OBJECTIVES: This study aims to explore the application of Generative Pre-trained Transformers (GPT) in developing early health economic models and predicting cost-effectiveness in the context of early-stage Health Technology Assessment (HTA) submission planning and discussions. The focus is on assessing the feasibility and accuracy of GPT for generating foundational data, crucial in informing early pricing strategies and go/no-go decisions before proceeding to comprehensive HTA-compliant models.

METHODS: GPT's advanced natural language processing capabilities using Open AI’s GPT 4.0 were employed to analyze and synthesize data from a diverse range of sources, including clinical trials published report, statistical analysis plan, epidemiological research, and existing economic evaluations. The study primarily used GPT to construct an early health economic model, focusing on identifying cost-effectiveness scenarios, key economic indicators and evidence gaps. Comparative analysis was conducted between GPT-generated early-stage health economic models with a published model to evaluate the consistency and reliability of the AI-driven approach.

RESULTS: GPT-enhanced models exhibited an increased capacity to process diverse datasets, yielding more nuanced early-stage economic assessments. The scenarios identified by GPT closely mirrored those obtained from subsequent, detailed HTA-compliant models. Notably, GPT was effective in pinpointing critical variables impacting cost-effectiveness and providing potential evidence gap that need to be filled prior to HTA submission.

CONCLUSIONS: The study indicates that GPT could significantly contribute to early health economic modeling, enhancing the precision and efficiency of initial cost-effectiveness analyses. This is especially valuable for early decision-making and resource allocation in healthcare technology development. Despite its potential, the study underscores the need for expert oversight in interpreting and applying GPT outputs, ensuring their applicability and relevance to real-world healthcare scenarios. Future research should focus on further integrating GPT into health economic modeling workflows, improving its interpretability, and optimizing its application in the dynamic field of HTA.

ISPOR 2024

May 5-8, 2024

Program