Hong Shibin, Liu Chang, Kang Xin, Lio Ka U, Le Yiping, Zhang Ting, Shi Haoting, Dai Lan, Di Wen, Zhang Ning
Department of Obstetrics and Gynecology, Renji Hospital, Shanghai Jiao Tong University School of Medicine, 160 Pujian Road, Pudong New District, Shanghai, 200127, China.
Shanghai Key Laboratory of Gynecologic Oncology, Renji Hospital, Shanghai Jiao Tong University School of Medicine, 160 Pujian Road, Shanghai, 200127, China.
BMC Pregnancy Childbirth. 2025 Sep 2;25(1):925. doi: 10.1186/s12884-025-08066-1.
The objective of this study is to establish and validate models that can accurately predict postpartum hemorrhage (PPH) in women with placenta previa totalis prior to undertaking cesarean delivery.
A retrospective cohort study was conducted on 306 pregnancies with placenta previa totalis delivered between January 2011 and June 2022. The pregnancies were classified into two groups, PPH group and non-PPH group, based on bleeding volume and red blood cell transfusion. Clinical features and pre-operative coagulation function indexes were recorded. The entire cohort was randomly divided into a development cohort (n = 214) and a test cohort (n = 92). Least absolute shrinkage and selection operator (LASSO) was implemented to select significant predictors, followed by step-wise logistic regression analysis to build the prediction model. Additionally, machine learning-based models were compared with the proposed model.
Among 306 participants, 115 (53.74%) and 50 (54.35%) cases of PPH were observed in the development and test cohorts, respectively. The LASSO-Logistic regression model incorporated preoperative serum fibrinogen level, history of prior cesarean delivery and history of antepartum bleeding as predictors. The model yielded an area under the receiver operating characteristic (ROC) curve of 0.721 (95% CI 0.652-0.790) in the development cohort and 0.706 (95% CI 0.600-0.813) in the test cohort. Additionally, the model demonstrated a specificity of 70.7% (95% CI 61.7-79.7%) and a positive predictive value of 72.1% (95% CI 63.5-80.7%) for distinguishing between PPH and non-PPH cases. The LASSO-Logistic regression model outperformed the machine learning based model in the test cohort, confirming its efficiency in predicting PPH in patients with placenta previa totalis.
This study successfully developed and validated a LASSO-Logistic regression model incorporating coagulation indicators to predict PPH in patients with placenta previa totalis. Further large-scale prospective studies are warranted to externally validate the three-variate-based model and assess its practical application in real-time practice.
本研究的目的是建立并验证能够在剖宫产术前准确预测完全性前置胎盘孕妇产后出血（PPH）的模型。
对2011年1月至2022年6月间分娩的306例完全性前置胎盘妊娠进行回顾性队列研究。根据出血量和红细胞输注情况将这些妊娠分为两组，即产后出血组和非产后出血组。记录临床特征和术前凝血功能指标。将整个队列随机分为一个开发队列（n = 214）和一个测试队列（n = 92）。采用最小绝对收缩和选择算子（LASSO）来选择显著预测因素，随后进行逐步逻辑回归分析以建立预测模型。此外，将基于机器学习的模型与所提出的模型进行比较。
在306名参与者中，开发队列和测试队列中分别观察到115例（53.74%）和50例（54.35%）产后出血病例。LASSO逻辑回归模型纳入术前血清纤维蛋白原水平、既往剖宫产史和产前出血史作为预测因素。该模型在开发队列中的受试者操作特征（ROC）曲线下面积为0.721（95%CI 0.652 - 0.790），在测试队列中为0.706（95%CI 0.600 - 0.
