Value of an obstetric intelligent assistant in predicting postpartum hemorrhage after vaginal delivery
10.3760/cma.j.cn113903-20250821-00443
- VernacularTitle:产科智能助手预测阴道分娩产后出血的价值
- Author:
Lin YU
1
;
Huilan WANG
1
;
Yanmei ZHOU
1
;
Lin LIN
1
;
Yanhong CHEN
1
;
Yong WANG
1
;
Xianqin YIN
1
;
Dunjin CHEN
1
Author Information
1. 广州医科大学附属第三医院妇产科，广东省产科重大疾病重点实验室，广东省妇产疾病临床医学研究中心，广东省母胎医学工程技术研究中心，广州 510150
- Publication Type:Journal Article
- Keywords:
Postpartum hemorrhage;
Machine learning;
Obstetrics;
Artificial intelligence;
Intelligent assistant
- From:
Chinese Journal of Perinatal Medicine
2025;28(10):829-834
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To evaluate the clinical value of an obstetric intelligent assistant in predicting postpartum hemorrhage (PPH) after vaginal delivery.Methods:This retrospective cohort study included 4 832 women who delivered vaginally at ≥26 weeks of gestation at the Third Affiliated Hospital, Guangzhou Medical University between May 2023 and April 2025. Participants were categorized into PPH (382 cases, blood loss ≥500 ml within 24 h after delivery) and non-PPH groups (4 450 cases). Using traditional statistical methods combined with machine learning approaches, including support vector machines and extreme gradient boosting, supplemented with deep learning techniques, we developed a novel artificial neural network model—the obstetric intelligent assistant. This model provides a refined classification of PPH occurrence and estimated blood loss volume. The obstetric intelligent assistant integrates 70 antenatal and intrapartum risk factors through hospital information system interfacing to generate visualized risk probability outputs. Predictive performance was compared between the obstetric intelligent assistant and four conventional prediction tools (Chinese Labor Room Traffic Light System; Association of Women's Health, Obstetric and Neonatal Nurses; American College of Obstetrics and Gynecology Safe Motherhood Initiative; and California Maternal Quality Care Collaborative prediction tools) using receiver operating characteristic curve.Results:(1) For antenatal prediction, the obstetric intelligent assistant achieved an area under the curve of 0.826 (95% CI: 0.774-0.838), with sensitivity of 0.794 and specificity of 0.712, while the four conventional prediction tools showed area under the curve ranging from 0.569 to 0.586. (2) For intrapartum prediction, the obstetric intelligent assistant achieved an area under the curve of 0.786 (95% CI: 0.751-0.820), with sensitivity of 0.837 and specificity of 0.762, whereas the conventional tools showed area under the curve between 0.600 and 0.613. Conclusion:The obstetric intelligent assistant demonstrates superior performance in predicting PPH compared to conventional prediction tools.
