应用分类与回归树筛选慢性非传染性疾病健康素养影响因素的研究  

Study on health literacy factors of chronic non-communicable diseases by using classification and regression trees

作　　者：崔梦晶[1] 郭海健[2] 李洋[1] 曲晨[2] 陈宇[1] 胡洁[1] 庞媛媛[1] 李小宁[2] 张徐军[1] 

机构地区：[1]东南大学公共卫生学院,江苏南京210009 [2]江苏省疾病预防控制中心健康教育所,江苏南京210009

出　　处：《东南大学学报（医学版）》2016年第5期704-708,共5页Journal of Southeast University(Medical Science Edition)

基　　金：全国健康素养现状评估与结果推广项目(1311206600002)

摘　　要：目的:探讨分类与回归树(C&RT)在筛选慢性非传染性疾病健康素养相关影响因素中的应用。方法:收集2013年宿迁市105例慢性病患者组成病例组,对地区、年龄、性别进行成组匹配选取210例非慢性病患者组成对照组。结果:分类树模型从纳入慢性病健康素养相关的19个变量中筛选出就医行为素养、运动素养、对健康的理解素养、心理调节素养、家庭年收入、BMI值是否正常以及成瘾行为素养等7个有统计学意义的影响因素,并且说明了不同人群各自的影响因素。模型的错分概率Risk值为0.270,ROC下曲线面积为0.763,模型拟合效果较好。结论:应用分类与回归树能较好地筛选出慢性病健康素养影响因素,同时能显示变量之间的相互作用,还可以研究变量科学定义分界点。Objective： To explore CRT methods in screening health literacy factors of chronic noncommunicable diseases. Methods： Database was created from Suqian in 2013. Case group composed by 105 patients with chronic diseases. Control group composed of 210 staff without chronic diseases matched by region,age and sex. Results： Seven out of 19 affecting factors were selected,which were willingness to see a doctor,exercise,understanding of health,psychological adjustment,annual family income,BMI index and addictive behaviors.Influence factors of different groups were also explained. The Risk value of model error probability was 0. 270,and the area under the ROC curve was 0. 763,suggesting that the classification tree model fit the actuality well.Conclusion： The classification tree model can screen out the major affecting factors quickly and effectively and could also identify the cutting- points for continuous and ordinal variables,as well as revealing the complex interaction among the factors at many levels.

关 键 词：分类树 分类与回归树 慢性非传染性疾病 健康素养 

分 类 号：R195[医药卫生—卫生统计学]