JMIR Publications

ABSTRACT

Background:

Artificial intelligence (AI) is rapidly being integrated into clinical care, yet structured opportunities for medical trainees and healthcare professionals to develop foundational AI and health data literacy remain limited. Psychiatry is a particularly relevant setting because clinical assessment relies heavily on complex, multidimensional information and objective biomarkers remain limited. Educational approaches that provide hands-on experience with clinically relevant healthcare data may help address this gap.

Objective:

To describe two distinct but complementary educational initiatives developed by The Ottawa Hospital and the University of Ottawa: the Artificial Intelligence Learning Opportunities in Healthcare Analytics (AILOHA) workshop for medical students and psychiatry residents and an Epic Cosmos datathon for clinicians and researchers, and to explore participant perspectives within and across initiatives.

Methods:

We conducted a descriptive case study of the development, implementation, and preliminary evaluation of the two initiatives. AILOHA was an accredited, two-session workshop in which small groups used synthetic datasets derived from electronic health records (EHRs) to develop psychiatric research questions, conduct analyses, and present findings. The Epic Cosmos datathon was a one-day event using a federated EHR network, co-led by faculty, data scientists, and healthcare executives, in which multidisciplinary teams addressed clinical questions. Evaluation focused on project completion, post-session participant ratings, and free-text feedback analyzed thematically within and across initiatives.

Results:

Twenty-five participants completed AILOHA, with four groups presenting projects addressing diagnostic delay in bipolar disorder, ethnocultural differences in delirium outcomes, obsessive-compulsive versus generalized anxiety disorder, and hospitalization following a first bipolar diagnosis. The workshop quality was rated 3.9/5 (SD 0.7), 15/25 participants (60%) would recommend it to peers, and it was accredited for 6.5 hours of continuing professional development. The Epic Cosmos datathon included 52 clinicians and researchers who developed 12 multidisciplinary projects. Of 38 attendees who provided feedback, 25 provided quantitative item-level responses. Of these, 21/25 (84%) self-rated content as good or excellent, 18/25 (72%) would recommend the datathon, and mean comfort with further exploring Cosmos was 3.7/5 (SD 0.8). Five datathon projects subsequently progressed to independent research. Thematic analysis identified three themes per initiative and four cross-initiative themes, including authentic, clinically anchored data work, interdisciplinary collaboration, and support needs that differed by learner stage.

Conclusions:

AILOHA and the Epic Cosmos datathon illustrate complementary models for experiential AI and health data education for learners at different stages of professional training. Feedback suggests that support needs shift from structured guidance toward sustained analytic partnership as experience increases. Project-based learning using privacy-aware healthcare data may provide a feasible approach to introducing practical analytic and AI concepts within medical education. Objective pre- and post-training assessments and longitudinal follow-up are needed to determine whether such initiatives produce measurable and sustained competency gains.