Presentation (CTI)

OBJECTIVES: Complex cost-effectiveness models often present a steep learning curve for new project team members, delaying onboarding and complicating communication across technical and non-technical stakeholders. This study evaluated an AI-based model explainer designed to facilitate structured understanding and interrogation of existing economic models during team onboarding.
METHODS: A comparative case study was conducted during the onboarding of two newly formed project teams (three members each) to an existing multi-state Markov cost-effectiveness model. The control group relied on standard onboarding approaches, including review of technical documentation and manual cell-tracing within Excel. The intervention group used an AI model explainer that generates a structured representation of model components (worksheets, formulas, linkages, and outputs) and enables natural-language querying.A standardized knowledge transfer assessment was administered covering model structure, key assumptions and drivers, inter-sheet linkages, interpretation of outputs, and ability to implement predefined model modifications (e.g., time horizon, discounting, scenario inputs). Outcomes included time-to-competency (hours to pass a model logic assessment), query latency for targeted information retrieval, and ability to communicate model results across different technical audiences).
RESULTS: The intervention group reached predefined model competency in 6 hours compared with 20 hours in the control group, representing a 70% reduction in onboarding time. For targeted technical questions, the AI explainer reduced information retrieval time by approximately 95% relative to manual cell-tracing. Team members using the AI tool successfully implemented predefined model modifications and correctly explained expected output changes in all assessed scenarios. The explainer also supported re-framing of model outputs for different stakeholder profiles, from clinical to quantitative audiences.  
CONCLUSIONS: In this exploratory evaluation, an AI-based model explainer substantially reduced onboarding time and improved accessibility of complex Excel-based cost-effectiveness models. Such tools may support more efficient knowledge transfer and communication within distributed HEOR teams, while preserving transparency of model structure and logic.