Risk stratification of childhood infection using host markers of immune and endothelial activation in Asia (Spot Sepsis): a multi-country, prospective, cohort study

Background: Prognostic tools for febrile illnesses are urgently required in resource-constrained community contexts. Circulating immune and endothelial activation markers stratify risk in common childhood infections. We aimed to assess their use in children with febrile illness presenting from rural communities across Asia. Methods: Spot Sepsis was a prospective cohort study across seven hospitals in Bangladesh, Cambodia, Indonesia, Laos, and Viet Nam that serve as a first point of contact with the formal health-care system for rural populations. Children were eligible if aged 1–59 months and presenting with a community-acquired acute febrile illness that had lasted no more than 14 days. Clinical parameters were recorded and biomarker concentrations measured at presentation. The primary outcome measure was severe febrile illness (death or receipt of organ support) within 2 days of enrolment. Weighted area under the receiver operating characteristic curves (AUC) were used to compare prognostic accuracy of endothelial activation markers (ANG-1, ANG-2, and soluble FLT-1), immune activation markers (CHI3L1, CRP, IP-10, IL-1ra, IL-6, IL-8, IL-10, PCT, soluble TNF-R1, soluble TREM1 [sTREM1], and soluble uPAR), WHO danger signs, the Liverpool quick Sequential Organ Failure Assessment (LqSOFA) score, and the systemic inflammatory response syndrome (SIRS) score. Prognostic accuracy of combining WHO danger signs and the best performing biomarker was analysed in a weighted logistic regression model. Weighted measures of classification were used to compare prognostic accuracies of WHO danger signs and the best performing biomarker and to determine the number of children needed to test (NNT) to identify one additional child who would progress to severe febrile illness. The study was prospectively registered on ClinicalTrials.gov, NCT04285021. Findings: 3423 participants were recruited between March 5, 2020, and Nov 4, 2022, 18 (0·5%) of whom were lost to follow-up. 133 (3·9%) of 3405 participants developed severe febrile illness (22 deaths, 111 received organ support; weighted prevalence 0·34% [95% CI 0·28–0·41]). sTREM1 showed the highest prognostic accuracy to identify patients who would progress to severe febrile illness (AUC 0·86 [95% CI 0·82–0·90]), outperforming WHO danger signs (0·75 [0·71–0·80]; p<0·0001), LqSOFA (0·74 [0·69–0·78]; p<0·0001), and SIRS (0·63 [0·58–0·68]; p<0·0001). Combining WHO danger signs with sTREM1 (0·88 [95% CI 0·85–0·91]) did not improve accuracy in identifying progression to severe febrile illness over sTREM1 alone (p=0·24). Sensitivity for identifying progression to severe febrile illness was greater for sTREM1 (0·80 [95% CI 0·73–0·85]) than for WHO danger signs (0·72 [0·66–0·79]; NNT=3000), whereas specificities were comparable (0·81 [0·78–0·83] for sTREM1 vs 0·79 [0·76–0·82] for WHO danger signs). Discrimination of immune and endothelial activation markers was best for children who progressed to meet the outcome more than 48 h after enrolment (sTREM1: AUC 0·94 [95% CI 0·89–0·98]). Interpretation: sTREM1 showed the best prognostic accuracy to discriminate children who would progress to severe febrile illness. In resource-constrained community settings, an sTREM1-based triage strategy might enhance early recognition of risk of poor outcomes in children presenting with febrile illness. Funding: Médecins Sans Frontières, Spain, and Wellcome. Translations: For the Arabic and French translations of the abstract see Supplementary Materials section.