Anesthesiology

Single-centre deployment of an anaesthesia mortality-prediction model, with operational tailoring.

A single-centre case study describes implementing a mortality-prediction model to trigger enhanced preoperative review by a float anaesthesiologist. Key operational decisions included selecting a decision threshold, refreshing input data every 6 hours, and reducing input features for real-time use. The model was tailored to the local workflow to support add-on cases rather than as a standalone triage replacement. This report demonstrates practical implementation steps but does not provide multicentre performance or impact on mortality. 1

Reinforcement-learning automated anesthesia for GI endoscopy was non-inferior to clinician management.

In a multicentre randomized trial of adults undergoing GI endoscopy, an RL-based system (AAS-GE) produced similar hypoxemia rates to clinicians (14.42% vs 14.29%; OR 1.01). AAS-GE shortened induction time (median 1.55 vs 1.90 minutes) without increasing total drug dose or recovery time. Intraoperative patient movement was more common with AAS-GE, consistent with a lighter anesthesia depth. The system supported safety and efficiency in ASA I–II adults but applicability to higher-risk patients and other agents remains untested. 2

Mixed reality intubation simulation improved nursing student performance and readiness.

In a randomized trial of 92 nursing students, mixed reality intubation training increased intubation performance (effect size 0.65) versus conventional simulation. Readiness and satisfaction improved modestly in the MR group (ES 0.24 for both outcomes). Knowledge improved in both groups, indicating MR augmented skill and confidence rather than basic knowledge. Because the MR group had additional contact time, some between-group gains may reflect more practice rather than the technology alone. 3