目的 基于机器学习分析产妇自然阴道分娩后创伤后应激障碍(posttraumatic stress disorder，PTSD)的危险因素并构建预测模型，探讨预测模型的预测价值。方法 采用便利抽样法选取2023年12月至2024年8月安徽省某院的产妇400例为研究对象，根据其是否发生PTSD分为未发生PTSD组（n=346）和发生PTSD组（n=54）。基于随机森林（random forest,RF）、支持向量机（support vector machine,SVM）、K-最近邻（K-nearest neighbor,KNN）、逻辑回归（Logistic regression,LR）4个机器学习模型构建产妇分娩后PTSD预测模型。采用准确度、精确度、接受者操作特性曲线（receiver operating characteristic curve，ROC）曲线下面积 (area under the curve，AUC) 等评价模型的稳健型及临床实用性。另纳入同期安徽省另一所医院的200例产妇进行外部验证。结果 产妇产后PTSD的发生率为13.50%。年龄、喂养方式、计划怀孕、母婴皮肤接触、急产、妊娠糖尿病等均是产后是否发生PTSD的独立影响因素（均P＜0.05）。RF模型对产妇PTSD的发生率的预测效果最好(AUC=0.882)、综合预测效能最佳。在外部验证中，RF、SVM、KNN和LR模型的AUC值分别为0.842、0.734、0.725、0.767。结论 基于RF算法构建的预测模型具有较优质的预测性能，可以更加简单、快速、有效地识别产后PTSD。

Objective To analyze the risk factors of posttraumatic stress disorder (PTSD) after natural vaginal delivery in parturients and develop a predictive model based on machine learning，and to explore its predictive value.Methods A total of 400 parturients from a hospital in Anhui Province from December 2023 to August 2024 were selected by the convenience sampling method，and were divided into the group without PTSD (n=346) and the group with PTSD (n=54) according to the occurrence of PTSD.Based on 4 machine learning models，namely random forest (RF)，support vector machine (SVM)，K-nearest neighbor (KNN)，and Logistic regression(LR)，a post-delivery PTSD prediction model was developed.The robustness and clinical practicability of the model were evaluated by using accuracy，precision，area under the curve (AUC) of the receiver operating characteristic curve (ROC)，etc.Another 200 parturients from another hospital during the same period were included for external validation.Results The incidence of PTSD in parturients was 13.50%.Age，feeding method，planned pregnancy，skin-to-skin contact between mother and baby，acute labor，gestational diabetes，etc.were all independent influencing factors for the occurrence of PTSD after delivery (all P＜0.05).The RF model has the best predictive effect on the incidence of PTSD in parturients (AUC=0.882) and the best comprehensive predictive efficacy.In external validation，the AUC values of the RF，SVM，KNN and LR models were 0.842，0.734，0.725 and 0.767，respectively.Conclusions The prediction model developed based on the RF algorithm has superior prediction performance and can identify postpartum PTSD more simply，quickly and effectively.

R473.71

安徽省妇幼健康科研课题(2020FY14)