Presentation (CTI)

OBJECTIVES: Gen AI and LLMs have recently emerged as prominent focal points of innovation and discussion across diverse fields. Current HEOR practices rely on human expertise. Although Gen AI and LLMs have gained traction in HEOR, their applications are not yet standardized. Continuous tracking of their advancements is vital for effective utilization by health economists. The objective of this study was to assess the current applications of Gen AI and LLM in HEOR.
METHODS: A targeted literature review was conducted on EMBASE database to identify publications on application of Gen AI or LLMs in HEOR. A simple targeted search using an appropriate search strategy utilizing AI, LLM, HEOR, and economic modelling related key words was developed. The searches were done for last five years and categorized into evidence synthesis or economic modeling based on where Gen AI and LLMs were primarily used.
RESULTS: A total of 1045 hits were retrieved for title and abstract screening, which was conducted by three reviewers. After full text review, 5 publications for evidence generation and 13 for economic modelling were included. AI application in evidence generation focused on creating key value messages, though it struggled with generating high-quality HTA briefing books. Other uses included qualitative and quantitative data extraction for modelling and network meta-analysis. In modelling, Gen AI and LLMs were used for cost-effectiveness analysis, metamodels, adapting Excel models, and simplifying formulas with high accuracy. Techniques like zero-shot prompting and GPT-4 automated model coding were utilized. Despite promising results, AI tools faced syntax errors and factually inconsistent information. Human expertise was highlighted as essential for refining models and ensuring accuracy in health-economic analyses.
CONCLUSIONS: Gen AI and LLMs show potential in HEOR by enhancing efficiency, automating tasks, and generating insights. Accuracy was still a concern in both; however, replication tasks were likely to be performed with better quality.