脑卒中后抑郁风险预测模型的系统评价

A systematic review on risk prediction models of post stroke depression

目的系统评价脑卒中后抑郁(PSD)风险预测模型。方法计算机检索Web of Science、The Cochrane Library、PubMed、Embase、CINAHL、知网、中国生物医学文献服务系统、万方和维普数据库中建库至2022年6月1日收录的PSD风险预测模型相关文献,采用预测模型构建研究数据提取和质量评价清单(CHARMS)对纳入文献中的模型进行质量评价,并对纳入模型中具有共性的预测因子的预测价值采用RevMan 5.3软件进行Meta分析。结果共纳入9篇文献,包含11个PSD风险预测模型,所有模型的建模时受试者工作特征曲线下面积(AUC)为0.726~0.854,其中7个模型的AUC≥0.8,预测效能较高,但仍存在偏倚风险,主要原因包括未报告缺失数据的处理、模型效果评价不完整以及未对模型进行内外部验证。Meta分析结果显示抑郁或其他精神疾病病史(OR=6.73,95%CI:3.87~11.73)、艾森克人格问卷(EPO)评分(OR=1.13,95%CI:1.03~1.23)、高血压(OR=0.47,95%CI:0.30~0.74)、Barthel指数(OR=0.98,95%CI:0.98~0.99)均是PSD的有效预测因子。结论PSD风险预测模型整体预测性能良好,但也仍存在一定的偏倚风险,未来应对建模方法进行改进;PSD风险预测模型的建立可重点关注抑郁或其他精神疾病病史、EPQ评分、高血压、Barthel指数等预测因子。

Objective To systematically evaluate the risk prediction models of post-stroke depression(PSD).Methods Web of Science,The Cochrane Library,PubMed,Embase,CINAHL,CNKI,SinoMed,WanFang Data,and VIP database were searched for literature related to PSD risk prediction models from inception to June 1,2022.The quality of the included models was evaluated by Checklist for Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modelling Studies(CHARMS),and Meta-analysis was performed on influencing factors enjoyed generality in the included models by RevMan 5.3 software.Results A total of 9 pieces of literature were included,analyzing 11 risk prediction models.The area under the curve(AUC)for all models ranged from 0.726 to 0.854,and the AUC of 7 models was≥0.8,enjoying a high prediction efficiency but a risk of bias;and the main reasons included not reporting the processing of missing data,incomplete evaluation of model effect,and lack of internal and external validation of the models.Meta-analysis results showed depression or other mental illness(OR=6.73,95%CI:3.87-11.73),Eysenck Personality Questionnaire(EPQ)scores(OR=1.13,95%CI:1.03-1.23),hypertension(OR=0.47,95%CI:0.30-0.74),and Barthel index(BI,OR=0.98,95%CI:0.98-0.99)were predictors for PSD.Conclusions PSD risk prediction models have good predictive performance but with a risk of bias,therefore,the modeling method should be improved in the future.The establishment of PSD risk prediction models should focus on the predictors as history of depression or other mental disorders,EPQ scores,hypertension,and BI.