Mental Health AmericaYusen Zhai, Ph.D.
Photography: Andrea Mabry predicts that nearly 60 million Americans suffered from a mental illness in 2024, and research from the University of Alabama at Birmingham shows that post-traumatic stress disorder and acute stress disorder diagnoses have risen significantly in American college students since 2017.
To address the rise of mental health issues in the United States is seeing, researchers from the UAB School of Education and Human Sciences have developed a tool to assist counselors in identifying college students at heightened risk of anxiety and depressive disorders — before their conditions intensify.
In his recently published article, Yusen Zhai, Ph.D., director of the UAB Community Counseling Clinic, used machine learning, a subtype of artificial intelligence, to spot patterns in information that schools already collect, such as age, biological sex, years in school, race and ethnicity, and majors, that could be indicators of a higher risk of mental health conditions.
“Current predictive models used to find students struggling with anxiety or depression rely on information from students who have already visited a health provider,” Zhai said. “This can overlook students who might not even realize they need help, are reluctant to seek care and experience barriers to services. Our AI tool can find students who might need help, and the exciting and fascinating part is that it works well among the general student population, not just those who have already used health services.”
Zhai further explains that many of the current models are built using clinical samples and data and are not exposed to the general student population during the training process. To bridge this gap, he and his team aimed to develop machine learning predictive models that do not rely on clinical samples or health-related information, but rather socioeconomic demographics that research has shown to be associated with more severe anxiety and depression. Factors such as gender, race and ethnicity, financial stress, a sense of belonging on campus, disability status, and age are considered in the model’s evaluation.
Without constraints and reliance, Zhai’s AI models demonstrate strong predictive accuracy and can help identify at-risk individuals from a broad college student population.
“Timely treatment is crucial as delays can lead to more severe health outcomes,” Zhai said. “By leveraging the potential of AI, mental health professionals can make more informed decisions and develop plans for students. It is important to note that these models are not designed to replace the critical role of human health providers, but to complement their expertise and enhance mental health professionals’ ability to make informed decisions while maintaining the relational and human-centered nature of counseling.”
Looking ahead, Zhai and his research team want to build on the knowledge and experience gained from this project, and work to develop machine learning models designed to identify a broader range of mental health issues, including substance use disorders and suicide risk, with the end goal of extending the application of the tools to K-12 settings and the general population.
“Mental health disparities are associated with serious health outcomes, and with the rapid advancement of technology, we want to continue our research efforts to create tools that assist mental health professionals in addressing gaps in mental health care and improving outcomes.”
This study was supported by the UAB Faculty Development Grant Program in the Office of the Provost. Co-authors of this study included members from the UAB Department of Computer Science, Baocheng Geng, Ph.D., and Yixin Zhang, M.S. and Xue Du, Ph.D., in the UAB Heersink School of Medicine.