An estimation of the clinical mastitis incidence per 100 cows per year based on routinely collected herd data

The aim of this study was to evaluate whether it was possible to (1) estimate the clinical mastitis incidence rate (CMI) for all Dutch dairy herds and (2) to detect farms with a high CMI based on routinely collected herd data. For this study, 240 dairy farms with a conventional milking system that participated in the milk recording program every 4 to 6 wk were randomly selected and agreed to participate. From the initial 240 herds, data of clinical mastitis (CM) registrations and routinely collected herd data of 227 herds were complete and could be used for analysis. Routinely collected herd data consisted of identification and registration records, antimicrobial usage, test-day records from the milk recording program, bulk tank milk (BTM) somatic cell count data and results of diagnostic tests on BTM samples. For each of the 227 herds, the CMI per 100 cows per year was calculated per quarter of the year and was combined with the available herd data. Two models were developed to predict the CMI for all dairy herds and to detect individual herds that belonged to the 25% herds with the highest CMI. Records of 156 (67%) herds were used for development of the models and the remaining 71 (33%) were used for validation. The model that estimated the CMI in all herds consisted of 11 explanatory variables. The observed and predicted averages of the validation herds were not significantly different. The model estimated a CMI per 100 cows per year of 32.5 cases (95% confidence interval=30.2-34.8), whereas the farmers registered 33.4 cases (95% confidence interval=29.5-37.4). The model that aimed at detecting individual herds with a high CMI contained 6 explanatory variables and could correctly classify 77% of all validation herds at the quarter-year level. The most important variables in the model were antibiotic usage for treating CM and BTM somatic cell count. In conclusion, models based on routinely collected herd data gave an accurate prediction of CMI for all Dutch dairy herds and could detect individual dairy herds with a high CMI. With these models it is possible to periodically monitor CMI both at the herd and at the national level, which is valuable for monitoring purposes and can motivate farmers to continuously improve udder health in their herds.