Data-driven categorization of on-farm animal welfare and slaughter outcomes in slow growing chickens with outdoor access: When data speak

Animal welfare is a crucial ethical and economic aspect in modern poultry production, even in alternative systems based on slow growing strains and outdoor access. Welfare assessment is commonly performed using standardized and quantitative protocols, such as the AWIN® assessment protocol, which provide objective animal-, resource-, and management-based indicators at flock level. However, the evaluation of multiple indicators requires trained assessors and time, and flock categorization depends on the application of pre-established thresholds. This study aimed to determine whether assessed on-farm welfare indicators allow the a priori, data-driven classification of commercial flocks according to their welfare, and whether such classification corresponds with slaughter outcomes. 98 slow growing chicken flocks from 22 farms in Northern Spain were assessed using animal-based, resource-based, and management-based indicators from the AWIN® protocol. Hierarchical clustering was used to group flocks according to these indicators. Generalized linear mixed models were used to characterize the welfare profile of the identified clusters, and to check whether on-farm welfare profiles corresponded with slaughter outcome differences. The hierarchical clustering identified 2 welfare profiles comprising 89 (Cluster 1) and 9 (Cluster 2) flocks. Cluster 2 exhibited the poorest welfare profile, characterized by a higher prevalence of lame, immobile, tail-damaged, terminally ill and dead birds, and the highest house NH3 concentrations. Importantly, the poorer welfare profile of Cluster 2 was associated with lower slaughter performance, including a lower percentage of Category A carcasses, a higher percentage of Category 0 carcasses, and a trend towards higher total mortality. These findings demonstrate that raw on-farm welfare data can be used to classify slow growth chicken flocks into clusters reflecting distinct welfare profiles associated with slaughter performance and carcass quality. This data-driven pre-classification provides an objective framework for flock stratification, enabling the identification of flocks with welfare deficiencies and supporting targeted management interventions to improve animal welfare and production outcomes, with potential application to other poultry production systems and livestock species.