Predicting Care Times at PACU.

Area of Science:

• Anesthesiology
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Post-Anesthesia Care Unit (PACU) patient flow is critical for hospital operations.
• Traditional methods for predicting PACU stay are often inadequate due to complex patient data.
• Improving patient throughput in PACU is essential for efficient healthcare delivery.

Purpose of the Study:

• To develop a machine learning (ML) tool for predicting patient care times in the PACU.
• To enhance patient throughput by optimizing PACU resource allocation.
• To increase clinical acceptance of ML tools through interpretable individual predictions.

Main Methods:

• Utilized a large dataset of over 84,000 patients with more than 170 variables.
• Developed and integrated a machine learning model for time prediction.
• Incorporated local-explanation models to provide insights into prediction drivers.

Main Results:

• The ML tool demonstrated potential for accurate PACU time prediction.
• Local-explanation models provided transparency into individual patient predictions.
• The approach is designed to facilitate better clinical decision-making and workflow management.

Conclusions:

• Machine learning offers a promising approach to predict PACU care times.
• Interpretable ML models can improve clinical trust and adoption in healthcare settings.
• This tool has the potential to significantly improve patient throughput in PACU.