Duke Secures $15M for AI in Teen Mental Health

Duke Researchers Awarded $15 Million Federal Grant to Advance AI Model Predicting Teen Mental Illness

Duke University School of Medicine researchers have secured a substantial $15 million grant from the National Institute of Mental Health (NIMH) to enhance and broaden the capabilities of an innovative artificial intelligence (AI) tool designed to predict mental illness in adolescents. This cutting-edge project aims to improve early detection of mental health risks among youth, particularly in underserved rural areas, potentially revolutionizing how mental health care is delivered to young populations.

Background: The Duke Predictive Model of Adolescent Mental Health (Duke-PMA)

The AI model at the center of this initiative, known as the Duke Predictive Model of Adolescent Mental Health (Duke-PMA), analyzes a combination of behavioral data, emotional indicators, and brain function metrics to identify adolescents at high risk of developing mental illness before clinical symptoms manifest. Unlike traditional approaches that primarily focus on diagnosing existing symptoms, Duke-PMA targets root causes such as sleep disturbances and family stress to enable earlier intervention.

In its initial iteration, the model demonstrated an impressive 84% accuracy in predicting worsening mental health outcomes up to a year in advance among children aged 10 to 15. This performance highlights its potential as a transformative tool for primary care providers—including pediatricians and family doctors—to identify at-risk youth early and facilitate timely referrals to mental health services.

Objectives of the Expanded Program

With the new federal funding, the Duke research team plans to:

Enroll 2,000 adolescents from rural clinics located in North Carolina, Minnesota, and North Dakota.
Test and refine the AI model under real-world clinical conditions, focusing on areas where access to mental health resources is typically limited.
Enhance the model’s algorithms by incorporating additional data streams and improving predictive accuracy.
Facilitate integration of the tool into primary care workflows to support frontline healthcare providers in early mental illness detection.

The project leadership includes Dr. Jonathan Posner, J.P. Gibbons Distinguished Professor of Psychiatry and vice chair for research at Duke’s Department of Psychiatry & Behavioral Sciences, and Dr. Matthew Engelhard, assistant professor of biostatistics & bioinformatics.

Scientific and Clinical Significance

Published in Nature Medicine earlier this year, the Duke-PMA study represents a significant advance in psychiatric research by shifting the focus from symptom-based diagnosis to predictive prevention. By identifying modifiable risk factors early—such as poor sleep and family conflict—this approach could enable targeted interventions that may prevent the progression of mental illness in vulnerable adolescents.

This proactive framework aligns with growing public health priorities emphasizing early mental health care to reduce the burden of psychiatric disorders, which are leading causes of disability among young people worldwide. The ability to predict mental illness with such high accuracy could transform clinical practice by:

Reducing delays in diagnosis.
Personalizing treatment plans based on individual risk profiles.
Improving long-term outcomes through early support and resources.

Broader Impact and Future Directions

Expanding the model’s application to rural clinics is particularly important, given the persistent disparities in mental health service availability in these regions. By embedding Duke-PMA into routine primary care screenings, the project aims to bridge gaps in mental healthcare access and reduce inequities in health outcomes.

Furthermore, the AI tool’s data-driven insights could inform policy decisions and resource allocation for adolescent mental health programs. The Duke team’s interdisciplinary approach, combining psychiatry, biostatistics, and bioinformatics, exemplifies the potential of AI to advance precision medicine in mental health.

Visuals and Related Resources

Duke University School of Medicine logo and campus images highlight the institution leading this research.
Photos of Dr. Jonathan Posner and Dr. Matthew Engelhard provide identification of key project leaders.
Illustrations of AI data analysis and adolescent brain imaging contextualize the technological foundation of the Duke-PMA model.
Maps showing rural clinic locations in North Carolina, Minnesota, and North Dakota demonstrate the geographic scope of the study.

This $15 million federal grant marks a pivotal moment for mental health innovation, positioning Duke researchers at the forefront of integrating AI into adolescent psychiatry. Their work promises to enhance early detection, intervention, and ultimately, the mental wellbeing of future generations.

References:

[1] Duke University Department of Psychiatry & Behavioral Sciences, “With $15 Million Grant, Duke Team Expands AI Tool to Predict Teen Mental Illness,” September 18, 2025.
https://psychiatry.duke.edu/news/15-million-grant-duke-team-expands-ai-tool-predict-teen-mental-illness