A predictive tool for vaginal birth after caesarean success in an Australian cohort

Abstract

Women who have previously had a caesarean section often face the choice between planning for a vaginal birth after caesarean (VBAC) or an elective repeat caesarean section (CS) for future pregnancies. Informing a woman of her individualised chance of a successful VBAC can aid her decision making. The aim is to create two VBAC prediction models using an Australian cohort - one for use in labour when labour variables are known, and one for use antenatally when labour characteristics are unknown. This study was a retrospective analysis of perinatal data in Victoria, Australia, over a 10-year period. During this time, 22 062 women were recorded as attempting a VBAC with a term singleton live birth. The data were separated into three parts. A 'training set' was used to build the complete VBAC prediction model and the antenatal VBAC prediction model using multivariate logistic regression. The models were then adjusted to only include the variables that contributed to model performance. The models were validated by testing the receiver operating characteristic (ROC) area under the curve within the 'validation set'. Then the models were tested for accuracy within the 'test set'. Using a 'test set' of data, the models demonstrated an area under the ROC curve of 0.7887 and 0.7384 for the complete and antenatal models respectively, showing adequate performance of both models. With these models, Australian women can be counselled about their predicted chance of VBAC success.