Validating a predictive model for caesarean section in low-risk nulliparous pregnancies

Abstract

ProblemCaesarean birth (CS) rates are steadily increasing. BackgroundIn 2017 Janssen et al. developed a model which could predict CB in nulliparous healthy woman with 71 % accuracy based on factors measurable on admission to the hospital. AimTo validate the predictive model for risk of caesarean birth among low-risk, nulliparous women in a new setting. MethodsA retrospective chart study in Abbotsford Regional Hospital (British Columbia, Canada) of healthy nulliparous women in spontaneous labour, at term, with a singleton fetus in cephalic position. Sociodemographic, pregnancy and labour-related characteristics were collected and independent predictors of CS were determined using multivariate logistic regression. The Janssen model was tested in the Abbotsford sample and additionally novel predictors were tested in an effort to improve the model. The area under the ROC curve (C-statistic) was computed and model calibration, sensitivity and specificity evaluated for the final model. Findings and discussionOf 348 women, 106 (30.5 %) had a CB. Applying the Janssen predictive model to the Abbotsford data resulted in a C-statistic of 0.77. No new predictors were added to the model. The mean predicted risk score for CS in the cohort was 0.30 ± 0.20. A risk score cut-off of 0.32 was determined resulting in a sensitivity and specificity of 69 %. The model had acceptable calibration. ConclusionA model with variables easily accessible at admission can predict caesarean birth in nulliparous women. The results from this study can guide provision of more intensive care during labour to women at higher risk, with the overall goal of reducing CB rates.