Machine learning-enhanced prediction of operating room occupation time and length of stay: a retrospective cohort study on emergency surgery care pathways
Authors' conclusion
As for elective cases, MLMs can improve OT and LOS predictions in emergency surgery, though predictive performance remains moderate.
Study context
- Population
- Emergency procedures at the Pitié-Salpêtrière hospital, a major trauma center, 2015–2018
- Participants
- 3,117 emergency procedures
- Exposure
- ML models (Ridge Regression, Random Forest, XGBoost, Multi-Layer Perceptron) vs manual surgeon estimates for OR occupation time and length of stay (LOS) prediction
- Outcome
- Prediction performance for OR occupation time (OT) and LOS (Mean Absolute Error, Root Mean Square Error, Mean Absolute Percentage Error, OT within 20% of actual value, LOS within fixed-day thresholds)
Methodological detailssingle-center, machine learning prediction models
- Single-center
- Machine learning prediction models
Main finding
As reported in the abstract
OT and LOS prediction
| Group | Occupation time (OT) MAE | OT Within20 | LOS MAE | LOS RMSE |
|---|---|---|---|---|
| RF | 32 min | 60% | ||
| XGB | 5 days | 12 days |
Occupation time (OT) MAE
RF32 min
XGB— not stated in the abstract
OT Within20
RF60%
XGB— not stated in the abstract
LOS MAE
RF— not stated in the abstract
XGB5 days
LOS RMSE
RF— not stated in the abstract
XGB12 days
- RF improved surgeon OT estimates by
- 13%
Observational data: associations, not effects.
Further findings
Prediction performance by specialty (MAPE)
- Performance varied across specialties, better accuracy in digestive and maxillofacial procedures
Original publication
Machine learning-enhanced prediction of operating room occupation time and length of stay: a retrospective cohort study on emergency surgery care pathways. J Clin Monit Comput. Published 18 August 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →