Developing Non-linear K-nearest Neighbors Classification Algorithms to Identify Patients at High Risk of Increased Length of Hospital Stay Following Spine Surgery

Introduction: Spondylolisthesis is a common operative disease in the United States, but robust predictive models for patient outcomes remain limited. Development of models that accurately predict postoperative outcomes would be useful to help identify patients at risk of complicated postoperative courses and determine appropriate healthcare and resource utilization for patients.

Methods: The Quality Outcomes Database (QOD) Spondylolisthesis module was queried for patients receiving either decompression alone or decompression plus fusion for degenerative spondylolisthesis. Pre- and peri-operative variables were queried, and Mann-Whitney U testing was performed to identify which variables would be included in the machine learning models. Two k-nearest neighbors (KNN) models were implemented (k=25) with a standard 60%/20%/20% training, validation, and testing set respectively, one with arthrodesis status (Model 1) and the other without (Model 2). Feature scaling was implemented during the pre-processing stage to standardize the independent features.

Results: Of 608 enrolled patients, 544 met prespecified inclusion criteria. Mean age of all patients was 61.9 (SD ±12.1) years and 309 (56.8%) patients were female. The Model 1 KNN had an overall accuracy of 98.1%, sensitivity of 100%, specificity of 84.6%, positive predictive value (PPV) of 97.9%, and negative predictive value (NPV) of 100%. Additionally, a receiver operating characteristic (ROC) curve was plotted for Model 1, showing an overall area under the curve (AUC) of 0.998. Model 2 had an overall accuracy of 99.1%, sensitivity of 100%, specificity of 92.3%, positive predictive value (PPV) of 99.0%, and negative predictive value (NPV) of 100%, with the same ROC AUC of 0.998.

Conclusion : Overall, our findings demonstrate that non-linear KNN machine learning models have incredibly high predictive value for LOS. Patient demographics, socioeconomic status variables, and comorbidities served as the most robust predictors within our models. In addition, we show that models agnostic to arthrodesis status perform similar to those with arthrodesis as a variable.