Identifying gram-negative and gram-positive clinical mastitis using daily milk component and behavioral sensor data

Area of Science:

• Animal Science
• Dairy Science
• Veterinary Medicine

Background:

• Automated animal health monitoring and early disease detection are crucial for dairy farms.
• Precision technologies offer opportunities for improved on-farm disease management, specifically for mastitis.

Purpose of the Study:

• To evaluate time-series changes in milk and activity variables for predicting clinical mastitis (CM).
• To differentiate CM cases caused by gram-negative (GN) or gram-positive (GP) pathogens, or those with no pathogen isolated (NPI).

Main Methods:

• Developed predictive algorithms using milk (yield, conductivity, somatic cell count, lactose, protein, fat) and activity (steps, lying time, lying bout duration, number of lying bouts) parameters.
• Analyzed data from 14 days preceding CM events (n=170) and matched controls (n=166).
• Estimated slope changes over 7-day periods relative to CM detection using linear regression.

Main Results:

• No single parameter significantly predicted all clinical mastitis (ACM).
• Significant predictors were identified for GN, GP, and NPI models.
• The best models achieved superior sensitivity (Se) and specificity (Sp) for GP (Se=82%, Sp=87%) and NPI (Se=80%, Sp=94%) compared to ACM (Se=73%, Sp=75%) and GN (Se=71%, Sp=74%).
• Optimal baseline for GN prediction was closer to CM detection (d -3), while GP and NPI were best predicted further back (d -10).

Conclusions:

• Milk and activity sensor data show potential for use in clinical mastitis detection systems.
• Differentiating mastitis types based on sensor data is feasible.
• Algorithm performance varies depending on the mastitis-causing pathogen or lack thereof.