Predicting risk of postpartum haemorrhage during the intrapartum period in a general obstetric population

Abstract

ObjectiveTo develop and validate (both internally and externally) a prediction model examining a combination of risk factors in order to predict postpartum haemorrhage (PPH) in a general obstetric Irish population of singleton pregnancies. Study designWe used data from the National Maternal and Newborn Clinical Management System (MN-CMS), including all singleton deliveries at Cork University Maternity Hospital (CUMH), Ireland during 2019. We defined PPH as an estimated blood loss of ≥ 1000 ml following the birth of the baby. Multivariable logistic regression with backward stepwise selection was used to develop the prediction model. Candidate predictors included maternal age, maternal body mass index, parity, previous caesarean section, assisted fertility, gestational age, fetal macrosomia, mode of delivery and history of PPH. Discrimination was assessed using the area under the receiver operating characteristic curve (ROC) C-statistic. We used bootstrapping for internal validation to assess overfitting, and conducted a temporal external validation using data from all singleton deliveries at CUMH during 2020. ResultsOut of 6,077 women, 5,807 with complete data were included in the analyses, and there were 270 (4.65%) cases of PPH. Four variables were considered the best combined predictors of PPH, including parity (specifically nulliparous), macrosomia, mode of delivery (specifically operative vaginal delivery, emergency caesarean section and prelabour caesarean section), and history of PPH. These predictors were used to develop a nomogram to provide individualised risk assessment for PPH. The original apparent C-statistic was 0.751 (95% CI: 0.721, 0.779) suggesting good discriminative performance. There was minimal optimism adjustment to the C-statistic after bootstrapping, indicating good internal performance (optimism adjusted C-statistic: 0.748). Results of external validation were comparable with the development model suggesting good reproducibility. ConclusionsFour routinely collected variables (parity, fetal macrosomia, mode of delivery and history of PPH) were identified when predicting PPH in a general obstetric Irish population of singleton pregnancies. Use of our nomogram could potentially assist with individualised risk assessment of PPH and inform clinical decision-making allowing those at highest risk of PPH be actively managed.