many scales in the Guide are straightforward, others, such as the Addiction Severity Index, the Comprehensive Drinker Profile, the Alcohol Timeline Followback procedure, and several diagnostic scales, require extensive training. Beyond adequate preparation in administration, training in interpretation of results is essential. This requires at least a basic academic foundation in psychometric principles (Moreland et al. 1995) as well as familiarity with research on the specific instruments used. To help satisfy this latter need, the fact sheets included in this volume provide some key references for each measure. Other citations for research may be obtained by searching computerized reference data bases such as Psych Info, ETOH, and Medline.

Availability of Computerized Scoring

A few tests noted in the Guide can be scored by computer. This is particularly convenient for measures such as the Alcohol Use Inventory and the Addiction Severity Index, which are lengthy and comprise multiple scales.

Fee for Use

Many of the instruments included in this volume, especially those developed under Government grants, may be used free of charge. Often, test authors generously make their scales available to clinicians and researchers at no cost. When imposed, fees are usually modest.

Reliability and Validity

Evaluation of how alternative measures fare on validity and reliability, the two primary psychometric characteristics of an assessment instrument, can assist in choosing one scale over another. Several different types of reliability and validity may be considered. They vary in importance depending on the nature of the measure and its intended application.

Reliability deals with generalizability of the instrument across different times, settings, scale versions, evaluators, and so forth (Cronbach et al. 1972). Reliability may be seen as a particular type of validity in which the relationship of performance on the measure with itself is evaluated. Measures low in reliability (i.e., those that cannot even predict themselves well) must of necessity also be low in other types of validity where the test is attempting to predict other performance. On the other hand, while a necessary condition, reliability is not a sufficient cause of validity. Measures may be consistent while not accurately measuring what the author intended.

Test-retest reliability refers to similarity of scores for administration of the measure at two points in time. As a rule, the interval between tests needs to be long enough that similarity in responses at the repeat administration is not largely due to the client simply remembering earlier answers. One would expect high test-retest reliability on measures that tap stable client characteristics, such as family history of alcoholism, age of onset of problem drinking, and general expectancies of alcohol effects. Scales for more transient client characteristics, such as craving and treatment motivation, would be expected to have lower test-retest reliability.

Split-half and internal consistency reliability reflect agreement of content coverage within the scale itself. Split-half reliability is calculated by correlating the score on half of the items with that on the other half. Internal consistency assesses how well responses on individual items correlate with those of other items of the scale. For instruments designed to measure a single phenomenon, such as severity of the alcohol dependence syndrome, these correlational coefficients should be high. The relationship between degree of internal consistency and clinical significance has been discussed by Cicchetti (1994).

A fourth type, parallel forms reliability,

refers to two sets of questions that address the same issues and produce comparable results. While equivalent forms of tests are useful-for example, to allow pretreatment and posttreatment functioning to be compared without risk of the potential confounding effect of client memory-for the most part, equivalent forms for alcoholism measures have yet to be developed.

The three common types of validity are content, criterion, and construct. Content validity refers to the degree to which items comprehensively and appropriately sample the domain of interest. For example, a checklist of alcohol consequences should comprise the multiplicity of adverse effects of drinking rather than singling out certain negative consequences to the minimization or exclusion of others that are equally damaging. Content validity is not quantified. Rather, it must be built into the test by careful construction and selection of test items (Nunnally 1978).

Criterion validity deals with how well scores on a measure relate to important, relevant nontest (real world) behaviors, such as initial motivation for treatment and longterm maintenance of sobriety. Criterion validity is a major concern in evaluating screening tests and is gauged by the extent to which individuals who score positive on them actually receive a diagnosis of alcoholism and, conversely, the extent to which those who score negative on the screen do not meet diagnostic criteria. Predictive, concurrent, and "postdictive" validity are all types of criterion validity. The distinctions among them reflect the temporal relationship between the test results and the phenomenon of interest.

Finally, construct validity refers to the degree to which a measure actually taps a meaningful hypothetical construct and a nondirectly observable, underlying causal or explanatory dimension of behavior. Scales purporting to measure hypothetical constructs in the alcoholism field, such as "craving," "loss of control," "denial," and "high

risk drinking situation," should yield high levels of construct validity. Scores on these measures should correlate well with other manifestations of the construct. At the same time, they should correlate only minimally or not at all with scores on scales that measure constructs distinct from them.

