Factor structure and predictive validity of the obsessive compulsive drinking scale

Background: The Obsessive Compulsive Drinking Scale (OCDS) is a 14-item, self-report instrument developed to measure obsessive thoughts about alcohol use and compulsive behaviors toward drinking. The objective of this study was to ascertain the factor structure underlying responses to the OCDS, and to further assess whether subscale scores derived from this structure were distinctive, internally consistent, predictive of future drinking, and able to differentiate between patients receiving naltrexone versus placebo in a controlled alcoholism treatment trial. Methods: OCDS data were collected from a total of 132 alcohol-dependent subjects at up to 15 assessment points during the study. Interitem correlations were pooled across assessment periods, and an iterated principal axis factor analysis with oblique promax rotation was performed. The factor analysis suggested that three primary factors could parsimoniously account for the common variance in item responses. Subscale scores were formed by summing responses to the most salient items on each factor. Results: The three common factors were interpreted as "resistance/control impairment," "obsession," and "interference." The subscale scores corresponding to these three factors were internally consistent, and their correlation with other baseline measures of alcohol use and severity suggested that they were distinct. Scores on each subscale reliably distinguished between subjects who remained abstinent, exhibited "slip" drinking, or relapsed to heavy drinking during the 12 weeks of active treatment. Additionally, scores on the resistance/control impairment subscale distinguished between those patients receiving treatment with naltrexone or placebo. Scores from each subscale also were able to predict the hazard for heavy drinking in the following week of treatment. Conclusions: The three OCDS factors are easily estimated with the summated scoring approach, and the resulting subscales appear to be internally consistent and distinctive. Moreover, the group differentiation capability and predictive utility of the subscale scores suggest that they might be useful as either predictor or outcome variables in alcoholism treatment trials. The duration of time for which a given OCDS assessment maintains its predictive utility awaits further confirmation.