脓毒症患者死亡预警评估工具准确性的网状Meta分析

Performance of early warning assessment tools in predicting the mortality among emergency patients with sepsis:a network Meta-analysis

目的 系统评价常见脓毒症死亡预警评估工具对脓毒症死亡的预测价值。方法 计算机检索Cochrane Library、Pubmed、Embase、Web of Science、EBSCO、CNKI、万方数据和维普数据库中关于脓毒症死亡预警的评估工具,检索时限均从建库至2023年9月30日。由2名研究者独立筛选文献、提取资料并评价纳入研究的偏倚风险,采用Stata 16.0软件进行网状Meta分析。结果 共纳入16项研究,共计患者16 616例,涉及7种死亡预警评估工具。网状Meta分析结果显示,敏感度排序为全身炎症反应综合征(SIRS)>急诊科脓毒症病死率评分(MEDS)>改良早期预警评分(MEWS)>国家预警评分(NEWS)>序贯器官衰竭评估(SOFA)>急性生理学与慢性健康状况评价Ⅱ(APACHEⅡ)>快速序贯器官衰竭评估(qSOFA);特异度排序为qSOFA>APACHEⅡ>MEWS>NEWS>SIRS>MEDS>SOFA。亚组分析结果显示,MEDS与qSOFA预测脓毒症患者死亡的准确性分别在不同年龄、不同疾病中差异有统计学意义(P均<0.05)。结论 SIRS与MEWS敏感度较高,但特异度低,qSOFA和APACHEⅡ的特异度较强但敏感度不佳。年龄可能会影响MEDS对脓毒症患者死亡的预警价值,诊断可能会影响qSOFA对脓毒症患者死亡的预警价值。

Objective To systematically evaluate predictive value of early warning assessment tools in the death of patients with sepsis.Methods The Cochrane Library,Pubmed,Embase,Web of Science,EBSCO,CNKI,WanFang Data and VIP databases were electronically searched to collect the early warning assessment tools for sepsis mortality from inception to September 30,2023.Two reviewers independently screened the literatures,extracted the data and assessed the risk of bias of included studies,then network Meta-analysis was performed by using Stata 16.0 software.Results A total of 16 studies with 16 616 patients involving 7 early warning assessment tools for the mortality were included.The results of the network Meta-analysis showed that sensitivity sorting was SIRS>MEDS>MEWS>NEWS>SOFA>APACHEⅡ>qSOFA,and the specificity sorting was qSOFA>APACHEⅡ>MEWS>NEWS>SIRS>MEDS>SOFA.Subgroup analysis showed that the predictive value of MEDS for septic patients was significantly affected by the age,and the predictive value of qSOFA was affected by the patients′ diagnosis.Conclusions Current evidence suggests that SIRS and MEWS have high sensitivity but poor specificity,while qSOFA and APACHEⅡhave strong specificity but poor sensitivity.The predictive value of MEDS in septic patients′ mortality is affected by age,and the diagnosis of patients may affect the predictive value of qSOFA.