In stark contrast to the proliferation of alcohol screens, there is a dearth of validated brief self-report drug screens suitable for use in generalist health care settings (Berman, Bergman, Palmstierna, & Schlyter, 2003; Brown & Rounds, 1995; Dawe et al., 2002; Fleming, 2002; Murphy & Impara, 1996; McPherson & Hersch, 2000; Roche & Freeman, 2003; WHOAWG, 2002). Moreover, those that are available are inadequate to specifically target problematic or at-risk cannabis use.
Limitations of existing drug screens for cannabis
As with alcohol screens, most of the small number of adolescent and adult drug screens originated and were validated among clinical samples in the United States. Experience using existing drug screening instruments in other cultures has often been unsatisfactory (Room, 1995, 1998; WHOAWG, 2002). There is remarkably little evidence, if any, of their sensitivity/specificity, or even acceptability, among general adult and adolescent populations, within or outside of the USA - notwithstanding the many ethnic and cultural subgroups in these populations (ALAC, 1996; Dawe et al., 2002; Knight, Sherrit, Shrier, Harris, & Chang, 2002; Leccese & Waldron, 1994; Martino, Grilo, & Fehon, 2000; McPherson & Hersch, 2000; Winters, 2003; WHOAWG, 2002). As with alcohol, ‘off the shelf’ screens originating elsewhere may not be relevant for dissimilar
cultures (Room, 1995, 1998), a critical consideration for use in New Zealand among Māori, Pacific Peoples, Asian, and the many diverse minority and subcultural groups. Adolescent screens must be developmentally appropriate, psychometrically sound, simple and practical for use in busy social services, educational, medical, or pediatric settings. Items must also be as innocuous as possible to ensure the adolescent’s relationship with their counsellor is not compromised (McPherson & Hersch, 2000; Monti et al., 2001; Silber, 1987a, 1987b; Winters, 2003). Existing drug screens vary widely in focus, content, group/s targeted, and score utility. Various operational characteristics further limit the applicability of these instruments as rapid screening devices among cannabis users in health care populations. A brief review of the most widely known screens available for use among adult and adolescent drug users will elucidate the current status of drug screening, highlighting the gap in instrumentation for a brief cannabis-specific screen suitable for use in generalist health contexts.
Clinical drug screens
While thought to be useful as brief screening tools, many clinical instruments are too long, complex, and expensive to be practical in busy primary health care settings. Prominent among these is the WHO Composite International Diagnostic Interview- Substance Abuse Module (CIDI-SAM; Robins, Cottler, & Babor, 1990), a standardised, structured interview providing definitive lifetime and 12-month DSM-IV and ICD-10 diagnoses across 10 drug classes, including cannabis (Andrews & Peters, 1998; Cottler et al., 1995). Extensive, expensive training is needed for administration, scoring, and interpretation of this complex, lengthy instrument requiring 30-60 minutes administration time. Another well-known clinical tool is the 149-item Drug Use Screening Inventory (DUSI; Tarter & Hegedus, 1991). Adolescent counterparts include the 139-item Problem Oriented Screening Instrument for Teenagers (POSIT; Gruenewald & Klitzner, 1990; Rahdert, 1991), the 81-item Substance Abuse Subtle Screening Inventory (SASSI; Miller, 1985), and the 149-item adolescent version of the DUSI (DUSI-A; Tarter & Kirisci, 2001; Tarter, Laird, Bukstein, & Kaminer, 1992; Kirisci, Mezzich, & Tarter, 1995). Although described as screening tools, all these protocols assess only one aspect of drug use, and are more appropriately diagnostic instruments for the second stage of assessment, being overly long and complex for rapid
preliminary assessment of problematic or potentially problematic cannabis use (Dawe et al., 2002; McPherson & Hersch, 2000). Rigorous diagnostic criteria are not necessary to efficiently recognize harmful or potentially harmful drug/cannabis use in a generalist health context (Maly, 1993).
Brief generic drug screens
Several standardised drug screens are generic (i.e., not drug-specific). The well known Drug Abuse Screening Test (DAST-28, DAST-20; DAST-10; Gavin, Ross, & Skinner, 1989; Skinner, 1982) provides a quantitative index of drug-related problems in a lifetime or 12-month timeframe. However, the problem severity score gives no indication which drug to target in treatment, and the primary focus is on clinically important drug abuse or dependence, rather than currently problematic or risky use (McPherson & Hersch, 2000). Another variant is the 16-item Simple Screening Instrument for Alcohol and Other Drug Abuse (SSI-AOD; Winters & Zenilman, 1994). A shared drawback of these and most other screens is a dichotomous ‘yes/no’ response option, limiting score utility by restricting information on the frequency of the behaviours tapped by each item.
