Predicting Hospital Length of Stay: A Feature Selection Approach to Predictive Modeling in Lumbar Fusion Surgery

Introduction: Predictive modeling in lumbar fusion surgery can assist in clinical decision making. Herein, we have identified 20 key variables that were used to train machine learning (ML) models to identify patients at risk of longer hospital length of stays (LOS) after lumbar fusion.

Methods: The American College of Surgeons database was queried for all lumbar fusions from 2012–2022. Anterior Lumbar Interbody Fusion, Posterolateral Interbody Fusion (PlatIF), Posterior Lumbar Interbody Fusion (PLIF), and combined PLIF+PlatIF codes were used to identify patients. Multivariate methods with Unbiased Variable selection in R (MUVR) and Boruta were used to select 20 variables that demonstrated the greatest importance in predicting hospital LOS. Hierarchical clustering and a 5-fold cross-validation of the methods including our selected variable were used to ensure the robustness and reliability of our findings. These 20 features were used to train the following ML classifiers: tree-based (random forest, XGBoost, CatBoost, LightGBM), kernel-based (SVM), neural networks, and ensemble methods (voting, stacking), along with a linear regression model, to compare predictive performance.

Results: 114,892 patients were included. The neural network marginally outperformed all other models, with both an accuracy and discriminative ability (AUC) of 71.2%. All models achieved an accuracy and AUC within 0.6% of the logistic regression model. The neural network also outperformed all other models in recall and F1 score, with all models being within 5.0% and 2.0% of each other respectively. However, the neural network performed the poorest of all models in terms of precision (68.9%), with all models being within 1.2% of each other.

Conclusion : These results demonstrate that feature selection is the key to predictive modeling, regardless of the ML model or approach that is used. We show that when selecting the variables that demonstrate the greatest importance in predicting hospital LOS, all models approached a similar level of accuracy and discriminative ability.