急性缺血性脑卒中患者早期神经功能恶化风险预测模型的系统评价

Systematic evaluation of risk prediction models for early neurological deterioration in patients with acute ischemic stroke

目的系统评价急性缺血性脑卒中(acute ischemic stroke,AIS)患者早期神经功能恶化风险预测模型。方法系统检索中国知网、万方、CBM、PubMed、Embase和Cochrane Library等数据库,检索时限为2010年7月1日-2022年7月31日。由2名研究员严格按照纳入与排除标准进行文献筛选和资料提取,运用预测模型研究的风险偏倚评估工具对纳入文献进行质量评价。结果共纳入7篇风险预测模型研究,AIS患者早期神经功能恶化的发生率为19.8%。1篇进行了外部验证,4篇进行了内部验证,2篇同时进行了外部和内部验证。7篇模型建模组与验证组的受试者工作特征曲线下面积均>0.7,校准度较好,整体预测性能较优。7篇研究偏倚性高,但适用性风险低。预测模型包含最常见的预测因子是血糖与炎性反应细胞。结论AIS早期神经功能恶化风险预测模型的构建及验证研究较少,且处于初级发展阶段。虽然各模型预测性能较优,但整体偏倚风险高,缺乏外部验证。提示未来应开发性能优良、适用价值高和同时进行内外部验证的预测模型。

Objective To systematically review the risk prediction model of early neurological deterioration in patients with acute ischemic stroke(AIS).Methods The databases of CNKI,Wanfang,CBM,PubMed,Embase,and Cochrane Library were systematically searched,and the retrieval time was from July 1,2010 to July 31,2022.Two researchers conducted literature screening and data extraction according to the inclusion and exclusion criteria,and used the PMROBAST tool to evaluate the quality of the included literature.Results Seven risk prediction model studies were included,and the incidence of early neurological deterioration in patients with AIS was 19.8%.In this paper,1 literature was verified externally,4 literatures were verified internally,and 2 literatures were simultaneously verified internally and externally.The area under the working characteristic curve of the subjects in the modeling group and the validation group of the 7 literatures was>0.7,with pretty good calibration and good overall prediction performance.Seven studies had high bias,but low risk of applicability.The most common risk factors in the prediction model were blood glucose and inflammatory cytokines.Conclusion There are few studies on the construction and verification of risk prediction models for early neurological deterioration in patients with AIS,which are still in the primary stage of development.Although the prediction performance of each model is better,the overall risk of bias is high and there is a lack of external validation.It suggests that the prediction model for the high value and internal external validation should be developed in the future.