Classifying High-Risk Patients for Persistent Opioid Use After Major Spine Surgery: A Machine-Learning Approach

Area of Science:

• Neurosurgery
• Data Science
• Pharmacology

Background:

• Persistent opioid use is a significant concern following major spine surgery.
• Prolonged opioid exposure can lead to dependence and escalation.
• Predictive modeling aims to identify patients at risk for long-term opioid use.

Purpose of the Study:

• To develop and evaluate machine-learning models for predicting persistent opioid use after major spine surgery.
• To identify key predictors of persistent opioid use in this patient population.

Main Methods:

• Five classification models were assessed: logistic regression, random forest, neural network, balanced random forest, and balanced bagging.
• Synthetic Minority Oversampling Technique (SMOTE) was employed to address class imbalance.
• Performance was measured using F1 score and area under the receiver operating characteristics curve (AUC), with feature importance determined by SHapley Additive exPlanations (SHAP).

Main Results:

• The balanced random forest classifier achieved the highest AUC (0.877) and F1 score (0.758).
• Key predictors identified by SHAP analysis included age, preoperative opioid use, preoperative pain scores, and body mass index.
• The model demonstrated strong performance with an accuracy of 0.780, specificity of 0.883, sensitivity of 0.700, and precision of 0.836.

Conclusions:

• The balanced random forest classifier is the most effective model for identifying persistent opioid use post-spine surgery.
• Accurate prediction can aid in early intervention and management of opioid use disorder.