JMIR Publications

ABSTRACT

Background:

Applications of Artificial Intelligence (AI) in healthcare have emerged as a pivotal aspect of the ongoing digital transformation of health systems worldwide. Despite the potential benefits across diverse medical domains, one significant barrier to successful AI adoption remains prevailing low trust by users.

Objective:

Overcoming this challenge requires addressing complexities in defining trust and trustworthiness and applying these concepts to real-world medical scenarios. In response to these challenges and to foster warranted trust, this paper advocates for a more refined and nuanced conceptual understanding of trust in the context of AI in healthcare.

Methods:

Employing a combination of framework analysis and an iterative modified Delphi process involving international experts from diverse disciplines, we present and discuss a comprehensive consensus-based framework for trust in AI in healthcare. This framework is developed through the examination of five case studies encompassing diagnosis, clinical risk assessment, public health surveillance, assistive technology, and healthcare resource allocation.

Results:

Our analysis reveals that certain consequences of trust, such as user acceptance, and certain aspects of trustworthiness, such as system accuracy, are applicable across all discussed cases. However, a multitude of crucial aspects is case-specific, influenced by the respective system’s environment, involved actors, and pertinent framing factors.

Conclusions:

By synthesising insights into commonalities and differences from these case studies, this paper establishes a foundational basis for future debates and discussions on trust in medical AI. The outcomes of this analysis aim to inform the improvement of AI system design, adoption processes, and governance within the healthcare domain.