How can the management of clinical and subclinical mastitis be supported by sensor systems?

Current sensor systems aim to detect cows with abnormal milk or mastitis. Although they may be less accurate than visual detection in detecting clinical mastitis, sensor systems have the advantage of multiple measurements per day. Mastitis detection, however, should be approached from a need to intervene (management support) perspective rather than based on clinical mastitis paradigms.

Cows with severe clinical mastitis need to be identified and treated properly as fast as possible. Sensor systems should have a very high sensitivity (at least 95%), combined with a high specificity (at least 99%) within a narrow time window (maximum 12 hours) to ensure that close to all cows with true cases are detected quickly. Since very sick animals may not visit a milking robot, detection algorithms need to take additional data into account, not only milk sensor data.

Cows that do not need immediate attention have a risk of progressing into severe clinical mastitis. However, they should get the chance to cure spontaneously under close monitoring. Intervention is needed for cows at risk of developing chronic mastitis, leading to production losses and increased risk of pathogen transmission. Sensor alerts should have a reasonable sensitivity (at least 80%) and a high specificity (at least 99.5%). The time window may be relatively long (around 7 days). Additional actions may contain further diagnostic testing.