Additional file 2: of Maternal and fetal predictors of fetal viral load and death in third trimester, type 2 porcine reproductive and respiratory syndrome virus infected pregnant gilts

Predictability of the fetal death probability logistic model. Double receiver operating characteristic (ROC) curves demonstrate the effects of “probability cutoff” on the sensitivity and specificity of the final multi-level logistic regression model. By default, a probability cutoff of 0.5 was used to generate the ROC curves; fetuses with a predicted probability <0.5 were classified as live, whereas fetuses with a predicted probability >0.5 were classified as dead. The sensitivity curve (solid black line) represents the percentage of fetuses that were correctly classified as dead by the model across a range of probability cutoffs. The specificity curve (grey dashed line) represents the percentage of fetuses that were correctly classified as live by the model. At a probability cutoff of 0.5, the model is superior at predicting live fetuses than dead, which implies that factors associated with fetal death, apart from those included in the final model, exist. The contingency table of raw data and summary of model performance are shown in the bottom left and right, respectively.