AI tools to help unveil patterns between social determinants of health, mental health risks, among autistic youth
November 10, 2025

By Mike McNulty

A new USC-led study will apply artificial intelligence tools to electronic health records (EHRs) in order to better understand how social determinants of health — non-medical factors impacting health, such as education, income, housing and neighborhood resources like grocery stores and transportation access — are related to mental health of autistic children and teens.

Social determinants of health account for up to 50 percent of health outcomes, yet they are mostly unstudied in autism research, which has traditionally focused on racial and ethnic variables for understanding care disparities. That limitation is especially problematic because autistic children and youth, compared to neurotypical populations, have disproportionately higher rates and frequencies of psychiatric care, such as hospital stays and emergency room visits. That suggests their mental health care needs were not adequately managed at lower levels of care, like primary care or psychiatry.

Using an AI method known as natural language processing, the research team will flag and extract keywords and phrases from anonymized clinical notes already stored within the EHR systems at Children’s Hospital Los Angeles and the University of Florida Health System. Then, using another AI approach called machine learning, the researchers will develop a model for predicting which social determinants pose relatively higher mental health risks, and for whom. That type of individualized risk assessment will empower clinicians and health systems to customize the mental health care and resources they offer to autistic patients.

The study is funded by a $3.6 million grant from the National Institutes of Health (NIH) National Institute of Mental Health.

“We know there is valuable data about the social and contextual drivers of mental health hidden in plain sight within patient records,” said Assistant Professor Amber Angell, the study’s principal investigator. Angell holds a joint appointment in the Department of Pediatrics at the Keck School of Medicine of USC.

“By using AI to analyze large datasets of clinical records, we will use an existing resource to gain an evidence-based understanding of what social determinants protect autistic people’s mental health, and what increases vulnerability,” Angell said. “That will help individuals, families, providers and policymakers to enact innovative solutions that enhance mental health in autism by decreasing adverse outcomes.”

At the USC Chan Division, Angell directs the Disparity Reduction and Equity in Autism Services, or “DREAmS” lab, a multidisciplinary group that includes neurodivergent team members working together to identify, measure, understand and reduce disparities in autism diagnosis and services. For this particular project, they are joined by collaborators from the USC Viterbi School of Engineering, Children’s Hospital Los Angeles, the University of Texas Health Science Center at San Antonio, the University of Florida and the University of Indiana.

The team will also work closely with a community advisory board of autistic adults and caregivers of autistic children and youth. The board’s participation ensures the project is relevant, responsive and impactful in light of their own lived experiences, and its members will help spread and share the study’s findings as they become available.

The study team will also gather insight from a range of people who work in systems of care that provide services for autistic people — including physicians, therapists and public school and Regional Center administrators — to collectively determine how the study findings can be practically implemented within health systems to identify the most vulnerable autistic children and proactively provide appropriate supports.

“We know that, for an autistic teen experiencing a mental health crisis, the emergency department is one of the worst, most dysregulating environments to get care, so we want to minimize that by providing robust, lower-level care that meets individual needs,” Angell said. “Every year in the United States, at least half-a-million autistic children turn 18 and become autistic adults. This project will give us a much clearer idea of ways to optimize mental health for today’s autistic youth, who are tomorrow’s autistic adults.”

“Machine learning prediction of persistent adverse mental health outcomes for autistic children: Leveraging social determinants of health from clinical data” (PI: A. Angell; 1R01 MH135867) is funded by the NIH National Institute of Mental Health.

⋯