Maximum

9.69

KEY: * = p < 0.05; ** = p < 0.001; SCID = Structural Clinical Interview for DSM-III-R.

Drake and colleagues that reflects the following degrees of alcohol use over the previous 6 months: no use, mild use but no problems, moderate use with some resulting problems, severe problems, and extremely severe problems probably resulting in hospitalization. All but one subject received ratings in the first three categories, so that only the 63 subjects assigned to the no use, mild, and moderate use categories were used for analysis. As shown in table 4, subjectgenerated TLFB data covering the same 6-month interval seem to be consistent with therapists ratings. While finding predictable group differences on a measure does not indicate an accurate reflection of reality, it does suggest that subjects who are perceived by professionals as having more alcohol involvement do report more nonabstinent days and heavier alcohol consumption. Enhancing the Accuracy of Substance Use Assessment

By integrating the above data and the literature (see Drake et al. 1993; Nurco 1985; Skinner 1984; Sobell and Sobell 1980), it is possible to offer several recommendations for enhancing the accuracy of

substance use assessment among persons with major mental disorders. These are organized below according to the four-factor scheme presented earlier. Self-reported alcohol use by subjects groups according to therapists' ratings.

TABLE 4.

Self-reported alcohol use from timeline follow-back Case manager rating scale determination

Abstinent Light days a days

Moderate Heavy days days b

Maximum quantityc

No use of alcohol (N = 35)

178.00

0.83

1.17

0.05

1.49

Mild use, no problems (N = 17)

159.71 10.29

2.35

8.64

4.29

Moderate use, some problems (N = 11)

124.67 22.56

5.89

26.89

12.44

KEY: a = F (2,55) = 10.77, p < 0.0001; moderate use < mild use and no use; b = F (2,55) = 6.62, p < 0.003; moderate use > no use; c = F (2,55) = 13.50 , p < 0.0001; moderate use > mild use and no use.