基于人工智能机器学习决策树算法的上海某社区中老年群体衰弱表型分析

Analysis of frailty phenotype of a middle-aged and elderly population in a community in Shanghai based on artificial intelli-gence machine learning decision tree algorithm

中文关键词: 中老年  衰弱  社区  人工智能  机器学习  决策树算法  重要度分析

英文关键词: middle-aged and elderly  frailty  community  artificial intelligence  machine learning  decision tree algo-rithm  importance analysis

中文摘要:

      目的 旨在利用人工智能技术分析上海某社区中老年群体的衰弱表型特征,并评估各特征对衰弱状态的贡献度,以期为老年人衰弱健康管理提供量化依据.方法 研究选取2023年7月30日-9月30日在上海市某社区卫生服务中心就诊的55岁及以上人群为研究对象.采用电子化问卷收集研究对象一般健康状况和日常生活状况,采用智能手环测量日常活动数据,收集实验室检查数据,并通过Fried衰弱表型量表进行衰弱评估.利用机器学习决策树算法对数据进行预处理和模型训练,分析各项表型特征对衰弱结果的重要程度.结果 共纳入556例样本,其中8.3％处于衰弱期,46.0％处于衰弱前期,45.7％无衰弱.研究发现,年龄、教育程度、婚姻状况和收入水平等因素均与衰弱的发生有显著关联.智能手环监测睡眠、步数、活动时间等日常活动数据显示衰弱组在这些指标上显著低于非衰弱组.机器学习算法的定量分析结果表明,年龄是对衰弱贡献最大的特征,其次为平均睡眠时间、丙氨酸氨基转移酶(alanine transarninase,ALT)和身体运动情况.结论 本研究通过人工智能技术揭示了影响老年人衰弱状态的多种因素,并识别出年龄、睡眠时间等多种重要度较高的衰弱表型.研究结果有助于提高老年人衰弱管理的效率,并为医疗保健系统资源的合理分配提供了依据.

英文摘要:

      Objective To analyze the frailty phenotype characteristics of the middle-aged and elderly population in a community in Shanghai based on artificial intelligence technology,and evaluate the contribution of each characteristic to frailty status,so as to provide quantitative basis for frailty health management of the elderly.Methods The participants aged 55 and above who visited a community health service center in Shanghai from July 30 to September 30,2023 were selected as the study subjects.Electronic questionnaires were used to collect the general health status and daily life conditions of the study subjects,daily activity data were measured by smart bracelets,and laboratory examination data were collected.Frailty was evaluated by Fried Frailty Phenotype Scale.The machine learning decision tree algorithm was used to preprocess the data and train the model,and the importance of various phenotypic characteristics on the frailty results was analyzed.Results A total of 556 samples were included,of which 8.3％were in the frailty stage,46.0％were in the pre-frailty stage,and 45.7％were without frailty.It was found that those factors such as age,education level,marital status and income level were significantly associated with the occurrence of frailty.The data of daily activities such as sleep,step count,and activity time monitored by smart bracelet showed that these indicators of the frailty group were significantly lower than those of the non-frailty group.Quantitative analy-sis by the machine learning algorithm showed that age was the characteristic that contributed most to frailty,followed by aver-age sleep time,alanine transarninase(ALT)and physical activity.Conclusion This study determines multiple factors that af-fect the frailty state of the elderly based on artificial intelligence technology,and identifies a variety of frailty phenotypes with high importance such as age and sleep time.The results of this study are helpful to improve the efficiency of the management of frailty in the elderly and provide a basis for the rational allocation of health care system resources.