Artificial intelligence and machine learning applications in ambulatory surgery – A systematic review

Background: We aimed to systematically review applications of artificial intelligence (AI) technologies for ambulatory surgical patients. Methods: We systematically searched PubMed, Scopus, Web of Science, and EBSCOhost (2015–2025). Studies were included if they used artificial intelligence in ambulatory surgical populations. Results: Of 26 studies identified, machine learning was used in 25, with a predominantly orthopaedic (65.3 %) focus. Except for two, all were originated in the USA. We found four themes: (1) Preoperative patient selection (n = 10) – Random forest (RF) and eXtreme gradient boost (XGBoost) algorithms predicted appropriateness with an area under curve (AUC) 0.72–0.85, (2) Same-day discharge prediction (n = 8) – Ensemble models demonstrated the highest AUC values (3) Postoperative management and complications (n = 3) – Artificial neural network incorporating intra- and postoperative features predicted opioid refill needs (4) Cost prediction (n = 4) – Ensemble models consistently outperformed single-model approaches. Conclusions: Our review underscores the promising potential of machine learning applications in ambulatory surgery, particularly with ensemble methods. We observed inconsistencies in the models; data related issues and a lack of external validation.