (No. 1-12). They were the items on knowledge. Because two items had the standardized factor loading value of less than 0.4, the two items were excluded. The scale on knowledge was made using 10 included in the third component.

The fourth component was comprised of 8 items (No. 23-30). There were the items on outcome expectation. Because two items had the standardized factor loading value of less than 0.4, the two items were excluded. The scale on outcome expectation was made using 6 items included in the fourth component.

In conclusion, all scales; knowledge, self-efficacy, outcome expectation and behavior to prevent chemical hazard of pesticide were 40 items.

3.5 Discriminative validity

Table 4 shows the mean scores of items on knowledge, self-efficacy, outcome expectation and behavior between those who used PPE and those who did not. The mean score of items on knowledge, selfefficacy and outcome expectation were higher in those who used PPE than in those who did not.

Table 4: Mean scores of items on knowledge, self-efficacy, outcome expectation and behavior of participants with those who used personal protective equipment (PPE) and those who did not (n = 353).

3.6 Internal consistency (Reliability)

Cronbach’s Alpha of 10 items on knowledge was 0.882. That of 10 items on self-efficacy was 0.965. That of 6 items on outcome expectation was 0.868. That of 14 items on behavior to prevent the chemical hazard of pesticide was 0.927. The values exceeded the acceptable standard (>0.80) (Table 5).

Table 5: Cronbach’s alpha of the knowledge, self-efficacy, outcome expectation and behavior on pesticide exposure prevention (n = 353).