MT6 Oversight of Artificial Intelligence in Medicine: A Review of Frameworks

Abstract

Artificial intelligence (AI) is rapidly expanding in medicine even while lacking formal oversight. 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. 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’, ‘implementation 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. Six frameworks were included in the review, and were either published as peer reviewed journal articles or 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. Frameworks provided broad guidance for the oversight of AI in medicine, but notably offered less input on the role engagement approaches for oversight, and for and any strategies for general surveillance stage of translation. Identifying and optimize strategies for engagement is essential to ensure that AI can meaningfully benefit patients and other end-users.

Ähnliche Arbeiten