脑卒中患者认知障碍风险预测模型的系统评价

Systematic review of risk prediction models for cognitive impairment in stroke patients

目的系统评价脑卒中患者认知障碍风险预测模型,以期为临床医护人员选择或构建高质量的风险评估工具提供参考与借鉴。方法计算机检索PubMed、Embase、Web of Science、OVID、Cochrane Library、SinoMed、中国知网、万方数据库、维普网中与脑卒中患者认知障碍风险预测模型相关的研究,检索时间限定为建库至2023年8月1日。由2名研究者独立筛选文献和提取数据,并使用PROBAST对纳入文献的偏倚风险与适用性进行评价。结果共纳入26篇文章。研究的适用性均较好,但所有研究均存在一定的偏倚风险,偏倚主要来自研究设计不合理、预测因子评估和结果确定的时间间隔不合理、样本量不足、连续性变量处理方式不合理、忽略缺失数据、未报道校准度、对模型过度拟合等原因。Meta分析结果显示,年龄(OR=0.05,95%CI:0.033~0.057)、受教育程度(OR=-0.13,95%CI:-0.171~-0.082)、糖尿病史(OR=2.32,95%CI:1.867~2.881)、高血压病史(OR=0.67,95%CI:0.420~0.918)、NIHSS评分(OR=0.40,95%CI:0.331~0.469)是脑卒中患者发生认知障碍的影响因素。结论现有的脑卒中患者认知障碍风险预测模型种类繁多,但存在方法学缺陷与高偏倚风险,预测因子存在一定共性与争议。未来应遵循透明报告个体预后或诊断的多变量预测模型开发偏倚风险低、适用性好的本土化预测模型,并进行内、外部验证,证明其在临床实践中的适用性与可行性。

ObjectiveTo systematically review the risk prediction models for cognitive impairment in stroke patients,aiming to provide references for clinical healthcare professionals in selecting or constructing high-quality risk assessment tools.MethodsA computerized search was conducted in PubMed,Embase,Web of Science,OVID,Cochrane Library,SinoMed,CNKI,Wanfang Database,and VIP to identify studies related to risk prediction models for cognitive impairment in stroke patients.The search was limited to articles published up to August 1,2023.Two researchers independently screened the literature,extracted data,and assessed the risk of bias and applicability of the included studies using PROBAST.ResultsA total of 26 articles were included.The applicability of the studies was generally good,but all studies had some degree of bias risk,mainly arising from unreasonable study designs,inappropriate time intervals between predictor assessment and outcome determination,insufficient sample sizes,unreasonable handling of continuous variables,omission of missing data,lack of reporting of calibration,and overfitting of the models.Meta-analysis results showed that age(OR=0.05,95%CI:0.033-0.057),education level(OR=-0.13,95%CI:-0.171--0.082),history of diabetes(OR=2.32,95%CI:1.867-2.881),history of hypertension(OR=0.67,95%CI:0.420-0.918),and NIHSS score(OR=0.40,95%CI:0.331-0.469)were factors for cognitive impairment in stroke patients.ConclusionsWhile various risk prediction models for cognitive impairment in stroke patients exist,they suffer from methodological flaws and high bias risks,with some commonalities and controversies in predictors.Future research should adhere to the principles of transparent reporting of individual prognosis or diagnosis of multivariate prediction models,develop localized prediction models with low bias risk and good applicability,and conduct internal and external validations to demonstrate their applicability and feasibility in clinical practice.