注意缺陷多动障碍7~16岁儿童影响因素分析及Nomogram预测模型的建立与验证

Analysis of influencing factors of attention deficit hyperactivity disorder in children from 7 to 16 years old and the establishment and verification of Nomogram prediction model

目的分析儿童注意缺陷多动障碍(ADHD)的影响因素并构建Nomogram(列线图)预测模型。方法采取随机整群抽样的方法,对新疆地区5所学校共5409名7~16岁儿童开展调查,采用斯诺佩(SNAP-Ⅳ)评估量表和影响因素调查问卷进行调查。利用最小绝对收缩选择算子(LASSO)回归和多因素Logistic回归分析调查儿童ADHD影响因素并建立Nomogram预测模型。结果1.调查儿童ADHD的检出率为7.3%;2.LASSO-Logistic回归模型显示,高热惊厥疾病史(OR=5.97,95%CI:3.52~9.86)、癫痫疾病史(OR=11.86,95%CI:7.83~17.89)、头部外伤疾病史(OR=10.0,95%CI:7.27~13.71)、母亲分娩方式(OR=2.53,95%CI:1.99~3.23)、母亲文化程度(OR=2.26,95%CI:1.45~3.67)、母亲吸烟超过1年以上(OR=12.65,95%CI:8.30~19.34)、家庭环境是否安静(OR=1.27,95%CI:1.00~1.63)、打骂的教育方式(OR=3.05,95%CI:2.13~4.31)为儿童ADHD的独立风险因素;3.将以上指标建立Nomogram预测模型,经Bootstrap重采样1000次进行内部验证,C-index指数为0.81(95%CI:0.78~0.83),提示Nomogram预测模型有较好的预测能力、精准度以及区分度。临床决策(DCA)曲线提示Nomogram模型的预测概率阈值>0.2时,患者的临床净收益水平高。结论调查儿童ADHD检出率为7.3%,比全国平均水平较高。高热惊厥、癫痫、头部外伤疾病史、母亲分娩方式、母亲文化程度、母亲吸烟超过1年以上、家庭环境是否安静、打骂的教育方式为儿童ADHD的独立风险因素,依此绘制的Nomogram预测模型能够简洁、直观地为儿童提供个体化的ADHD患病风险预测。

Objective To analyze the influencing factors of attention deficit hyperactivity disorder(ADHD)in children and construct a Nomogram prediction model.Methods A total of 5409 children aged 7 to 16 from 5 schools in Xinjiang were investigated by using SNAP-Ⅳassessment scale and influencing factors questionnaire.Least absolute shrinkage and selection operator(LASSO)regression and multivariate Logistic regression were used to analyze and investigate the influencing factors of ADHD in children,and then Nomogram prediction model was established.Results(1)The detection rate of ADHD was 7.3%.(2)The LASSO-Logistic regression model showed that the history of febrile convulsions(OR=5.97,95%CI:3.52-9.86),the history of epilepsy disease(OR=11.86,95%CI:7.83-17.89),the history of head trauma disease(OR=10.0,95%CI:7.27-13.71),mother′s delivery method(OR=2.53,95%CI:1.99-3.23),mother′s education level(OR=2.26,95%CI:1.45-3.67),mother′s smoking more than 1 year(OR=12.65,95%CI:8.30-19.34),whether the family environment is quiet(OR=1.27,95%CI:1.00-1.63),and the education method of beating and scolding(OR=3.05,95%CI:2.13-4.31)was an indepen-dent risk factor for children with ADHD;(3)The Nomogram prediction model was built and verified by Bootstrap for 1000 samples.The C-index was 0.81(95%CI:0.78-0.83),suggesting that the Nomogram prediction model has good prediction ability,accuracy,and distinction.Decision curve analysis(DCA)of the clinical decision curve suggested that patients with Nomogram model with a predictive probability threshold greater than 0.2 had a higher clinical net benefit.Conclusions The detection rate of ADHD was 7.3%,which was higher than the national average.The Nomogram prediction model drawn here can provide individualized ADHD risk predictions for children based on the history of hyperthermia,epilepsy,and head trauma,maternal mode of childbirth,maternal education level,maternal education level,maternal smoking for more than 1 year,quiet family environment,and scolding education methods.