JMIR Publications

ABSTRACT

Background:

Adults living with serious mental illness and co-occurring chronic medical conditions experience substantial barriers to chronic disease self-management. Artificial intelligence-supported digital interventions may offer scalable approaches to addressing this gap in care.

Objective:

This pilot study examined the feasibility, acceptability, and preliminary effectiveness of My AI Companion, an AI-supported text and call-based intervention designed to support adults with serious mental illness and co-occurring chronic medical conditions.

Methods:

Nine participants enrolled in the study, and seven completed post-intervention assessments (with a retention rate of 77.8%). Participants completed baseline and one-month follow-up assessments evaluating anxiety, depression, loneliness, self-efficacy, social functioning, and health-related quality of life. Measures included the Generalized Anxiety Disorder-7, Geriatric Depression Scale, PROMIS-29, PROMIS Depression, PROMIS Anxiety, PROMIS Physical Function, PROMIS Social Roles, UCLA Loneliness Scale, Self-Efficacy for Managing Chronic Disease Scale, and Net Promoter Score. Wilcoxon signed-rank tests were used to examine pre- and post- intervention changes. Engagement was assessed through text message frequency over the one-month intervention period.

Results:

Participants ranged in age from 31 to 71 years (M = 50.1, SD = 14.5). The majority were female (85.7%) and White (85.7%). Participants demonstrated substantial psychiatric and medical comorbidity, including generalized anxiety disorder (85.7%), major depressive disorder (57.1%), diabetes (42.9%), and arthritis (42.9%). Over the one-month intervention period, participants exchanged a total of 918 text messages with My AI Companion, with individual engagement ranging from 27 to 330 messages. Directional improvements were observed across several mental health and psychosocial measures. Mean Generalized Anxiety Disorder-7 scores decreased from 6.57 (SD = 5.88) to 4.71 (SD = 3.59) (p = 0.41), while PROMIS Depression scores decreased from 57.59 (SD = 12.01) to 52.96 (SD = 8.17) (p = 0.34). Self-efficacy for managing chronic disease improved from 5.83 (SD = 3.06) to 7.90 (SD = 1.42) (p = 0.19), and PROMIS Social Roles scores increased from 47.59 (SD = 10.34) to 50.83 (SD = 9.62) (p = 0.75). PROMIS-29 overall scores remained relatively stable over time. Participants reported high satisfaction with the intervention, with a mean Net Promoter Score of 7.71 (SD = 2.36).

Conclusions:

My AI Companion was feasible and acceptable among adults with serious mental illness and chronic health conditions, with participants demonstrating sustained engagement throughout the intervention period. Findings suggest potential improvements in anxiety, depressive symptoms, self-efficacy, and social functioning. These findings support the potential of AI-enabled companions as a scalable approach to augment behavioral health support and warrant evaluation in larger, adequately powered effectiveness trials.