Benchmarking bias: Expanding clinical AI model card to incorporate bias reporting of social and non-social factors

TL;DR

This paper advocates for expanding clinical AI model cards to include comprehensive bias reporting, covering social and non-social factors like disease, anatomy, and instruments, to enhance safety and transparency.

Contribution

It introduces an expanded framework for clinical AI model cards that incorporates non-social bias factors alongside social ones, addressing a gap in current reporting practices.

Findings

Proposes a comprehensive bias reporting framework for clinical AI models.

Highlights the importance of including non-social factors such as disease and instrument effects.

Aims to improve safety and transparency in clinical AI deployment.

Abstract

Clinical AI model reporting cards should be expanded to incorporate a broad bias reporting of both social and non-social factors. Non-social factors consider the role of other factors, such as disease dependent, anatomic, or instrument factors on AI model bias, which are essential to ensure safe deployment.