Cluster analysis of the pain outcomes questionnaire.

Objective: The purpose of this study was to derive and describe subtypes of Pain Outcomes Questionnaire (POQ) profiles produced by a large, heterogeneous multisite sample of chronic pain patients (N = 672).
Methods: The POQ core scales were subjected to Ward's hierarchical agglomerative cluster analyses in order to establish the optimal number of clusters in validation (N = 336) and cross-validation (N = 336) subsamples. Based on the consistency of the solutions derived from the two subsamples, a third Ward's procedure was performed on the entire sample and, finally, a K-means iterative partitioning procedure was used to assign cases to the cluster groups. The resulting cluster solution was validated using a broad range of demographic, pain, and treatment outcomes data.
Results: The cluster analytic procedure revealed the presence of eight unique subgroups of POQ profiles. The adequacy of the final solution was demonstrated using a multivariate analysis of variance and a discriminant function analysis. Group differences on the demographic and pain variables were generally consistent with expectations based on mean profiles, and therefore, provided evidence of the validity of the cluster solution. Brief descriptions of each cluster group were generated based on the correlate data. The clinical utility of the solution was in part demonstrated by differential treatment response rates among the groups.
Discussion: The results of this study are generally consistent with those of previous cluster analytic investigations of pain patients and suggest that psychosocial/behavioral classification systems may serve a useful heuristic function in the assessment and treatment of chronic pain that is not provided by current diagnostic taxonomies.