Presentation (CTI)

OBJECTIVES: Increasing use of artificial intelligence (AI) in health research raises questions around transparency, ethics, and methodological integrity. We assessed how journals adopted AI-guidance for manuscript preparation and publication, compared the scope and depth of such guidance, and identified gaps for HEOR communication.
METHODS: A comprehensive review was conducted of author instructions, editorial/ethics policies from the top 10% of journals per Journal Citation Reports in two categories: Medicine, General & Internal (‘clinical journals’) and Health Care Sciences & Services (‘HEOR journals’). AI-guidance was assessed for presence, format, and content, including authorship, disclosure/reporting, copy-editing, images, peer review, and journal/editor use. Findings were compared between HEOR and clinical journals.
RESULTS: Fifty-two journals were included (HEOR: 19; clinical: 33), published by 20 publishers (mean Impact Factor 14.4: HEOR 7.7, clinical 18.3). AI-guidance was present in 50/52 journals: HEOR 19/19, clinical 31/33; journal-level 31/52, publisher-level 19/52. Guidance was embedded within author instructions (26/50) or editorial/ethics policies (24/50). Nearly all guidance prohibited AI in authorship (48/50; HEOR: 18/19, clinical: 30/31) and all required disclosure of AI-use (50/50). AI for copy-editing without disclosure was permitted by 24/50 journals (HEOR: 11/19; clinical: 13/31). Guidance on AI-generated images was provided by 20/50 journals (HEOR: 11/19; clinical: 9/31), mostly prohibiting use with defined exceptions or allowing use with limitations. Peer reviewer guidance was addressed by 19/50 journals (HEOR: 6/19; clinical: 13/31), generally allowing use with disclosure (15/19). Only 5/50 specified AI-use by journals/editors within internal editorial processes (HEOR: 1/19; clinical: 4/31). Overall, HEOR journals emphasized practical, use-oriented guidance, including ISPOR’s encouragement of responsible AI application, whereas clinical journals emphasized limitations, transparency, and risk mitigation.
CONCLUSIONS: AI-guidance for publication is widespread, but gaps remain in consistency, granularity, and HEOR-specific applicability, particularly regarding analytical workflows, model development, and value communication. Tailored HEOR-guidance could support responsible AI-use while preserving transparency, credibility, and decision relevance in publications.