Adolescent generic screens include an adapted version of the adult DAST (DAST-A; Martino et al., 2000); the 42- and 30-item Drug and Alcohol Problem Quick Screen (DAP; Klitzner, Schwartz, Gruenewald, & Blasinsky, 1987; Schwartz & Wirtz, 1990); the 12-item (expanding to 53 items) Adolescent Drug Involvement Scale (ADIS; Wisniewski, Glenwick, & Graham, 1985; Moberg & Hahn, 1991); and the 40-item Personal Experience Screening Questionnaire (PESQ; Winters, 1992). A recent model is the 45-item GAIN-Quick Screener (GAIN-Q; Titus & Dennis, 2005). As do their adult counterparts, however, several of these tools fail to query the amount, frequency, recency, type, and nature of adolescents’ drug use (e.g., Martino et al., 2000), while others take a lifetime window with no focus on current drug-related problems. With reported administration times ranging between 10-20 minutes, these tools are still too long for use in primary care settings (McPherson & Hersch, 2000). Limited, and in some cases no, validity data are yet reported (Leccese & Waldron, 1994; Martino et al.,
2000). Again, indices of severity of drug involvement with a variety of drugs gives no specific indication of the adolescent’s current, or potential, cannabis use problem level. Other brief generic screens contain a few embedded items that may, or may not be, applicable to cannabis. These include alcohol screens modified to simultaneously screen for other drugs (‘conjoint’ screens), such as the CAGE-Adapted to Include Drugs (CAGE-AID), the SMAST-Adapted to Include Drugs (SMAST-AID) (Brown & Rounds, 1995), and the (briefest) Two-Item Conjoint Screening Test (TICS, Brown, Leonard, Saunders, & Papasouliotis, 1997). With diminished sensitivity for the earliest, least severe drug problems (Brown & Rounds, 1995), these ‘conjoint’ screens hold little promise for detecting early stage cannabis problems and/or risky use. Their adolescent counterparts specifically designed for use in generalist health contexts include the 4- item Drug and Alcohol Problem (DAP) Quickscreen (DAP-4; Knight et al., 2000) and CAGE-AA (Knight et al, 2000). Sensitivity of the CAGE-AA proved unsatisfactory, however, and neither screen is recommended for use until extensive psychometric testing is completed (Knight et al., 2000).
Among adolescent medical patients, the developmentally appropriate 6-item CRAFFT (Knight et al., 1999, 2002), has been shown to reliably and accurately discriminate adolescents requiring further intensive assessment from those at risk and amenable to early intervention. Further validation studies are required on larger, more diverse populations (Knight et al., 2002). Another screen, the 9-item self-administered Substance Misuse in Adolescence Questionnaire (SMAQ; Swadi, 1997) demonstrated good psychometric properties and concordance with DSM-IV diagnosis among a small clinical sample of 12-17 year-olds using cannabis, cocaine, and volatile substances (Swadi, 1997). However, this screen makes no attempt to measure either drug-related social problems (abuse) or sub-clinical drug problems. Subsequently modified to increase its suitability for younger people and drug non-using 11-16 year-old students (ASMA; Assessment of Substance Misuse in Adolescents; Willner, 2000), the 8-item version is considered to have potential for detecting both problem drug use and at risk use among adolescents in the community. Extensive validation research is needed (Willner, 2000).
Screens for multiple drugs and severity indices
Several instruments combine screening for problem use of different drugs used simultaneously, as well as providing drug-specific problem severity indices. Two models intended for use in generalist health care settings internationally are the 8-item Alcohol, Smoking and Substance Involvement Screening Test (ASSIST, Version 3; World Health Organization ASSIST Working Group, 2002) and the 11-item Drug Use Disorders Identification Test (DUDIT), with a companion test (DUDIT-E) for more comprehensive problem assessment (Berman et al., 2003). To varying degrees, both were designed to parallel (DUDIT) or to complement (ASSIST) the AUDIT. Targeting adults and adolescents, the DUDIT assesses all drug categories, but excludes alcohol and tobacco. By contrast, the ASSIST targets alcohol, tobacco, prescription drugs and illicit drugs among adults. Items in both screens encompass consumption, dependence, and problems domains. The DUDIT adopts a 12-month window to screen for both lifetime and current drug problems, while the ASSIST uses a 3-month window. Both incorporate a dimensional scoring system.
