首页|产后输血量的高危因素评估和风险评分模型的建立

产后输血量的高危因素评估和风险评分模型的建立

扫码查看
目的 评估产后输血量相关的高危因素并构建风险评分模型。方法 选取2021年5月31日至2023年5月31日于西北妇女儿童医院进行输血的557例分娩产妇为研究对象,根据输血量分为<800mL(n=345)、≥800mL(n=212)两组,并随机选择389例(70%)作为训练集、168例(30%)作为验证集。根据文献及临床实际收集产妇信息,构建Logistic回归模型和随机森林模型进行变量分析,筛选并验证可能与产妇分娩后出血相关的变量。结果 多因素Logistic回归结果显示,孕期重度贫血、低蛋白血症、凝血功能障碍,以及存在凶险性前置胎盘、胎盘植入为影响孕产妇输血量的危险因素,其OR值及95%CI分别为3。25(1。66~6。36)、2。64(1。60~4。35)、3。12(1。12~5。28)、2。55(1。19~4。63)、1。98(1。02~3。82);训练集及验证集受试者工作特征(ROC)曲线下面积(AUC)分别为0。739、0。694,交叉验证、决策曲线分析(DCA)和临床影响曲线(CIC)均展示了良好的临床指导功能;随机森林模型验证结果显示训练集中输血的危险因素前5位包括贫血、低蛋白血症、凝血功能障碍、凶险性前置胎盘、胎盘植入(验证AUC=0。803),与多因素回归结果一致。结论 贫血、低蛋白血症、凝血功能障碍、凶险性前置胎盘、胎盘植入为产后输血量的相关独立危险因素,可指导临床备血及选择合适的手术方式。
Evaluation of risk factors and establishment of risk score model for postpartum blood transfusion volume
Objective To evaluate the risk factors related to postpartum blood transfusion volume and establish a risk scoring model.Methods A total of 557 postpartum patients who received transfusions at the Northwest Women and Children's Hospital from May 31,2021,to May 31,2023,were selected as study subjects,and were divided into<800mL(n=345)and≥800mL(n=212)groups according to blood transfusion volume.A random sample of 389 cases(70%)was used as the training set,and 168 cases(30%)as the validation set.Maternal information was collected according to literature and clinical practice,Logistic regression model and random forest model were constructed for variable analysis,and variables that might be related to postpartum hemorrhage were screened and verified.Results Multivariate Logistic regression analysis results indicated that severe anemia,hypoproteinemia,coagulopathy,dangerous placenta previa and placenta implantation during pregnancy were the risk factors influencing postpartum transfusion volume,with OR values and 95%CI were 3.25(1.66-6.36),2.64(1.60-4.35),3.12(1.12-5.28),2.55(1.19-4.63),1.98(1.02-3.82),respectively.The area under receiver operating characteristic(ROC)curve(AUC)of training set and validation set were 0.739 and 0.694,respectively.Cross-validation,decision curve analysis(DCA)and clinical impact curve(CIC)all demonstrated good clinical guidance function.The random forest model validation results showed that the top five risk factors for transfusion in the training set included anemia,hypoproteinemia,coagulopathy,placenta previa,and placenta implantation(validation AUC=0.803),consistent with the multivariate regression results.Conclusion Anemia,hypoproteinemia,coagulopathy,dangerous placenta previa and placenta implantation are independent risk factors related to postpartum blood transfusion volume,which can guide clinical blood preparation and selection of appropriate surgical methods.

postpartum hemorrhageblood transfusionmachine learningrisk factorsrandom forest

徐慧双、李敏、韩香、魏旸、屈明利

展开 >

西北妇女儿童医院输血科,陕西 西安 710061

西安市东方医院中医科,陕西 西安 710043

西北妇女儿童医院产科,陕西西安 710061

产后出血 输血 机器学习 危险因素 随机森林

2024

中国妇幼健康研究
西安交通大学,中国疾病控制中心妇幼保健中心

中国妇幼健康研究

CSTPCD
影响因子:0.942
ISSN:1673-5293
年,卷(期):2024.35(12)