Prediction of Clinically Significant Improvement after Lumbar Fusion Surgery Based on Machine Learning

Area of Science:

• Spine Surgery
• Machine Learning in Healthcare
• Patient Outcomes Research

Background:

• Lumbar fusion surgeries have seen a significant rise, increasing the importance of patient satisfaction as a quality metric.
• Traditional statistical methods struggle to predict clinically significant improvement (CSI) effectively.
• Accurate prediction of CSI is crucial for informed clinical decision-making and patient counseling.

Purpose of the Study:

• To develop and validate a machine learning model for predicting CSI after lumbar fusion.
• To utilize only preoperative patient factors for model development, ensuring applicability in early decision-making.
• To enhance preoperative decision support for both clinicians and patients undergoing lumbar fusion.

Main Methods:

• A cohort of 359 patients undergoing lumbar fusion was analyzed.
• Twenty-two preoperative variables were used to train multi-label classification models.
• Six machine learning algorithms were evaluated using 5-fold cross-validation, with performance metrics including F1-score and AUROC.

Main Results:

• The Extra Trees machine learning model demonstrated superior performance (F1-score: 0.850, AUROC: 0.835).
• Key preoperative predictors identified by SHAP analysis included hypertension, diabetes mellitus, BMI, bone mineral density, and revision status.
• The study also explored associations between comorbidities, postoperative complications, and revision surgery outcomes.

Conclusions:

• A machine learning model effectively predicts clinically significant improvement following lumbar fusion surgery.
• The model leverages preoperative data, offering a valuable tool for preoperative decision support.
• This approach can assist clinicians and patients in anticipating surgical outcomes and optimizing care planning.