Factor-derived groups of OCD symptoms: Predictors of treatment outcome

Abstract

Objective. Prior studies suggest that clinical subtypes of obsessive compulsive disorder (OCD) might be associated with treatment outcome. To determine this association, a factor analysis was performed on the 16 subscales of the 58-item Yale-Brown Obsessive Compulsive (Y-BOCS) Symptom Checklist. This instrument, the gold standard for diagnosing OCD and its various subtypes, was intended to classify different categories of OCD, such as aggressive obsessions, cleaning compulsions, hoarding, etc. However, the items in each subscale were grouped together based on clinical observation rather than statistical derivation. Thus, the Y-BOCS Symptom Checklist has yet to be validated. Consequently, a factor analysis of the individual items of the instrument was performed to determine whether the item composition of the clinically-derived subscales has a valid scheme. The next objective was to determine which of the factors (those derived from the clinical categories or those derived from the individual items) were the strongest predictors of outcome. Methods. A principal component factor analysis was performed on the clinical categories of the checklist using baseline data from 160 OCD patients and results were compared to clinically-derived categories of prior studies. Next, individual items of the checklist were factor analyzed using baseline data from the same 160 patients. Lastly, linear regression analyses were performed on baseline Y-BOCS Symptom Checklists of 30 patients with severe refractory OCD who were treated with intravenous clomipramine. Predictor variables were factor scores derived from clinically-derived and factor-analytically derived clusters. Outcome variables were percent change in Y-BOC severity scale and change in CGI scale at one month follow-up. Results. A four-factor model best explained the clinical presentation of symptoms derived from the clinical categories and overlapped somewhat with clinically-derived categories of prior studies. A four-factor and nine-factor model best explained the overall variance of the item-by-item factor analysis. The only factors with predictive value after treatment were two factors from the four-factor model of the item-by-item factor analysis. Scores on Factor 2 (disgust with contaminants/cleaning compulsions) were positively associated with percent improvement on the Y-BOCS and scores on Factor 4 (hoarding/symmetry/repeating) were associated with a worsening of clinical symptoms. Discussion. Evidently, an association between OCD symptom subtype and treatment outcome was found after reevaluating the item composition of the Y-BOCS Symptom Checklist. If replicated on a study may have implications for clinician's future treatment selection and OCD research.