Calf-Level Factors Associated with Bovine Neonatal Pancytopenia – A Multi-Country Case-Control Study

Jones, Bryony A.; Sauter-Louis, Carola; Henning, Joerg; Stoll, Alexander; Nielen, Mirjam; Van Schaik, Gerdien; Smolenaars, Anja; Schouten, Matthijs; den Uijl, Ingrid; Fourichon, Christine; Guatteo, Raphael; Madouasse, Aurélien; Nusinovici, Simon; Deprez, Piet; De Vliegher, Sarne; Laureyns, Jozef; Booth, Richard; Cardwell, Jackie M.; Pfeiffer, Dirk U.

Multivariable analysis was conducted using conditional logistic regression with farm as the group variable. The model was built using forward stepwise regression starting with the variable with the highest odds ratio and lowest p value on univariable analysis. Variables were retained if the likelihood ratio test indicated that inclusion led to a better model fit (p,0.05). For variables from the same risk factor group (e.g. colostrum management, dam vaccination) that were likely to be collinear, the variable with the highest number of observations was added first. It was then removed and the other collinear variables were added and removed one by one. If more than one of the collinear variables improved the model then the one with the most observations and/ or that was most biologically relevant was retained, as described below.