OBJECTIVES: Artificial intelligence (AI) and machine learning (ML) methods have potential applications in Health Economics and Outcomes Research (HEOR) and Health Technology Assessment (HTA). However, the use of AI/ML is still very new and not much research has been conducted on its use in HTA. This study aimed to examine the current use of AI/ML in HTA and how this might evolve in the future.

METHODS: We conducted a search using the relevant keywords (machine learning, artificial intelligence, neural network, deep learning, supervised learning and LASSO) to identify HTA reports that incorporated AI/ML methods. The keywords were also translated to German, French, Italian and Spanish. The analysis focused on HTAs from seven countries (Australia, Canada, France, Germany, Italy, Spain and United Kingdom) and EUnetHTA published between 2012 and 2023. Additionally, HTA policies and methodology guidelines were reviewed.

RESULTS: Analysis identified 10 HTA reports (2 from the UK, 6 from Canada, 1 from Italy, 1 from France) where AI/ML methods were mentioned. Six identified HTAs from Canada mentioned principal component analysis, considered a traditional approach, used to validate patient-reported outcomes (PRO) instruments. Remaining HTAs mentioned AI/ML in the context of post-hoc subgroup analyses, determining the probability of cure in economic models or were assessing technologies based on AI/ML. In the review of HTA policies and methodology guidelines, we found France, Italy and Germany mentioning the use of AI/ML for medical devices (France), in clinical trials (Italy, France) or in conducting systematic literature reviews of clinical evidence (Germany, EUnetHTA). In Canada, horizon scanning reports providing an overview of clinical applications of AI/ML have been published.

CONCLUSIONS: To date, the usage of AI/ML in HTA remains very limited. The significance of AI/ML can be expected to gradually increase as the HTA agencies begin to consider these methods in their methodology guidelines.