Back

Referral of febrile children in resource-constrained community settings in Asia (Spot Sepsis): a multi-country, prospective, cohort study

Chandna, A.; Koshiaris, C.; Mahajan, R.; Ahmad, R. A.; Van Anh, D. T.; Choudhury, K. S.; Keang, S.; Nguyen, P. N. T.; Rattanavong, S.; Vannachone, S.; Spot Sepsis Investigator Group, ; Painter, C.; Yosia, M.; Waithira, N.; Abdad, M. Y.; Thaipadungpanit, J.; Turner, P.; Phuc, P. H.; Mondal, D.; Mayxay, M.; Liem, B. T.; Ashley, E. A.; Arguni, E.; Perera-Salazar, R.; Richard-Greenblatt, M.; Lubell, Y.; Burza, S.

2025-10-14 pediatrics
10.1101/2025.10.12.25337822 medRxiv
Show abstract

In resource-constrained community settings, distinguishing which febrile children require referral is a major unmet need. Current WHO danger signs lack accuracy, resulting in missed severe illness and unnecessary referrals. We developed and validated simple clinical prediction models using data from 3,405 children aged 1-59 months presenting with community-acquired acute febrile illnesses to seven hospitals across Bangladesh, Cambodia, Indonesia, Laos, and Viet Nam. Cambodian data were held-out for external validation. All models outperformed WHO criteria to predict progression to severe febrile illness (death or organ support) within two days (sensitivity=0.56, 95%CI=0.42-0.69; specificity=0.83, 95%CI=0.78-0.87). Incorporating pulse oximetry or the host biomarker sTREM1 further enhanced sensitivity (0.89, 95%CI=0.79-0.97) vs. clinical features alone (0.75, 95%CI=0.62-0.86). The pulse oximetry-based model achieved these gains while improving specificity, concomitantly reducing referral rates three-fold. These approaches appear cost-effective and could transform referral practices for febrile children in resource-constrained community settings. They warrant evaluation in randomised controlled trials.

Matching journals

The top 6 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.