Artificial intelligence–assisted perioperative risk prediction, decision support, and supervised automation in anesthesia: A narrative review
- 1Department of Anesthesiology, Peking University Third Hospital, China
- 2Senior Department of Anesthesiology, PLA General Hospital (PLA Medical School), China
Abstract
Artificial intelligence is increasingly being investigated as a tool for supporting personalized perioperative care. This narrative review, guided by the Scale for the Assessment of Narrative Review Articles principles, examines clinically relevant applications of artificial intelligence and related automation in anesthesiology, with emphasis on three linked functions: risk prediction, clinical decision support, and supervised automation. We performed a structured search of PubMed, Web of Science, and Scopus databases and identified English-language literature published from January 2010 through June 2026. We prioritized clinical studies, guidelines, systematic reviews, and representative methodological studies. A central distinction of this review is that machine-learning prediction, rule-based decision support, conventional feedback control, and genuinely adaptive artificial intelligence–enabled control are not treated as equivalent technologies. Current evidence is strongest for selected prediction tasks and supervised systems that improve process measures such as time within a physiological target range. However, evidence that these technologies improve patient-centered outcomes remains less consistent. Model discrimination alone is insufficient; calibration, actionable thresholds, false-positive burden, external validation, clinical utility, dataset shift, missing data, and workflow consequences must also be considered. The proposed three-level framework is intended to provide anesthesiologists and perioperative teams with a practical way to connect technical capability with bedside action. Near-term implementation should prioritize interpretable prediction and low-burden decision support, whereas closed-loop systems should remain under clinician supervision. Future studies should emphasize prospective multicenter validation, clinically meaningful endpoints, transparent reporting, and post-deployment monitoring rather than technical performance alone.
Related
- Artificial intelligence and machine learning for prediction of postoperative complications: A systematic review focused on anesthesiology, perioperative risk stratification, and clinical applicability.
- Automation and artificial intelligence in the clinical microbiology laboratory
- Bridging the Artificial Intelligence Translation Gap in Perioperative Pulmonary and Respiratory Complications: A Practical Framework for Institution-Level Implementation
- Perioperative Blood Management in General Surgery: Recent Advances, Current Limitations, and the Emerging Role of Artificial Intelligence-A Narrative Review.
- Artificial intelligence in perioperative pressure injury management: a scoping review of evidence, gaps, and implications for population health and digital public health policy
BibTeX
@article{an2026artificial,
title = {{Artificial intelligence–assisted perioperative risk prediction, decision support, and supervised automation in anesthesia: A narrative review}},
author = {Ran An and Bingbing Meng and Xun Lu},
journal = {The Journal of International Medical Research},
year = {2026},
month = sep,
doi = {10.1177/03000605261491422},
url = {https://doi.org/10.1177/03000605261491422},
note = {CC BY-NC 4.0},
}
RIS
TY - JOUR
TI - Artificial intelligence–assisted perioperative risk prediction, decision support, and supervised automation in anesthesia: A narrative review
AU - Ran An
AU - Bingbing Meng
AU - Xun Lu
PY - 2026
DA - 2026/09/29
JO - The Journal of International Medical Research
DO - 10.1177/03000605261491422
UR - https://doi.org/10.1177/03000605261491422
AB - Artificial intelligence is increasingly being investigated as a tool for supporting personalized perioperative care. This narrative review, guided by the Scale for the Assessment of Narrative Review Articles principles, examines clinically relevant applications of artificial intelligence and related automation in anesthesiology, with emphasis on three linked functions: risk prediction, clinical decision support, and supervised automation. We performed a structured search of PubMed, Web of Science, and Scopus databases and identified English-language literature published from January 2010 through June 2026. We prioritized clinical studies, guidelines, systematic reviews, and representative methodological studies. A central distinction of this review is that machine-learning prediction, rule-based decision support, conventional feedback control, and genuinely adaptive artificial intelligence–enabled control are not treated as equivalent technologies. Current evidence is strongest for selected prediction tasks and supervised systems that improve process measures such as time within a physiological target range. However, evidence that these technologies improve patient-centered outcomes remains less consistent. Model discrimination alone is insufficient; calibration, actionable thresholds, false-positive burden, external validation, clinical utility, dataset shift, missing data, and workflow consequences must also be considered. The proposed three-level framework is intended to provide anesthesiologists and perioperative teams with a practical way to connect technical capability with bedside action. Near-term implementation should prioritize interpretable prediction and low-burden decision support, whereas closed-loop systems should remain under clinician supervision. Future studies should emphasize prospective multicenter validation, clinically meaningful endpoints, transparent reporting, and post-deployment monitoring rather than technical performance alone.
SN - 0300-0605
SN - 1473-2300
ER -