Associations between routinely collected Dairy Herd Improvement data and insemination outcome in UK dairy herds

Milk constituent concentrations in samples taken during early lactation are often used to generate proxy measures for energy balance in dairy herds. This study aimed to explore associations between these and other measures routinely recorded by Dairy Herd Improvement schemes and insemination outcome, with an emphasis on the likely predictiveness of such measures for conception risk (the proportion of inseminations that are successful) at herd level. Data from 312 UK dairy herds were restructured so that each unit of data represented an insemination at less than 100 d in milk (DIM). Milk constituent concentrations from the first and second test days (corrected for the effects of season and DIM at sampling) were used as potential predictors of insemination outcome in a logistic regression model. Other predictors included representations of milk yield and other information routinely collected by Dairy Herd Improvement Associations; random effects were used to account for clustering at cow and herd levels. The final model included a large number of predictors, with several interaction and nonlinear terms. The relative effect sizes of the measures of early lactation milk constituent concentrations were small. The full model predicted just under 64% of observed variation in herd-year conception risk (i.e., the proportion of inseminations that were successful in each herd in each calendar year); however, around 40% was accounted for by the herd-level random effect. The predictors based on early lactation milk constituent concentrations accounted for less than 0.5% of observed variation, and representations of milk yield (both overall level of yield and early lactation curve shape) accounted for around 7%; DIM at insemination, parity, interservice interval, year, and month accounted for the remaining 15%. These results suggest that early lactation milk constituent information is unlikely to predict herd conception risk to a useful extent. The large proportion of observed variation explained by the herd-level random effect suggests that there are unmeasured (in this study) or unmeasurable factors that are consistent within a herd and are highly influential in determining herd conception risk.