Abstract: Objective To construct a risk prediction scoring model for tic disorders (TD) in Chinese children using a two-stage approach:Meta-analysis for updated prevalence and risk factors followed by external validation, in order to establish an appropriate predicting model for TD in Chinese children. Methods An extensive search in Chinese and international databases (CNKI, Wanfang, VIP, PubMed, Embase, Web of Science) was conducted to identify observational studies on TD prevalence and risk factors in Chinese children published before October 2022.Meta-analysis in R yielded updated prevalence estimates and pooled odds ratios (ORs) for risk factors.A risk prediction model was developed using Logistic regression based on these findings.A total of 644 children (TD patients and healthy controls) from Dongzhimen Hospital, Beijing, from October 2022 to June 2023 were recruited for model validation.Predictive performance was assessed by the area under the receiver operating characteristic (ROC) curve (AUC) and clinical utility via decision curve analysis. Results Finally 34 studies were included in this Meta-analysis, encompassing 9 955 patients.The overall prevalence of TD in Chinese children diagnosed using DSM-IV criteria was 1.22% (95%CI:0.79% - 1.86%).After filtering, eight risk factors were incorporated into the model, including psychiatric abnormality during pregnancy (OR=2.50, 95%CI:1.78 - 3.52), other perinatal factors (OR=3.05, 95%CI:2.21 - 4.21), recurrent respiratory infections (OR=2.51, 95%CI:2.12 - 2.97), family history of TD (OR=4.86, 95%CI:2.98 - 7.93), corporal punishment (OR=2.78, 95%CI:1.73 - 4.47), poor dietary habits (OR=2.27, 95%CI:1.50 - 3.42), excessive screen time (OR=2.29, 95%CI:1.91-2.73), and single-parent/left-behind child/family disharmony (OR=2.44, 95%CI:1.53 - 3.89).The model score ranged from 0 to 80, with an AUC of 0.726 (95%CI:0.675 - 0.776).The optimal cutoff was 11 points, yielding 65.4% sensitivity and 70% specificity.The model demonstrated significant clinical net benefit within a 10% - 40% probability threshold. Conclusions This TD risk prediction scoring system based on Meta-analysis shows promising performance as a clinical assessment tool.However, further validation and refinement are warranted.

Key words: tic disorders, Meta-analysis, Logistic regression, risk prediction model

摘要： 目的 基于Meta分析及外部验证构建中国儿童抽动障碍(TD)风险预测评分模型,旨在建立适用于中国儿童的TD临床预测模型。方法 检索中国知网、万方数据知识服务平台、中国生物医学文献数据库、维普数据库、PubMed、Embase和Web of Science数据库中从建库至2022年10月期间所有关于中国儿童TD患病率及危险因素的观察性研究。采用R语言进行Meta分析,获得最新患病率数据和TD患病危险因素的合并风险值,基于Logistic回归模型构建风险预测评分系统。选择2022年10月—2023年6月于北京中医药大学东直门医院儿科招募的TD患儿及健康体检儿童共644例进行模型验证,采用受试者工作特征曲线(ROC)的曲线下面积(AUC)评价模型的预测性能,运用决策曲线分析评价预测模型的临床实用性。结果 纳入34项观察性研究,共9 955例患儿。Meta分析结果显示,以DSM-Ⅳ为诊断标准的中国儿童TD患病率为1.22%(95%CI:0.79%~1.86%)。经过数据筛选,最终有8个危险因素被纳入风险预测模型,包括:母孕期精神或情绪异常(OR=2.50,95%CI:1.78~3.52)、围生期其他不利因素(OR=3.05,95%CI:2.21~4.21)、反复呼吸道感染(OR=2.51,95%CI:2.12~2.97)、抽动障碍家族史(OR=4.86,95%CI:2.98~7.93)、家教严厉或打骂体罚的教育方式(OR=2.78,95%CI:1.73~4.47)、不良饮食习惯(OR=2.27,95%CI:1.50~3.42)、长时间看电子产品(OR=2.29,95%CI:1.91~2.73)、单亲家庭或留守儿童或家庭不和睦(OR=2.44,95%CI:1.53~3.89)。预测模型总分为0~80分,AUC为0.726(95%CI:0.675~0.776),最佳截断值为11分,灵敏度为65.4%,特异度为70%,当阈值范围在10%~40%时,预测模型具有较高的临床净获益。结论 基于Meta分析构建的TD风险预测评分系统具有一定的预测性能,有望作为TD的风险评估工具应用于临床,但未来仍需进一步验证及优化。

关键词: 抽动障碍, Meta分析, Logistic回归, 风险预测模型