(Newswise) — DURHAM, N.C. — Attention-deficit/hyperactivity disorder (ADHD) is a condition impacting millions of children, yet many remain undiagnosed for years, missing opportunities for early intervention that can significantly improve their long-term outcomes. A new study by researchers at Duke Health reveals that artificial intelligence (AI) tools can analyze routine electronic health records to accurately predict a child’s risk of developing ADHD even before a formal diagnosis typically occurs.
This innovative approach leverages patterns in everyday medical data to identify children who might benefit from earlier evaluations and follow-ups. Published in Nature Mental Health on April 27, the study emphasizes how valuable insights can be gleaned from information collected during regular healthcare visits, thereby enabling primary care providers to make informed decisions sooner.
“We have this incredibly rich source of information sitting in electronic health records,” stated Elliot Hill, the lead author and a data scientist at the Department of Biostatistics & Bioinformatics at Duke University School of Medicine. “The aim was to uncover patterns within that data to help predict which children might be diagnosed with ADHD much earlier than is typically the case.”
In reaching these conclusions, researchers examined electronic health records from over 140,000 children, both those diagnosed with ADHD and those without. They developed a specialized AI model trained to analyze medical histories from birth through early childhood, allowing it to recognize the combinations of developmental, behavioral, and clinical events that often predate an ADHD diagnosis by several years.
This model demonstrated high accuracy in assessing the risk of ADHD in children aged 5 and older, consistently performing well regardless of the child’s sex, race, ethnicity, or insurance status.
It’s crucial to note that the AI tool does not provide a diagnosis. Instead, it identifies children who might benefit from additional attention from their pediatric primary care provider or an earlier referral for assessment by a specialist.
“This is not an AI doctor,” clarified Matthew Engelhard, M.D., Ph.D., a senior author of the study and a member of Duke’s Department of Biostatistics & Bioinformatics. “It’s meant to support clinicians in directing their time and resources effectively, ensuring that children in need don’t slip through the cracks or wait for years to get answers.”
The researchers believe that early identification through screening can lead to quicker diagnoses, ultimately allowing for timely support, which is linked to improved academic, social, and health outcomes for children with ADHD. They stress the importance of further research before these tools are integrated into clinical practice.
“Children with ADHD often face significant challenges when their needs are not recognized and sufficient support is absent,” remarked study author Naomi Davis, Ph.D., an associate professor in the Department of Psychiatry and Behavioral Sciences. “Connecting families with timely, evidence-based interventions is vital for helping them reach their goals and establishing a foundation for future success.”
Hill and Engelhard have additionally explored the use of AI models in identifying potential risks and causes of mental health issues in adolescents. Joining Hill, Engelhard, and Davis in the study were De Rong Loh, Benjamin A. Goldstein, and Geraldine Dawson.
The research received support from grants provided by the National Institute of Mental Health (K01-MH127309, UL1 TR002553) and the National Center for Advancing Translational Sciences.
