基于家长问卷法预测学龄前注意缺陷多动障碍诊断的决策树模型  

Decision tree model for predicting diagnosis of attention-deficit hyperactivity disorder among preschoolers based on parental questionnaire

作　　者：张劲松[1,2] 潘景雪 帅澜[1] 张慧凤[1,2] 夏卫萍 王周烨[1] 李伟[1,2] ZHANG Jinsong;PAN Jingxue;SHUAI Lan;ZHANG Huifeng;XIA Weiping;WANG Zhouye;LI Wei(Department of Medical Psychology, Xinhua Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai 200092, China;Ministry of Education and Shanghai Key Laboratory of Children s Environment Health, Shanghai 200092, China)

机构地区：[1]上海交通大学医学院附属新华医院临床心理科,中国上海200092 [2]教育部与上海市环境与儿童健康重点实验室,中国上海200092

出　　处：《教育生物学杂志》2018年第4期184-190,共7页Journal of Bio-education

基　　金：上海市科学技术委员会科研计划项目(14411968600)

摘　　要：目的通过观察学龄前注意缺陷多动障碍(attention-deficit hyperactivity disorder,ADHD)的诊断,探讨影响学龄前ADHD诊断的影响因素,建立基于临床数据的ADHD诊断模型。方法经过临床符合《美国精神障碍诊断与统计手册(第5版)》(Diagnostic and Statistical Manual of Mental Disorders,fifth edition,DSM-5)诊断标准和学龄前婴幼儿诊断性访谈(diagnostic infant and preschool assessment,DIPA),纳入学龄前ADHD儿童263例和正常儿童105名为研究对象,使用中文版SNAP-Ⅳ量表(父母版)(Swanson,Nolan,and PelhamⅣRating Scale,parent version,SNAP-Ⅳ)、长处和困难问卷(strengths and difficulties questionnaire,SDQ)及学龄前儿童执行功能行为评定量表(父母版)(behavior rating scale of executive function-preschool version parent form,BRIEF-P)评估儿童ADHD症状和执行功能。在相关分析筛选出关系显著的影响因素后,运用Logistic回归方法建立了全局模型。以决策树方法和Logistic回归方法在不同节段下对目标变量ADHD进行分型和预测。结果 (1)影响学龄前ADHD诊断的因素主要为抑制、工作记忆、抑制自我调控指数、元认知指数和量表BRIEF-P总分,多动问题,注意缺陷和多动冲动因子;(2) Logistic回归预测全局ADHD诊断为88. 8%;决策树分析发现,影响ADHD诊断预测节点的因素主要为注意缺陷、抑制和抑制自我调控因子;(3)进一步Logistic回归及决策树分析预测ADHD分型诊断,得出各分型的影响因素和诊断预测模型。结论学龄前儿童的ADHD症状及执行功能可预测ADHD诊断,同时对ADHD不同分型及共患对立违抗障碍(oppositional defiant disorder,ODD)者具有一定诊断预测,提供临床应用参考。Objective To discuss the correlation between executive function and attention-deficit hyperactivity disorder (ADHD) symptoms in preschool children, and to establish the ADHD diagnostic prediction model based on clinical data. Methods Parents of kindergarten students ( n =105), non-reference clinical patients who were diagnosed as having ADHD ( n =263) by the psychiatrists based on Diagnostic and Statistical Manual of Mental Disorders, fifth edition (DSM-5) and diagnostic infant preschool assessment (DIPA) completed the (Swanson, Nolan, and Pelham Ⅳ Rating Scale,parent version,SNAP-Ⅳ), strengths and difficulties questionnaire (SDQ), and behavior rating scale of executive function-preschool version parent form(BRIEF-P) to evaluate children s symptoms and executive function through different perspectives. Significant factors which contribute to relationship were chosen by correlation analysis, and then whole prediction was made by Logistic regression. To improve the prediction of ADHD, subsections prediction was made using decision tree analysis and Logistic regression conditionally to classify ADHD types. Results ① Factors which were relative to ADHD mainly included inhibition, working memory, inhibitory self-control, emergent metacognition and total BRIEF-P score, hyperactivity, inattention and hyperactivity/impulsivity.② Logistic regression predicted 88.8% of ADHD, and decision tree analysis showed that the prediction nodes influencing diagnosis of ADHD included inattention, inhibition and inhibitory self-control factors.③ Logistic regression and decision tree analysis were conducted on ADHD subtype diagnostic prediction and related significant factors. Conclusion The ADHD symptoms and executive function can predict the diagnosis of ADHD in preschool children. Moreover, it predicts the diagnosis of ADHD subtype and comorbidity with oppositional defiant disorder(ODD), providing the references for clinic application.

关 键 词：注意缺陷多动障碍 家长问卷 决策树模型 学龄前儿童 

分 类 号：R749.94[医药卫生—神经病学与精神病学]