Machine Learning Prediction Models to Reduce Length of Stay at Ambulatory Surgery Centers Through Case Resequencing.

Area of Science:

• Anesthesiology
• Health Informatics
• Machine Learning in Healthcare

Background:

• Post-anesthesia care unit (PACU) length of stay is a critical metric for perioperative efficiency.
• Prolonged PACU stays can impact staffing and resource utilization, particularly for ambulatory surgery centers.

Purpose of the Study:

• To develop machine learning models for predicting prolonged PACU length of stay (≥ 3 hours) in ambulatory surgery patients.
• To utilize pre-operative factors exclusively for risk prediction.
• To simulate the impact of these models on reducing after-hours PACU staffing needs through optimized case sequencing.

Main Methods:

• Development and evaluation of various machine learning classifier models on a training dataset.
• Identification of the best-performing model (XGBoost with SMOTE) based on predictive accuracy (AUC = 0.712).
• A case resequencing simulation on a test set, reordering historic cases by predicted risk of prolonged PACU stay.

Main Results:

• Analysis included 10,928 ambulatory surgical patients, with 5.31% experiencing a PACU stay of 3 hours or longer.
• The XGBoost model demonstrated strong predictive performance (AUC = 0.712).
• Case resequencing using the XGBoost model significantly reduced instances of after-hours PACU occupancy (12% vs. 41%, P<0.0001).

Conclusions:

• Pre-operative patient characteristics can be effectively used by machine learning models to predict prolonged PACU stays.
• Optimized case sequencing based on these predictions can substantially mitigate the impact of extended PACU durations on after-hours staffing.
• This approach offers a promising strategy for improving operational efficiency in ambulatory surgical settings.