Establishment of a predictive model for postpartum hemorrhage in twins: a retrospective study

To explore the risk factors and develop a predictive model for postpartum hemorrhage in twin pregnancies.

All patients who gave birth at Ningbo Women and Children's Hospital from January 2018 to August 2022 were recruited. Patients were randomly allocated to a training cohort (n[Formula: see text]1395) validation cohort (n[Formula: see text]650) at a 7:3 ratio. In the training cohort, LASSO regression for screening variables and multifactorial logistic regression analysis were performed to identify independent risk factors for postpartum hemorrhage in twin pregnancies. A nomogram was established based on the results of multiple logistic regression analysis. Nomogram performance was quantified using the receiver operating characteristic curve, Hosmer- Lemeshow test and decision curve analysis.

A total of 2045 patients were included in this study. Multifactorial Logistic regression analysis showed maternal age, assisted reproduction, platelet count, fibrinogen level, albumin level, hypertensive disorders of pregnancy, placenta praevia, number of previous cesarean deliveries, number of previous intrauterine manipulation, and neonatal weight were independent risk factors for postpartum hemorrhage in twin births. The area under curve (AUC) for the training cohort was 0.810 [95[Formula: see text] CI (0.781, 0.839)], with a sensitivity of 76.5[Formula: see text], specificity of 71.0[Formula: see text], and positive and negative predictive values of 0.358 and 0.935, respectively, while the AUC for the validation cohort was 0.821 [95[Formula: see text] CI (0.781, 0.860)], with a sensitivity of 80.9[Formula: see text], specificity of 69.49[Formula: see text], and positive predictive value and negative predictive value of 0.426 and 0.929, respectively.

The predictive model can effectively and quantitatively assess the risk of postpartum hemorrhage in twin pregnancies and help clinicians to take personalized preventive measures.