|
Article Excerpt
This study extends the Michigan Alcoholism Screening Test (MAST; M. L. Selzer, 1971) literature base by examining 4 issues related to the validity of the MAST scores. Specifically, the authors examine the validity of the MAST scores in light of the presence of impression management, participant demographic variables, and item endorsement infrequency.
**********
According to the National Mental Health Association (NMHA; 2001), of the more than 54 million Americans who have a diagnosable mental illness, approximately 8 million will seek counseling this year. Fifteen percent of these treatment seekers will experience a comorbid substance disorder (Beeder & Millman, 1992; NMHA, 2001). Given such percentages, it is likely that many counselors, regardless of their practice areas, will encounter clients who abuse or are dependent on alcohol. Thus, it is essential that counselors have the necessary assessment skills to screen clients to determine who is and who is not experiencing alcohol-use-related problems (McLellan & Dembo, 1992). Although not all counselors can be experts at diagnosing and treating alcohol abuse and dependence, Horrigan, Piazza, and Weinstein (1996) asserted that the failure to screen for the presence of alcohol dependence may lead counselors to misdiagnose their clients' presenting concerns and subsequently provide less than optimum care.
Counselors interested in using a brief, easy to administer and score screen for alcohol abuse and dependence are faced with a wide variety of choices. Some of the most well-known and well-researched instruments currently available include the CAGE (asks about respondents' needing to cut down, being annoyed by criticism, feeling guilty over drinking, and using drink as an eye-opener; Ewing, 1984), the Alcohol Use Disorders Identification Test (AUDIT; Saunders, Aasland, Babor, De La Fuente, & Grant, 1993), the TWEAK (similar to CAGE but asks about tolerance, worried, eye-openers, amnesia, and k for cut down; Russell et al., 1994), the T-ACE (tolerance, annoyance, cut down, and eye-opener; Sokol, Martier, & Ager, 1989), the Substance Abuse Subtle Screening Inventory-3 (SASSI-3; Miller & Lazowski, 1999), and the Michigan Alcoholism Screening Test (MAST; Selzer, 1971). Of these various choices, the MAST and the SASSI-3 are the most often and widely selected screening instruments (Brady, Foulks, Childress, & Pertschuk, 1982; Juhnke, Vacc, Curtis, Coll, & Paredes, 2003; Myerholtz & Rosenberg, 1997; Piazza, Martin, & Dildine, 2000; Thurber, Snow, Lewis, & Hodgson, 2001).
The MAST was designed to serve as a quantifiable measure of alcohol dependence (Selzer, 1971). The MAST contains 25 true/false questions about alcohol consumption and related behaviors. It is estimated that the MAST can be completed in 10 to 15 minutes (Hedlund & Vieweg, 1984). Selzer assigned item weights of 0, 1, 2, or 5 based on his perception of each item's contribution toward an overall indication of alcohol abuse severity. The MAST's total score, the sum of each individual item, ranges from to 53. Selzer's original scoring system classified total scores of through 3 as representative of social drinking, 4 as borderline or suggestive of alcohol abuse, and a score of 5 or more as a clear indication of alcohol abuse. Later, however, the MAST scoring system was adjusted such that the following range of scores is generally used by clinicians: 0-4, not alcohol dependent; 5-6, may be alcohol dependent; 7 or more, alcohol dependent (Hedlund & Vieweg, 1984; Selzer, Vinokur, & Rooijen, 1975).
Since its inception, the MAST has been studied extensively. Articles summarizing the populations in which the MAST has been studied appeared in 1982 (Brady et al., 1982) and 1984 (Hedlund & Vieweg, 1984). While representing exhaustive reviews of the literature at the time, none of the referenced studies compared the MAST's performance against other psychological measures of alcohol dependence. Thurber et al. (2001), using confirmatory factor analysis, reported results supportive of the notion that the MAST reflects a single latent variable of alcohol dependence. This finding is contrary to exploratory factor analytic findings (Parsons, Wallbrown, & Myers, 1994; Zung, 1978, 1980a, 1980b) that suggested that the MAST measures many facets of alcohol dependence.
