Evaluation framework to guide implementation of AI systems into healthcare settings

Abstract

ObjectivesTo date, many artificial intelligence (AI) systems have been developed in healthcare, but adoption has been limited. This may be due to inappropriate or incomplete evaluation and a lack of internationally recognised AI standards on evaluation. To have confidence in the generalisability of AI systems in healthcare and to enable their integration into workflows, there is a need for a practical yet comprehensive instrument to assess the translational aspects of the available AI systems. Currently available evaluation frameworks for AI in healthcare focus on the reporting and regulatory aspects but have little guidance regarding assessment of the translational aspects of the AI systems like the functional, utility and ethical components.MethodsTo address this gap and create a framework that assesses real-world systems, an international team has developed a translationally focused evaluation framework termed ‘Translational Evaluation of Healthcare AI (TEHAI)’. A critical review of literature assessed existing evaluation and reporting frameworks and gaps. Next, using health technology evaluation and translational principles, reporting components were identified for consideration. These were independently reviewed for consensus inclusion in a final framework by an international panel of eight expert.ResultsTEHAI includes three main components: capability, utility and adoption. The emphasis on translational and ethical features of the model development and deployment distinguishes TEHAI from other evaluation instruments. In specific, the evaluation components can be applied at any stage of the development and deployment of the AI system.DiscussionOne major limitation of existing reporting or evaluation frameworks is their narrow focus. TEHAI, because of its strong foundation in translation research models and an emphasis on safety, translational value and generalisability, not only has a theoretical basis but also practical application to assessing real-world systems.ConclusionThe translational research theoretic approach used to develop TEHAI should see it having application not just for evaluation of clinical AI in research settings, but more broadly to guide evaluation of working clinical systems.

Automating Rey Complex Figure Test scoring using a deep learning-based approach: A potential large-scale screening tool for congnitive decline

“…Third, applying AI systems in the medical eld requires rigorous evaluation [35,36]. Therefore, validity of predicted RCFT scores using our model was veri ed in two ways.…”

Multi-cohort transcriptomic subtyping of B-cell acute lymphoblastic leukemia

“…Notably, the Australian samples were collected from routine health care settings, which provides a realistic spread of sample quality and leukemia burden as encountered in clinical practice. Furthermore, we used additional datasets to help assess the quality of the samples and safety against misclassification, which is an important practical consideration outside research settings [31]. The data were obtained from three Western countries, and caution is warranted if applying the findings in different ethnic or socioeconomic context.…”

Moving towards vertically integrated artificial intelligence development

“…Increasing attention is being paid to translational aspects of clinical AI 27 . Recent frameworks 28 and maturity classifications in literature reviews 1,29 adopt a high-level view of where an algorithm sits in its development roadmap. These supplement checklists for risk of bias and reporting that are internal to algorithm training and evaluation, for prediction 30 and diagnostic accuracy 31 , which focus on model-building 32 and generalisability 33 .…”