External validation of vaginal birth after cesarean delivery calculator in Israeli cohort of parturients and construction of an improved model

Objective To validate the Maternal Fetal Medicine Unit's (MFMU) vaginal birth after cesarean delivery (VBAC) calculator in an Israeli cohort, and to detect other variables associated with VBAC and construct an improved VBAC calculator. Methods A retrospective cohort study was performed at a single university-affiliated medical center. Women carrying a singleton, term, cephalic-presenting fetus, with previous one low transverse cesarean delivery who opted for trial of VBAC were included. Demographic and obstetric characteristics were incorporated into the MFMU's calculator, to predict probabilities of VBAC and compare prediction performance with the original publication utilizing receiver operating characteristic (ROC) statistics. Logistic regression analysis was used to investigate other variables and construct an improved model for success of VBAC. Results Of 490 parturients, 396 (80.8%) had a successful vaginal delivery. Compared to the original publication, the MFMU's calculator underperformed: area under the ROC curve (AUC) was 0.709 (95% confidence interval [CI] 0.652-0.766, P < 0.001). Sensitivity, specificity, positive and negative predictive values, and overall accuracy were 67.42%, 65.96%, 89.30%, 32.46%, and 32.46%, respectively. An improved model that included previous VBAC, prior vaginal delivery, spontaneous onset of delivery, and maternal diabetes resulted in improved prediction performance with an AUC of 0.771 (95% CI 0.723-0.82, P < 0.001). Conclusion MFMU's VBAC calculator needs to be validated in different populations before implementation.

Automated summary

The study aimed to validate the MFMU's VBAC calculator in an Israeli cohort and discover other variables associated with VBAC, resulting in an improved VBAC calculator. The results showed that the MFMU's calculator underperformed compared to the original publication, but an improved model including additional variables demonstrated better prediction performance.