首页|期刊导航|兰州大学学报(医学版)|早期先兆流产孕妇早期流产风险的预测模型构建与验证

早期先兆流产孕妇早期流产风险的预测模型构建与验证OA

Risk factors and prediction model for pregnancy outcomes in women with early threatened miscarriage

中文摘要英文摘要

目的 探讨早期先兆流产孕妇发生早期流产的独立预测因素,构建列线图预测模型,为临床早期识别高危患者提供量化工具.方法 回顾性纳入中山大学孙逸仙纪念医院妇产科2018年1月—2022年1月单胎早期先兆流产孕妇,妊娠结局以妊娠12周为界分为自然流产组和继续妊娠组,采用单因素和多因素Logistic回归筛选独立预测因素并构建预测模型,结果以调整后比值比(OR)和95%CI评价.采用受试者操作特征(ROC)曲线、校准曲线及决策曲线(DCA)评估模型性能,采用5折交叉验证法对模型进行内部验证.结果 纳入208例先兆流产患者,其中自然流产组为39例(18.8%)、继续妊娠组为169例(81.2%),208例患者中83例合并绒毛膜下血肿(39.9%).总人群多因素Logistic回归分析最终纳入出现症状就诊时孕龄(d)(OR=0.958,95%CI:[0.928,0.989],P=0.000 9)、慢性高血压(OR=7.807,95%CI:[0.867,73.785],P=0.073)及孕期监测曾有胎心(OR=0.006,95%CI:[0.000 313,0.119],P<0.001)3个变量构建预测模型.该模型曲线下面积(AUC)为0.808,最佳截断值0.144(敏感度82.1%,特异度68.0%),校准良好(Brier=0.097),DCA分析显示其具有临床净收益.五折交叉验证平均AUC为(0.806±0.109),提示模型稳定性良好.绒毛膜下血肿亚组多因素回归分析最终纳入出现症状就诊时孕龄(d)(OR=0.913,95%CI:[0.841,0.991],P=0.030)和孕期监测曾有胎心(OR=0.035,95%CI:[0.001,1.200],P=0.063)2个变量构建预测模型.该模型AUC为0.821,最佳截断值0.188(敏感度72.7%,特异度80.6%),校准良好(Brier=0.078).DCA分析显示具有临床净收益.五折交叉验证平均AUC为(0.828±0.101).结论 出现症状就诊时孕龄(d)小及超声监测无胎心搏动是早期先兆流产孕妇发生早期流产的独立风险因素.所构建的两个列线图模型区分度及校准度良好,可为临床提供直观的个体化风险评分工具,但仍需多中心研究并进行外部验证.

Objective To investigate the independent predictors of early miscarriage in women with early threatened abortion and to construct a nomogram-based prediction model,providing a quantitative tool for ear-ly identification of high-risk patients in clinical practice.Methods Singleton pregnant women with early threatened abortion who were admitted to the Department of Obstetrics and Gynecology,Sun Yat-sen Memo-ri-al Hospital,Sun Yat-sen University between January 2018 and January 2022 were retrospectively enrolled.Pregnancy outcomes were divided into a spontaneous miscarriage group and a continuing pregnancy group using 12 weeks of gestation as the cutoff.Univariate and multivariate Logistic regression were used to identify independent predictors and construct prediction models.Results were reported as adjusted OR with 95%CI.Model performance was assessed using receiver operator characteristic(ROC)curves,calibration curves,and decision curve analysis(DCA).Five-fold cross-validation was performed for internal validation.Results A total of 208 patients with threatened miscarriage were enrolled,including 39 cases(18.8%)in the spontaneous miscarriage group and 169 cases(81.2%)in the continuing pregnancy group.Among them,83 cases(39.9%)were complicated with subchorionic hematoma.In the overall population,multivariate Logistic regression analysis ultimately identified three variables for model construction:gestational days at presentation(OR=0.958,95%CI:[0.928,0.989],P=0.000 9),chronic hypertension(OR=7.807,95%CI:[0.867,73.785],P=0.073),and presence of fetal heartbeat during pregnancy(OR=0.006,95%CI:[0.000 313,0.119],P<0.001).The model achieved an area under the curve(AUC)of 0.808,with an optimal cutoff of 0.144(sensitivity 82.1%,specificity 68.0%),and good calibration(Brier=0.097),with a positive net clinical benefit on DCA.The five-fold cross-validation yielded a mean AUC of(0.806±0.109),indicating satisfactory model stability.In the subchorionic hematoma subgroup,multivariate regression analysis ultimately identified two variables for model construction:gestational days at presentation(OR=0.913,95%CI:[0.841,0.991],P=0.030)and presence of fetal heartbeat during pregnancy(OR=0.035,95%CI:[0.001,1.200],P=0.063).The model achieved an AUC of 0.821,with an optimal cutoff of 0.188(sensitivity 72.7%,specificity 80.6%),and good calibration(Brier=0.078),with a positive net clinical benefit on DCA.The five-fold cross-validation yielded a mean AUC of(0.828±0.101).Conclusions Shorter gestational days at presentation and absence of fetal heartbeat on ultrasound are independent risk factors for early miscarriage in women with early threatened abortion.The two nomogram-based prediction models demonstrate good discrimination and calibration,pro-viding practical and intuitive tools for individualized risk assessment in clinical practice,though external validation in multicenter studies is warranted.

郑琤;冀晓慧;汤景;区绮云;陈慧

中山大学孙逸仙纪念医院,超声科,广东 广州 510120中山大学孙逸仙纪念医院,妇科生殖内分泌专科,广东 广州 510120中山大学孙逸仙纪念医院,妇科生殖内分泌专科,广东 广州 510120中山大学孙逸仙纪念医院,超声科,广东 广州 510120中山大学孙逸仙纪念医院,妇科生殖内分泌专科,广东 广州 510120

医药卫生

早期先兆流产绒毛膜下血肿Logistic回归列线图风险预测模型胎心孕龄高血压

early threatened abortionsubchorionic hematomaLogistic regressionnomogrampredictive fac-torsfetal heart rategestational agehypertension

《兰州大学学报(医学版)》 2026 (7)

30-38,9

国家中医药管理局综合司、国家卫生健康委办公厅、中央军委后勤保障部卫生局,重大疑难疾病中西医临床协作—先兆流产项目(ZDYN-2024-A-073)

10.13885/j.issn.2097-681X.T20260064

评论