JMIR Publications

ABSTRACT

Artificial intelligence has expanded the ability to observe, detect, and structure clinical performance, but it has not resolved how observed performance should be judged against standards of competence. This distinction is especially important in protocol-driven care, where expected actions, timing, sequence, and decision conditions are often specified in advance. In this Viewpoint, we propose computable clinical competence as a framework for protocol-based adjudication in AI-enabled clinical assessment. The framework separates the perception layer, in which AI systems detect and structure clinical events, from the adjudication layer, in which those events are evaluated against encoded protocol expectations through transparent and reproducible rules. Advanced Cardiac Life Support serves as the initial test domain because it combines explicit protocol structure with clinically consequential challenges of timing, coordination, exception handling, and team performance. We propose that this adjudication framework can support real-time deviation detection, longitudinal competency records, team coordination assessment, system-level protocol reliability analysis, and future learner digital twins. At the same time, its use requires expert governance, validation, auditability, and safeguards against overextension beyond protocol-defined domains. Computable clinical competence therefore does not imply automation of clinical judgment in general; it defines a bounded and testable approach to representing protocol-defined competence as transparent, evidence-preserving computational adjudication.