Inter-observer reliability of a scoring system to evaluate bruises on turkey carcasses

Introduction Traumatic injuries such as bruises have been considered an important indicator to assess animal welfare in livestock farming. The possibility of assigning the injury to a particular stage or moment in the production process may allow judgments on possible causes and thus reduce its prevalence. Currently, there is no consistent definition or scoring system for bruised poultry carcasses in German meat inspection and the prevalence is affected by the variability of scoring systems as well as observer bias. Therefore, the objective of this study was to determine the inter-observer reliability of bruise characteristics at the slaughter line and to validate the findings with measurements of bruises and photographed carcasses. Methods & results Inter-observer reliability was assessed with two observers who each scored 10,880 turkey carcasses simultaneously at a running slaughter line after a short training session. The strength of agreement was “good” for the total number of detected bruises and the number of bruises on breasts, wings, and legs per flock (ICC = 0.81 – 0.88). Agreement (ICC) on the number of small, medium, and large bruises ranged between “good” and “moderate” values (0.70 – 0.84), whereas the number of bruises in different colors showed “moderate to poor” reliability (0.04 – 0.64). Additionally, agreement on bruise characteristics was tested using photographs (n= 513 carcasses) without a time limit. The highest agreement between observers was found for the location of bruises (Kw = 0.98). Again, the color of the bruises showed the lowest agreement (Kw = 0.36), whereas it was “moderate” for the size of the bruises (Kw = 0.45). When comparing each observer’s scoring values for size with size measurements (digital analysis) of the bruises, the observers tended to underestimate the actual size. Discussion & conclusion Overall, the total number of detected bruises and the location of the bruises showed the highest agreement between observers at the slaughter line and from photographs, indicating they were reliable variables. However, as the color variable showed a low agreement, a standardized objective method should be developed to objectively assess bruise prevalence and characteristics.