Determinants of Cesarean Section among Primiparas: A Comparison of Classification Methods

Background: Over the last few decades, Cesarean section (CS) rates have increased significantly worldwide par-ticularly in Iran. Classification methods including logistic regression (LR), random forest (RF) and artificial neural network (ANN) were used to identify factors related to CS among primipars. Methods: This cross-sectional study included 2120 primipars who gave singleton birth in Tehran, Iran between 6 and 21 July 2015. To identify factor associated with CS, the classification methods were compared in terms of sensitivity, specificity, and accuracy. Results: The CS rate was 72.1%. Mother’s age, SES, BMI, baby’s head circumference and infant weight were the most important determinant variables for CS as identified by the ANN method which had the highest accuracy (0.70). The association of RF predictions and observed values was 0.36 (kappa). Conclusion: The ANN method had the best performance that classified CS delivery compared to the RF and LR methods. The ANN method might be used as an appropriate method for such data.