Artificial intelligence (AI) is being increasingly applied in surgical healthcare to support perioperative risk assessment, enhance diagnostic accuracy, refine clinical workflows, assist intraoperative decision-making and reduce postoperative complications. Reported AI approaches include machine learning predictive analytics, computer vision, robotic assistance and clinical decision support systems with promising clinical performance across surgical specialties. Despite developing evidence on technical performance, knowledge about real-world implementation — including system-level outcomes, barriers and facilitators — remains fragmented and lacks comprehensive synthesis.
This systematic review protocol aims to collect and synthesise global evidence on how AI is implemented in surgical settings, focusing on implementation determinants, strategies used, adoption outcomes and reported impacts on care processes and outcomes.
The primary objective is to identify, categorise and synthesise evidence on the barriers and facilitators influencing implementation of AI applications in surgical healthcare. Secondary objectives comprise characterising implementation outcomes (for example, adoption, feasibility, acceptability, fidelity, penetration and sustainability), cataloguing implementation strategies, and summarising reported effects on surgical quality metrics, workflow efficiency, and cost and resource utilisation.
The review will include a broad range of study designs to capture implementation evidence: randomised controlled trials, observational studies, implementation studies, qualitative and mixed-methods research, and programme evaluations. Eligible studies will report on AI applications used within surgical healthcare contexts and provide data relevant to the primary or secondary outcomes described.
A comprehensive electronic search will be performed across multiple bibliographic and grey literature sources for the period January 2000 through December 2025. Databases to be searched include PubMed/Medline, CINAHL (EBSCO), Web of Science, Scopus, Global Index Medicus, medRxiv, IEEE Xplore and the ACM Digital Library. The search strategy is designed to capture clinical, technical and implementation-oriented literature relating to AI in surgery.
The primary outcome is the set of barriers and facilitators to AI implementation identified in the literature; these will be synthesised using framework synthesis based on Carroll et al. No quantitative meta-analysis will be performed for the primary outcome.
Secondary outcomes include implementation outcomes such as adoption, feasibility, acceptability, fidelity, penetration and sustainability. These secondary implementation outcomes will be characterised using the RE-AIM (Reach, Effectiveness, Adoption, Implementation, Maintenance) framework. Additional secondary outcomes include reported effects on surgical quality metrics, workflow efficiency, and cost and resource utilisation.
Qualitative and mixed-methods findings pertinent to barriers and facilitators will undergo framework synthesis to identify and categorise implementation determinants. For quantitative secondary outcomes, where three or more studies report comparable outcome data with effect estimates, a random-effects meta-analysis will be performed. If fewer than three comparable quantitative studies exist for a given outcome, meta-analysis will not be undertaken for that outcome.
Risk of bias for included studies will be appraised using the Mixed Methods Appraisal Tool to accommodate diverse study designs. Certainty of quantitative findings will be evaluated using GRADE (Grading of Recommendations Assessment, Development and Evaluation). Certainty of qualitative evidence will be assessed using GRADE-CERQual.
Where data permit (minimum three studies per subgroup), subgroup analyses are planned by geographic setting, surgical specialty and AI modality. Sensitivity or additional exploratory analyses will be contingent on the volume and heterogeneity of available data and will be specified based on the characteristics of included studies during review conduct.
This protocol is reported in accordance with the Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols (PRISMA-P) 2015 guidelines and is registered with PROSPERO (CRD420261305972). Ethical approval is not required for the review. Findings will be disseminated via peer-reviewed publication and presentations at relevant conferences to inform clinicians, implementers and researchers interested in deploying AI in surgical care.