Reliability and validity of a questionnaire survey in canine anxiety research

Canine behavior can be assessed through behavioral testing or questionnaire surveys. Behavioral tests, often observed by independent experts, can provide objective data, but capture only a short behavioral sequence in a defined situation and are practical only for small study cohorts. Questionnaires can be targeted to much larger populations and provide long-term perspectives on behaviors, including potential correlated environmental factors. However, questionnaires are sensitive for misinterpretations and the subjectivity of the participant may affect the outcome. Therefore, particular attention has to be paid to structure and form of the questions to ensure that the questionnaire measures what it is supposed to measure. We have initiated a program on genetic studies in canine anxiety and have developed a questionnaire for behavioral phenotyping. The aim of the questionnaire is to identify the most fearful dogs (cases) and dogs not showing fear (controls) within and across breeds for gene mapping purposes. Knowing the challenges in the phenotyping of complex neurobehavioral traits, the aim of this study was to assess the external and convergent validity and re-test reliability of the developed questionnaire. External validity was assessed by comparing questionnaire data against short behavioral test of selected dogs (N=54), and convergent validity was estimated using known group comparisons, i.e. whether the dogs with separation anxiety have also more noise phobia. Re-test reliability was evaluated from the subsample (N=36) of the owners who filled the questionnaire twice within an 8-month period. We found an excellent correlation between questionnaire data and the behavioral test, and the most important questions on the fearful reactions toward strangers, had the highest correlation with the behavioral test in which the dog met a stranger (mean r(s) = 0.63). Convergent validity was also well demonstrated by showing that a known co-morbid anxiety trait, separation anxiety, was more common in noise sensitive dogs as expected (Wilcoxon rank-sum test chi(2)(1.1732) = 9.8, p < 0.01). Similarly, the test-retest reliability was good (mean values: 0.69 Spearman correlation; 0.59 Kappa value). Most importantly, the questions asking the frequency of fear reactions had excellent repeatability (mean frequency of fear reactions toward humans, dogs and situations 0.78 Spearman correlation, toward thunder, gunshot, fireworks, 0.76 Spearman correlation). These results encourage a conclusion that our questionnaire may serve as a reliable phenotyping tool for fearfulness in dogs, which, in turn, should help us to establish sufficiently large study cohorts for successful genetic analyses. (C) 2014 Elsevier B.V. All rights reserved.