Benefits of Assessment

From the clinician's perspective, the primary benefit of assessment is to accurately and efficiently determine the treatment needs of an alcoholic client. Carefully selected assessment procedures can quickly and validly evaluate severity of dependence, adverse consequences resulting from problematic drinking, contributing roles of other emotional and behavioral problems to drinking, cognitive and environmental stimuli for drinking, and so forth. These variables all have major significance in suggesting the intensity and nature of intervention needed.

Assessment, however, also yields valuable secondary clinical benefits (Allen and Mattson 1993). For example, giving clients individualized feedback based on test results may enhance their motivation for change (Miller and Rollnick 1991) and help them formulate personal goals for improvement. Also, research indicates that clients themselves highly value assessment (Sobell 1993) and that programs with formal assessment procedures are better able to retain clients in treatment (Institute of Medicine 1990).

If the core battery is administered to all clients, the data base of results can be periodically analyzed to determine, at a program level, needs for additional services, types of clients treated, and so on. This information can target efforts to modify the core treatment regimen to more specifically address needs of the clientele. These positive benefits of formal assessment can be fully realized only if the scales are properly administered, interpreted, and utilized by the clinician.

Setting, Timing, and Sequencing of Formal Assessment

The Guide is organized according to a framework of sequencing of care for clients. The physical settings for assessment also likely reflect this sequencing. Screening is generally performed in a primary health care unit, diagnosis and triage in a general inpatient or outpatient medical facility, and specific treatment planning assessment within a facility or by a provider offering alcoholspecific services.

More research needs to be done to determine optimal timing for alcohol assessment. For the tests to be maximally useful, they need to be conducted soon enough after treatment entry that results from them can help shape the individualized treatment plan. At the same time, it should be borne in mind that following recent heavy alcohol usage, clients may be so impaired in neuropsychological and emotional functioning that they are unable to give an accurate picture of themselves (Nathan 1991; Goldman et al. 1983; Grant 1987).

While various guidelines have been offered for time following admission necessary for valid psychological testing (e.g., Sherer et al. 1984; Nathan 1991), insufficient research has been done on this critical issue to offer firm guidance. Time guidelines may be specific to the nature of the measure (e.g., tests requiring a high level of neuropsychological functioning may need to be delayed longer than trait-focused personality measures). Common practice and clinical judgment suggest that, to the extent practicable, most tests should be deferred at least until the client has stabilized following alcohol withdrawal.

Granted the large number of measures available to clinicians, but also considering limitations in time and resources available, the strategy of assessment must be clearly thought through.

The underlying assumption is that "more is better." However, such a comprehensive approach may not be feasible because of the constraints often experienced within many clinical settings. Furthermore, Morganstern (1976) has suggested that such an approach may not be appropriate, and presents a somewhat more limited perspective: "The answer to the question 'What do I need to know about the client?' should be: 'Everything that is relevant to the development of effective, efficient, and durable treatment interventions.' And from an ethical (and economical) consideration, one could add, 'And no more.' (Donovan 1988, p. 52)

Finally, it is important to not regard assessment as a single activity performed at a single point in time. Assessment should be seen as ongoing since it supports clinical decisionmaking throughout the course of treatment (Donovan 1988).

Approaching the Client

Regardless of the setting for psychometric evaluation, it is important to establish rapport with the client by adopting an empathetic approach. The client should also be assured of confidentiality, and any institution-mandated limitations on confidentiality should be clearly articulated.

In introducing measures, it is important to elicit clients' full cooperation by explaining that they will receive feedback on results and that this information will assist in developing a treatment plan maximally helpful to them. The tenor for the assessment enterprise should be characterized as collaborative, with the assessor and client jointly committed to discovering those client features that will contribute to important decisions about future clinical management.

Also, to increase the likelihood of test results being valid, particularly as regards level of alcohol consumption, it is important

to assure that the client is not currently under the influence of alcohol (Sobell and Sobell 1990). A handheld breathalyzer can provide such confirmation.

Giving Clients Feedback

Research suggests that feedback on results of assessment can reinforce commitment for behavior change. While little research has been done on how feedback process variables specifically influence its motivational impact, some general guidelines can be offered on how to give feedback (Miller and Rollnick 1991; Allen and Mattson 1993).

Both rapport and objectivity should characterize the feedback process. Providing feedback should be a positive experience for both the client and the clinician. Clients are intensely interested in what tests can tell them about themselves, a topic of considerable interest to most people. As in the testing activity itself, the process of giving feedback should be seen as collaborative. The clinician is professionally and objectively sharing the findings, the client is sizing-up the implications of these results and, together, they will use this information to design an optimal treatment program.

