A simple model to predict vaginal deliveries
We performed a forward multivariate logistic regression analysis model that included the factors associated with the delivery mode in the univariate analysis (P<0.1). In this model, the AoP and maternal height were the independent factors associated with the mode of delivery after adjustment for confounders, including maternal age, gestational age, overweight (BMI≥25 Kg/m2), premature rupture of the membrane, labor induction and oxytocin augmentation (Table S4). The combination of maternal height and the AoP could evaluate whether women were to undergo vaginal delivery, yielding an AUC of 0.738 (95% CI, 0.763-0.793; P<0.001) at T1, 0.820 (95% CI, 0.752-0.888; P<0.001) at T2 and 0.940 (95% CI, 0.885-0.995; P<0.001) at T3 (Figure 4), respectively. Figure 3(b) showed the optimal calibration of the prediction model. This model was also suitable for women with fetuses in the occiput posterior or transverse position at the latent phase, yielding AUC 0.725 (95% CI, 0.649-0.801; P<0.001) at T1 and 0.790 (95% CI, 0.705-0.874; P<0.001) at T2, respectively (Figure S2).