GLOBAL LANDSCAPE OF ARTIFICIAL INTELLIGENCE IN HEALTH TECHNOLOGY ASSESSMENT: TRENDS, GOVERNANCE, AND EMERGING PRACTICES
Author(s)
Eon Ting, MBA, MSc1, Matthew Badin, MSc, MBA1, Vivian Vuong, MSc1, Nishu Gaind, MBA2, Mir-Masoud Pourrahmat, MSc2, Luka Ivkovic, MSc2, Thomas Haugli-Stephens, BSc, MPhil, PhD3, Johanna Jacob, PH Lic4, Mir Sohail Fazeli, PhD, MD2.
1AstraZeneca Canada, Mississauga, ON, Canada, 2Evidinno Outcomes Research Inc., Vancouver, BC, Canada, 3AstraZeneca, Oslo, Norway, 4AstraZeneca, Stockholm, Sweden.
1AstraZeneca Canada, Mississauga, ON, Canada, 2Evidinno Outcomes Research Inc., Vancouver, BC, Canada, 3AstraZeneca, Oslo, Norway, 4AstraZeneca, Stockholm, Sweden.
OBJECTIVES: Artificial intelligence (AI) and machine learning (ML) are increasingly applied in health technology assessments (HTAs) to enhance efficiency, rigor, and reproducibility in evidence generation, systematic reviews, economic modeling, real-world data analysis, and internal operations. This study provides a comprehensive overview of AI/ML adoption, guidance, and governance across international HTA agencies.
METHODS: A targeted review of HTA agency websites, policy documents, guidance statements, gray literature, and Embase was conducted from inception to October 01, 2025, covering 17 countries including the United States, Netherlands, England, Canada, France, Germany, Belgium, Norway, Sweden, Finland, Italy, Spain, Denmark, Scotland, Japan, Singapore, and Australia, as well as EUnetHTA/JCA.
RESULTS: Thirty-seven publications from nine HTA agencies were identified, showing that AI/ML is primarily applied to systematic literature reviews, study screening, data extraction, evidence synthesis, economic modeling, and real-world evidence processing. Guidance documents were available from NICE (England), CDA-AMC (Canada), IQWiG (Germany), HAS (France), NIPH (Norway), FIMEA (Finland), KCE (Belgium), and EUnetHTA/JCA. NICE led global efforts with a 2024 position statement detailing the use of ML classifiers, generative AI in modeling, and large language model-assisted reporting, emphasizing human oversight. CDA-AMC’s 2025 guidance aligned with NICE while incorporating principles from Canada’s Artificial Intelligence and Data Act. European agencies piloted ML tools for literature screening, data extraction, and model validation, often applying high sensitivity thresholds (>95%) and supporting staff training and ethical governance. Pilot projects by KCE, NIPH, and the NICE HTA Laboratory advanced AI/ML use in abstract screening, evidence synthesis, and health economic modeling, focusing on best-practice principles, tool evaluation, and human-in-the-loop oversight.
CONCLUSIONS: AI/ML integration in HTAs is growing but remains inconsistent. Standardized frameworks, cross-agency collaboration, and continuous evaluation are essential to ensure ethical, transparent, and reliable implementation. While NICE and CDA-AMC offer structured approaches, broader guidance is needed to support scalable and trustworthy adoption of AI/ML in HTAs worldwide.
METHODS: A targeted review of HTA agency websites, policy documents, guidance statements, gray literature, and Embase was conducted from inception to October 01, 2025, covering 17 countries including the United States, Netherlands, England, Canada, France, Germany, Belgium, Norway, Sweden, Finland, Italy, Spain, Denmark, Scotland, Japan, Singapore, and Australia, as well as EUnetHTA/JCA.
RESULTS: Thirty-seven publications from nine HTA agencies were identified, showing that AI/ML is primarily applied to systematic literature reviews, study screening, data extraction, evidence synthesis, economic modeling, and real-world evidence processing. Guidance documents were available from NICE (England), CDA-AMC (Canada), IQWiG (Germany), HAS (France), NIPH (Norway), FIMEA (Finland), KCE (Belgium), and EUnetHTA/JCA. NICE led global efforts with a 2024 position statement detailing the use of ML classifiers, generative AI in modeling, and large language model-assisted reporting, emphasizing human oversight. CDA-AMC’s 2025 guidance aligned with NICE while incorporating principles from Canada’s Artificial Intelligence and Data Act. European agencies piloted ML tools for literature screening, data extraction, and model validation, often applying high sensitivity thresholds (>95%) and supporting staff training and ethical governance. Pilot projects by KCE, NIPH, and the NICE HTA Laboratory advanced AI/ML use in abstract screening, evidence synthesis, and health economic modeling, focusing on best-practice principles, tool evaluation, and human-in-the-loop oversight.
CONCLUSIONS: AI/ML integration in HTAs is growing but remains inconsistent. Standardized frameworks, cross-agency collaboration, and continuous evaluation are essential to ensure ethical, transparent, and reliable implementation. While NICE and CDA-AMC offer structured approaches, broader guidance is needed to support scalable and trustworthy adoption of AI/ML in HTAs worldwide.
Conference/Value in Health Info
2026-05, ISPOR 2026, Philadelphia, PA, USA
Value in Health, Volume 29, Issue S6
Code
MSR96
Topic
Methodological & Statistical Research
Topic Subcategory
Artificial Intelligence, Machine Learning, Predictive Analytics
Disease
No Additional Disease & Conditions/Specialized Treatment Areas