Presentation (CTI)

OBJECTIVES: Health economic (HE) models, especially in chronic and progressive diseases, often evolve into highly complex structures that can be challenging to interpret, validate, and communicate - particularly for non-technical stakeholders such as market access and payer teams. This study explores how artificial intelligence (AI), when used in a hybrid intelligence framework, can help simplify model structures without compromising clinical integrity or decision relevance.
METHODS: We developed an AI-powered workflow integrating prompt engineering, graph-based retrieval-augmented generation (GraphRAG), and structured reasoning. The process was applied to a diverse set of HE models across Parkinson’s disease, Alzheimer’s disease, type 2 diabetes, and selected rare conditions. Existing models were deconstructed into their structural elements: health states, transitions, and assumptions and redesigned through AI-supported iterations. Subject matter experts guided each step to ensure preservation of clinical and economic nuances.
RESULTS: The AI-supported redesign led to more streamlined model structures with reduced complexity. The revised models retained critical clinical pathways and economic logic, while becoming more transparent and easier to validate. Market access teams reported improved understanding and confidence in presenting simplified schematics to stakeholders.
CONCLUSIONS: This case series demonstrates the potential of combining AI algorithms with expert oversight to simplify HE model structures across therapeutic areas. The use of GraphRAG and prompt-based reasoning enables intelligent pruning of overly granular states, while retaining fidelity to disease progression and treatment effects. This hybrid intelligence approach enhances model usability, improves communication with decision-makers, and may support faster, more confident HTA submissions. Further work will explore automation scalability and integration with HEOR workflows.