While both screens are still only in a preliminary stage of development, several characteristics invite comment. Both screens are subject to the criticisms directed (above) at generic measures attempting to simultaneously screen for multiple drugs. Drug categorization was identified as a “consistent problem” and an area of “confusion” in the international ASSIST reliability study (WHOAWG, 2002), with implications for reliability/validity of findings reported. Given the terminology problems reported, confusion between culture-specific perceptions of ‘harmful’ cannabis use and medicinal cannabis use was possible. Translation issues (i.e., semantics, literal and idiomatic) apply for use in local languages/cultures. Confusion was also reported between the ‘last 3 months’ versus the ‘lifetime’ timeframe. Moreover, the 3-month ASSIST reference period may be inadequate for detection of both a drug disorder and a developing or latent drug problem. Diagnostic systems (DSM-IV, ICD-10) for current PSUDs are based on a ‘past 12-month’ assessment window. Preliminary reliability data from a small adult sample indicate only fair test-retest (1-3 days) kappas for cannabis ranging from 0.52 (harmful use/ problems with work, home or school) to 0.95 (ever used), with
an average kappa of 0.64. Given the brief test-retest interval, these results are unsatisfactory. Since values greater than 0.7 are considered satisfactory (Bishop et al., 1975; Fleiss, 1991), a screen should aim for higher reliability estimates. A critical prerequisite to use of the ASSIST in primary care settings, however, is evidence of its concurrent and predictive validity. Neither have yet been reported. The 8 final items were selected on the basis of their association with drug use frequency, and not with ICD-10 diagnostic criteria (WHOAWG, 2002). Given that results reported derive largely from drug treatment clients, the ASSIST may prove even less reliable and feasible under typical conditions of busy primary health care practices.
In a high-risk sample (n=154) the DUDIT evidenced excellent sensitivity and specificity for DSM-IV and ICD-10 cannabis dependence (n=35), but performed poorly in diagnosing cannabis abuse/harmful use, a result confirmed by ROC curves (Berman et al., 2003). That is, in the high-risk population DUDIT diagnostic validity for cannabis abuse/harmful use was “no better than chance” (Berman et al., 2003). Among a general population sample (n=1099) only 33 (3.1%) respondents scored one or more on the DUDIT, with the remaining 96.9% scoring zero. A selective response bias was apparent (Berman et al., 2003). Given there was no validation interview or follow-up, these data are non-interpretable, and the predictive utility of the DUDIT unknown. The authors claim the DUDIT has “good potential” for screening for drug use and dependence for different drugs in both clinical and general population samples (Berman et al., 2003). In fact, these results suggest the DUDIT potential for identifying cannabis dependence and abuse/harmful use among primary care populations to be minimal. For example, appropriate cut-off scores for cannabis dependence in this population remain unknown. Further - and of paramount importance - given its poor performance in identifying harmful use/abuse, it is difficult to envisage the DUDIT having any relevance whatsoever in detection of risky cannabis use. After all, identification of this alcohol group (‘hazardous’ consumption) was the primary stated aim of the AUDIT (Saunders & Aasland, 1987; Saunders et al., 1993).
Brief dependence screens
Complementing this small arsenal of self-report drug screening tools are several brief dependence screens adaptable to the specific drug of interest. Most well known are the 10-item Leeds Dependence Questionnaire (LDQ; Raistrick et al., 1994) and the 5-item Severity of Dependence Scale (SDS; Gossop et al., 1992, 1995) used for screening for various drugs, including cannabis (Martin, Copeland, Gates, & Gilmour, 2006; Swift et al., 1998a). With their clinical focus on severity of dependence symptoms, however, these screens were not designed for early detection of potentially harmful or risky drug use. Insensitive to low-level misuse, therefore unable to capture the spectrum of cannabis-related problems along the continuum of use with or without dependence symptoms, their utility in identifying both early-stage and more advanced cannabis use problems - the focus of this paper - is minimal.
Summary
In different ways and to varying degrees, in their present form all of the drug screening tools reviewed above are clearly either inefficient or otherwise unsuitable for use in generalist health contexts as quick and accurate screens specifically for cannabis use problems (McPherson & Hersch, 2000). Limitations derive from their: (a) complexity and comprehensiveness (b) timeframe (c) generic focus on use of “drugs” (d) insensitivity to low-level misuse (e) mere listing of cannabis among multiple other drugs or (f) reliance on only one or two items about cannabis use, (g) lack of concurrent and/or predictive validity and/or (g) lack of evidence of validity for use among both general adult and adolescent (i.e., developmentally appropriate) populations, and/or various ethnic/cultural subgroups within these populations. Confounded by other drug or alcohol use, scores on these screens are clearly insufficient for expeditious detection of problematic or risky cannabis use among both adult and adolescent users. Albeit, as chapter five shows, individual items carefully selected from among these sources may have potential as candidate items in the pool for such a screen (McPherson & Hersch, 2000).