(204) Machine Learning Prediction of risk for Cesarean delivery among nulliparous women in spontaneous labor

We aimed to design a machine-learning algorithm to offer real-time insights for mode of delivery, which may assist obstetric decision-making during labor, focusing on the most significant parturient population, i.e., nulliparous women with spontaneous onset of labor of a singleton gestation.

Study Design:

A dataset of electronic medical-obstetrical records was utilized, including all nulliparous women with a singleton gestation and spontaneous onset of labor, collected during a 10-year period (2012-2022). Each record contained 102 maternal and fetal features available as either baseline characteristics at labor admission (59 features) and real-time data acquired during labor process (43 features). These data were associated with an outcome of mode of delivery. We trained a machine-learning model to predict the outcome defined as vaginal delivery (vaginal or assisted delivery).

Results:

A total of 9,723 (8,210 vaginal, 1,100 assisted and 413 cesarean deliveries) women were included. The most influential characteristic available at delivery ward admission was maternal age; while the most influential features during labor were the first cervical dilation examination and the partogram’s curve. A model utilizing only baseline features yielded an area under the curve of 0.74 (95% confidence interval 0.67-0.81), while a model which - in addition - included real-time data accumulating throughout different stages of labor, increased dramatically the model’s performance, yielding an area under the curve of 0.98 (95% confidence interval 0.94-1.0).
As shown in figure 1 we found a significant difference in the labor partogram between the 2 groups divided by mode of delivery, whether vaginal or cesarean. 
 

Conclusion:

Combination of data gathered from time of labor and delivery admission with real-time data throughout the process of labor enables successful and effective prediction, based on a machine learning tool, of mode of delivery. Such a tool, can aid the obstetrician assessing the risk of intra-partum cesarean delivery, as it evolves during labor.