JMIR Publications

ABSTRACT

Background:

Artificial intelligence (AI) is increasingly integrated into health care systems and health professions education, yet practical guidance for responsible implementation remains limited. Existing guidance and educational resources often emphasize foundational knowledge or individual competencies, with less attention to real-world decision-making, governance, trust, equity, and the organizational processes required to move ethical principles into practice.

Objective:

Building on the previously published Health CARE-AI Framework, this tutorial describes the development, iterative refinement, and formative evaluation of a practical implementation guide and toolkit designed to operationalize responsible AI principles across educational, clinical, research, and administrative contexts in the health professions.

Methods:

The Health CARE-AI Implementation Guide and Toolkit (Navigator) was developed through a multiphase, user-informed process. Phase 1 used a scenario-based workshop and retrospective pretest-posttest survey to examine perceived readiness, confidence, trust in ethically implemented AI, and ability to address ethical concerns while identifying resource needs. Phase 2 used domain-based focus groups to examine clarity, relevance, usability, and practical applicability. Phase 3 involved patient partner co-construction, and Phase 4 comprised a dedicated equity-focused review and member checking. Quantitative changes were assessed using 2-tailed paired tests with Bonferroni correction, with Wilcoxon signed-rank tests conducted as sensitivity analyses. Qualitative feedback was examined deductively in relation to CARE-AI domains and intended resource functions and inductively for emergent concerns and recommendations.

Results:

The scenario-based workshop was attended by 120 individuals; 18 consented to research participation and contributed survey data. Phase-specific participation included domain-based focus groups (n=12), patient partner co-construction (n=4), and an equity-focused review involving 13 returning participants from the preceding phases (9 domain-based focus group participants and 4 patient partners). Retrospective ratings increased across four applied domains, including confidence working in AI-integrated settings, trust in ethically implemented AI, professional readiness to engage with AI, and ability to address ethical concerns. All four increases were significant in paired t tests after Bonferroni correction (adjusted P=.03-.04). In Wilcoxon signed-rank sensitivity analyses, gains in professional readiness and ability to address ethical concerns remained significant after correction. Successive phases led to clearer domain-specific navigation, expanded facilitation and reflection supports, stronger integration of patient trust and transparency, and more explicit guidance addressing accessibility, structural barriers, accountability, and community participation. The final Navigator integrates self-assessment, case-based scenarios and facilitation models, implementation checklists, a strategic roadmap, and action-planning resources.

Conclusions:

The Health CARE-AI Navigator provides a practical, user-informed approach to translating responsible AI principles into context-specific implementation planning across the health professions. The findings suggest that principles-based guidance is strengthened by structured tools for interdisciplinary dialogue, reflection, and action, and by integrating trust and equity throughout resource development rather than treating them as end-stage considerations. Further evaluation should examine implementation behavior, organizational change, and the measurement properties of the Navigator progress tools.