Design and study population
The study was designed as a cross-sectional study. Three thousand sheep farmers were randomly selected from the Register of Production Subsidies using a pseudo random number procedure [27]. To exclude hobby farmers, only sheep farmers who had ten or more breeding sheep were considered for selection. Altogether 18,404 sheep farmers were eligible.
Data collection
The data were collected by an anonymous questionnaire survey. The four-page questionnaire contained questions concerning location and flock characteristics, contact patterns and management, and the farmer's knowledge of scrapie and his attitude towards the Norwegian scrapie surveillance and control programme. The questionnaire (in Norwegian) is available from the corresponding author.
The questionnaire was accompanied by a covering letter and a free-post return envelope. It was mailed to all participants in March 2002 followed by a reminder card in April 2002. The data from the questionnaires were entered into an MS access database (Microsoft® Access 2000 Version 9. Microsoft Corporation, WA, USA, 1992–2001).
The farmers' willingness to report scrapie
The farmers were asked how they would act if they discovered an animal showing signs suggestive of scrapie. Those answering "Report to DVO" or "Submit the material to a diagnostic laboratory" were considered to be "Willing to report scrapie suspects", while those only marking "Send to slaughter" or "Cull the animal" were considered not willing to report these cases. The answer "Don't know" were treated as missing.
Furthermore, the farmers were asked for their reporting behaviour in the situation where an animal had shown clinical signs of listeriosis and did not recover after treatment (hereafter "non-recovering listeriosis cases"). The answers "Contact DVO for examination regarding scrapie" (in short "Notifying") and "Contact veterinary practitioner for further examination" (in short "Re-examine") were kept as separate categories, while the answers "Send to slaughter" or "Kill the animal" were grouped together in the category "Not report" giving three different categories in total. The answer "Don't know" was treated as missing.
Explanatory variables
Demographic information
The farms were located in four geographical regions: South-Eastern, Western, Middle and Northern Norway, based on county. The farmers were asked to specify the flock size within one of four groups: 1–10, 10–49, 50–99 or ≥ 100 breeding sheep.
The farmers' knowledge of scrapie signs
Using check-boxes, the farmers were asked to mark the clinical signs they considered as associated with scrapie among ten different clinical signs, where emaciation, hair loss, lip smacking, pruritus and trembling were considered as scrapie-associated signs, and abortion, coughing, diarrhoea, fever, and frequent urinating were considered as not associated with scrapie. The question was considered answered when at least one of the check-boxes was marked. The information was collated into one variable by summing the number of correctly checked scrapie-associated signs. In the multivariate analysis, the variable was treated as a numerical variable.
The farmers' anticipated reactions if scrapie should be detected
The information that the detection of a scrapie-positive animal would lead to the destruction of all the sheep in the flock was presented in the questionnaire. The farmers were then asked to give their anticipated reaction to seven concerns regarding the potential detection of scrapie in their flock (Table 1). The reactions were graded in four levels: not important, of minor importance, important, and very important. For the multivariate analysis, each variable was dichotomised to derive two evenly-sized groups (Table 1).
The farmers' opinion on factors potentially important for reporting behaviour
The farmers were asked about the importance of four factors which might affect whether they would report an animal with signs suggestive of scrapie. These were: "I need more knowledge of scrapie symptoms", "Having easy access to a DVO", "Being offered free examination of scrapie suspects", and "The Government compensates for the cost of the control measures when scrapie is detected". The answers were graded in four levels: not important, of minor importance, important, and very important. For the multivariate analysis, each variable was dichotomised to derive two evenly-sized groups (Table 1).
Statistical analyses
The farmers' reporting of non-recovering listeriosis was used as a measurement of the farmers' vigilance in reporting scrapie. The reporting behaviour of non-recovering listeriosis was analysed in a multinomial logistic regression with the farmer as the statistical unit and the response variable categorized in the three nominal categories: Notifying, Re-examine and Not report.
The explanatory variables region, flock size, knowledge of scrapie-associated signs, each of the seven anticipated concerns of the farmer if scrapie were to be detected in their flock, and each of the four factors of potential importance for the reporting of scrapie suspects were considered as candidates for the multivariate analysis. All candidate variables were included in the initial model. Maximum likelihood estimation was used for model fitting, and the likelihood ratio test was used to assess the overall significance of the model. The best model fit was found by using backward stepwise deletion of insignificant terms. For each step the single least explanatory variable was removed until there were no significant difference between the full and all possible reduced models when using the likelihood ratio test with significance level of 0.05. Thereafter, all two-way interaction terms between the remaining explanatory variables were tested one by one for significance in the final multivariate model.
The adjusted RRR (corresponding to the adjusted Odds ratio estimated in binomial logistic regression) were used as measures of association between the response variable and the explanatory variable [28]. For each explanatory variable a separate estimate of the RRR was given for the responses "Notifying" and "Re-examine" relative to the category "Not report" (reference group). For completeness, the RRR for "Notifying" relative to "Re-examine" was calculated from the two other contrasts.
All variable processing and statistical analyses were performed in SAS-PC System® for Windows (SAS Institute Inc., Cary, NC, USA). The descriptive analyses were conducted by using PROC FREQ and missing observations were included when calculating the population percentages. The multinomial logistic regression was conducted by using PROC CATMOD.