老年轻度认知障碍风险预测模型的系统评价  

Risk Prediction Models for Mild Cognitive Impairment in Elderly Patients: a Systematic Review

作　　者：覃淑红 李睿琳 郑展航 伍陈幸子 王雯娟 QIN Shuhong;LI Ruilin;ZHENG Zhanhang;WU Chenxingzi;WANG Wenjuan(Faculty of Nursing,Guangxi University of Chinese Medicine,Nanning 530200,China)

机构地区：[1]广西中医药大学护理学院,广西壮族自治区南宁市530200

出　　处：《实用心脑肺血管病杂志》2024年第10期88-93,共6页Practical Journal of Cardiac Cerebral Pneumal and Vascular Disease

基　　金：2024年度广西高校中青年教师科研基础能力提升项目(2024KY0315);2023年广西中医药大学引进博士科研启动基金项目(2023BS056)。

摘　　要：目的系统评价国内外现有的老年轻度认知障碍(MCI)风险预测模型的偏倚风险、适用性。方法计算机检索中国知网、万方数据知识服务平台、中国生物医学文献数据库、PubMed、Cochrane Library、Web of Science、Embase中关于老年MCI风险预测模型的研究,检索时限为建库至2024-05-31。剔除重复文献后,由两位研究人员严格按照文献纳入与排除标准独立进行文献筛选;对于最终纳入的文献,由两位研究人员根据系统评价的批判性评估和数据提取清单(CHARMS)进行资料提取;由两位研究人员使用预测模型偏倚风险评估工具(PROBAST)评估纳入模型的偏倚风险和适用性。结果最终纳入文献14篇,共构建了21个老年MCI风险预测模型,本研究仅选择各文献中性能表现最佳的模型。14个模型的候选预测变量数为4~42个,每个自变量的事件数(EPV)为2.5~195.5;在建模方法方面,主要采用Logistic回归分析、Cox回归分析;在模型性能评估结果方面,12个模型的AUC为0.676~0.96;在模型校准方法方面,11个模型未进行校准;在模型验证方法方面,8个模型未进行验证。14个模型的整体偏倚风险均为高风险,整体适用性均为低风险。结论现有的老年MCI风险预测模型多数具备一定的区分能力,整体适用性较好,但其校准度尚不明确,整体偏倚风险较高。Objective To systematically assess the risk of bias and applicability of existing risk prediction models for mild cognitive impairment(MCI)in elderly patients,both domestically and internationally.Methods Databases including CNKI,Wanfang Data,CBM,PubMed,Cochrane Library,Web of Science,and Embase were systematically searched to collect the studies on the risk prediction models for MCI in elderly patients.The search time was from the establishment of the database to May 31,2024.After removing duplicate entries,literature screening was conducted independently by two researchers in strict accordance with the literature inclusion and exclusion criteria.Data from the final included literature was extracted by two researchers using the Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modelling Studies(CHARMS).The prediction model risk of bias assessment tool(PROBAST)was employed by two researchers to assess the risk of bias and applicability of included models.Results In the end,14 literatures were included,and a total of 21 risk prediction models for MCI in elderly patients were constructed,and only the best-performing model in each literature was selected for this study.The number of candidate predictor variables of the 14 models ranged from 4 to 42,and the events per variable(EPV)ranged from 2.5 to 195.5.In terms of modeling methods,Logistic regression analysis and Cox regression analysis were mainly used.In terms of model performance evaluation results,the AUC of 12 models was 0.676 to 0.96.In terms of model calibration methods,11 models had not been calibrated,and in terms of model validation methods,8 models had not been validated.The overall risk of bias for all 14 models was high,and the overall risk of applicability was low.Conclusion Most of the existing risk prediction models for MCI in elderly patients have certain discriminative ability and good overall applicability,but their calibration degree is not yet clear,and the overall risk of bias is high.

关 键 词：认知功能障碍 轻度认知障碍 老年人 预测模型 系统评价(主题) 

分 类 号：R741[医药卫生—神经病学与精神病学]