OBJECTIVES: The growing adoption of artificial intelligence (AI) technologies raises important ethical challenges in healthcare. Current health technology assessment (HTA) frameworks are not adequately responsive to the (bio)ethics challenges introduced by AI-based technologies. We aim to build an ethics framework using a lifecycle approach in HTA, to adequately incorporate responsible AI principles and sociotechnical ethics perspectives in value assessments.

METHODS: We conducted a scoping review of published reviews on HTA addressing ethical dimensions of AI technologies in PubMed, HTA databases, and grey literature. The structure of our scoping review follows the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) guidelines. The search strategy was built focusing on the “HTA”, “AI” and “ethics” terms. Studies were screened according to a priori eligibility criteria.

RESULTS: As of January 12, 2023, we included 16 full-text, English-language, review articles. The extracted data were summarized in evidence tables. A multi-faceted understanding of complex ethical issues of AI technologies was documented, including the following considerations: beneficence, patient integrity, privacy (i.e., patient confidentiality), equity (i.e., fair use and access to AI), trust, transparency, responsibility, accountability, and autonomy (i.e., human oversight of AI). None of the included studies reported an iterative approach to systematically address ethical issues of AI technologies and their responsible integration into health systems.

CONCLUSIONS: Building a lifecycle ethics framework for AI technologies in healthcare can be achieved by documenting ethical issues across the lifecycle of each technology, including the phases of design, development, regulation, market access, evaluation, and real-world evidence generation. We encourage further research on the ethical integration of AI technologies in healthcare and an update on the HTA frameworks that systematically include additional (bio)ethics dimensions in future AI technology assessments.