基于"治未病"理论与 XGBoost算法的膝骨关节炎患者下肢深静脉血栓风险预测模型的构建与验证OA

Objective:Based on the theory of"eliminating impending risks and treating diseases before they manifest"(a core concept of preventive healthcare in traditional Chinese medicine),this study aims to establish a risk prediction model for lower extremity deep vein thrombosis(DVT)in patients with knee osteoarthritis(KOA),thereby providing a quanti-tative tool for early prevention and intervention.Methods:A retrospective analysis was conducted on the clinical data of 883 hospitalized KOA patients treated from March 2018 to March 2023.All patients were divided into the DVT group and non-DVT group based on the results of lower extremity deep vein color Doppler ultrasound examination.A total of 25 fea-tures,including routine blood test indicators,coagulation function indicators,and biochemical indicators,were extracted.SPSS 25.0 software was used for univariate analysis to screen out factors with significant differences.Subsequently,the XGBoost algorithm was adopted to construct the prediction model.The 883 patients were randomly divided into a training set(n=618)and a validation set(n=265)at a ratio of 7∶3.Recursive feature elimination(RFE)was used to screen key features and explore the risk factors of lower extremity DVT in KOA patients.Results:The incidence of lower extremity DVT in KOA patients was 8.61%.For the XGBoost model,the area under the curve(AUC)in the validation set was 0.804,with a specificity of 0.877 and a sensitivity of 0.643.Fourteen key features were screened out,among which prothrombin time(PT),hematocrit(HCT),and neu-trophil percentage(NEU%)were the core risk factors.The prediction model based on the XGBoost algorithm could effec-tively identify the DVT risk in KOA patients.Conclusion:This study can effectively predict the risk factors of DVT in KOA patients and provide a reference for the early prevention of thrombosis.