Durham, N.C.-based Duke University has secured a $15 million federal grant to expand an AI model designed to predict mental illness in adolescents, according to an Oct. 7 report from The Chronicle.
The Duke predictive model of adolescent mental health, or Duke-PMA, was co-developed by Jonathan Posner, MD, a professor of psychiatry, Matthew Engelhard, MD, PhD, an assistant professor of biostatistics and bioinformatics, and Elliot Hill, an AI health fellow. The tool uses questionnaire-based assessments to identify 10- to 15-year-olds at risk of developing a mental illness within one year and pinpoints the key factors contributing to those predictions, according to the report.
The model achieved 84% accuracy in identifying at-risk adolescents and performed consistently across race, sex and socioeconomic status. Researchers said Duke-PMA remained accurate even when limited to modifiable risk factors like sleep disturbances and family conflict, which may help guide early interventions.
The next phase of the project will enroll 2,000 adolescents from rural clinics in North Carolina, Minnesota and North Dakota. The team will use the model to generate risk assessments, then follow up with psychiatric evaluations a year later to measure predictive accuracy, according to the report.
“This is exactly the pathway to get it in [the clinicians’] hands and actually identify people early and connect them with services and support that can hopefully bend that trajectory,” Dr. Engelhard said.
The grant marks a turning point for the initiative’s growth, with researchers emphasizing the tool’s scalability and potential to support clinician decision-making in underserved areas.
