基于支持向量机的股骨骨折患者术中输血预测模型分析  

Analysis of intraoperative blood transfusion prediction model for femoral fracture patients based on support vector machine

作　　者：杜垚强 杨叶晓青 蒋璐茜[1] 王震[1] 陈钦宏 陈秉宇[1] DU Yao-qiang;YANG Ye-xiao-qing;JIANG Lu-xi;WANG Zhen;CHEN Qin-hong;CHEN Bing-yu(Department of Blood Transfusion,Zhejiang Provincial People’s Hospital,People’s Hospital Affiliated to Hangzhou Medical College,Hangzhou,Zhejiang 310014,China)

机构地区：[1]浙江省人民医院杭州医学院附属人民医院输血科,浙江杭州310014 [2]温州医科大学检验医学院、生命科学学院,浙江温州325035

出　　处：《中国卫生检验杂志》2021年第5期513-517,共5页Chinese Journal of Health Laboratory Technology

基　　金：教育部“云数融合科教创新”基金(2017A11036);浙江省自然科学基金(LQ21H200007);浙江省人民医院优秀科研启动基金(ZRY2019C008)。

摘　　要：目的研究股骨骨折手术患者术中输血的有效影响因素,进一步构建输血预测模型并评估分析。方法收集到231例股骨骨折手术患者病历信息,根据术中输注悬浮红细胞分组,对患者各项指标采用单因素变量分析筛选,然后采用支持向量机(SVM)算法构建输血预测模型,并与传统的Logistic回归方法评估比较。结果单因素分析结果显示,麻醉方式、术前是否输注红细胞、收缩压、红细胞计数、血红蛋白计数、红细胞压积、总蛋白含量、白蛋白含量和球蛋白含量9个变量与股骨骨折患者术中输注悬浮红细胞显著关联(P<0.05)。采用SVM算法构建模型并优化参数可使预测准确率达95.0%,而采用传统的Logistic模型预测术中输血准确率仅为90.0%,评估方法也显示SVM模型更优。结论模型涉及的9个临床变量对于预测骨折患者术中输血具有重要借鉴意义,而SV M等智能型算法推广应用于指导临床输血实践,能更有效地调配血液资源,实现合理备血。Objective To study the effective influencing factors of intraoperative blood transfusion in patients with femoral fracture surgery,and further construct a transfusion prediction model and evaluate the analysis.Methods The medical records of 231 patients with femoral fracture surgery were collected and grouped according to the suspended red blood cells transfused during the operation,and univariate analysis was used to screen the patients,various indicators.The Support vector machine(SVM)algorithm was used to construct a blood transfusion prediction model,and compared to the Logistic regression model.Results The results of univariate analysis showed that there were 9 significant variables including method of anesthesia,preoperative red blood cells transfusion,systolic blood pressure,red blood cells count,hemoglobin count,hematocrit,total protein content,albumin content,and globulin content.They were significantly correlated to suspended red blood cells transfusion(P<0.05).Using SVM algorithm to construct the model and optimize the parameters can make the prediction accuracy rate reach 95.0%,while using the Logistic model to predict the accuracy of intraoperative blood transfusion is only 90.0%,the evaluation methods also indicated that SVM model was better.Conclusion The 9 clinical variables involved in the model are of great significance for predicting intraoperative blood transfusion in fracture patients,and intelligent algorithms such as SVM are widely used to guide clinical transfusion practice,which can more effectively allocate blood resources and achieve reasonable blood preparation.

关 键 词：股骨骨折 术中输血 单因素分析 支持向量机 LOGISTIC回归 

分 类 号：R446.11[医药卫生—诊断学]