the concept of well-designed clinical trial-based assessment of AI-based interventions is beginning. Worth a review as this forms another nice tool to evaluate AI in medicine. s41591-020-1034-x.pdf (1.05 MB)
I’ve found the SPRINT and CONSORT AI extensions to be limited to Academic/Research AI. We need to be careful not to use these frameworks for production level AI (i.e. those used in day to day clinical support)
I would completely concur. However, one of the issues from a medical technology implementation perspective, is that traditional applications have relied substantially on clinician-level critical appraisal. This is substantially at odds with most of AI-in-healthcare framework. So while academic, things like CONSORT are routinely used by job-a-day clinicians as a framework to help them assess the merits of a drug, surgical approach, etc.