基于体检数据的肌少症患病风险预测

Risk Prediction for Sarcopenia Based on Physical Examination Data

利用体检数据预测肌少症的患病风险,探索预测肌少症的重要因素,以实现肌少症的早期诊断.基于2019年上海某医院的5641名人员的体检数据,利用Wilcoxon秩和检验和卡方检验找出差异显著的特征,使用8种机器学习方法对是否患有肌少症进行预测.采用受试者操作曲线下面积(AUC)评价模型预测效果,建立较优的肌少症预测模型,并利用特征评分寻找预测肌少症的重要因素.LightGBM(Light-gradient boosting machine)、随机森林和逻辑回归预测效果较优,测试集AUC值达到0.93以上.模型确定了年龄、体质量、身高、身体质量指数(body mass index,BMI)、腰围、臀围、舒张压以及平均红细胞血红蛋白量、高密度脂蛋白、平均红细胞体积、红细胞、甘油三酯是预测肌少症的重要因素,体格检查、血检指标、血常规、肝肾功能、生活习惯和一般信息是预测肌少症重要体检项目.文章建立了有效的肌少症患病风险预测模型,确定了预测肌少症的重要因素和体检项目,在一定程度上有助于肌少症患者的管理.

In this paper,the physical examination data is adopted to predict the risk of sarcopenia,explore the influencing factors for sarcopenia prediction,and realize the early diagnosis of sarcopenia.Basing on the medical examination data of 5641 medical examiners in a hospital in Shanghai in 2019,Wilcoxon rank sum test and chi-square test are used to find significant differences.And eight machine learning methods are proposed to predict whether they have sarcopenia.The area under curve(AUC)is used to evaluate the performance of the prediction models.Furthermore,a better prediction model for sarcopenia is proposed,and the rank of feature scores is used to find important influencing factors for sarcopenia prediction.LightGBM(Light-gradient boosting machine),random forest and logistic regression have better prediction performance,and the AUC values on the test set are above 0.93.The models indicate that age,weight,height,BMI,waist circumference,hip circumference,diastolic blood pressure,and mean corpuscular hemoglobin,high-density lipoprotein,mean corpuscular volume,red blood cells,and triglycerides are important factors for sarcopenia prediction.Physical examination,blood test index,blood routine,liver and kidney function,living habits and general information are important physical examination items for sarcopenia prediction.In summary,this study establishes a predictive model for the risk of sarcopenia,and determines the important factors and physical examination items for sarcopenia prediction,which may help to the management of the patients with sarcopenia.