基于机器学习算法构建中老年腹膜透析患者肌少症发生的预测模型OA北大核心CSTPCD
A predictive model of sarcopenia occurrence in middle-aged and elderly peritoneal dialysis patients based on machine learning algorithm
目的 基于机器学习算法构建中老年腹膜透析(peritoneal dialysis,PD)患者肌少症发生的预测模型,为肌少症的诊治测量提供参考依据.方法 选取 2020 年 5 月至 2023 年 10 月期间于海军军医大学第一附属医院(上海长海医院)住院治疗的 648 例中老年PD患者作为研究对象,收集患者临床资料,根据是否发生肌少症分为肌少症组 169 例和无肌少症组 479 例.基于机器学习算法,分别采用极限梯度提升树(extreme gradient boosting,XGBoost)和 Logistic回归、随机森林(random forest,RF)、支持向量机(support vector machine,SVM)方法构建不同PD患者发生肌少症风险模型,并对模型进行评价比较.结果 共采集患者 26 项指标,单因素与 Logistic回归筛选出 9 项肌少症的影响因素.测试集上验证后的 XGBoost、Logistic、RF、SVM模型构建的预测中老年PD患者发生肌少症的曲线下面积(area under the curve,AUC)分别为 0.807、0.788、0.804、0.791,准确度分别为 0.829、0.813、0.855、0.819,F1 分数为 0.733、0.659、0.728、0.653.结论 基于机器学习算法 XGBoost预测模型在敏感性与准确度上优于 RF、Logistic、SVM模型,有助于指导临床医务人员识别PD发生肌少症高风险患者,有利于临床尽早制定干预策略.
Objective To construct a prediction model of sarcopenia in middle-aged and elderly peritoneal dialysis(PD)patients based on machine learning algorithm,and to provide a reference for the diagnosis and treatment of sarcopenia.Methods A total of 648 middle-aged and elderly PD patients who were hospitalized between May 2020 and October 2023 were selected as the study subjects.Clinical data of the patients were collected,and they were divided into a sarcopenia group of 169 cases and a non sarcopenia group of 479 cases based on the occurrence of sarcopenia.Based on machine learning algorithms,extreme gradient boosting(XGBoost)and logistic regression,random forest(RF),and support vector machine(SVM)methods were used to construct risk models for sarcopenia in different PD patients,and the models were evaluated and compared.Results A total of 26 indicators were collected,and 9 risk factors of sarcopenia were screened out by single factor and Logistic regression.After verification on the test set,the area under the curve(AUC)constructed by XGBoost,Logistic,RF and SVM models to predict the occurrence of sarcopenia in middle-aged and elderly PD patients was 0.807,0.788,0.804 and 0.791 respectively,and the accuracy was 0.829,0.813,0.855 and 0.819,respectively.F1 scores were 0.733,0.659,0.728 and 0.653.Conclusions The prediction model based on machine learning algorithm XGBoost is superior to RF,Logistic and SVM models in sensitivity and accuracy.It is helpful to guide clinical staff to identify PD patients at high risk of developing sarcopenia and to facilitate intervention strategies as early as possible.
季亚平;王璇;沈玢;王登台
海军军医大学第一附属医院 (上海长海医院) 肾病内科,上海 200433海军军医大学第一附属医院 (上海长海医院) 急诊科,上海 200433
腹膜透析肌少症预测模型机器学习算法
Peritoneal dialysisSarcopeniaPrediction modelMachine learning algorithm
《中国医学前沿杂志（电子版）》 2024 (006)
31-37 / 7
评论