Utilization of Machine Learning Models to More Accurately Predict Case Duration in Primary Total Joint Arthroplasty

Area of Science:

• Orthopedic Surgery
• Data Science
• Health Informatics

Background:

• Accurate operating room scheduling is crucial for resource allocation.
• Predicting surgical case duration for total hip arthroplasty (THA) and total knee arthroplasty (TKA) is essential for efficient operating room utilization.

Purpose of the Study:

• To implement and evaluate machine learning (ML) models for predicting primary THA and TKA case times.
• To compare the accuracy of ML models against traditional scheduling methods.

Main Methods:

• Retrospective analysis of 10,590 THAs and 12,179 TKAs (July 2017-December 2022).
• Development and comparison of four ML algorithms: linear ridge regression, random forest, XGBoost, and explainable boosting machine.
• Evaluation of model performance based on reducing "underbooking" (wait time) and "overbooking" (excess time) in 15-minute blocks.

Main Results:

• The XGBoost model demonstrated superior predictive accuracy for both TKA and THA case times.
• The ML model significantly reduced TKA "excess time blocks" by 85 and "wait time blocks" by 96.
• The ML model significantly reduced THA "wait time blocks" by 134, improving overall operative booking by 181 blocks for TKA and 138 blocks for THA.

Conclusions:

• Machine learning models significantly outperform traditional methods in scheduling total joint arthroplasty cases.
• The median duration of the prior 30 surgical cases was the most critical factor for accurate scheduling.
• ML utilization in surgical case scheduling is recommended as models continue to advance, with prior case data serving as a viable alternative.