Clinical Deployability: A Socio-technical Construct for Evaluating Real-World Readiness of Medical AI

Medical AI evaluation still largely emphasizes model performance under controlled conditions. However, whether an AI-enabled clinical capability can be introduced, used safely and appropriately, and sustained in routine care depends on additional socio-technical conditions. We propose clinical deployability as a second-order, multi-level construct capturing this readiness for real-world clinical work. Deployability integrates three coordinated domains: (A) task usability and safety-in-use; (B) implementation workability (embedding, routinization, sustainment); and, (C) accountability (in-use decision performance and risk governability). We outline operational criteria and a pragmatic measurement strategy based on a deployability dashboard rather than a single scalar score. This paper presents an ongoing research program; future work will apply the framework in real-world implementation case studies to produce comparative deployability assessments and refine the indicator set.