When AI Enters the OR, Patients Win

A review of 21 studies finds machine learning cuts surgical risk before, during, and after the operation

8 May 2026

Machine learning is now demonstrating measurable patient safety gains across every stage of complex surgery. Results are substantial. Published in Patient Safety in Surgery in November 2025, a narrative review of 21 peer-reviewed studies mapped AI and ML performance across preoperative, intraoperative, and postoperative care, delivering the clearest evidence picture US hospital systems have had to date.

Before an incision is made, findings are consistent. ML models outperformed conventional risk scores in identifying patients most likely to face serious complications, drawing on electronic health record variables ranging from laboratory values to socioeconomic indicators. Earlier identification reshapes how surgical teams prepare, shifting resource allocation from reactive to genuinely anticipatory.

Reaching further, intraoperative evidence is compelling. A randomized trial found AI decision support directly reduced patient exposure to dangerous drops in blood pressure, a condition linked to acute kidney injury and elevated postoperative mortality. Computer vision systems also verified safety-critical procedural steps and tracked instruments in real time, addressing two of surgery's most persistent sources of preventable harm.

Postoperatively, results extend the picture. Multimodal models combining structured records, imaging, and smartphone wound photographs predicted complications including surgical site infections with clinically meaningful accuracy, pointing toward a future where postoperative monitoring follows patients into home recovery and cuts costly readmissions.

Challenges persist. Most studies are retrospective, single-center, or prototype-stage, and external validation across diverse US hospital populations remains limited. Algorithmic bias, interoperability with existing clinical infrastructure, and FDA risk classification frameworks are unresolved at scale.

For US hospitals managing surgical backlogs and workforce pressure, the evidence is accelerating. Targeted AI investment across the perioperative continuum is no longer speculative. Safer surgery, powered by data, is arriving faster than many anticipated.