OBJECTIVE: To externally validate a prediction model for the risk of a cesarean section after induction of labor (IOL) using a web-based cesarean risk calculator in a tertiary perinatal center and to compare the performance of three calculators in predicting the cesarean risk after IOL based on their sensitivity and specificity. METHODS: This was a prospective observational study of 577 women over a period of 1 year at a tertiary perinatal center in India. Women with singleton-term pregnancies with intact membranes that underwent induction were included. We used three prediction tools
Levine's, Rossi's and Irwinda's calculators for predicting cesarean rates. The variables were entered directly into the calculators at the start of IOL, producing an individualized risk of cesarean delivery. The mode of delivery was the primary outcome variable. Area under the ROC curve (AUC), calibration plots and decision making curve analysis were used for comparison. RESULTS: Out of 577 mothers who underwent IOL, 345 (59.79%) women had a vaginal birth and 232 (40.21%) underwent cesarean section. The Levine calculator reached the maximum discriminative capacity (AUC: 0.785) for our population, followed by Rossi (AUC: 0.7723) and Irwinda (AUC: 0.6608). Levine's calculator slightly overestimated the risk of cesarean section at lower thresholds but underestimated the risk at higher threshold probabilities whereas Rossi's calculator underestimated the risk of cesarean section at all threshold probabilities above 11%. CONCLUSION: The Levine and Rossi calculators have the greatest potential for use in clinical settings. However, these cannot be used individually for clinical decision making without prospective studies evaluating their clinical impact on pregnant women undergoing IOL.