Cows and herds constitute distinct hierarchical levels of heterogeneity in the variability of and association between milk yield and pregnancy outcome in dairy cows

Abstract

The objectives of this study were to investigate heterogeneity in the variance of and the association between milk yield (MY) and pregnancy outcome (PO) in dairy cows, formally separating the within-herd (i.e., cow-level residual effects) from the between-herd (i.e., herd-level random effects) components. Based on a recently developed extension to bivariate generalized hierarchical linear mixed models, we specified functions of residual and random effect variances and covariances as linear combinations of fixed and random effects to infer upon heterogeneity in the variation of and the association between MY and PO at first postpartum insemination. As potential sources of heterogeneity, we evaluated various management practices and herd attributes of interest by assessing model fit using the deviance information criterion. Our data consisted of 89,105 Dairy Herd Improvement Association cow records from 379 dairy herds in Michigan. Within herds, no evidence of a cow-level (residual) association between MY and PO was observed, as the corresponding association parameter did not significantly depart from zero. However, the herd-level (random effects) relationship between MY and PO was antagonistic and depended on management practices that determine the baseline level of fertility for a herd. In other words, herds with greater average MY at the time of first postpartum insemination had lower pregnancy rates, but within such herds, cows with higher daily yields did not seem to be any more or less likely to become pregnant than lower-yielding herdmates. Nevertheless, Michigan counties differed in the magnitude of the herd-level association between MY and PO, thus indicating that regional environmental conditions or management practices may partially alleviate the herd-level antagonism between MY and PO. The heterogeneity in variability of MY was substantial and primarily explained at the cow level by management conditions and other herd-specific attributes. In summary, the nature of the variability of and the association between MY and PO in dairy cows is complex due to the heterogeneous contributions of both cow- and herd-level components. Further research should be pursued to investigate additional management scenarios that ameliorate or even enhance the association between MY and PO in commercial dairy cows during first postpartum insemination.