0145 Metritis severity score misclassification underpredicts consequence cost of disease

The objective of this research was to determine the impact of disease misclassification on the estimated impact of metritis on milk production and time to pregnancy. Differential misclassification introduces bias that usually results in underestimating the true association. A convenience sample of DairyComp305 data representing 1 yr of calvings (n = 3485) from one midwestern Holstein herd was used. This herd was chosen because it had good recorded incidence of both mild and severe metritis cases. The original dataset included metritis recorded as mild or severe, or not recorded (NR) where no metritis was observed, and was considered to contain the metritis true severity (TS). First, to evaluate the impact of misclassification bias, we retrospectively randomized 40% of mild metritis to be classified as NR to represent inconsistent disease recording (IR); then, all mild metritis cases were changed to NR to represent poor disease recording (PR). The TS, IR, and PR datasets were analyzed separately in JMP. ANOVA was conducted for second test 305 d mature equivalent (2nd305ME), and a multivariate Cox proportional hazards model was conducted for time to pregnancy, censoring at 300 d in milk. Nonsignificant variables were removed when P > 0.10, but the variable metritis was forced into all models. Based on the TS model, adjusting for effects of lactation group, month fresh, early lactation mastitis and displaced abomasum, a case of mild metritis was associated with 405 kg less 2nd305ME and a case of severe metritis was associated with 1106 kg less 2nd305ME compared with no metritis. For the IR model, a case of mild metritis was associated with 376 kg less 2nd305ME and a case of severe metritis was associated with 1050 kg less 2nd305ME compared with no metritis. For the PR model, severe metritis was associated with 990 kg less 2nd305ME compared with NR. The IR and PR models underestimated 2nd305ME loss for severe metritis cases by 56 and 116 kg/cow, resulting in 8721 and 18,007 kg of milk loss unaccounted for at the herd level, respectively, compared with TS. For the TS model, cows that did not have metritis were 1.31 times more likely to get pregnant than cows with severe metritis (P = 0.01). The risk ratio difference in IR and PR models were 0.03 and 0.08, respectively. Overall, misclassification of metritis cases results in greater bias and largely underestimates the true association between metritis and consequence costs of the disease.