An individual modelling tool for consecutive clinical mastitis during the same lactation in dairy cows: a method based on a survival model.

The high number of clinical mastitis recurring within the same lactation in dairy cows constitutes one of the factors of overdispersion in standard Poisson models. Our method, based on biological parameters, i.e., recurrence hazard in relation to udder exogenous infection (Rex) or recurrence hazard and rate in relation to endogenous infection (Ren), produced a model capable of integrating a possible change of state in the udder after clinical mastitis. This model was based on a study of the time intervals between successive clinical episodes, both types of risk being considered in the form of a distribution mixture in the survival model. The modelling tool allowed to determine the factors that specifically act on either one of the potential risks and estimated the distribution of the number of clinical mastitis per lactation, as well as the distribution of when mastitis occurs. Estimation results obtained by this method in an experimental herd were compared with those from more classical models with or without random individual effects. The distribution of the number of mastitis per lactation estimated by our method was well-fitted to the data and the method identified variation factors which were relatively standard in this type of study: lactation number, lactation stage and calving month. Prediction results obtained in another experimental herd with more recent data without parameter re-estimation demonstrated the adequacy of the model in fitting observed data. This modelling method based on biological parameters in a mixture of survival distributions was interesting to model clinical mastitis recurring within the same lactation. However in the future it will also be important to integrate the possible relationship between successive lactations and to apply this model to other types of farming systems