Operating Room Usage Time Estimation with Machine Learning Models.

Area of Science:

• Healthcare Management
• Operations Research
• Machine Learning in Medicine

Background:

• Operating rooms are expensive, limited resources crucial for hospital efficiency.
• Inefficient scheduling leads to increased case-time duration and significant idle time.
• Accurate surgery duration estimation is vital for optimizing operating room utilization.

Purpose of the Study:

• To develop and evaluate a machine learning model for accurate surgery duration prediction.
• To enhance operating room scheduling and reduce overall idle time.
• To provide a more universally usable model with fewer features compared to previous studies.

Main Methods:

• Development of department-specific XGBoost models utilizing a dozen predictive features.
• Evaluation of model performance using metrics such as RMSE, MAE, R2, and MAPE.
• Analysis of the proportion of predictions within a 10% variation.

Main Results:

• The best performing department-specific XGBoost model achieved an R2 of 0.71.
• Key performance metrics included RMSE of 31.6 min, MAE of 18.71 min, and MAPE of 28%.
• The model demonstrated a 27% proportion of estimated results within a 10% variation.

Conclusions:

• Machine learning models, particularly XGBoost, can accurately predict surgery durations.
• Improved prediction accuracy facilitates better operating room scheduling and resource management.
• The developed models offer comparable performance to prior studies with enhanced usability due to fewer features.