Generative AI for Health Technology Assessment: An ISPOR Report

This ISPOR Working Group Report, titled "Generative Artificial Intelligence for Health Technology Assessment: Opportunities, Challenges, and Policy Considerations," offers a comprehensive review of the emerging landscape of generative artificial intelligence (AI) and foundation models, including large language models (LLMs), within the field of Health Technology Assessment (HTA)14. Authored by Rachael L. Fleurence and colleagues, the article aims to fill a critical gap by exploring the opportunities, limitations, and risks associated with developing, evaluating, and deploying these novel AI tools for supporting evidence generation in Health Economics and Outcomes Research (HEOR)5.

The report delves into three key methodological areas where generative AI shows significant promise for transforming HTA processes:

Systematic Literature Reviews (SLRs) and Evidence Synthesis: The authors highlight generative AI's potential to automate various aspects, such as proposing search terms, screening abstracts and full texts, extracting data, and even generating code for meta-analyses6....

Real-World Evidence (RWE): The article discusses how generative AI can enhance efficiency in data processing and analysis, particularly by facilitating information extraction from unstructured clinical notes and images, and improving accuracy through domain-specific LLMs69.

Health Economic Modeling: Generative AI is explored for its capacity to support different phases of model development, from conceptualization and parameterization to implementation, evaluation, and validation. It also shows potential for automating resource-intensive tasks like structural uncertainty analysis6....

Despite these promising applications, the report emphasizes that current generative AI tools are in their early stages and present significant limitations413. These include concerns regarding scientific validity and reliability (e.g., inaccuracies, "hallucinations," and reproducibility challenges)4..., the potential for bias, equity, and fairness issues (e.g., propagation of biases from training data, underrepresentation of marginalized groups)4..., and crucial regulatory and ethical considerations (e.g., data privacy, security, informed consent, and the risk of patient harm from inaccurate information)4....

Crucially, the authors stress that generative AI tools should augment human activities rather than autonomously replace them, with humans retaining full accountability for generated results22.... The article concludes by providing concrete suggestions for HTA agencies and policymakers on responsibly integrating generative AI into their workflows. These suggestions include developing clear guidance, harmonizing standards, investing in workforce training, ensuring health equity considerations, and fostering transparency and rigorous reporting on LLM usage2.... The report underscores the rapidly evolving nature of this technology and the continued necessity for careful evaluation and human oversight in its application to HTA for the foreseeable future

Reference: Fleurence, R. L., Bian, J., Wang, X., Xu, H., Dawoud, D., Higashi, M., Chhatwal, J., on behalf of the ISPOR Working Group on Generative AI. (2025). Generative Artificial Intelligence for Health Technology Assessment: Opportunities, Challenges, and Policy Considerations: An ISPOR Working Group Report. Value in Health, 28(2), 175–183. https://doi.org/10.1016/j.jval.2024.10.3846

Видео Generative AI for Health Technology Assessment: An ISPOR Report канала Health topic