Different assessment systems fail to agree on the evaluation of dairy cattle welfare at farm level

As different animal welfare assessment schemes are based on different assumptions but all have similar welfare claims, we studied the degree of agreement among three systems, two mainly based on resource-based measures: Animal Needs Index (ANI) and Lombardy Extension Service Index (IBS) and one mainly based on animal basedmeasures (i.e. Welfare Quality (R)). Thirty-three dairy cattle farms were assessed by one trained assessor. In order to make the welfare categories of the different welfare assessment systems comparable with one another the first three welfare categories of the ANI scheme were merged into a not acceptable category so that a score scale with 4 welfare categories was obtained. Then, we calculated the Cohen's weighted kappa (k) to assess the degree of agreement between the three welfare assessment systems in terms of welfare categories. In addition, the correlation between pair of systems in terms of total scores (the mean of the four principles of the Welfare Quality (R) was used for this calculation) was calculated using the Spearman correlation coefficient (r(s)). Our results showed that only few farms were scored in the same welfare category of different welfare assessment systems as only 12.1% of the farms were scored as enhanced and 3% of the farms were scored as acceptable according to all three welfare assessment systems. Based on k statistics, the degree of agreement in terms of welfare categories was not significant for the pairs IBS - ANI (k = 0.128; P = 0.130) and WQ - ANI (k = 0.014; P = 0.850), whereas it tended to be significant for the pair WQ - IBS (k = 0.256; P = 0.070). The Spearman correlation coefficient of the total scores confirmed the low agreement among the three systems. In particular, the correlation coefficients between ANI-IBS, IBS-WQ and ANI-WQ were r(s) = 0.022, 0.208 and - 0,068, respectively, and none of them was significant (P = 0.905, P = 0.246 and P = 0.707, respectively). Significant correlations were only observed between the Welfare Quality (R) principle "Good health" and the somatic cell counts (r(s) = - 0.500; P = 0.003), the number of lactating animals (r(s) = - 0.360; P = 0.040), the total number of cows per herd (r(s) = - 0.339; P = 0.053), whereas the correlation of the same principle with mortality (r(s) = - 0.315; P = 0.074) and calving interval (r(s) = - 0.330; P = 0.070) tended to be significant. We conclude that different schemes based on different assessment criteria yield different results albeit making similar welfare claims.