The study in Nature Mental Health shows that a specialised AI model can use routine electronic health records to estimate a child’s risk of attention-deficit/hyperactivity disorder years before a typical diagnosis. The researchers trained the model on records from birth through early childhood, using data from more than 140,000 children, and taught it to recognise combinations of developmental, behavioural and clinical events that often precede an ADHD diagnosis.
The tool was highly accurate at estimating future ADHD risk in children age 5 and older. Its performance remained consistent across patient groups defined by sex, race, ethnicity and insurance status. The authors emphasise that the AI does not make a diagnosis; rather, it flags children who may benefit from closer attention by their paediatric primary care provider or from an earlier referral for specialist assessment. The intended use is to help clinicians focus time and resources so children who need help do not wait years for answers.
Lead author Elliot Hill said the study tested whether hidden patterns in records could predict later ADHD, and senior author Matthew Engelhard described the tool as a way to prioritise care. Study author Naomi Davis highlighted the importance of connecting families with timely, evidence-based interventions to improve academic, social and health outcomes. Hill and Engelhard have also researched AI models for predicting risks and causes of mental illness in adolescents. The team calls for further studies before these tools are used in routine clinical care; the research was supported by grants from the National Institute of Mental Health and the National Center for Advancing Translational Sciences.
- Large EHR sample trained the AI model
- High accuracy for children aged five and older
- Consistent performance across demographic groups
- AI flags risk but does not diagnose
Difficult words
- specialised — designed for a particular purpose or task
- electronic health record — digital file of a patient's medical historyelectronic health records
- attention-deficit/hyperactivity disorder — a neurodevelopmental condition affecting attention and activity
- recognise — identify a pattern or person in information
- precede — happen or exist before something else
- consistent — showing the same result in different cases
- flag — mark something for further attention or reviewflags
- intervention — a planned action to improve health or behaviourinterventions
Tip: hover, focus or tap highlighted words in the article to see quick definitions while you read or listen.
Discussion questions
- What are the main benefits of using an AI tool to flag children at risk of ADHD in primary care?
- What concerns might parents or clinicians have about an AI that flags risk but does not make a diagnosis?
- What further studies would you suggest to check the tool's accuracy and fairness before clinical use?
Related articles
How Long-Term Singlehood Affects Young Adults
A study tracked more than 17,000 young people in Germany and the UK from ages 16 to 29. It found that long periods of singlehood are linked to falling life satisfaction and rising loneliness, while starting a first relationship improves well-being.
Low-cost cooling could help Bangladesh garment workers
A University of Sydney study, published in The Lancet Planetary Health on Monday (20 October), tested simple low-cost cooling in a chamber that mimicked extreme factory heat. Fans and water partly restored productivity; a reflective roof cut indoor temperature by 2.5°C.
