Validity and reliability of animal-based measures of welfare protocols in finisher pigs

Animal-based measures are a key part of welfare protocols, but to what extent these measures can report differences in welfare among pigs provided e.g. with different accesses to valued resources (i.e., whether the measures are valid) and whether a single measurement point is representative of the entire production period (i. e., whether the measures are reliable) is not well documented. In this study, a total of 814 finishing pigs were assigned to either 1 of 4 welfare improving treatments (increased space: 1.4 [S9] or 2.1 m(2) [S6] per pig, provision of roughage [R], extra provision of enrichments [E]) or to a control treatment (C). Animal-based measures of a refined version of the Welfare Quality (R) protocol were assessed weekly (Batch 1) or once during the period (Batch 2). All parameters showed poor reliability throughout the 10 weekly assessments of Batch 1. Among the clinical parameters, only few indicators showed differences between welfare improving treatments. Among the behavioural indicators, differences were recorded for manipulation of the pen (C > S6) and of enrichment materials (E > all other treatments). When aggregating indicators into domains (nutrition, health, environment, behaviour), the reliability did not improve, and treatment effects were recorded within the domain of nutrition (R > all other treatments), and behaviour (E > S9). When aggregating domains into a final welfare score, both reliability and validity, as defined for the study, were poor. These results call for attention over the interpretation of the animal-based measures obtained from a single welfare assessment.