A comparison of udder health trait definitions in German Holstein dairy cattle based on mastitis treatment data

Udder diseases are a major problem in dairy cattle herds. Apart from the improvement of hygiene and food quality, udder health can be improved by genetic selection. This study used disease treatment data of 29,698 Holstein cows from 21 German contract testing herds to evaluate different trait definitions with respect to variance components, breeding values, and genetic correlations to somatic cell score (SCS) and milk yield. Considering the first 305 days of a lactation, four traits were defined: a binary trait, the number of udder health treatments (TRS), the number of sickness periods (NUMS) and the days a cow was considered to be sick. Additionally the number of treatments and the number of sickness periods defined from the treatments were analyzed based on a model with an underlying Poisson distribution. Results from analyses with single and multiple trait sire and animal models showed low heritabilities, ranging from 0.05 to 0.13. The genetic correlation of 305 day average somatic cell score with all disease treatment traits defined in this study was high but differed clearly from one. Genetic correlations with 305 day milk yield were unfavorable and remarkably higher than genetic correlations of SCS with milk yield. Reliabilities of estimated breeding values (EBV) were the highest for NUMS. Spearman correlations of EBV for binary and linear trait definitions differed considerably from unity. Correlations of EBV between the disease treatment traits defined in this study and SCS were similar to genetic correlations. An exception was found for the correlation between EBV of TRS and SCS (0.46) which was remarkably lower than the corresponding genetic correlation. In summary, this study demonstrates the differences between various disease treatment traits as well as the benefits of using a more precise trait definition instead of the binary option. Despite the lower heritabilities, the discrepancy between the correlations of disease treatment traits and SCS to milk yield showed the necessity to include disease treatment traits, i.e. case data on mastitis, into breeding objectives and selection indices.