O2077: The Digital Colleague: Implementation of an AI-Enabled Clinical Assistant for Pediatric Neurology

Background: Graduate medical education in pediatric neurology presents unique challenges, requiring residents to quickly assimilate complex knowledge while delivering patient care. Although valuable, traditional resources such as handbooks and protocols can be cumbersome to access during time-sensitive clinical situations. Additionally, the specialized nature of pediatric neurology creates substantial knowledge gaps among trainees, affecting the quality of care, patient safety, and operational efficiency. The complexity of neurological disorders and rapid advances in diagnostic and therapeutic approaches further compound these challenges. Digital solutions that provide on-demand, context-specific guidance present a promising approach. We developed Brainy McSynapse, an AI-enabled clinical assistant utilizing retrieval-augmented generation technology, integrated into existing clinical communication platforms to provide rapid access to institutional knowledge resources and support clinical decision-making in pediatric neurology.

Methods: Brainy McSynapse was developed using Microsoft Copilot Studio with orchestration capabilities that manage dataflow between knowledge sources and user queries. Our approach was guided by three complementary theoretical frameworks: Design-Based Research methodology, Self-Determination Theory, and the Theory-Technology Alignment Framework. Development began with a comprehensive needs assessment using semi-structured interviews with faculty, residents, and staff to identify key knowledge gaps and workflow constraints. An interdisciplinary advisory committee including clinical informaticians, medical education experts, ethics consultants, nurses, and pharmacists provided oversight throughout development, ensuring the tool addressed multidisciplinary perspectives and use cases. The agent's knowledge base was constructed from the institutional pediatric neurology handbook, evidence-based algorithms, and professional society guidelines, with each response including primary resource citations. Alongside deployment, we implemented a parallel educational curriculum on AI literacy to improve adoption and proper utilization, focusing on appropriate prompt construction and interpretation of AI-generated responses.

Results: Primary outcomes included (1) perceived time efficiency through structured surveys, (2) user satisfaction scores, and (3) adoption rates across user roles. Secondary outcomes included (1) impact on resident confidence in managing complex neurological presentations, (2) practice with effective query formation as measured by analysis of user prompts, (3) impact on resident wellness and cognitive load, (4) response accuracy as validated against expert consensus, and (5) source attribution transparency. The analytics dashboard tracks session metrics, user engagement patterns, and frequency of topic utilization, providing real-time insights into knowledge gaps and educational needs across the division.

Conclusion: Brainy McSynapse demonstrates how generative AI agents can transform graduate medical education by bridging knowledge gaps and enhancing clinical decision support at the point of care. Our methodical approach provides a comprehensive framework for institutions seeking to deploy similar AI tools in specialized medical education contexts. This work extends beyond a single specialty, offering an adaptable model for AI integration that balances technological innovation with educational theory, clinical relevance, and ethical oversight—ultimately advancing patient care through improved access to specialized knowledge when and where it's needed most.