Responsible AI-Assisted Early Pediatric Vision Risk Detection Before Age 3: A Hospital-Integrated Framework for Early Identification and Referral Support for Strabismus and Amblyopia

Early childhood vision disorders such as amblyopia and strabismus remain significant and largely preventable causes of long-term visual impairment, collectively affecting an estimated 100 million children globally. Treatment outcomes depend heavily on early identification, particularly during the first years of life when the visual system undergoes critical neurodevelopment. However, current pediatric vision screening programs are often implemented at preschool or school-entry age, which may be beyond the most sensitive period for optimal intervention. This white paper presents a conceptual framework for responsible AI-assisted pediatric vision risk identification within routine healthcare systems, with a specific focus on children under the age of three. The proposed approach emphasises integration within pediatric care visits, utilisation of electronic health record (EHR) signals, clinician-in-the-loop decision support, and structured referral pathways to support timely ophthalmologic evaluation. Rather than proposing an autonomous diagnostic system, this work highlights the importance of aligning AI-based risk assessment with healthcare system workflows to address gaps in care continuity, adherence to follow-up, and early intervention opportunities.