According to the TPB (Figure 1), behavioral intention is the best prediction of future behavior (i.e., to perform or not to perform a certain behavior) [(14), p. 179–180]. The theory proposes that a behavioral intention is determined by three psychological factors, namely, AT, SN, and PBC, as shown in Figure 1 [(14), p. 179–180].

When applying the TPB to the context of changing farmers' behavior regarding HPAI control, AT can be defined as a farmer's beliefs about the outcomes of performing measures against HPAI [i.e., behavioral beliefs (BBs)] weighted by their evaluation of these outcomes [i.e., outcome evaluation (OE)]. For example, farmers who strongly believe and highly value the outcomes of vaccination are expected to have a more positive AT toward vaccination.

Similarly, SN can be defined as a farmer's beliefs on social pressure or other people's opinions about the implementation of HPAI prevention and control measures on their farms (i.e., normative beliefs) weighted by their motivation to comply (MC) with these pressures or opinions (i.e., MC). Influential opinions could originate from technical support (TS) and veterinarians of the integrated company, government veterinarians, TS of animal medicine companies, buyers or customers, broiler farmer peers, family members, neighbors or friends, people who live nearby, and role-model farmers.

Finally, PBC can be defined as a farmer's beliefs about factors such as money, time, and skills required to implement HPAI measures [i.e., control beliefs (CBs)], weighted by their confidence in the power of each control factor to facilitate or inhibit the decision to implement the measures [i.e., perceived power of control (PPC)].

This article studies the intention of small-scale commercial broiler farmers to implement any of the following preventive, monitoring, and control measures against HPAI: [1] improved biosecurity [i.e., routine cleaning and disinfection (C&D) of the farm area or barn], [2] vaccination, [3] reporting, and [4] stamping-out.

Biosecurity is defined as a set of isolation and sanitation measures with the aim of preventing the introduction as well as the spreading of diseases on the farm. Biosecurity measures will reduce the general risk of avian disease introduction, including HPAI. Ensuring or maintaining the sanitation of the farm, barn, and equipment is recognized as one of the appropriate and practical biosecurity measures on poultry farms (18, 19). In this study, sanitation measures, defined as routine C&D of the farm area or barn for every 2 days, are used as a proxy for biosecurity measures. The term biosecurity was not used in interviews with farmers because the term is not well-recognized and is interpreted differently among farmers. Avian influenza vaccination is defined as the implementation of AI vaccination of 7-day-old chickens via subcutaneous injection. This definition was tailored to the Western Java context in which having chickens slaughtered in multiple batches in every rearing cycle is a common practice. Farmers usually start selling their chickens when they are 21–25 days old, and AI vaccines require 14 days to provide sufficient protection from HPAI [(20), 146, p. 18,145–18,146].

Reporting is defined as declaring the observation of HPAI symptoms in one or more chickens to the authority or TS. Surveillance is based on participatory disease surveillance and response due to the lack of surveillance capacity of the veterinary and laboratory services [(21), p. 750]. In theory, reporting is suggested to be an effective early detection tool [(22), p. 435].

Stamping-out is an effective way to eradicate the virus at the source and to prevent further spreading. We included stamping-out in this study because it has been suggested to be effective and efficient in regions with low HPAI endemicity, although the program has been terminated since 2007 [(21), p. 752]. In this study, two scenarios of stamping-out were used, namely, with and without 50% compensation for culled healthy chickens. As such, the approach is adapted to the conditions of limited budgets for HPAI mitigation in Indonesia.

These four measures can be further categorized into two groups based on their scope. Biosecurity and vaccination can be seen as measures that are more in line with the prime interest of farmers for their farm, whereas reporting and stamping-out are in the interest of the sector at large (i.e., as not to be a liability in the farmer community), the government, and the public.

The questionnaire was developed based on the TPB framework and the list of HPAI measures explained above. The questionnaire contained two parts. The first part collected information about farmers' sociodemographic characteristics (e.g., age, education) and farm characteristics (e.g., chicken population). The second part collected information about intentions, AT, SN, and PBC regarding the four measures defined above.

The first part of the questionnaire used multiple-choice, open-ended and closed questions to collect information related to respondents' and farm characteristics, such as age, gender, education, poultry farming experience, chicken population per cycle, awareness of HPAI and its signs, and dependency level on broiler farming. In the second part of the questionnaire, Ajzen's TACT principle, which stands for target, action, context, and time, was used to define intentions. For example, “If HPAI (target) were to occur in the environment where my farm is located (context) within 1 year (time), I would vaccinate all my chickens once in every production cycle on the seventh day (action).” The target and context were used as follows: “If HPAI were to occur in the environment where my farm is located within 1 year” in the sections of BBs, OE, MC, and CBs to emphasize the hypothetical situation to respondents. No additional phrasing was used in the sections of normative beliefs and PPC because these sections are about the opinions of referents and farmers' resources in general.

