ABSTRACT
Objective:To evaluate the prognostic accuracy of the sequential organ failure assessment (SOFA), quick sequential organ failure assessment (qSOFA) and systemic inflammatory response syndrome (SIRS) criteria in predicting the mortality in patients with infection or suspected infection by using network Meta-analysis.Methods:Five databases including Wanfang Data, China National Knowledge Infrastructure (CNKI), China Science and Technology Journal Database (VIP), PubMed, Web of Science were searched from February 23, 2016 to September 5, 2020 to identify the relevant literatures comparing the prognostic accuracy of two or more scores for mortality in patients with infection or suspected infection. The literatures screening, data extraction and the quality assessment of the included studies were all conducted independently by two reviewers. Stata 14.0 software was used to test the heterogeneity between the original studies of pairwise comparison of each of the three scoring systems. Ring inconsistency test was used to judge the consistency between direct comparison and indirect comparison. Then network Meta-analysis was performed and the results were ranked. The predictive ability of the three scoring systems was evaluated by surface under cumulative ranking curve (SUCRA). A "comparison-correction" funnel plot was drawn to assess whether there was publication bias in the included studies.Results:A total of 38 studies were enrolled, the overall quality was high. Network meta-analysis showed that SOFA had a great prognostic performance in predicting mortality for patients with infection or suspected infection, which was followed by qSOFA [mean difference ( MD) = 0.07, 95% confidence interval (95% CI) was 0.05-0.09] and SIRS scores ( MD = 0.16, 95% CI was 0.14-0.18), and the qSOFA score was better than SIRS score ( MD = 0.09, 95% CI was 0.07-0.11). In the order of predicting the death risk of patients with infection or suspected infection, SOFA score had higher predictive value, followed by qSOFA score, and SIRS score was the lowest, with SUCRA values of 1.0, 0.5 and 0, respectively. Funnel plot showed that all the studies were distributed on both sides of the midline, but the distribution was not symmetrical, suggesting that there was a high possibility of publication bias and small sample effect. Conclusions:SOFA score had the best prognostic performance in predicting mortality of patients with infection or suspected infection as compared with qSOFA score and SIRS score. However, the funnel plot showed that included literatures may exist small sample effects or publication bias. So the final results should be validated by more prospective studies with multicenters and large samples.