基于随机森林算法的中学生焦虑、抑郁影响因素分析及预测模型构建  

Analysis of factors influencing anxiety and depression in secondary school students and construction of predictive model based on random forest algorithm

作　　者：高红霞 陈靖[1,2,3] 张晓红 侯丽明[3] 王丽洁 孟亚清[3] 高雯[3] 黄亚欣[3] GAO Hong-xia;CHEN Jing;ZHANG Xiao-hong;HOU Li-ming;WANG Li-jie;MENG Ya-qing;GAO Wen;HUANG Ya-xin(Academy of Medical Sciences,Shanxi Medical University,Taiyuan,Shanxi 030001,China;不详)

机构地区：[1]山西医科大学医学科学院,山西太原030001 [2]山西医科大学公共卫生学院流行病学教研室 [3]山西省疾病预防控制中心学校卫生科

出　　处：《现代预防医学》2025年第2期267-273,368,共8页Modern Preventive Medicine

基　　金：山西省卫生健康委员会卫生健康科研课题(2024155)。

摘　　要：目的为基于随机森林算法分析2023年山西省中学生发生焦虑、抑郁情绪的影响因素并构建预测模型,为防控中学生焦虑和抑郁等负性情绪的发生提供科学依据。方法于2023年9月采用分层随机整群抽样方法,在山西省11个市抽取104826名中学生进行问卷调查,使用广泛性焦虑量表(GAD-7)和流调中心抑郁量表(CES-D)调查焦虑抑郁情况。采用Epidata建立数据库,组间比较采用卡方检验,用logistic回归进行影响因素分析,用Python3.7建立随机森林预测模型及自变量重要性排序。结果2023年山西省中学生焦虑和抑郁情绪检出率分别为33.26%和23.72%,女生高于男生,普通高中和职高学生高于初中生。多因素回归分析结果显示,女生(0R:1.846、1.655,95%CI:1.794~1.901、1.602~1.710)、高中(0R:2.019、1.670,95%CI:1.956~2.083、1.612~1.731)、职高(0R:1.167、1.356,95%CI:1.085~1.254、1.254~1.466)、非常态家庭(OR:1.105、1.248,95%CI:1.072~1.140、1.206~1.292)、住校(0R:1.102、1.117,95%CI:1.070~1.136、1.080~1.155)、吃油炸食物≥1次/天(0R:1.296、1.375,95%CI:1.227~1.369、1.297~1.457)、受到严重伤害(OR:1.636、1.688,95%CI:1.524~1.755、1.569~1.816)、遭受校园欺凌(OR:2.315、3.411,95%CI:2.150~2.492、3.165~3.677)、被家长打骂(0R:2.881、3.191,95%CI:2.766~3.001、3.059~3.328)、吸烟(0R:1.344、1.435,95%CI:1.266~1.427、1.349~1.527)、饮酒(0R:2.062、2.113,95%CI:1.978~2.151、2.022~2.208)、网瘾(0R:3.635、3.976,95%CI:3.289~4.016、3.611~4.377)是中学生焦虑、抑郁情绪检出的危险因素;吃新鲜水果口≥1次/天(0R:0.824、0.758,95%CI:0.800~0.849、0.733~0.783)、吃早餐频次高(有时吃:0R:0.843、0.715,95%CI:0.781~0.910、0.661~0.773;天天吃:0R:0.538、0.384,95%CI:0.500~0.579、0.356~0.414)、中高强度运动频次高(1~3天/周:0R:0.888、0.821,95%CI:0.854~0.923、0.787~0.857;≥4天/周:0R:0.931、0.84895%CI:0.893~0.971、0.810~0.887)、体育课时数多(2~3节/周:0R:0.826、0.859,95%CI:0.801~0.853、0.829~0Objective To analyze the influencing factors on the occurrence of anxiety and depression among secondary school students in Shanxi Province in 2023 based on the random forest algorithm and to construct a prediction model to provide a scientific basis for preventing and controlling the occurrence of negative emotions such as anxiety and depression among secondary school students.Methods In September 2023,104826 secondary school students in 11 cities of Shanxi Province were selected by stratified random whole-cluster sampling method for questionnaire survey,and the Generalized Anxiety Scale(GAD-7)and the Center for Streamlined Depression Scale(CES-D)were used to investigate anxiety and depression.Epidata was used for database establishment,chi-square test was used for between-group comparisons,logistic regression was used for influencing factor analysis,and Python 3.7 was used to establish a random forest prediction model and rank the importance of each variable.Results The detection rates of anxiety and depression among secondary school students in Shanxi Province in 2023 were 33.26%and 23.72%,respectively,and were higher for girls than for boys,and higher for students in general and vocational high schools than for junior high school students.The results of multifactorial regression analysis showed that female students(OR:1.846、1.655,95%CI:1.794-1.901、1.602-1.710),high school(OR:2.019、1.670,95%CI:1.956-2.083、1.612-1.731),vocational high school(OR:1.167、1.356,95%CI:1.085-1.254、1.254-1.466),non-conventional family(OR:1.105、1.248,95%CI:1.072-1.140、1.206-1.292),living in school(OR:1.102、1.117,95%CI:1.070-1.136、1.080-1.155),eating fried food≥1 time/day(OR:1.296、1.375,95%CI:1.227-1.369、1.297-1.457),being seriously injured(OR:1.636、1.688,95%CI:1.524-1.755、1.569-1.816),being bullied in school(OR:2.315、3.411,95%CI:2.150-2.492、3.165-3.677),being scolded by parents(OR:2.881、3.191,95%CI:2.766-3.001、3.059-3.328),smoking(OR:1.344、1.435,95%CI:1.266-1.427、1.349-1.527),drinking alcohol(OR

关 键 词：焦虑 抑郁 影响因素 随机森林 预测模型 

分 类 号：R749.72[医药卫生—神经病学与精神病学] R749.4[医药卫生—临床医学]