AI Model Assesses Patient Engagement in Tele-Mental Health Sessions  

Patient engagement remains critical in behavioral health care, yet assessing engagement in telehealth settings presents unique challenges. Traditional, in-person therapy allows therapists to gauge patient involvement by observing nonverbal cues, such as body posture and gestures. However, in virtual sessions, these cues are either diminished or absent, requiring therapists to rely heavily on verbal interactions.

To address this challenge, a team of researchers from the University of Maryland, in collaboration with the University of Maryland’s School of Medicine, has developed a machine learning model to estimate patient engagement during telemental health sessions. The study, led by Pooja Guhan, a Ph.D. student in the Department of Computer Science, was recently published in a top medical journal.

The research, “Developing a Machine Learning–Based Automated Patient Engagement Estimator for Telehealth,” examines the potential of artificial intelligence to assess patient engagement levels and assist therapists in fostering a stronger therapeutic relationship with clients during virtual mental health sessions.

The study estimates engagement levels using AI-driven analysis of behavioral and contextual signals, providing therapists with additional insights beyond traditional verbal communication. By offering clinicians data-driven feedback on engagement patterns, the model aims to enhance therapeutic interactions in telehealth settings and improve patient outcomes.

“Our research bridges the gap between in-person and virtual mental health therapy,” Guhan said. “Providing additional insights into patient engagement can help therapists tailor their approaches for more effective interactions.”

The COVID-19 pandemic saw a surge in telehealth adoption, underscoring the necessity of alternative ways to measure patient involvement without in-person interactions. Guhan and her team aimed to address this by developing a psychology-informed multimodal learning approach to estimate the perceived patient engagement levels in remote therapy.

The researchers also developed the MEDICA dataset, incorporating multimodal data sources to train machine learning models in engagement estimation. This dataset serves as a foundation for further studies on AI applications in mental health care. By analyzing verbal and contextual patterns, the model provides therapists with real-time feedback on the perceived patient responsiveness, allowing them to adjust strategies accordingly.

“This interdisciplinary work between AI researchers and psychiatry researchers presents a novel application of machine learning for estimating patient engagement,” said Distinguished University Professor of Computer Science Dinesh Manocha, a co-author of the study. “It introduces techniques to assist psychotherapists, which could spur further research in this area.”

The study was supported by the MPower initiative, which facilitates research collaborations between University of Maryland institutions. Other contributors to the research include Professor Gloria Reeves and Professor Kristin Bussell from the University of Maryland School of Medicine and Aniket Bera, an adjunct associate professor in UMD’s Department of Computer Science.

The interdisciplinary nature of this research highlights the intersection of artificial intelligence and mental health care. Guhan emphasized the importance of cross-field collaboration in advancing AI applications beyond traditional computing.

“Interdisciplinary research in computer science expands innovation by integrating knowledge from domains like psychology, healthcare, and the arts,” Guhan said. “It enables the development of AI-driven solutions that are technically robust and socially relevant.”

Guhan expressed appreciation for the opportunity to work on the project and for the support from her colleagues and advisors.

“This project has been a rewarding journey, pushing the boundaries of how we can improve tele-mental health care through technology,” she said. “I’m grateful to my team of co-authors and our advisors for their support.”

As AI continues to integrate into various aspects of health care, studies like this provide insights into its potential to enhance patient experiences and treatment outcomes.

"We hope this work enhances telehealth platforms," Guhan said. "Improving accessibility and effectiveness can help ensure that more patients receive the mental health support they need to achieve better outcomes."

—Story by Samuel Malede Zewdu, CS Communications