Cluster analysis in adult spinal deformity unsupervised machine learning to guide classification and management

Adult spinal deformity (ASD) encompasses heterogeneous conditions associated with abnormal spinal alignment, pain, and disability, presenting substantial classification and management challenges. This review examines how cluster analysis identifies distinct ASD phenotypes and whether these classifications support cohort stratification and assessment of outcomes, surgical treatment, and complications.

This narrative review describes applications of k-means, hierarchical, and two-step clustering in ASD. It examines the use of shared clinical, radiographic, and demographic features to characterize patient subgroups and explores associations between these subgroups and patient-reported outcomes, surgical treatment, and complications.

Cluster analysis provides a framework for identifying ASD phenotypes that reflect the heterogeneity of clinical presentation and spinal alignment. Reported applications support cohort description, comparisons of surgical treatment and outcomes, and investigation of complication patterns. However, data quality and interpretive complexity remain important limitations. Current evidence primarily supports cohort stratification and hypothesis generation; cluster assignments have not been validated for individual treatment decisions.

Cluster analysis is a promising approach for characterizing ASD heterogeneity and informing stratification research. Further validation is needed to establish the clinical utility of cluster-derived phenotypes before their incorporation into individualized treatment planning.