Purpose: Estimates of the proportion of individuals with substance use disorders who also have co-occurring serious mental illnesses (SMI) range from 20 to 70 percent depending on the population and measures. Because of the high rates of co-occurring SMI among those with drug dependencies, it is clinically important to screen for and treat SMI when present. However, we are unaware of any screening scales specifically developed for detecting SMI in substance-dependent populations, although a number of recently developed scales have been validated for the general population and for those in primary care. Because distinguishing between substance and non-substance related psychiatric symptoms can be difficult, however, it is not clear if the scales validated for populations with low rates of substance dependence work equally well in clinical or criminal justice populations that have high rates of substance dependence. This study assessed the performance of a brief, 6-item screening scale, the K-6, for detecting psychiatric disorders in a sample of substance-dependent individuals.

Methods: We analyzed data from the combined 2001 and 2002 National Household Survey on Drug Use and Health (NSDUH) for participants 18 years of age or older (N = 74,502). The gold-standard for SMI was based on responses to the embedded Composite International Diagnostic Interview – Short Form (CIDI-SF), which allowed for classification of respondents as likely having a psychotic disorder, bipolar disorder, major depressive episode, generalized anxiety disorder, post-traumatic stress disorder, simple phobia or social phobia. The subsample of substance dependent participants (N=10,068) were those meeting the criteria for past-year abuse or dependence on alcohol or other drugs based on responses to sequences of DSM-IV diagnostic questions. Responses to the K6 screening scale, also included in the NSDUH questionnaire, were scored and summed yielding a range of possible scores of 0 to 24. Separate receiver operating characteristic (ROC) curves were generated and area under curves (AUC) calculated for each diagnosis for the entire sample and for the subsample that met the criteria for past-year substance dependence. The AUC is an indicator of the overall accuracy of a scale with higher percentages corresponding to greater accuracy. For each diagnosis, we compared the difference in the AUC for the entire sample and for the substance dependent sample.

Results: Although there was typically a decline of 4 to 5 percentage points in the AUCs from the entire sample to the substance-dependent subsample, the K6 scale performed well for both groups across all diagnoses. AUCs ranged from .96 to .88 for the entire sample and from .92 to .83 for the substance dependent subsample, indicating good to excellent performance even with the modest declines.

Implications: Among available screening scales for detecting SMI, the K6 appears to be suitable for use with populations where the rate of substance dependence is high such as in criminal justice settings or drug treatment. The K6 is briefer than screens such as the CIDI-SF, but performs as accurately. Use of the K6 in clinical samples should aid in improved detection and treatment of co-occurring SMI and substance use disorders.