AI Helps Physicians Revisit Rare Childhood Disease Cases
AI reasoning model boosts rare-disease diagnosis with 18 new pediatric cases See how reanalysis of genomic data may uncover answers and improve care
Researchers from Boston Children’s Hospital, Harvard University, and OpenAI used an AI reasoning model to review 376 previously unsolved raredisease cases in children and adolescents.
The study, published June 18, 2026 in NEJM AI, found that expert review and followup testing confirmed 18 additional diagnoses, increasing the diagnostic yield by 4.8%. The model did not diagnose patients directly; instead, it produced evidencelinked hypotheses for clinicians to examine through standard medical and laboratory processes.
The work suggests that periodic reanalysis of old genomic data can uncover new answers as genedisease knowledge expands and clinical records are better connected. The researchers said larger, prospective studies are still needed to compare AIassisted review with standard practice and to measure effects on accuracy, time, and workload.