An Ensemble Learning Approach to Improving Prediction of Case Duration for Spine Surgery Using Patient and Surgical Features (Preprint)

Abstract

Estimating surgical case duration accurately is an important operating room efficiency metric.

The primary objective of this 4-year, single academic center retrospective study was to utilize an ensemble learning approach to improve the accuracy of scheduled case duration for spine surgery. The primary outcome measure was case duration.

We compared machine learning models using surgical and patient features to our institutional method, which used historic averages and surgeon adjustment as needed. We implemented multivariable linear regression, random forest, bagging, and XGBoost and calculated average R2, root mean squared error (RMSE), and mean absolute error (MAE) using stratified k-folds cross-validation. We then used the Shapley Additive exPlanations (SHAP) explainer model to determine feature importance.

3,315 patients who underwent spine surgery were included. The institution’s current method of predicting case times had poor coefficient of determination with actual times (R2 = 0.19). On k-folds cross-validation, the linear regression model had an R2 of 0.34, RMSE of 165.3, and MAE of 128.4. Among all models, the XGBoost regressor performed the best with an R2 of 0.70, RMSE of 110.9, and MAE of 75.8. Based on SHAP analysis of the XGBoost regression, body mass index, spinal fusions, surgical procedure, and number of spine levels involved were the features with the most impact on the model.

Utilizing ensemble learning-based predictive models, specifically XGBoost regression, can improve accuracy of the estimation of spine surgery times.

N/A