Clustering Analysis of 71,000 Patients with Traumatic Lumbar Spinal Cord Injuries using Machine Learning

Introduction: Lumbar spine trauma is a major cause of morbidity and complex injuries leading to significant long-term complications. Machine learning clustering analysis could provide insight into diverse presentations of lumbar trauma and optimize patient management strategies. This study utilizes clustering techniques to identify profiles based on complications, comorbidities, and covariates.

Methods: The 2015-2019 NIS was queried using ICD-10 PCS coding to identify patients with lumbar vertebral fracture and spinal cord injury. Machine learning clustering analysis evaluated the population based on 50 comorbidities, complications and clinical covariates. Optimal number of clusters was determined using the Davies-Bouldin Index and Calinski-Harabasz Index. Between-cluster multivariate logistic regression analysis was performed to assess risk of mortality and non-routine discharge. Kruskal-Wallis H-Testing was performed to assess variance in length-of-stay between clusters. Statistical analysis was performed using Python.

Results: 71,623 patients were clustered into 4 groups ranging from 816 to 60,221 patients. Mortality ranged from 2.39% in Cluster 1 to 28.10% in Cluster 4. Cluster 4 displayed higher mortality risk [OR 14.30, p< 0.001] relative to Cluster 1. Cluster profiles are visualized in a heatmap. Cluster 1 had the lowest prevalence of comorbidities. Cluster 4 had the greatest prevalence of acute kidney failure, sepsis, pneumonia, and mechanical ventilation use. Risk of non-routine discharge was lowest in Cluster 3 [OR 1.57, p< 0.001] and highest in Cluster 4 [OR 7.40, p< 0.001] when compared to Cluster 1. Kruskall-Wallis H-testing and post-hoc pairwise testing of length-of-stay distributions showed significant (p < 0.001) differences between all clusters. The greatest test statistics occurred when comparing group 4 to all other groups.

Conclusion : The analysis of patients with traumatic lumbar spine injuries highlights four uniquely comorbid patient clusters. Identifying these subgroups may enhance personalized decision-making in terms of surgical interventions and managing long-term care.