Despite earning a reputation in the addictions treatment community as the "gold standard" against which all other assessment instruments are compared (Martin, Liepman, & Young, 1990), our review of the literature revealed several issues about the MAST that have been either partially addressed or wholly unanswered. First, due to item transparency of intent, some (e.g., Friedrich & Loftsgard, 1978; Kaplan, Pokorny, Kanas, & Lively, 1975) suggest that MAST scores are more reflective of a willingness to self-identify as having alcohol-related problems than a pure measure of alcohol dependence. In addition, due to the MAST's obvious nature, this instrument has been criticized as being vulnerable to positive dissimulation, a test-taking set in which participants attempt to minimize or hide their problems (Graham, 2000; Groth-Marnat, 2003; Otto & Hall, 1988; Stinnett, Benton, & Whitfill, 1991). In addition to minimization, client denial about the scope and depth of their alcohol-related problems may limit the MAST's ability to accurately represent a client's presenting problems (Shedler, Mayman, & Manis, 1993). These concerns raise questions regarding the MAST's psychometric accuracy with clients who may be motivated to minimize the nature of their troubles (Carver & Scheier, 1996; Corsini & Wedding, 1995; Royce, 1989; Wiseman, Souder, & O'Sullivan, 1996) or who may be in denial of their alcohol-related problems (Friedrich & Loftsgard, 1978; Otto & Hall, 1988; Selzer, Vanosdall, & Chapman, 1971; Tulevski, 1989).
Second, the MAST's results may be sensitive to demographic variables. Previous researchers (Gomberg, 1993; Lemle & Mishkind, 1989; Lex, 1994; McCreary, Newcomb, & Sadava, 1999; Wilke, 1994) have demonstrated that men and women experience alcohol dependence and the concomitant symptoms and consequences differently. The MAST's original standardization sample was composed primarily of males convicted of driving under the influence (Selzer, 1971). As such, MAST may be measuring different traits in men than it does in women. In addition, diversity of educational history and age may affect MAST results. For instance, education may be linked to awareness and ability to identify the signs and symptoms of alcohol dependence (Friedrich & Loftsgard, 1978). Likewise, alcohol-dependence symptom development may be a function of age (Hirata, Almedia, Funari, & Klein, 2001; Milam & Ketcham, 1983). Thus, the MAST's usefulness with young and undereducated clients is suspect (Friedrich & Loftsgard, 1978; Hirata et al., 2001). The MAST's ability to detect alcohol-related problems independent of gender, age, and education is critical considering the nondiscriminating nature of substance abuse in general and the fact that counselors likely will work with people from all demographic backgrounds.
Third, the MAST's scoring rubric is based on a total score. This practice presupposes that the instrument's items are homogenous and represent a single latent variable. Thurber et al. (2001) raised concerns regarding the homogeneity of MAST scores due to the frequency with which certain items are endorsed. Infrequently endorsed items could influence the internal consistency of MAST scores and therefore compromise the MAST's ability to discriminate between clients' presenting concerns (DuBois, 1965). As such, it is important to examine the internal consistency of MAST scores and determine which, if any, items limit the overall reliability estimate.
Responding to concerns that the MAST's global score may pose limitations by erroneously classifying persons in a homogenous fashion, the MAST's underlying factor structure has been examined via exploratory factor analysis (EFA; Friedrich, Boriskin, & Nelson, 1978; Skinner, 1979; Zung, 1978, 1980a, 1980b, 1982; Zung & Ross, 1980). To our knowledge, only one study (Thurber et al., 2001) subjected the MAST to confirmatory factor analytic techniques (CFA). CFA is the preferable statistical method for use when the research either tests hypotheses or when an empirical foundation has been laid by previous research (Stevens, 1996; Tinsley & Tinsley, 1987). While the MAST's construction was atheoretical, previous researchers (Parsons et al., 1994; Snowden, Nelson, & Campbell,...
|
