Development and Evaluation of an Operative Case Length Prediction Model in Adult Surgical Patients

Area of Science:

• Health Informatics
• Machine Learning in Healthcare
• Surgical Workflow Optimization

Background:

• Surgical care represents a significant portion of U.S. healthcare costs.
• Existing surgical case length prediction models are often inaccurate or too specialized.
• Inefficient operating room (OR) scheduling leads to dissatisfaction among patients and providers.

Purpose of the Study:

• To develop a machine learning (ML) model for predicting surgical case length.
• To compare the ML model's performance against an embedded electronic health record (EHR) model.

Main Methods:

• Retrospective analysis of 55,495 surgical cases (January 2022 - April 2024).
• Temporal data split for training (46,767 cases) and validation (8728 cases).
• ML models utilized patient, provider, operation, and hospital data available pre-surgery.

Main Results:

• The ML model significantly outperformed the EHR model in accuracy (lower RMSE and MAE, higher R²).
• ML model achieved 61.0 min RMSE vs. 91.0 min for EHR (P < 0.01).
• ML model improved prediction accuracy by 35% in cases lacking historical time data.

Conclusions:

• A comprehensive ML model significantly enhances surgical case length prediction over EHR systems.
• Implementation of this ML model can optimize OR efficiency.
• Improved prediction accuracy can lead to greater patient and provider satisfaction.