Psychometric properties of the MAST and two briefer versions.

Abstract

The discriminative value and internal consistency of items of the Michigan Alcoholism Screening Test (MAST) and of 2 briefer versions (MAST-10 and MAST-13) and the predictive validity of the 2 briefer versions, were studied by analyzing the test scores of 2 groups of 100 motorists each, arrested in Harris County, Texas, USA, for driving while intoxicated. The 2 groups did not differ in race (approximately 2/3 of each were white), in mean age (41 yr) or education (10 yr); 96 of group 1 and 90 of group 2 were men. Based on recommended cut-off scores, the arrestees were classified as either problem drinkers or adjustive drinkers (i.e., relatively well-adjusted drinkers). Approximately half the items on the MAST discriminated between the problem and adjustive drinkers; the coefficients of internal consistency reliability were 0.88 for group 1 and 0.83 for group 2. In groups 1 and 2, 7 and 9 of the MAST-10 items distinguished between the problem and adjustive drinkers; the coefficients of internal consistency reliability were 0.80 and 0.60 for groups 1 and 2. When scores on the MAST and MAST-10 were correlated, the validity coefficients (adjusted for overlapping measurement error) were 0.84 and 0.74. In both groups more than 75% of the MAST-13 items discriminated between problem and adjustive drinkers; the coefficients of internal consistency reliability were 0.81 and 0.77, and when the MAST and MAST-13 scores were correlated, the corrected validity coefficients were 0.85 and 0.81. In sum, 66 and 63% of the MAST-10 classifications and 85 and 67% of the MAST-13 classifications were supported by the MAST. These levels of agreement do not justify the use of these shorter versions of the MAST.