产后出血输血的影响因素分析及预测模型构建

Construction of a prediction model for blood transfusion in postpartum hemorrhage

目的探讨影响产后出血输血的危险因素并构建相关预测模型。方法选取2015年6月至2022年6月湖州市妇幼保健院产后出血患者322例为研究对象,其中产后出血有输血者89例(输血组),产后出血未输血者233例(未输血组)。比较两组患者年龄、孕周、子宫手术史、前置胎盘、胎盘植入、凝血相关疾病等临床指标的差异,采用多因素logistic回归分析影响产后出血输血的独立危险因素,根据各危险因数的回归系数权重构建预测模型,采用多因素ROC曲线评估各独立危险因素及联合预测因子预测产后出血输血的效能。结果单因素分析发现,患者年龄、子宫手术史、前置胎盘、胎盘植入、凝血相关疾病发生率与产后出血输血均有关(均P<0.05)。多因素logistic回归分析显示前置胎盘、胎盘植入、子宫手术史、凝血相关疾病均是影响产后出血输血的独立危险因素(均P<0.05)。联合预测因子=2.233×子宫手术史+0.961×前置胎盘+2.045×胎盘植入+2.428×凝血相关疾病,联合预测因子的灵敏度为0.730,特异度为0.888,AUC=0.855,Hosmer-Lemeshow拟合优度检验显示,χ^(2)=13.675,P>0.05。联合预测因子的AUC均大于各独立危险因素(均P<0.01)。结论多因素logistic回归模型构建联合预测因子可以对各因素直接进行量化分析,且有较好的预测效能。

Objective To construct a prediction model of blood transfusion for postpartum hemorrhage.Methods A total of 322 patients with postpartum hemorrhage admitted in Huzhou Maternal&Child Health Care Hospital from June 2015 to June 2022 were enrolled,including 89 cases with blood transfusion(blood transfusion group)and 233 cases without blood transfusion(non-transfusion group).The age,gestational age,uterine surgery history,placenta previa,placenta accreta,and coagulation-related diseases were analyzed with logistic regression and the independent risk factors affecting postpartum hemorrhage and blood transfusion were determined.A prediction model was developed based on the risk factors,the efficacy of the model in predicting postpartum hemorrhage blood transfusion was evaluated with receiver operating characteristic curve(ROC).Results Univariate analysis showed that patients’age,history of uterine surgery,placenta previa,placenta accreta,and coagulation-related diseases were all related to postpartum hemorrhage(all P<0.05).Multivariate logistic regression analysis showed that placenta previa,placenta accreta,history of uterine surgery,and coagulation-related diseases were independent risk factors for postpartum blood transfusion.The prediction model was constructed with the combination of risk factors as follows:2.233×uterine surgery history+0.961×placenta previa+2.045×placenta accreta+2.428×coagulationrelated diseases.The area under the ROC curve(AUC)of the model in predicting postpartum blood transfusion was 0.855,the sensitivity and specificity was 0.730 and 0.888,respectively.Hosmer-Lemeshow goodness of fit test wasχ^(2)=13.675,P>0.05.The AUC of the model was greater than that of the single risk factors(all P<0.01).Conclusion In this study a logistic regression model has been developed,which has good predictive performance for postpartum blood transfusion.