We compared the diagnostic accuracy of seven early warning tools for predicting short-term mortality in adult sepsis patients using a network meta-analysis. Early warning tools trade sensitivity for specificity; combining direct and indirect evidence in an NMA and ranking by SUCRA permits concurrent comparison of relative performance while assessing transitivity and inconsistency. Following PRISMA 2020, we searched major English and Chinese databases to June 30, 2025, and included cohort studies of adult sepsis reporting in-hospital, 28-day, or 30-day mortality with extractable 2 × 2 data. Two reviewers independently screened studies, extracted data and assessed bias with the Newcastle–Ottawa Scale. A mixed-effects NMA was performed in Stata 16.0; primary outcomes were sensitivity and specificity, ranked by SUCRA. Sixteen studies enrolling 16,616 patients evaluated seven tools. Sensitivity ranking (highest→lowest) was SIRS > MEDS > MEWS > NEWS > SOFA > APACHE II > qSOFA; specificity ranking was qSOFA > APACHE II > MEWS > NEWS > SIRS > MEDS > SOFA. Subgroup analyses showed MEDS and qSOFA accuracy varied by age and diagnosis (all p 0.05) and no major publication bias was evident. SIRS and MEWS demonstrated high sensitivity but limited specificity, favoring screening, whereas qSOFA and APACHE II showed high specificity but poor sensitivity, favoring confirmation. MEDS performance appeared age-dependent and qSOFA may be affected by sepsis definitions; tool selection should match clinical purpose and setting.
Li et al. (Tue,) studied this question.