The evolving literature on the ethics of artificial intelligence for healthcare: a PRISMA scoping review

This scoping review analyzes the literature on the ethics of artificial intelligence (AI) tools in healthcare to identify trends across populations and shifts in published research between 2020 and 2024. We conducted a PRISMA scoping review using structured searches in PubMed and Web of Science for articles published from 2020 to 2024. After removing duplicates, the study team screened all sources at three levels for eligibility (title, abstract, and full text). We extracted data from sources using a Qualtrics questionnaire. We conducted data cleaning and descriptive statistical analyses using R version 4.3.1. A total of 309 sources were included in the analysis. While most sources were conceptual articles, the number of empirical studies increased over time. Commonly addressed ethical concerns included bias, transparency, justice, accountability, privacy/confidentiality, and autonomy. In contrast, disclosure of AI-generated results to patients was infrequently addressed. There was no clear trend indicating greater attention to this topic within our period of review. Among all eligible sources, the proportion addressing legal and policy issues broadly has shown a declining trend in recent years. There was an uptick in the number of sources discussing legal liability, patient acceptability, and clinician acceptability. Yet, these three topics remained infrequently addressed overall. Significant gaps in research on the ethics of AI applications in healthcare include disclosure of results to patients, legal liability, and patient and clinician acceptability. Future research should focus more on these ethical issues to facilitate the responsible and appropriate implementation of AI in healthcare.