Study: Machine learning improves OR efficiency, but challenges remain

Editor's Note

Machine learning (ML) is enhancing OR efficiency by optimizing scheduling, predicting surgical durations, and reducing delays, according to a systematic review published February 21 in Cureus. However, privacy concerns, data access limitations, and the need for further validation remain barriers to widespread implementation.

The review analyzed 21 studies on ML applications in OR management, selected from an initial pool of 608 research articles across five major databases. Studies focused on ML’s role in surgical scheduling, workflow optimization, PACU resource allocation, and patient safety. Various ML models, including neural networks, XGBoost, and random forests, were evaluated for their predictive accuracy and efficiency improvements.

Key findings from the studies include:

• ML models significantly improved the accuracy of estimated case durations, allowing for better scheduling and reduced OR inefficiencies. Predictive accuracy varied across models, with ensemble learning techniques such as XGBoost and balanced bagging classifiers demonstrating superior performance.
• ML-based models successfully identified key predictors of extended PACU stays, such as patient BMI, sex, and procedural complexity. AI-driven scheduling adjustments led to measurable reductions in PACU wait times.
• AI models demonstrated potential in predicting and mitigating case cancellations by analyzing patient and procedural factors, helping to minimize costly last-minute disruptions.
• Several studies found that ML applications enhanced OR scheduling, reducing inefficiencies and optimizing case prioritization. Some models outperformed human schedulers in forecasting surgery duration within 10% accuracy.

Despite these benefits, Privacy concerns, data standardization issues, and limitations in access to high-quality datasets hinder ML adoption in OR management, researchers write. Additionally, most studies relied on retrospective data, necessitating further prospective validation in real-time clinical settings.