Changes in Feeding Behavior as Possible Indicators for the Automatic Monitoring of Health Disorders in Dairy Cows

Abstract

Changes in short-term feeding behavior of dairy cows that occur with the onset of the health disorders ketosis, acute locomotory problems, and chronic lameness were investigated using data collected during previous experiments. The objective of the study was to describe and quantify those changes and to test their suitability as early indicators of disease. Feed intake, feeding time, and number of daily feeder visits were recorded with computerized feeders. Ketosis in 8 cows was characterized by rapid daily decreases in feed intake [−10.4kg of fresh matter (FM)], feeding time (−45.5min), and feeding rate (−25.3g of FM/min) during an average of 3.6 d before diagnosis by farm staff. Acute locomotion disorders in 14 cows showed smaller daily decreases in feed intake (−1.57kg of FM) and feeding time (−19.1min), and a daily increase in feeding rate (+21.6g of FM/min) during an average of 7.7 d from onset to diagnosis. The effects of chronic lameness on short-term feeding behavior were assessed by analyzing changes during the 30 d before and 30 d after all cows were checked for foot lesions and trimmed, and cows were classified as either lame (n = 81) or not lame (n = 62). During the 30 d before trimming, cows classified as lame showed significant changes in daily feeding time, number of daily visits, and feeding rate, but nonlame cows did not. In lame cows, the observed daily changes (slope) for the 30 d before and the 30 d after trimming were −0.75 and +0.32 min/d for daily feeding time, −0.35 and +0.31 for daily number of visits, and +0.77 and −0.35 g/min for feeding rate, respectively. These changes in feeding behavior were not different among cows consuming low or high forage rations. Daily feeding time was the feeding characteristic that changed most consistently in relation to the studied disorders. A simple algorithm was used to identify cows whose daily feeding time was lower than the previous 7-d rolling average minus 2.5 standard deviations. The algorithm resulted in detection of more than 80% of cows with acute disorders at least 1 d before diagnosis by farm staff. Short-term feeding behavior showed very characteristic changes with the onset of disorders, which suggests that a system that monitors short-term feeding behavior can assist in the early identification of sick cows.