(170) Successful Trial of Labor After Cesarean Prediction Utilizing Machine Learning Frameworks

Vaginal birth after cesarean section (VBAC) is associated with reduced maternal morbidity and improved outcomes in future pregnancies. Planned c-section leads to better outcomes over an unplanned c-section following failed TOLAC. 

Recent literature has shown that clinicians overestimate the likelihood of successful VBAC (AUC=0.60).  Identifying best candidates for TOLAC can guide clinical decision making and reduce emergency cesarean section. We aimed to develop and validate a machine learning framework utilizing non-invasive data to accurately predict successful VBAC.

Study Design: All patients from 2018-2020 in the National Vital Statistics System who had undergone a previous cesarean section met the inclusion criteria. Clinical variables studied included patient demographics, obstetric history, and clinical risk factors. The primary outcome predicted was successful VBAC. An extreme gradient boosting machine learning model was developed on the data. Weight scaling algorithms were utilized to improve representation of patients who underwent sucessful VBAC. The model was trained on all patients from 2018-2020, as well as VBAC patients from years 2014, 2016, and 2017 and was tested on a hold-out set of patients from the year 2021.

Results:

43 variables across 1,948,912 obstetric patients were included in the training data, of which 453,085 successfully had a VBAC. 

The model was tested on 561,717 patients from the 2021 year, of which 79,558 patients had a successful VBAC. The extreme gradient boosting model had an area under the receiver operating characteristic curve of 0.85, specificity of 0.89, with an accuracy of 85.4%, and F1 score of 0.85.

The highest weighted factors in the model were interval since last live birth, prepregnancy weight, and paternal age.

Conclusion: Machine learning models using routinely collected clinical data can accurately predict the probability of successful TOLAC. This could optimize counseling for pregnant patients considering TOLAC versus repeat cesarean delivery and have the potential to reduce unnecessary operative deliveries and associated costs for hospital systems.