基于机器学习构建下肢骨折患者术后日常生活活动能力缺损的预测模型

打开文本图片集

中图分类号R493R683文献标识码A 文章编号2096-7721（2026）01-0159-08

AbstractObjective：Toconstruct predictionmodelsfopostoperativeimpaimentinactivitiesofdailyliving（ADL）inpatints with lowerlimbfractureusing five machine learningalgorithms，chosethebest performed modelandconduct interpretabilityanalysis. Methods：232patientswithlowerlimbfractureswhowere treatedatourhospitalfromJanuary2O21toJuly2O24 wereenroled.ADL impairment wasaessdat3monhsaftersurgeryusingteDLscale.Lassoegresonwasusdforvarableseectionndlected variableswerdtosctListiceoost tctoce（dBoosicio models.Modelpefomaneasompaedtodentifyeoptialodel，flowedyiterpretablityalysis.Results：Lasoio identfidneeyariables：hprtension，sleepsatus，jointijuryfracturetyeSAgrade，postoperativecogitiveiit complicatiosopieeitdtlaioistcodo modelswerecotuctedsdosevbls.IhodeligsetthrUCsre.7，92.8，.86dO8spciely validationset，theirUsere0.740.62，0.66，.77nd.63respectively.TeSVmodelhadtebestpeforanceinthealidation set.Conclusion：TheSVMmodelshowedsuperiorpredictiveperformanceforADLimpairmentinpatientsunderwentlowerlimfracture surgeryprovgsisarlydfafgsolat.epealitalyisaifdosedi of each variable on the outcome，laying a foundation for clinical application of the model.

Key WordsLower Limb Fracture;Activities ofDailyLiving; Machine Learning; Prediction; Interpretability Analysis

随着我国老龄化进程加剧，下肢骨折发生率逐年增加，且呈现复杂化趋势[]。（剩余8495字）