Fleming M F, Barry K L
Department of Family Medicine, University of Wisconsin, Madison 53715.
J Subst Abuse. 1988;1(2):173-82. doi: 10.1016/s0899-3289(88)80020-8.
The psychometric properties of the Short Michigan Alcoholism Screening Test (SMAST-13), a 13-question abbreviated form of the MAST, were determined on two samples of alcoholics, with their nonalcoholic family members serving as controls. The study hypothesized that the SMAST-13 has a low specificity when used to screen family members for alcohol problems and would exhibit a low internal reliability coefficient in this population. All participants were interviewed using DSM-III criteria for alcohol abuse or dependence. Only family members who failed to meet these criteria were enrolled in the study. When a weighted cut-off score of 5 or more was used as a positive score for alcohol problems, the sensitivity was .98 in Sample A and .94 in Sample B. Utilizing the scores of the family members as controls, the specificity was .58 in Sample A and .70 in Sample B. The internal reliability of the SMAST-13, using Chronbach's alpha on the total sample, produced a reliability coefficient of r = .57 in Sample A and r = .62 in Sample B. To determine if the specificity of the SMAST-13 could be improved, the weighted cut-off score for a positive response was raised from 5 to 10. In Sample A, the specificity changed from .58 to .90 and the sensitivity dropped slightly from .98 to .92. In sample B, the specificity increased from .70 to .95; the sensitivity decreased from .95 to .85. Suggestions for changing the wording of two questions which explain the majority of the false positives are discussed. Epidemiological studies which have used the SMAST-13 to estimate the prevalence of alcohol problems in family members of alcoholics may need to reconsider their methods of classifying alcoholics.
