基于Logistic回归构建急性白血病患者红细胞输注效果不良的风险预测模型

Construction of Risk Prediction Model for Poor Erythrocytes Transfusion Effect in Patients with Acute Leukemia Based on Logistic Regression

目的探讨Logistic回归构建急性白血病患者红细胞输注效果不良的风险预测模型。方法回顾性选取急性白血病患者200例,根据患者红细胞输注后效果分组为效果良好组(n=158)和效果不良组(n=42)。多因素Logistic回归模型分析急性白血病红细胞输注效果的危险,受试者工作曲线(ROC)检测模型的预测效能。结果急性白血病患者输注效果不良与输血史、妊娠史、合并感染、合并恶性肿瘤和脾肿大有关(P<0.05),性别、年龄、疾病种类与Hb无显著影响(P>0.05)。Logistic回归分析显示,输血史≥4次(OR=2.956,P<0.05)、合并感染(OR=2.683,P<0.05)和脾肿大(OR=2.643,P<0.05)均为影响急性白血病患者红细胞输注效果不良的独立危险因素。风险预测模型概率P=1/[1+e^((-7.638+1.084×(输血史≥4次)+0.987×(合并感染)+0.972×(脾肿大)))],Hosmer-Lemeshowχ^(2)=4.187,P=0.651。ROC分析显示,Logistic回归模型预测输注效果不良的AUC为0.605,95%CI为0.514~0.697。结论输血史≥4次、合并感染和脾肿大均为影响急性白血病患者红细胞输注效果不良的独立危险因素,临床可采用Logistic回归模型分析各变量与红细胞输注效果相关性并予以早期干预。

Objective To explore the construction of risk prediction model for poor erythrocytes transfusion effect in patients with acute leukemia based on Logistic regression.Methods 200 patients with acute leukemia were retrospectively selected and divided into good effect group(n=158)and bad effect group(n=42)according to erythrocytes transfusion effect.The risk of erythrocytes transfusion effect were analyzed by multivariate Logistic regression model.And its predictive efficiency was tested by receiver operating characteristic(ROC)curves.Results Poor erythrocytes transfusion effect was related to histories of blood transfusion and pregnancy,infection,malignant tumors and splenomegaly(P<0.05),while not significantly related to gender,age,disease types or Hb(P>0.05).Logistic regression analysis showed that blood transfusion≥4 times(OR=2.956,P<0.05),infection(OR=2.683,P<0.05)and splenomegaly(OR=2.643,P<0.05)were all independent risk factors of poor e-rythrocytes transfusion effect.The probability by risk prediction model was as follow:P=1/[1+e^((-7.638+1.084×(transfusion times≥4 times)+0.987×(infection)+0.972×(splenomegaly)))],Hosmer-Lemeshowχ^(2)=4.187,P=0.651.ROC analysis showed that AUC of Logistic regression model for predicting poor erythrocytes transfusion effect was 0.605(95%CI:0.514-0.697).Conclusion Blood transfusion times≥4 times,infection and splenomegaly are all independent risk factors of poor erythrocytes transfusion effect in patients with acute leukemia.Clinically,Logistic regression model can be applied to analyze the correlation between each variable and erythrocytes transfusion effect so as to provide early intervention.