基于机器学习的老年人高脂血症风险预测  

Risk prediction of hyperlipidemia in the elderly based on machine learning

作　　者：王俊 王培承[1] 郭琼 袁炳鑫 WANG Jun;WANG Pei-cheng;GUO Qiong;YUAN Bing-xing(School of Public Health,Weifang Medical University,Weifang,Shandong 261053,China)

机构地区：[1]潍坊医学院公共卫生学院,山东潍坊261053

出　　处：《现代预防医学》2023年第18期3408-3413,共6页Modern Preventive Medicine

基　　金：国家自然科学基金资助项目(71673202)。

摘　　要：目的探究机器学习方法对老年人高脂血症的预测价值,并从中分析老年人高脂血症的危险因素.方法采用分层随机整群抽样方法抽取潍坊市9个县市区27个社区5759位60岁及以上老年人,并收集这些老年人在2020年度的健康体检资料,排除缺失或异常资料,最终将4534位老年人纳入研究.将这些老年人按照7∶3的比例随机划分为训练集和验证集,并采用单因素分析筛选出10个显著性变量,用于构建支持向量机、决策树、XGBoost、CATBoost和LightGBM风险预测模型,随后采用AUC(ROC曲线下面积)、精确率、准确率、召回率、F1值评价其性能.结果CATBoost机器学习模型的综合性能最好,其AUC、精确率、准确率、召回率、F1值分别为0.82、76.49%、92.46%、78.68%、0.85.在CATBoost模型预测老年人高脂血症的风险过程中,9个变量对高脂血症风险预测较为重要,其重要程度由高到底排序依次为收缩压、腰围、空腹血糖、体质指数、舒张压、吸烟情况、年龄、饮酒情况和性别.而且经模型优化后发现,仅纳入这9个变量即可较好的预测风险.结论基于CATBoost机器学习方法构建的老年人高脂血症风险预测模型具有较好的预测效果,可为老年人高脂血症的防治工作提供参考.Objective To explore the predictive value of machine learning methods for hyperlipidemia in the elderly,and analyze the risk factors for hyperlipidemia in the elderly.Methods The stratified random cluster sampling method was used to select 5759 elderly people aged 60 and above in 27 communities in 9 counties and municipalities of Weifang City,and the health examination data of these elderly people in 2020 was collected.Missing or unusual data were excluded,and 4534 older adults were eventually included in the study.Ten important risk factors were screened out by univariate analysis,which was used to construct support vector machines,decision trees,XGBoost,CATBoost and LightGBM risk prediction models,and AUC(Area under ROC curve),accuracy,accuracy,recall,and F1 value were used to evaluate its performance.Results The comprehensive performance of the CATBoost machine learning model was the best and its AUC,precision,accuracy,recall and F1 values were 0.82,76.49%,92.46%,78.68%,and 0.85,respectively.In the process of CATBoost model predicting the risk of hyperlipidemia in the elderly,nine variables were more important for the prediction of hyperlipidemia risk,and their importance was in the order of high to bottom:systolic blood pressure,waist circumference,fasting blood glucose,body mass index,diastolic blood pressure,smoking,age,alcohol consumption,and gender.Moreover,after model optimization,it was found that only including these nine factors could better predict risks.Conclusion The risk prediction model of hyperlipidemia in the elderly based on the CATBoost machine learning method has a good prediction effect,which can provide a reference for the prevention and treatment of hyperlipidemia in the elderly.

关 键 词：高脂血症 老年人 机器学习 危险因素 风险预测 

分 类 号：R589.2[医药卫生—内分泌]