Background: Postoperative ambulation status after spinal metastasis surgery is currently difficult to predict. Improved ability to predict this important postoperative outcome would improve management decision-making and help in determining realistic goals of treatment. Accordingly, this study set forth to develop machine learning models to predict ambulation outcomes after surgery for spinal metastasis.
Methods: This retrospective study included patients who underwent spinal metastasis at a university-based medical center in Thailand during January 2009-November 2021. Collected data included preoperative parameters, and ambulatory status at 90 and 180 days after surgery. Seven machine learning algorithms, including decision tree, random forest, XGBoost, logistic regression, support vector machine (SVM), neural network, and stochastic gradient descent (SGD), were developed to predict ambulatory status at 90-days and 180-days postoperation. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC) and F1-score.
Results: A total of 178 patients were enrolled. The number of patients classified as ambulatory at 90-days and 180-days after surgery was 150 (84.3%) and 145 (81.5%) respectively. The XGBoost algorithm was found to most accurately predict 180-day ambulatory outcome (AUC: 0.92, F1-score: 0.92), and the random forest algorithm was shown to most accurately predict 90-day ambulatory outcome (AUC: 0.95, F1-score: 0.94).
Conclusion: Machine learning algorithms were shown to be effective for predicting ambulatory status after surgery for spinal metastasis. The XGBoost and random forest algorithms best predicted postoperative ambulatory status at 180-days and 90-days after spinal metastasis surgery, respectively.