Presentation (CTI)

OBJECTIVES: For health economic (HE) models to achieve clinical face validity, it is essential that each health state accurately reflects meaningful stages of disease to capture the full spectrum of data and natural history. In practice, however, many models are simplified for tractability and data availability. While simplification can aid technical implementation, it may limit meaningful engagement with clinicians and obscure key assumptions during health technology assessment (HTA). This study explores whether artificial intelligence (AI), in a hybrid intelligence framework, can help “re-expand” such models - starting from simplified structures and reconstructing the underlying clinical story - thereby clarifying what simplification omits and enabling better communication with stakeholders.
METHODS: We applied an AI-supported workflow incorporating large language model (LLM) prompt engineering, graph-based retrieval-augmented generation (GraphRAG), and structured reasoning. Starting from common simplified HE model templates, the framework deconstructed model elements, identified potential gaps in clinical representation, and proposed enriched structures. Expert reviewers iteratively assessed AI-suggested refinements to ensure alignment with clinical and economic reasoning. The goal was to expand model fidelity and visually and conceptually clarify the clinical meaning behind each health state.
RESULTS: The AI-guided expansions revealed several overlooked aspects of disease logic, such as treatment sequencing, complication pathways, and quality-of-life dynamics. These additions helped improve communication with clinical stakeholders and created more transparent structures for internal discussion and external submission. Importantly, the process also highlighted what was deliberately excluded in the original simplified models - making simplification itself a more conscious and explainable decision.
CONCLUSIONS: This work demonstrates the strategic value of hybrid intelligence in refining HE models - not just for technical accuracy, but for more effective communication, faster alignment with clinical experts, and better preparedness for payer engagement. It offers a scalable approach for organizations seeking to bridge the gap between operational simplicity and clinical credibility.