This Arkangel AI whitepaper is a narrative community evidence brief, not a peer-reviewed study or a formal scoping review, on what AI can support today in pediatrics in Latin America. It synthesizes 29 sources found in PubMed, Crossref, biomedical journals and institutional documents (cutoff June 14, 2026) across seven domains: clinical decision support, radiology, neonatal and pediatric intensive care, asthma, pediatric surgery, medication safety and clinical language, with regional evidence from Brazil, Mexico and Argentina and a PAHO information-systems framework.
The field is dominated by model development and early evaluation: a review of 112 pediatric surgery AI studies found only 6% of models both interpretable and externally validated and 40% at high risk of bias, and all 40 pediatric clinical-text studies identified were retrospective. Regionally, smartphone-video screening for retinopathy of prematurity in 512 neonates from Mexico and Argentina reached 93.3% patient-level sensitivity versus 73.3% for ophthalmologists, but with lower specificity and accuracy, and a São Paulo neonatal mortality model reached an AUC of 0.971 retrospectively. The authors conclude that the most defensible uses are supervised information review, prioritization and reviewable drafts, and that adult models cannot be transferred to children without local validation by age, weight and context.
