New research in the Journal of Dairy Science examined if accelerometers could be used to predict postpartum disease. In the trial 489 multiparous cows on a commercial Holstein dairy in Spain were fitted with accelerometers and monitored from 3 weeks prior to 30 days after calving. The accelerometers combined with a sensor to detect cows at the feed bunk were used to measure steps, time spent at the feed bunk, frequency of meals, number of lying bouts and total time spent lying.
During the first 30 days in milk 144 cows, 29.4%, were diagnosed with at least one disease. The remaining 345 cows had no disease diagnosed (NDD). When prepartum activity was analyzed as a 3-week block no differences were found. Cows averaged 1,613 steps, spent 181 minutes at the feed bunk, ate 8.3 meals and spent 742 minutes lying during 9.8 lying bouts.
But when researchers examined activity in just the week before calving and evaluated results for individual diseases, differences that were statistically significant and predictive emerged. Cows diagnosed with metritis spent 32 more minutes at the feed bunk than cows NDD, 188 vs. 156 minutes/day. Cows diagnosed with a displaced abomasums (DA) ate 24% fewer meals, 6.2 vs. 8.2 meals/day, for cows NDD. There was also a tendency for DA cows to take fewer steps, 1,395 vs. 1,708 for cows NDD. And for cows diagnosed with ketosis two predictors were identified: time at the feed bunk and number of meals per day. Ketotic cows spent 62 fewer minutes/day eating and ate fewer meals 6.4 vs. 8.2 meals/day when compared to cows that were not diagnosed with ketosis.
Researchers were able to develop prediction models to classify animals as high and low risk for DA and ketosis. But due to the small number of clinical cases the false discovery rate was higher than anticipated. Refinement will help improve accuracy before prediction models are used farm.