机器人外科学杂志（中英文）2025，Vol.6Issue(6)：1019-1024,6.DOI:10.12180/j.issn.2096-7721.2025.06.026
老年髋部骨折患者术前深静脉血栓形成的机器学习预测模型构建
Development of a machine learning prediction model for preoperative deep vein thrombosis in elderly hip fracture patients
摘要
Abstract
Objective:To identify risk variables for preoperative deep vein thrombosis(DVT)formation in elderly patients with hip fractures using machine learning algorithms and to construct a nomogram model for optimizing preoperative assessment and preventive measures.Methods:174 elderly patients with hip fractures from June 2023 to December 2024 were enrolled and divided into the thrombus group(n=57)and the non-thrombus group(n=117)based on preoperative lower extremity venous ultrasonography.Two machine learning algorithms,Least Absolute Shrinkage and Selection Operator(LASSO),and Support Vector Machine-Recursive Feature Elimination(SVM-RFE),as well as multivariate Logistic regression,were used to screen preoperative risk variables.The model with the highest AUC was selected to construct the nomogram,followed by validation.Results:Multivariate Logistic regression identified time from injury to hospital admission,platelet count(PLT),neutrophil count(NC),D-dimer,fibrinogen(FIB),platelet-to-lymphocyte ratio(PLR),and systemic immune-inflammation index(SII)as independent risk factors for preoperative DVT.The LASSO model generated nine features with non-zero coefficients,including time from injury to hospital admission,WBC,PLT,NC,D-dimer,FIB,monocyte-to-lymphocyte ration(MLR),PLR,and SII.The SVM-RFE algorithm identified five risk variables,including time from injury to hospital admission,PLT,D-dimer,FIB,and SII.ROC analysis showed that the AUC for LASSO,SVM-RFE,and Logistic regression were 0.869,0.978,and 0.933,respectively.The SVM-RFE model had the highest sensitivity(98.18%)and specificity(90.77%).The SVM-RFE model was selected to construct the nomogram,and the Hosmer-Lemeshow goodness-of-fit test demonstrated good calibration ability of the nomogram(χ 2=2.157,P=0.867).Conclusion:The constructed nomogram model can assist healthcare providers in accurately assessing the preoperative DVT risk in elderly patients with hip fractures,thereby enabling timely preventive measures,optimizing the allocation of medical resources,and improving the quality of care.关键词
髋部骨折/深静脉血栓/机器学习/列线图模型/风险预测Key words
Hip Fracture/Deep Vein Thrombosis/Machine Learning/Nomogram Model/Risk Prediction分类
临床医学引用本文复制引用
黄玉凤,仇燕冉..老年髋部骨折患者术前深静脉血栓形成的机器学习预测模型构建[J].机器人外科学杂志（中英文）,2025,6(6):1019-1024,6.基金项目
合肥市卫生健康应用医学研究项目(Hwk2022zd010)Hefei Municipal Health Applied Medical Research Project(Hwk2022zd010) （Hwk2022zd010）
