BACKGROUND
Artificial intelligence (AI) is rapidly expanding in medicine even while lacking formal oversight.
OBJECTIVE
We sought to identify and describe considerations for the oversight of AI in medicine. We also explored where along the translational process (i.e., AI development, reporting, evaluation, implementation, and surveillance) these considerations were targeted.
METHODS
We conducted a targeted review of frameworks for the oversight of AI in medicine. The search included key topics such as ‘artificial intelligence,’ ‘machine learning’, ‘guidance as topic’, ‘translational science’, ‘medical device legislation’, and ‘evaluation study,’ and spanned the time period 2014-2021. Frameworks were included if they described translational considerations for AI. The included frameworks were summarized descriptively. Content analysis was used to identify considerations for the oversight of AI in medicine. An evaluation matrix methodology was used to map each consideration across the different translational stages for each framework.
RESULTS
Six frameworks were featured in the review and included peer reviewed and white papers from consortium and professional organizations. Content analysis of the frameworks revealed five overarching considerations related to the oversight of AI in medicine, including: transparency, reproducibility, ethics, effectiveness, and engagement. All frameworks included discussions regarding transparency, reproducibility, ethics, and effectiveness, while only half of frameworks discussed engagement. The evaluation matrix revealed that frameworks were most likely to report AI considerations for the translational stage of development, and least likely to report considerations for the translational stage of surveillance.
CONCLUSIONS
Frameworks provided broad guidance for the oversight of AI in medicine, but notably offered less input on the role engagement approaches for oversight, and regarding the translational stage of surveillance. Identifying and optimizing strategies for engagement is essential to ensure that AI can meaningfully benefit patients and other end-users.