Predicting referral need for febrile children in low-resource community settings in South and Southeast Asia.

Summary

A multicountry model using simple clinical signs outperformed WHO danger signs for identifying febrile children at risk of death or early organ support, and adding pulse oximetry or sTREM1 increased sensitivity to ~89% while reducing referral rates threefold. The approach was externally validated and cost-effective, supporting evaluation in community-based trials.

Key Findings

• A simple clinical parameter model achieved sensitivity 74.7% and specificity 99.1%, outperforming WHO danger signs (55.5%/82.6%) for severe disease within 2 days.
• Adding pulse oximetry or sTREM1 raised sensitivity to 88.9% and 89.2%, respectively, with the pulse oximetry model reducing referral rates threefold.
• Cost-effectiveness analysis showed favorable ICERs: $26.28 for pulse oximetry and $196.46 for sTREM1; external validation on Cambodian data preserved performance.

Clinical Implications

Adopting simple clinical models augmented by pulse oximetry (and where feasible sTREM1) could improve triage accuracy, reduce unnecessary referrals, and better capture children at imminent risk, informing early antimicrobial therapy and escalation.

Limitations

• Hospital-based recruitment may limit generalizability to true community-level first-contact settings
• Biomarker (sTREM1) availability and implementation logistics vary; real-world impact requires community trials

Future Directions

Prospective community-based trials to assess mortality/morbidity impact, integration into digital triage tools, and evaluation of alternative low-cost biomarkers.