Prediction of energy balance in a high yielding dairy herd in early lactation: model development and precision

This study was conducted to (1) predict herd mean EB, and (2) investigate w hether herd size affects the precision of prediction. In order to achieve t he first goal, it was studied to what extent milk test day information, bod y condition scores, and blood and milk ketones can estimate energy balance at cow level. EB was calculated in 72 Holstein-Freisian cows of one researc h herd (10,500 kg milk per 305-day lactation) during lactation weeks 2-12. A repeated measures model was used for multiple regression of predictors on energy balance. In addition to a base model (including lactation week, par ity and milk yield), the fat-protein-ratio, milk fat and milk protein conce ntrations explained substantially more variability in EB than did milk lact ose concentration and body condition scores, or tests for ketone bodies. A simulation of random selections of milk test day and animal subsets showed that sufficiently precise prediction of energy balance would require either herds larger than 150 cows or pooling of the results of subsequent test da ys. It was concluded that milk test day information without ketone levels o r body condition scores is sufficient to estimate herd mean energy balance, but that herd size limits the precision of prediction. Model validation in other herds and rations would be the next phase of this research. (C) 2000 Elsevier Science B.V. All rights reserved.