Predicting antibiotic use in febrile children using explainable ML in Bangladesh

Logistic regression with a consensus feature set achieved moderate performance, but held-out validation on unseen PSUs showed weak discrimination (ROC-AUC 0.561), with confidence interval including 0.5. The model's transportability is limited, and due to reliance on self-reported data without clinical verification, the clinical necessity of antibiotic use cannot be confirmed.

Summarised from the original article by an open-source model, then checked automatically: every number and English name must appear in the source. The original article is authoritative.