Short communication: Evaluation of the sample size of individual indicators in gestating sows concerning the Welfare Quality® protocol applied to sows and piglets
Sample sizes of welfare assessment protocols must warrant to reflect prevalences on-farm properly - regardless of farm size. Still, solely a fixed sample size was specified for the Welfare Quality® protocol for sows and piglets. The present study investigated whether animals may be assessed from only one body side as applied in the protocol and whether the pre-set sample size of 30 animals mirrors the prevalences of the animal-based indicators on-farm in the gestation unit considering different farm sizes. All indicators were assessed for both sides of an animal's body by one observer on 13 farms in Germany, which were visited five times within 10 months. The farm visits were treated as independent since different animals were housed in the gestation units. The number of sows in the gestation units varied between 18 and 549 animals. The comparison of sides was carried out calculating exact agreement between animals' sides and a Wilcoxon signed-rank test (W). The results signified that it is sufficient to assess the animal from one side (exact agreement: 88.3% to 99.5%, except for bursitis (70.0%); W: P-values 0.14 to 0.92). However, if side preferences existed in the indicator bursitis a potential bias must be considered. In the following, the sample size was evaluated by comparing samples' prevalences against true prevalence, that is, the prevalence of all observed animals in the gestation unit in each farm visit. Therefore, subsets of data were generated by applying simple random sampling without replacement. The samples randomly included the animals' right or left sides. Linear regression was rated as appropriate provided: coefficient of determination R2 ≥ 0.90, slope = 1 and intercept = 0 signifying exact agreement. The results revealed that the sample size required by the protocol and the application of calculation formulas are solely appropriate to mirror the prevalences of frequent indicators in the gestation unit, for example, bursitis (mean prevalence 34.4%). Using a proportion of animals, for example, a sample of 30% of all observed animals in a farm visit, pointed out that proportions must increase with indicators' underlying prevalence narrowing 0.00%. Local infections (mean prevalence 13.3%) needed samples including 60% of all observed animals in each farm visit, whereas vulva lesions (mean prevalence 7.28%) only reached accuracy with the inclusion of 70% of the animals. Indicators with a mean prevalence of <1% were not analysed but can most likely only be ascertained by the assessment of all animals.