Artificial Intelligence for Perioperative Risk Prediction in Anesthesiology: A Bibliometric and Knowledge-Mapping Analysis

Background: Artificial intelligence (AI) is increasingly being investigated in perioperative medicine for risk stratification, complication prediction, patient safety, and clinical decision support. This study characterized the research landscape, intellectual structure, and thematic evolution of AI-based perioperative risk prediction in anesthesiology. Methods: Publications indexed in the Web of Science Core Collection between 2018 and 3 June 2026 were analyzed using Bibliometrix and Biblioshiny. Scientific productivity, citations, collaboration networks, keyword co-occurrence, thematic evolution, and Reference Publication Year Spectroscopy were evaluated. Partial 2026 data were excluded from the compound annual growth-rate calculation. Results: A total of 152 publications were included. Scientific production increased substantially, with a compound annual growth rate of 77.72% between 2018 and 2025. The United States and China led publication output, while the United States had the highest total citation count. Citation and thematic analyses identified perioperative outcome prediction, predictive hemodynamic monitoring, and explainable AI as important components of the field’s intellectual structure. Research attention increasingly shifted from methodological machine learning development toward clinically oriented topics, including mortality, postoperative complications, delirium, postoperative nausea and vomiting, and perioperative risk stratification. Emerging themes included explainable AI, large language models, natural language processing, and clinical decision-support systems. Conclusions: Research attention has increasingly shifted from exploratory model development toward clinically oriented risk-stratification and decision-support applications. Emerging directions emphasize interpretability, multimodal data integration, and individualized risk assessment. These bibliometric patterns reflect changes in research emphasis but do not demonstrate clinical adoption or effectiveness.