Psychometric properties of the Cannabis Abuse Screening Test (CAST) in a French sample of adolescents

Abstract Purpose Psychometric and screening properties of the Cannabis Abuse Screening Test (CAST) were investigated using DSM-IV diagnoses of cannabis dependence (CD) and cannabis use disorders (CUD) as external criteria. Performance of the binary and the full version of the CAST were compared. Methods The sample consisted of 2566 French adolescents aged 17 who reported cannabis use 12 months prior to the survey. The Munich Composite International Diagnostic Interview (M-CIDI) was used as a gold standard for DSM-IV diagnoses. Internal consistency (Cronbach's α ), construct validity (exploratory and confirmatory factor analyses, correlation of CAST scores with related variables), and criterion validity (Receiver Operating Characteristic analyses) were assessed. Results Both CAST versions were unidimensional and Cronbach's α was 0.748 for the binary and 0.775 for the full version. High and comparable AUC values indicate a good ability of both test versions to discriminate between individuals with and without a clinical diagnosis. Based on balanced sensitivity and specificity, the optimal cut-off scores for CD and CUD were 2 for the binary and 3 or 4 for the full version. While both versions largely overestimated CD prevalence, CUD prevalence was slightly underestimated. Conclusions The binary and the full version of the CAST are equally useful for screening for cannabis-related disorders. Both clinical and research applications of the scale are possible. The CAST may be used for estimating CUD prevalence rather than CD prevalence. The ultimate choice of the cut-off depends on the purpose of the specific study using the CAST.