Short communication: Variance and autocorrelation of deviations in daily milk yield are related with clinical mastitis in dairy cows.

Daily milk production, and fluctuations therein, can provide information on health and resilience of dairy cows. We studied variance and autocorrelation of deviations in daily milk yield in relation to the occurrence of clinical mastitis (no, early or later in lactation). Individual lactation curves were fitted to 305-d lactations of 414 dairy cows using quantile regression. Log-transformed variance (lnVar) and autocorrelation of the quantile residuals of daily milk yield (predicted - observed) were evaluated for intervals until 30 and until 305 days in milk (DIM). Cows were classified as having no mastitis (n = 249), early mastitis that first occurred before 30 DIM (n = 29); or later mastitis (n = 136). Subsequently, linear models were used to assess effects of mastitis and parity class (primiparous or multiparous) on lnVar and autocorrelations; and logistic regression analyses were performed to predict mastitis from lnVar or autocorrelation and parity. From 10 to 30 DIM, lnVar was greater for cows with early mastitis than for cows with no or late mastitis, and autocorrelation tended to be lower for cows with early mastitis than for cows with no mastitis. The lnVar and autocorrelation from 10 to 30 DIM were not predictive of late mastitis. From 10 to 305 DIM, lnVar was greater and autocorrelation was lower for both cows with early and late mastitis than for cows with no mastitis; and both were predictive of having mastitis in the 305-d lactation. Primiparous cows had lower lnVar than multiparous cows. In cows without mastitis, autocorrelation values were positively correlated with lnVar. Results confirm that increased lnVar is associated with clinical mastitis.