A comparison of cluster and factor analytic techniques for identifying symptom-based dimensions of obsessive-compulsive disorder. (August 2019)

• Record Type: Journal Article
• Authors: Cameron, Duncan H.; Streiner, David L.; Summerfeldt, Laura J.; Rowa, Karen; McKinnon, Margaret C.; McCabe, Randi E.
• Is Part Of: Psychiatry research. Volume 278(2019)
• Page Start: 86
• Page End: 96
• Publication Date: 2019-08
• Subjects: Obsessive-compulsive disorder -- Subtypes -- Dimensions -- Factor analysis -- Cluster analysis; Psychiatry -- Periodicals; Psychiatry -- periodicals; Psychiatrie -- Périodiques; 616.89
• DOI: 10.1016/j.psychres.2019.05.040
• Languages: English
• ISSNs: 0165-1781

Abstract

Highlights: Factor analysis of Y-BOCS unreliable with under-represented OCD symptoms. Cluster analysis grouping by case rather than symptom not ideal for this practice. Given major limitations, dimensions based on overt symptom are not recommended. Abstract: A growing body of literature suggests that obsessive-compulsive disorder (OCD) is a heterogeneous condition. The studies investigating symptom dimensions have been limited by numerous methodological differences and sample characteristics. The purpose of this study was to compare the two most commonly applied statistical techniques used in addressing this question in the same large cohort of individuals with OCD. Both cluster analysis and factor analysis were used to examine OCD symptom data as measured by the Yale-Brown Obsessive Compulsive Scale (Y-BOCS) Symptom Checklist for 355 individuals with a primary diagnosis of OCD. The factor analysis revealed a three-factor model best described as symmetry obsessions/ordering compulsions, contamination obsessions/cleaning compulsions and aggressive obsessions/checking compulsions. In contrast, the cluster analysis yielded a stable four-cluster solution best described as symmetry obsessions/ordering compulsions, contamination obsessions/cleaning compulsions, aggressive-somatic-religious obsessions/checking compulsions and a mixed symptom profile. Although there was overlap in the models resulting from these two statistical approaches, cluster analysis better captured the dimensional nature of OCD by demonstrating the prevalence of symptom categories in each subgroup. Though both analyses are capable of providing similar outputs, the validity of these results is limited given the input of a priori symptom categories from the Y-BOCS.