A five-point Likert scale was used in the second part of the questionnaire. Respondents were asked to state the extent of their agreement/disagreement (1 = strongly disagree, 2 = disagree, 3 = neutral, 4 = agree, 5 = strongly agree) to all statements in the sections of intentions, BBs, MC, CBs, and PPC. Respondents were also asked to state the extent of the importance (1 = very unimportant, 2 = unimportant, 3 = neutral, 4 = important, 5 = very important) of all statements in the sections of OE and normative beliefs. The option “do not know” was added to the five-point Likert scale in the normative beliefs section.

Attitude was evaluated by asking respondents about their BBs toward the outcome of implementing each HPAI measure and their subjective evaluation of the importance of different outcomes (OEs). To estimate BBs, respondents were asked about their agreement related to possible outcomes of each measure (e.g., “cleaning and disinfecting my chicken barn every 2 days will reduce the risk of HPAI infection on my farm”). Respondents were also asked to give their subjective evaluation of the importance of outcomes (e.g., “reducing the risk of AI infection on my farm is…”). The following outcomes were included in this study: reduction of HPAI introduction on the farm, prevention of HPAI infection to other chickens in the flock, reducing mortality rate in the flock, prevention of the spread of HPAI to other poultry farms, prevention of HPAI transmission to humans (i.e., high-risk groups), and increase in the likelihood of the farm to be included in stamping-out.

Subjective norms were evaluated by asking respondents about the normative beliefs (NBs) of referents on the importance of implementing HPAI prevention and control measures on their farms (e.g., “according to your knowledge, what is the opinion of your technical support about prevention and control of HPAI?”). Next, respondents were asked about their own MC to the opinions of relevant referents (e.g., “do opinions of your technical support influence you to implement prevention and control of HPAI on your farm?”). If respondents indicated they did not know the opinion of a referent in section NBs, they would not be asked about their MC to the respective referent.

Perceived behavioral control was evaluated based on the resources available to farmers (time, money, and skills) to implement HPAI measures. To measure CBs, respondents were asked to indicate whether implementing the measures is time-consuming, expensive, or difficult (e.g., “implementing HPAI vaccination once in every cycle is expensive”). Then, respondents were asked about their PPC through statements that imply whether respondents perceive they have the necessary skills, spare time, and financial resources to implement measures (e.g., “I can afford to pay the costs for implementing HPAI vaccination once in every cycle”).

The questionnaire was first written in English, then translated to Bahasa Indonesia, and translated back to English for verification and publication. The questionnaire was tested in a pilot study. Ten small-scale broiler farmers were interviewed to check their understanding of the statements, as well as their ability to answer the questionnaire. Statements and terminologies that were difficult to understand by the test-farmers were modified.

The data were checked for errors and missing values. A descriptive analysis was carried out on farmers' and farm characteristics, the intention to implement the four HPAI measures, and product composites of TPB factors. Product composites were created to measure the three TPB factors (i.e., AT, SN, and PBC) with as little variables per TPB factor as possible. According to the TPB framework (Figure 1), the product composite of AT is the product of BB and the corresponding OE. The product composite of SN is the product of normative belief and the corresponding MC. The product composite of PBC is the product of CB and the corresponding PPC. Theoretically, the scores for all product composites range from 1 to 25.

For each of the measures, the internal consistency among the product composites of AT, SN, and PBC was evaluated through Cronbach α. If Cronbach α > 0.7, internal consistency among the product composite for that factor was regarded as significant (22, p. 574). For those factors, the product composites were averaged to derive a single direct measure (i.e., mean score) as shown in equations 1 to 3 [(27), p. 405]. When the product composites were inconsistent (Cronbach α < 0.7), a subset of product composites within the factor was used to find a combination of product composites that was internally consist. The other product composites were used as separate TPB factors.

where i = the TPB factor i; j = the product composite item j; n = the number of items for AT, SN, or PBC.

The data of all the intention variables and TPB factors turned out to be skewed, and thus, logistic regression models were applied to explain the association between the factors and intention. For the logistic regression analyses, intention toward each of the measures against HPAI was divided into two categories based on the Likert scores given for each intention to identify respondents with low or high-level intentions. The responses “strongly disagree,” “disagree,” and “neutral” were considered to indicate a low-level intention to implement the measure, and the responses “agree” and “strongly agree” were considered to indicate a high-level intention to implement the measure [(28), p. 4,633–4,634; (15), p. 141–142]. The TPB factors were categorized into three levels, based on the distribution of product composites score, in order to identify respondents who had weak, moderate, or strong AT, SN, and PBC. Theory of Planned Behavior factors (i.e., AT, SN, and PBC) that score <12 were considered weak; TPB factors that score ≥ 12 but <16 were considered moderate, and those scoring ≥16 were considered strong. If the number of observations in a category was smaller than 15, that category was merged with the nearest other category, resulting in two categories, that is, weak and moderate or moderate and strong. For all the logistic regression analyses using the TPB framework, the categorized intention variables were used as dependent variables, and the categorized AT, SN, and PBC factors were used as independent variables (i.e., one model for each intention). First, a univariable analysis was carried out to check the association between farming scheme and intentions. In this univariable analysis, farming scheme was used as fixed effect, and intention as a dependent variable; however, the analysis did not indicate any associations. Thus, farming scheme was excluded as independent variable in the logistic regression models.

