STATUS OF EMPIRICAL LITERATURE ON SUBSTANCE USE

ASSESSMENT IN THIS POPULATION

Data from an ongoing study provide preliminary evaluations of the concurrent validity of self-reported alcohol and/or drug use. The subjects are outpatients at a state psychiatric center in a medium-sized city in upstate New York; all are voluntary participants in a longitudinal investigation of the psychosocial aspects of functioning of persons with severe mental disorders. The data provide information on the extent of agreement between self-reported substance use (obtained from the ASI) and information provided by collateral reports and urinalysis results, and also allow comparison of the concordance between self-reported drinking on two separate measures. Finally, the TLFB is evaluated with respect to known groups validity. ASI and Collaterals

Forty-five subjects provided data for this analysis. All were outpatients, three-quarter had been given schizophrenia-spectrum diagnoses, and one-quarter had been diagnosed with bipolar disorders. Most were male (84 percent), and 92 percent received pensions for a psychiatric disability. Each subject identified a collateral who could provide information about the subject's substance

use and other functioning. Collaterals were interviewed over the phone within a few days of an interview with the subject. A third of the collaterals interviewed were in daily contact with subjects, a third saw subjects weekly, and another third saw subjects at least once during the target month. Friends (often fellow patients) and treatment/residential staff were most often identified as collaterals. For 10 substance use categories, subjects were asked how many days in the past 30 they had used the substance, following standard ASI interview procedures. The collaterals were asked, "To the best of your knowledge, did (insert subject's name) use this substance in the last 30 days?" To make responses comparable, subjects' estimates of frequency were collapsed into a simple use/no use index. As shown in table 1, six categories were endorsed; for these the percentages of agreement ranged from 84 to 98, primarily due to the high levels of agreement that no use occurred. Inspection of the kappa coefficients (Cohen 1960) indicates no association above that expected for chance in the majority of drug categories. Subjects admitted 24 instances of substance use, which included multiple use by single subjects, but collaterals were able to corroborate subjects' recent use history accurately in only 7 instances. In the other