Self-Reported Symptoms of Cannabis Use Disorder (SRSCUD): Psychometric Testing and Validation

Findings from national surveys demonstrate that cannabis use peaks in young adulthood and that the annual prevalence of marijuana use among young adults (34.0%) is the highest it has been in decades (Johnston et al., 2016). We developed a 13 item measure designed to characterize the 11 symptoms of CUD as described in the DSM 5 (APA, 2013). To evaluate the performance of this Self Reported Symptoms of Cannabis Use Disorder (SRSCUD) measure, we examined its associations with other measures of CUD symptoms, negative cannabis related consequences, and other known risk factors for CUD (i.e., coping motives). Colleges students (n =7000) recruited from 9 universities in 9 states throughout the U.S. Our analyses focus on past month cannabis users (n = 2077). We split our sample in half to conduct exploratory factor analysis (EFA,n = 1011) and confirmatory factor analysis (CFA, n = 1012). All items loaded saliently on a single factor of CUD symptoms in both EFA (.553 = λ = 805) and CFA models (.524 = λ = 830) (see Table 1). In our final model, we allowed correlated errors between the two indicators of tolerance (items 10 and 11) and the two indicators of withdrawal (items 12 and 13), and obtained acceptable model fit across most indices: CFI = .941, TLI = .927, RMSEA = .059, SRMR = .042. As shown in Table 2, the total score of the SRSCUD was strongly correlated with other CUD symptoms measures (.617 < r s < .697), demonstrating convergent validity. SRSCUD was moderately positively correlated with a well known risk factor for CUD (coping motives) and moderately negative correlated with a well known protective (cannabis protective behavioral strategies). We conducted receiver operator characteristic (ROC) curve analyses to identify well how our continuous measure of CUD symptoms could identify individuals who exceed the cutoffs for probable CUD on these other symptom measures. For the most well validated measure (CUDIT R), we had excellent sensitivity/specificity (mean score of 1.5 on SRSCUD) for predicting probable CUD. Although more research evaluating performance of the SRSCUD compared to a clinical diagnosis is needed, we have preliminary evidence for construct validity of this measure.