For each of the four control measures, the following univariable and multivariable analyses were carried out. Before conducting multivariable logistic regression, univariable analysis was carried out to examine the association of each TPB factor with intention separately. Theory of Planned Behavior factors with a p < 0.25 were included in the multivariable logistic regression (29). Before conducting multivariable analyses, the presence of multicollinearity between TPB factors was checked using Spearman rank correlation coefficients (ρ). The multicollinearity check did not indicate high levels of correlation between TPB factors; all the correlation coefficients (ρ) were <0.8 [(30), p. 224]. Thus, all TPB factors were included in the multivariable analyses.

To evaluate the association of farmers' and farm characteristics with TPB factors, all the background information was included as binary variables in the multivariable logistic regression models. Farmers' and farm characteristics were used as independent variables, whereas TPB factors were used as dependent variables. Independent variables with a p < 0.15 by χ² test, or by Fisher exact test when there were fewer than five observations in a cell, were included in the multivariable logistic regression analysis.

All multivariable logistic regression analyses were carried out using a backward stepwise procedure. Independent variables that were not significant (p > 0.05) were excluded from the models one-by-one at each step. All the statistical analyses were performed using SPSS version 25.0 (31).

Table 2 provides the descriptive statistics of the sociodemographic and informational background of small-scale broiler farmers interviewed in this study. Of 203 small-scale broiler farmers interviewed, 141 (70%) were makloon farmers, 58 (28%) were contract farmers, and four (2%) were independent farmers. On average, participating small-scale broiler farmers were 45 years old, had 10 years of broiler farming experience, and had 3,000 birds on their farm. More than 75% depended on broiler farming activities as their main source of income.

Table 3 shows the descriptive statistics of intentions to implement HPAI measures. A large proportion of small-scale broiler farmers had a high level of intention to implement C&D (89%), vaccination (80%), and reporting (88%), while a smaller proportion of broiler farmers had a high level of intention for joining stamping-out with (67%) and without (53%) and compensation.

Table 4 shows descriptive statistics and Cronbach α's of the categorized AT, SN, and PBC for each intention. Most broiler farmers (>80%) had a strong SN to the opinions of the referents to implement HPAI prevention and control on their farms. More than half of broiler farmers had a strong AT toward all intentions. In contrast, a relatively low proportion of broiler farmers had strong perceived control over their time and money investments toward C&D, vaccination, and stamping-out.

Table 5 shows the statistically significant TPB factors for both the univariable and multivariable models for all intentions. Tables with detailed results for all the measures are provided in Appendices 1–5. The Nagelkerke R² scores suggest that the intentions to implement those HPAI measures that are within the prime interest of broiler farmers (i.e., sanitation and vaccination) are explained better by the model, compared to the intentions for HPAI measures that are more important for public interest (i.e., reporting and stamping-out).

From our models, the AT factor was significantly (P < 0.05) and positively related to the intention of broiler farmers to implement routine C&D [odds ratio (OR), 211] and AI vaccination (OR, 20.4). The opinion of veterinarians of the integrated company (OR, 106.1) and the technical adviser from animal health companies (OR, 2.88) were significantly (P < 0.05) and positively associated to the intention of broiler farmers to implement AI vaccination and to join stamping-out without any compensation, respectively. The money and time factor were significantly (P < 0.05) and positively related to the intention of broiler farmers to implement routine C&D (money: OR, 14.5; time: OR, 9.8), AI vaccination (money and time: OR, 38.4), and to join stamping-out without any compensation (OR, 8.2). None of the TPB factors was significantly (P < 0.05) associated with the intention to join stamping-out with 50% compensation in either the univariable or multivariable model.

Seven sociodemographic farmer characteristics, namely, age, gender, education, poultry farming experience, chicken population per cycle, awareness of HPAI and its signs, and dependency level on broiler farming, were regressed to TPB factors that were statistically significant in the models. Of these seven characteristics, only the contribution of broiler farming to farmers' income was found to be significantly associated with a strong AT toward the intentions to implement C&D and vaccination (Table 6). Farmers with a household income derived for 75% or more from broiler farming were more likely to have a strong AT toward C&D (OR, 7.3) and vaccination (OR, 20.2).