Milk yield dynamics as indicators for disease in dairy cows

Dairy cows experience the majority of diseases during early lactation, and diseases are linked to milk losses. Fluctuations in milk yield were previously suggested to add value in the health monitoring of dairy cows. The aim of this study was to investigate the relations between health status during the first 56 DIM and 2 types of milk yield dynamics: (1) lactation curve characteristics and (2) dynamic indicators of milk yield derived from deviations from the lactation curve. Data on the daily milk yield and disease treatments of 4 previous experiments were merged and examined. Cows were categorized in 3 ways based on disease treatments until 56 DIM: (1) into 3 groups based on the health status: no health problems (NHP), systemic or inflammatory health problems (SHP), other health problems (OHP), (2) by the number of disease treatments per cow (0, 1, or ≥2), (3) by the presence or absence of disease treatments for either mastitis, endometritis, claw and leg disorders, vaginal discharge, or cystic ovaries. The following lactation curve characteristics were calculated using the 5-d rolling average of daily milk yield: peak production, day of peak production, slope to peak, total milk yield until 56 DIM, 305-d milk production, and persistency. The following dynamic indicators of milk yield were calculated derived from daily deviations in milk yield during the periods of 10 to 56 DIM and 10 to 305 DIM: log-transformed variance (LnVar), lag-1 autocorrelation (R_auto), skewness, and mean negative residuals (MNR). Mixed models were used to investigate relations between disease treatments and milk yield dynamics. Peak production and total milk yield until 56 DIM were lower in SHP cows compared with NHP or OHP cows. Cows of parity ≥3 in SHP had a later day of peak production and lower slope to peak production compared with NHP and OHP cows of parity ≥3. Persistency and 305-d production were not affected by health status group. Cows with health problems (SHP and OHP) during the first 56 DIM had a higher LnVar than cows without treatment for health problems (NHP). More specifically, cows treated for endometritis or cystic ovaries had a higher R_auto through lactation compared with healthy cows. The MNR from 10 to 56 DIM was lower for cows treated for claw and leg disorders compared with healthy cows, but MNR was not of added value for other treatments types or health groups, and neither was skewness. The LnVar from 10 to 56 DIM was higher with increasing number of disease treatments. This indicated that cows with health problems in early lactation had more perturbations in milk yield than healthy cows, but the exact response was dependent on the type of disease and number of disease treatments in early lactation. In conclusion, although lactation curve characteristics specifically indicated cows with a systemic or inflammatory health problem, dynamic indicators of milk yield indicated both cows with a systemic or inflammatory health problem and cows with other health problems. The impact of specific diseases on milk yield dynamics should be further investigated to potentially allow for application in on-farm health monitoring tools.