OBJECTIVES: Interest in using large language models (LLMs) for health economics (HE) and informing health technology assessment (HTA) strategies has increased in recent years; however, feasibility of integration into existing workflows remains unclear. A targeted review was conducted to identify case studies and guidance on the use of LLMs for HE and understanding HTA body decision-making.

METHODS: Embase, congress abstracts (ISPOR, HTAi, Cochrane Colloquium), HTA guidance and ISPOR good practice guidelines were reviewed. Quantitative and qualitative studies (evaluation studies, observational studies, discussion papers and preprints) published after November 2022 were included.

RESULTS: Nine case studies and one commentary article were included. HE case studies (6/9) focused on economic model conceptualization, development and adaptation (4/6), input derivation (1/6) and reporting (1/6). One HE study used an LLM to build a de novo model; however, its reporting was limited to a conference abstract, precluding comprehensive evaluation of its validity. No HTA guidance for LLM usage in economic modelling was identified. HTA case studies (3/9) analyzed previous HTA decisions (2/3) and simulated HTA committee discussion (1/3). Barriers to LLM usage for HE (extracted from the commentary article and six cases studies) included issues with inaccurate or misleading model responses (5/7), code explainability (1/7), LLM data security (3/7), and HTA body acceptance (1/7). Across HE and HTA, study replicability was poor – conference abstracts (8/9) reported minimal methods and results.

CONCLUSIONS: Evidence of LLM usage for HE and understanding HTA decision-making was limited. Many potential use cases remain unexplored – no identified studies used LLMs for deterministic sensitivity analyses, model validation or adaptation beyond input updates. Better understanding of the amount of LLM response validation needed to maintain research quality is necessary before LLMs can be routinely used for economic modelling.