Reliability of the qualitative behavior assessment as included in the Welfare Quality Assessment protocol for growing pigs

Positive emotions constitute a very important part of animal welfare. They are, however, also the most challenging elements to be objectively measured. Due to its feasibility, the qualitative behavior assessment (QBA) is included in the Welfare Quality Assessment protocol for growing pigs as the animal-based measurement tool for positive emotions. Reliability testing on the QBA in the form as included in the protocols is, however, rare. Therefore, the present study aimed at the evaluation of the inter-and intraobserver as well as test-retest reliability of the QBA in growing pigs. This was done by trained observers based on 19 joint on-farm assessments, the repeated assessments of 24 farms during 2 growing periods, and 107 video sequences. The results were compared between the observers and the repeated farm visits. Therefore, millimeter values were directly compared by calculation of Spearman's rank correlation coefficients (RS), and furthermore, the results were subjected to a principal component analysis (PCA). The results identified 2 main principal components (PC; PC1 and PC2) together explaining from 42 to 75% of the variation in the recorded variables of the different PCA. The factor loadings that the adjectives reached on PC1 and PC2 were compared by calculation of RS between observers and farm visits, respectively. Reliability was interpreted as acceptable if at least a moderate correlation was detected; that is, RS was greater than or equal to 0.4. Regarding the on-farm assessments, and, therefore, under practical conditions, no sufficient interobserver reliability (RS = -0.16 for PC1 and RS = 0.13 for PC2) was found. In terms of the test-retest reliability, only 1 comparison of 2 farm visits showed a positive correlation for PC1 (RS = 0.79) as well as for PC2 (RS = 0.64). The other 5 comparisons presented negative to weak positive correlations. However, based on video sequences, good interobserver (RS = 0.67 for PC1 and RS = 0.60 for PC2) and intraobserver (RS = 0.94 for PC1 and RS = 0.44 for PC2) reliability was achieved. Therefore, the present study revealed good reliability for the QBA in the form as it is currently included in the Welfare Quality Assessment protocol for growing pigs based on video sequences but insufficient reliability for the application on the farm.