Related Experiment Video
Updated: Jun 29, 2026
05:28
A Mouse Model of Lumbar Spine Instability
Published on: April 23, 2021
Development and validation of a machine learning model utilizing the ACS-NSQIP database for predicting early
Mert Marcel Dagli1, Jaskeerat Gujral1, Connor A Wathen1
1Department of Neurosurgery, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, USA.
Brain & Spine
|June 15, 2026
Summary
A machine learning model accurately predicts 30-day re-operation after lumbar microdiscectomy due to surgical site infection (SSI). This advance aids in better risk assessment for this rare but serious complication.
Area of Science:
- Spine surgery outcomes research
- Clinical informatics
- Machine learning in healthcare
Background:
- Postoperative surgical site infection (SSI) following lumbar microdiscectomy can necessitate early re-operation, leading to significant patient morbidity and increased healthcare costs.
- Traditional statistical and machine learning (ML) models have shown variable performance in predicting this critical outcome.
Purpose of the Study:
- To develop and validate a highly accurate ML model for predicting 30-day re-operation after lumbar microdiscectomy in patients experiencing SSI.
- To identify key predictors of re-operation risk using interpretable ML techniques.
Main Methods:
- Analysis of a large cohort (79,870 cases) from the American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP).
- Development of an extreme gradient boosting (XGBoost) model integrated with Synthetic Minority Over-sampling Technique (SMOTE) for data augmentation.
- Utilized a nested cross-validation pipeline with Bayesian optimization and SHapley Additive Explanations (SHAP) for model interpretability.
Main Results:
- The final ML model demonstrated exceptional performance with an area under the receiver operating characteristic curve (AUC) of 0.996 and accuracy of 0.993.
- High sensitivity (0.924-0.985) and perfect negative predictive value (1.000) were achieved across evaluated thresholds.
- SHAP analysis identified surgical site infection, race, smoking status, diabetes, and revised Risk Analysis Index as significant predictors of re-operation.
Conclusions:
- An optimized ML model, enhanced with synthetic data, effectively predicts the infrequent but critical complication of re-operation following SSI in lumbar microdiscectomy.
- These findings highlight the potential of data-driven approaches for improving perioperative risk stratification and guiding targeted preventive interventions.
