Does Information Available at Delivery Improve the Accuracy of Predicting Vaginal Birth after Cesarean? Validation of the Published Models in an Independent Patient Cohort

Abstract

We sought to validate a proposed vaginal birth after cesarean (VBAC) prediction model that includes variables available "at or close to delivery" and compare its accuracy to one that only uses variables available at "entry to care." We performed a retrospective cohort study of term pregnant women with a vertex singleton gestation attempting a trial of labor (TOL) after a single prior low transverse cesarean delivery. VBAC rates, predicted using the "close to delivery" model, were partitioned into deciles. The observed VBAC rate in each partition was compared with the predicted one. The accuracy of the two models was compared using the receiver operating characteristics curve. The predicted VBAC probability was higher in patients who had VBAC compared with those who failed a TOL (median [interquartile range]: 74.9% [59.6 to 86.1] versus 48.6% [35.4 to 66.7]; P < 0.001). The correlation between the observed and predicted VBAC rates was high (R = 0.98; P < 0.001). In the subset of patients who had the complete set of variables available for the two models (N = 490), the "close to delivery" model was more accurate. We validated the proposed VBAC prediction model in an independent cohort. Incorporating information available at delivery improves its accuracy.