The Welfare Quality ® protocols provide a multidimensional assessment of welfare, which is lengthy, and hence limited in terms of practicality. The aim of this study was to investigate potential 'iceberg indicators' which could reliably predict the overall classification as a means of reducing the length of time for an assessment and so increase the feasibility of the Welfare Quality ® protocol as a multidimensional assessment of welfare. Full Welfare Quality ® assessments were carried out on 92 dairy farms in England and Wales. The farms were all classified as Acceptable or Enhanced. Logistic regression models with cross validation were used to compare model fit for the overall classification on farms. 'Absence of prolonged thirst', on its own, was found to correctly classify farms 88% of the time. More generally, the inclusion of more measures in the models was not associated with greater predictive ability for the overall classification. Absence of prolonged thirst could thus, in theory, be considered to be an iceberg indicator for the Welfare Quality ® protocol, and could reduce the length of time for a farm assessment to 15 min. Previous work has shown that the parameters within the Welfare Quality ® protocol are important and relevant for welfare assessment. However, it is argued that the credibility of the published aggregation system is compromised by the finding that one resource measure (Absence of prolonged thirst) is a major driver for the overall classification. It is therefore suggested that the prominence of Absence of prolonged thirst in this role may be better understood as an unintended consequence of the published measure aggregation system rather than as reflecting a realistic iceberg indicator.

This paper presents an account of a Welfare Quality® assessment of 92 dairy farms carried out by seven experienced assessors. The aim was to evaluate the potential of the Welfare Quality® assessment protocol with respect to its uptake by UK farm assurance schemes. Data collection, and measure aggregation were performed according to the Welfare Quality® protocol for dairy cows. This study examined the data itself, by the testing of how hypothetical interventions might be reflected in changes in the aggregated scores, and also investigated human-related aspects, through inter-assessor standardisation sessions to evaluate reliability, and an assessor focus group to collect feedback. Overall, three main ‘challenges’ were identified. The first challenge related to the large amount of missing data. Unexpectedly, this was such that it was only possible to calculate an overall classification for 7% of farms. The second challenge concerned the way in which aggregated scores did not always reflect hypothetical interventions. The final challenge was inter-assessor reliability, where not all assessors were found to achieve acceptable levels of agreement on a number of outcome measures by the third training session. Suggestions for managing these challenges included, follow-up to assessor training, the use of multiple imputation methods to fill in missing data, and, where applicable, not aggregating the scores. The conclusion of the study was that the protocol provided useful information from which to make an informed selection of measures, but that the challenges, combined with the lengthy assessment time, were too great for its use as a certification tool.

Lameness in dairy cows is an important welfare issue. As part of a welfare assessment, herd level lameness prevalence can be estimated from scoring a sample of animals, where higher levels of accuracy are associated with larger sample sizes. As the financial cost is related to the number of cows sampled, smaller samples are preferred. Sequential sampling schemes have been used for informing decision making in clinical trials. Sequential sampling involves taking samples in stages, where sampling can stop early depending on the estimated lameness prevalence. When welfare assessment is used for a pass/fail decision, a similar approach could be applied to reduce the overall sample size. The sampling schemes proposed here apply the principles of sequential sampling within a diagnostic testing framework. This study develops three sequential sampling schemes of increasing complexity to classify 80 fully assessed UK dairy farms, each with known lameness prevalence. Using the Welfare Quality herd-size-based sampling scheme, the first 'basic' scheme involves two sampling events. At the first sampling event half the Welfare Quality sample size is drawn, and then depending on the outcome, sampling either stops or is continued and the same number of animals is sampled again. In the second 'cautious' scheme, an adaptation is made to ensure that correctly classifying a farm as 'bad' is done with greater certainty. The third scheme is the only scheme to go beyond lameness as a binary measure and investigates the potential for increasing accuracy by incorporating the number of severely lame cows into the decision. The three schemes are evaluated with respect to accuracy and average sample size by running 100 000 simulations for each scheme, and a comparison is made with the fixed size Welfare Quality herd-size-based sampling scheme. All three schemes performed almost as well as the fixed size scheme but with much smaller average sample sizes. For the third scheme, an overall association between lameness prevalence and the proportion of lame cows that were severely lame on a farm was found. However, as this association was found to not be consistent across all farms, the sampling scheme did not prove to be as useful as expected. The preferred scheme was therefore the 'cautious' scheme for which a sampling protocol has also been developed.