Clients may be overwhelmed by test findings. Therefore, it is important that feedback be given in a clear, concrete, and organized fashion. Often, showing clients their standing on relevant dimensions by using visual displays such as plots or graphs can be informative. Review results slowly to assure that clients fully understand them. Periodically during the feedback session, clients may be asked to summarize test findings in their own words and to reflect on the meaning they ascribe to them. Asking clients to give concrete examples to illustrate the findings may also deepen their understanding of the information.

Often, test results are not totally positive. While remaining fully honest with them,

help clients understand that, with abstinence and behavior change, many of the negative findings should improve. If clients are treated for an extended time, the measures can be periodically repeated so that they can recognize positive changes in scores as well as identify areas in which further improvement is needed.

Finally, in reviewing test results with clients, it is important to show them how the findings influence development of treatment plans. Recognizing the coherence of treatment with their own personal needs should further motivate them to actively participate in treatment.

Use of Nonalcohol-Specific Measures

While the instruments in the Guide are alcohol-specific, other types of psychometric measures can also play a helpful role in clinical management of alcoholics. Considering the frequency of comorbidity of psychiatric problems in alcoholics in treatment (NIAAA 1993) and the implications of such conditions for treatment of alcoholism (Litten and Allen 1995), assessment of collateral psychopathology may be useful.

General personality measures may also assist in treatment planning (Allen 1991). Traits such as impulsivity, need for social support, insight, and so forth have important implications for choosing interventions and helping the clinician relate most effectively with the client.

A variety of treatment process measures, including scales to assess client satisfaction and treatment atmosphere, may provide guidance for periodic redesign of the treatment program.

Research Needs

Although substantial progress over the past decade has produced a rich array of

assessment instruments to inform alcoholism treatment, several areas remain inadequately explored and warrant further research. Foremost among these is development of computerized adaptive testing algorithms. Given the variety of available instruments, a computerized assessment program tailored to the needs of the individual client would greatly facilitate and economize the assessment process. Such a program would capitalize on advances in decision tree technology.

Expert systems, such as those employed in other areas of medical diagnostics, could be modified for alcoholism assessment programs. Computerized technology would offer the clear advantage of allowing easy, automated scoring and would permit comparability within and across individuals and treatment settings. Such a system could satisfy the dual needs of providing the busy clinician with information relevant to individual client treatment planning as well as providing data for subsequent program evaluation and modification. In addition, computerized testing may also yield significant advantages in eliciting more accurate information from younger clients who are not threatened by the technology and might well prefer the computer to a therapist's interview (Leccese and Waldron 1994).

A critical concern for treatment providers and researchers alike is establishing appropriate timing for administration of assessment instruments. Demands for quick turnaround to aid in triage and treatment planning compete with the clients' ability to provide accurate and reliable information after detoxification. Drastic reductions in clients' length of stay imposed by managed care decisions further complicate the dilemma. Applied research to identify the optimal times for test administration is badly needed. Objective indicators that document client readiness for administration of different tests must be operationalized in terms of client functioning.

Construction of subpopulation norms for individual assessment instruments also merits further research. A related, but often ignored, issue concerns the degree to which response surfaces and underlying factor structures for tests differ for women and various subpopulations. For example, does the construct of alcohol consequences fundamentally differ in men and women? Women typically score very low on alcohol conse quence inventories that include such items as violence and physical spousal abuse. Does this suggest that a scoring adjustment should be made or rather that a different set of items should be queried for women in evaluating the adverse effects of drinking?

While certain treatment-related issues are measured well by existing scales, other important dimensions are not. For example, assessing clients' motivation for treatment in general and specific treatment preferences have proved to be difficult for clinicians and alcoholism treatment researchers. The fre quently invoked construct of craving remains elusive, despite numerous attempts to operationalize it. Various scales purporting to measure craving often elicit conflicting and unresolvable information with little reliability or face validity.

Conclusion

As suggested by the sheer volume of instruments covered in this Guide, alcoholism clinicians and researchers now have available a variety of choices to assist in planning alcoholism treatment and better understanding the nature of the problem. In order to take full advantage of this resource, clinicians and researchers must clearly understand the nature of the questions they must answer and the strengths and weaknesses of alternative psychometric instruments that can assist them. It is hoped that this overview, the excellent chapters by subject matter experts, and the fact sheets for the alternative scales will assist this important venture.