Sepsis screening tools in resource-limited settings: a systematic review and meta-analysis of diagnostic accuracy in low- and middle-income countries

Abstract

Background – Sepsis causes disproportionate mortality in low- and middle-income countries (LMICs), yet evidence on screening tool performance in these resource-limited settings remains fragmented. Objective – This systematic review and meta-analysis aimed to evaluate and compare the diagnostic accuracy of sepsis screening tools—ranging from purely clinical assessments (qSOFA, NEWS, MEWS, UVA) to those incorporating laboratory parameters (SIRS, SOFA)—for sepsis identification in adult populations within LMICs. Methods – A comprehensive search of PubMed, Embase, Cochrane Library, Web of Science, and Global Index Medicus was conducted from inception through June 2025. Eligible studies evaluated sepsis screening tools ranging from purely clinical bedside assessments (qSOFA, NEWS, MEWS, UVA) to those requiring basic laboratory parameters (SIRS, SOFA), enabling comparison across the resource-availability spectrum in LMICs. Bivariate random-effects models were employed to calculate pooled sensitivity, specificity, and area under the hierarchical summary receiver operating characteristic curve (AUROC). Evidence quality was assessed using the GRADE approach. Results – Twenty-seven studies encompassing 30, 310 patients across 14 LMICs were included. qSOFA demonstrated pooled sensitivity of 0.51 (95% CI: 0.42–0.60) and specificity of 0.83 (95% CI: 0.77–0.88) with AUROC of 0.74. SIRS exhibited high sensitivity (0.86) but poor specificity (0.32). NEWS achieved the highest point estimate of discriminative ability (AUROC 0.77, 95% CI: 0.73–0.81), followed by SOFA (AUROC 0.75, 95% CI: 0.71–0.79) and UVA (AUROC 0.74, 95% CI: 0.70–0.78), although confidence intervals overlapped substantially across tools. qSOFA yielded a positive likelihood ratio (LR+) of 3.00 and a negative likelihood ratio (LR−) of 0.59, indicating moderate rule-in but limited rule-out utility. Substantial heterogeneity was observed across studies (I2 > 75%). Conclusion – No single screening tool demonstrates optimal performance across all metrics in LMIC populations. When analysis is restricted to purely clinical tools requiring no laboratory parameters, NEWS (AUROC 0.77) and qSOFA (AUROC 0.74) demonstrate comparable discriminative ability with broadly overlapping confidence intervals, supporting their consideration in the most resource-constrained settings. However, the limited rule-out capacity of qSOFA (LR − 0.59) suggests it should not be used as a standalone screening tool. Tool selection should be guided by local healthcare priorities and available laboratory capacity, with tiered screening strategies potentially optimizing sepsis recognition while ensuring efficient resource allocation.