ABSTRACT
This makes statements about how best to approach the evaluation of AI/ML in healthcare somewhat speculative. This, combined with knowledge of the potential shortcomings of AI and lessons learned from previous health information technology implementations, allows us to suggest an approach to evaluation that will likely remain relevant even as our understanding of AI/ML evolves. If the AI/ML claims align with our organizations purpose, then the learners should proceed to next stage of evaluation, which is plausibility of claims. If the claims are supported by the evidence from trials, learners should move to next phase of evaluation, a decision to implement. The principles of HIT evaluation can be applied to the evaluation of AI/ML in healthcare but there are some unique considerations relating to our adoption of development approaches and tools that have been matured and optimized to serve business needs outside of healthcare setting.
