Decision-centered artificial intelligence for perioperative care outside the operating room: a practical review for surgeons

It proposes a decision-centered framework to reorganize the surgical AI literature, classifying each study by the clinical decision point it was intended to support rather than by algorithmic type or predicted outcome. Existing reviews have organized prior studies primarily by algorithmic type or predicted outcomes; this review instead follows when and why AI is integrated into real-world clinical workflows, addressing the insufficiently developed structured understanding of how AI outputs inform surgical decisionmaking. This is a narrative review whose strength derives from organizing PubMed-indexed literature from 2015 to 2025 by decision point, rather than from comparative model performance.

It identifies a difference in emphasis between preoperative and postoperative AI: preoperative AI predominantly supports patient selection and treatment planning in the context of diagnostic uncertainty, while postoperative AI mainly supports time-sensitive management and prognostic assessment. This finding explicitly distinguishes the preoperative and postoperative phases, noting that although both involve risk prediction, they differ fundamentally in decision contexts, data characteristics, and modes of clinical application. The conclusion rests on categorizing the reviewed studies by decision point, a literature-level pattern synthesis; the text reports no specific effect sizes or sample sizes.

It reframes surgical AI as an integral component of phase-specific clinical decision pathways across the surgical care continuum, rather than a standalone predictive instrument. This reframing shifts attention from model performance to the functional position of AI within clinical workflows, offering surgeons a new lens on AI's practical uses. This is a conceptual framework proposed by the review, supported by its overall organization of the literature by decision point.

It clearly delimits scope: focusing on AI that supports preoperative and postoperative decision-making, covering decision points such as patient selection, treatment planning, surgical extent and approach planning, postoperative monitoring, discharge readiness, and surveillance planning, while explicitly excluding intraoperative AI such as surgical video analysis, robotic automation, and real-time image guidance. By drawing clear boundaries, the review treats perioperative care outside the operating room as a distinct topic and emphasizes decision points relevant to minimally invasive surgical practice. The scope definition is a methodological statement of the review, delimiting the clinical settings to which its conclusions apply.

The review's conclusions apply to preoperative and postoperative decision settings in perioperative care outside the operating room, targeting surgeons and perioperative care teams, and are especially relevant to minimally invasive surgical practice, covering decision points such as patient selection, treatment planning, surgical extent and approach planning, postoperative monitoring, discharge readiness, and surveillance planning; intraoperative AI applications such as surgical video analysis, robotic automation, and real-time image guidance are explicitly outside its scope.

As a review-level synthesis, it provides no sample sizes, effect sizes, or model performance metrics for the individual studies, so it cannot be used to judge how specific AI tools actually perform at each decision point; moreover, it does not elaborate on the specific differences among included studies in data characteristics and modes of clinical application, so readers assessing the maturity of evidence at a given decision point would still need to return to the original studies.