ARTIFICIAL INTELLIGENCE IN PREDICTING POSTOPERATIVE COMPLICATIONS IN GENERAL SURGERY AND PERIOPERATIVE CARE: A SCOPING REVIEW

Abstract

Postoperative complications are a significant cause of surgical morbidity and mortality worldwide, posing challenges to clinical planning and health resource allocation. Artificial intelligence (AI), particularly through machine learning techniques, has emerged as a promising tool for risk prediction and clinical decision support in the perioperative period. This study aimed to map and analyze scientific evidence on AI applications for predicting postoperative complications in general surgery through a scoping review. The search was conducted in PubMed and Virtual Health Library (BVS) databases using the DeCS/MeSH descriptors: "Artificial Intelligence", "Postoperative Complications", and "General Surgery". After sequential application of eligibility filters, 21 studies were included. Results showed that machine learning and deep learning models demonstrate superior accuracy compared to isolated clinical judgment in predicting outcomes such as ICU admission, acute kidney injury, anastomotic leakage, intraoperative hypotension, postoperative pain, and surgical mortality. It is concluded that AI represents a relevant advance in perioperative care, although gaps remain regarding prospective validation and ethical implementation in health systems.