Artificial intelligence in surgery research: Successfully implementing AI clinical decision support models.

Abstract

Artificial intelligence (AI) in surgery literature typically encompasses decision support models that aim to help clinicians make better decisions. Many studies report developing and validating models, yet few models are implemented at the bedside. Exceedingly few models achieve their intended goal upon implementation. While the TRIPOD-AI and DECIDE-AI guidelines outline separate reporting standards for the development/validation, and staged implementation of AI models, respectively, this article outlines how future implementation should be considered at the outset before model development. Building on lessons from high-performing AI decision support models that faced challenges upon implementation, we will discuss study design consideration for building trustworthy and actionable AI clinical decision support models that can cross the database-to-bedside gap and become successfully implemented.