Researchers at Duke Health have developed an artificial intelligence (AI) model that can predict when teenagers are at high risk for developing serious mental health issues—before their symptoms become severe.
Unlike previous models that mainly focus on existing symptoms, this AI system identifies hidden risk factors like sleep problems and family conflicts. This means doctors could step in earlier and help prevent mental health issues from escalating.
The tool could also make mental health care more accessible by allowing primary care doctors—not just specialists—to assess children’s mental health risks.
Addressing the Youth Mental Health Crisis
“The U.S. is facing a youth mental health crisis—almost half of all teenagers will experience a mental illness,” said Dr. Jonathan Posner, a professor at Duke University and senior author of the study, which was published in Nature Medicine.
At the same time, the U.S. has a shortage of mental health providers, making it difficult for many young people to get the help they need. This AI model could help pediatricians and family doctors quickly assess a child’s risk level and step in before the situation worsens.
How the AI Model Works
Posner and his team used data from the ABCD study, which has tracked the brain and mental health development of more than 11,000 children over five years.
The researchers trained an AI system modeled after the brain’s neural networks to predict which children would move from low risk to high risk for mental illness within a year. The AI then scores a questionnaire filled out by the child or their parents, ranking their behaviors, feelings, and symptoms to determine how likely they are to develop more serious mental health problems.
- The model was 84% accurate in predicting which children would experience worsening mental health within the next year.
- A second version of the model, which identifies the causes behind the risk, was 75% accurate.
Key Risk Factors for Worsening Mental Health
Instead of just predicting risk, the AI system also identified what might be causing a child’s mental health to decline. These underlying risks include:
- Sleep problems (the strongest predictor)
- Behavioral issues
- Family history of mental illness
- Difficult life experiences
- Family conflict
“It’s easier to predict mental illness in children who already show a lot of symptoms,” said researcher Elliot Hill. “But what’s truly valuable is predicting mental illness before symptoms become severe, so we can intervene early.”
A Game-Changer for Primary Care
Right now, most primary care doctors don’t have time for in-depth psychiatric evaluations. This AI tool could automate the process—analyzing the data in real time and giving doctors a clear risk assessment for each child.
“This model could help identify children who need early mental health support before they reach a crisis,” Dr. Posner said. “If we can catch these issues early, we can provide better care and improve mental health outcomes for millions of young people.”
The research findings can be found in Nature Medicine.
Copyright © 2025 Knowridge Science Report. All rights reserved.
