CURRENT APPROACHES TO SUBSTANCE USE ASSESSMENT

To borrow a scheme used by Skinner (1984), options available for sub- stance use assessment include (a) prospective methods, (b) retrospective methods, and (c) objective indicators. Prospective methods consist of variants on self-monitoring. Self-monitoring reduces reliance on memory, and is generally regarded as the most accurate alternative to direct observation. Successful self-monitoring does, however, require a subject with the skills and motivation to complete the task. Prospective information gathering also requires time. Retrospective methods involve asking the subject to report on past substance use over a designated time interval. Examples include the Addiction Severity Index (ASI) (McLellan et al. 1980), the Time Line Follow-back interview (TLFB) (Sobell and Sobell 1992), and various quantity-frequency methods (e.g., Cahalan and Room 1974; Polich et al. 1981). Retrospective self-report is practical for most settings and is the most frequently used. However, its drawbacks include the potential for memory failure or other sources of distortion. Objective indicators include blood- or urine-based drug screens, breathalyzer tests, laboratory tests (e.g., gamma-glutamyl transpeptidadase, high density lipoproteins, mean corpuscular volume), collateral reports, and official records. Each of these information sources has limitations. Breathalyzer tests and blood/urine screens yield information about recent use only (Schwartz 1988). Other laboratory tests identify medical consequences of substance use, but are generally sensitive only to prolonged high levels of use; furthermore, elevations are nonspecific to substance use. None of these indices yields data on the pattern of substance use. Collateral reports or other official records tend to be limited due to incomplete knowledge or representation of actual use history, and collaterals may be unavailable for some socially isolated subjects (Drake et al. 1993; O’Farrell et al. 1984). In the absence of a gold standard, confidence in the accuracy of assessment can be enhanced by adopting a convergent validity approach (Sobell and Sobell 1980). This involves using multiple indicators that will tend to converge on a consistent picture of actuality. Significant discrepancies must be evaluated from a methodological perspective as well as allowing for subject-specific factors. In any given population, consideration must be given to appropriate selection of measures as well as to ways in which their accuracy can be enhanced. Since retrospective self-reports continue to serve as the cornerstone of assessment, factors affecting the validity of self-reported substance use will be considered next. Factors Affecting Validity of Self-Reported Substance Use

For substance use assessment to be useful in a treatment context, measures must be both accurate and sensitive to change. The literature on accuracy indicates that acceptable levels of reliability and validity are found for alcoholics' self-reports when recommended procedures are followed (e.g., O'Farrell and Maisto 1987; Sobell and Sobell 1980). However, some samples and procedures have yielded less than impressive findings. Furthermore, the reliability and validity of self-reported drug use varies across both studies and types of drugs (Maisto et al. 1990). Test-retest reliability is infrequently reported and, when it is, shows only modest reliability coefficients. Validity coefficients tend to be similarly moderate. A reasonable conclusion is that self-report data are "inherently neither valid nor invalid, but vary with the methodological sophistication of the data gatherer and the personal characteristics of the respondent" (Babor et al. 1990, p. 8). In the substance abuse field, questions have thus moved beyond "Are self-reports valid?" and are framed more as "When, and under what conditions, are self-reports valid?" (Brown et al. 1992). It is this approach to evaluating the accuracy of self-reports that provides a framework for understanding the process of substance use assessment among persons with major mental disorders. In a discussion of the factors affecting the accuracy of self-reported substance use, Babor and colleagues (1990) highlighted four classes of variables. The first class yields characteristics of the respondent. These include the respondent's state of sobriety at the time of assessment and the possible influence of a social desirability response set. The second class of variables includes aspects of the task that might enhance or detract from accurate responding. These include the degree of rapport between assessor and respondent, whether assurances of confidentiality can be made, the likelihood that selfreports will be verified, the criterion interval reported on, and the clarity of the questions. Motivational factors constitute the third class of variables affecting accuracy of self-report. Obvious short-term goals (e.g., to obtain treatment or to avoid arrest) must be considered, as well as the fear of potentially judgmental attitudes or other threats to the respondent's self-esteem. Cognitive processes constitute the fourth type of variables influencing self-report accuracy. Impairment of attentional processes, verbal comprehension, or retrieval will interfere with the accuracy of an assessment. Cognitive processes may be impaired due to recent alcohol or drug use, to situational stress or anxiety, or to associated psychiatric syndromes such as depression. In addition, recall of past behavior is subject to information-processing biases, so that recall is disproportionately influenced by salient and/or very recent events. Application to Persons With Severe Mental Disorders

There are reasons for concern about the reliability and validity of substance use assessment in persons with severe mental disorders (e.g., schizophrenia, bipolar disorder, major depressive disorder). Several studies suggest significant underreporting of substance abuse in this population (e.g., Safer 1987; Shaner et al. 1993; Test et al. 1989). No formal reliability or validity studies have been published on substance use assessment, but the accuracy of screening measures has been empirically addressed. For example, the criterion validity of screening measures (e.g., the Michigan Alcoholism Screening Test) has been evaluated in psychiatric settings. Results indicate adequate sensitivity but low specificity (Drake et al. 1990; Teitelbaum and Carey, in press), suggesting that population differences may affect the psychometric properties of many of the standard industry tools. A recently completed literature review (Teitelbaum and Carey, in press) found surprisingly few studies addressing the reliability of alcohol/drug screening or diagnostic measures used with psychiatric patients. In this population, which is characterized by fluctuating mental status, the issue of test-retest stability is a fundamental psychometric concern. Clinical characteristics of persons with major mental disorders offer many potential sources of unreliability and/or invalidity. For example, in an emergency room setting, Barbee and associates (1989) found that significant numbers of schizophrenic patients were unable or unwilling to complete structured interviews. The presence of acute psychotic symptoms can impair reality orientation, thus invalidating responses. Common correlates of mental disorders such as hostility and psychosocial instability raise concerns about compliance and cooperativeness with assessments. Interactive effects of recent substance use, psychiatric symptoms, and prescribed medications increase the risk of cognitive impairment. In addition, some persons with mental disorders report concerns about the impact of accurate reporting on their psychiatric treatment. Fears about losing access to

medications or being transferred to another facility may motivate underreporting of substance use (Ridgely et al. 1990). Finally, anxiety about further stigmatization because of multiple disorders may come into play. Such clinical concerns suggest that important variables may include (a) the setting in which the assessment takes place (e.g., emergency room versus community mental health center); (b) the timing of the assessment (on admission versus later in treatment); and (c) institutional policies on providing integrated versus separated treatment. However, empirical data to evaluate these possibilities are unavailable. Clearly, there is a need for procedures paralleling the efforts by researchers in mainstream substance abuse settings to document the respondent, task, motivational, and cognitive factors that might affect the reliability and validity of self-reports in mental health settings. At the present time, however, even preliminary demonstrations supporting the accuracy of self-reported substance use patterns are lacking.