2.3. Data analysis
2.3.1. Prediction of vaccination compliance against LSD in 2019 by previous vaccination:
We examined the influence of the following factors on LSD vaccination in 2019 among 566 farmers: previous vaccination (divided into four categories: 1. No previous vaccination. 2. Vaccination only in 2017. 3. Vaccination only in 2018. 4. Vaccination in both 2017 and 2018.), farm type (family vs. cooperative vs. school), geographic area (North, Centre and South of Israel) and previous occurrence of the disease in the herd (yes/no). Univariable analysis was performed by using the Chi-square test. Multivariable analysis was performed by fitting a logistic regression model to the 2019 vaccination data. Variables were included in the model in a forward stepwise process with a p-value of 0.05 in the univariable analysis as a cut-off for inclusion and a p-value> 0.05 in the multivariable analysis as a cut-off value for variable exclusion.
2.3.2. Analysis of attitude, subjective norms, and perceived behavioural control among farmers:
We used the intersection data (Figure 1 ) for determining the association of the three constructs with actual vaccination among farmers who showed high intention to vaccinate. Thirty eight of the 56 dairy farmers intended to vaccinate their herd (scores 4-5) (Figure 1 ). Vaccination records for the period preceding questionnaire filling were available for half of these 38 (future behaviour). For the other half, we had data on recent vaccination behaviour (up to one year prior to questionnaire filling). Within this group, we compared the three constructs’ direct and indirect measurements between those who did and did not vaccinate.
For determining the association of the three constructs with intention among the 90 farmers (Figure 1, Dataset 2 ), we compared A, N and PBC between farmers with negative and positive intention to vaccinate. We performed univariable analysis for each construct’s direct and indirect measurement using T tests. The intention was analysed as a dichotomic variable, scores 1-2-3 were marked as a negative intention and scores 4-5 as a positive intention. The same was performed for each belief. Similarly, we determined the direct effect of the background variables (farm and behavioural) on the intention to vaccinate. In that univariable analysis we used Fisher/Chi-square tests.
2.3.3. Effect of distance in time from LSD epidemic on the intention to vaccinate and the three constructs’ direct and indirect measurements:
Ideally, we would have distributed the questionnaires each year and followed the change in these factors. However, the questionnaire study began only in 2018, and data on vaccination intention in preceding years was not available. In 2019, during the questionnaire study, another LSD epidemic took place in Israel. This epidemic enabled us to compare questionnaires filled in by farmers before the epidemic and after the epidemic. Fifty-seven questionnaires were answered before this epidemic while 33 questionnaires were answered during and after its occurrence (Figure 1 ). These 33 questionnaires represent a status resembling the situation after the epidemic of 2012-2013, when vaccination became voluntary. The questionnaires which were answered before the epidemic in 2019 (the ”before” group) represent the situation long after the epidemic in 2013 (Figure 2 ). The comparison between the before and during/after the epidemic in 2019, therefore, represent different time distances from an LSD epidemic. Comparing the intention to vaccinate and the three constructs’ direct and indirect measurements between the two groups was performed by using t tests after ensuring that the data were normally distributed.
Intention to vaccinate was modelled using the data of 90 RAA questionnaires. A general linear model of intetion to vaccinate was fitted to the farmers’ beliefs, the location of the herd and the time from the last epidemic (”before” or ”after”) as the explanatory variables. The model was fitted using a stepwize process with a p-value of 0.05 in the multivariable analysis as a cut-off value for variable exclusion.
All statistical analysis was performed using R version 3.6.0 (R Core Team, 2019), the ”dplyr” , ”ltm” , ”Hmisc” ,”olsrr” and ”lme4” packages (Bates et al., 2015; Harrell Jr et al., 2021; Hebbali, 2020; Rizopoulos, 2006 & Wickham et al., 2020).