SCREENING FOR ALCOHOL USE DISORDERS AND AT-RISK DRINKING IN THE GENERAL POPULATION: PSYCHOMETRIC PERFORMANCE OF THREE QUESTIONNAIRES

HANS-JRGEN RUMPF 0 2 ULFERT HAPKE 0 1 2 CHRISTIAN MEYER 0 1 2 ULRICH JOHN 0 1 2 0 Psychiatry and Psychotherapy, Research Group S:TEP, Medical University of Ratzeburger Allee 160 , 23538 Lbeck, Germany 1 University of Greifswald, Institute of Epidemiology and Social Medicine, Addiction Research Center , Greifswald, Germany 2 Medical University of Lbeck, Department of Psychiatry and Psychotherapy, Research Group S:TEP (Substance Abuse: Treatment, Epidemiology , and Prevention), Lbeck Most screening questionnaires are developed in clinical settings and there are few data on their performance in the general population. This study provides data on the area under the receiver-operating characteristic (ROC) curve, sensitivity, specificity, and internal consistency of the Alcohol Use Disorders Identification Test (AUDIT), the consumption questions of the AUDIT (AUDIT-C) and the Lbeck Alcohol Dependence and Abuse Screening Test (LAST) among current drinkers (n = 3551) of a general population sample in northern Germany. Alcohol dependence and misuse according to DSM-IV and at-risk drinking served as gold standards to assess sensitivity and specificity and were assessed with the Munich-Composite Diagnostic Interview (M-CIDI). AUDIT and LAST showed insufficient sensitivity for at-risk drinking and alcohol misuse using standard cut-off scores, but satisfactory detection rates for alcohol dependence. The AUDIT-C showed low specificity in all criterion groups with standard cut-off. Adjusted cut-points are recommended. Among a subsample of individuals with previous general hospital admission in the last year, all questionnaires showed higher internal consistency suggesting lower reliability in non-clinical samples. In logistic regression analyses, having had a hospital admission increased the sensitivity in detecting any criterion group of the LAST, and the number of recent general practice visits increased the sensitivity of the AUDIT in detecting alcohol misuse. Women showed lower scores and larger areas under the ROC curves. It is concluded that setting specific instruments (e.g. primary care or general population) or adjusted cut-offs should be used. INTRODUCTION Several screening questionnaires have been developed to detect individuals with alcohol dependence, alcohol misuse or high levels of alcohol consumption. Mostly, these screening tools are applied in medical care settings to detect subjects eligible for brief interventions. In general population studies, screening instruments serve as tools for case finding or to provide data to estimate prevalence rates of alcohol problems, and are used when other diagnostic procedures are too comprehensive. Screening questionnaires in general population studies are appealing, because of their inexpensive format. However, it might be argued that screening questionnaires developed in clinical settings are not automatically suitable in the general population and that a spectrum bias may occur, e.g. because of differences in the severity of dependence. Only a few studies have addressed the validity of screening questionnaires in the general population. The CAGE (Ewing, 1984; Mayfield et al., 1974) (acronym based on its four items: Cut down on drinking, Annoyed by criticism, Guilty feelings, and Eye opener) revealed a lower sensitivity in the general population than in primary-care patients (Chan et al., 1994a; Cherpitel, 1998) and in an emergency room setting (Cherpitel, 1998). Interestingly, among the general population sample of the latter study, the CAGE showed a tendency to perform better among individuals reporting an emergency room or primarycare visit in the previous 12 months and was significantly more sensitive in men with a previous emergency room visit in the last year (Cherpitel, 1999). The lower validity of the CAGE in general population samples corresponds with findings from a large scale Canadian study (Bisson et al., 1999). The Brief MAST (Pokorny et al., 1972), a shortened version of the Michigan Alcoholism Screening Test (MAST; Selzer, 1971), was also less sensitive in a general population sample compared with primary-care out-patients (Chan et al., 1994b). For the TWEAK test (Russell et al., 1994) (acronym based on its five items Tolerance, Worry about drinking, Eye opener, Amnesia (blackouts), and c(K)ut down on drinking), findings are not so clear. In one study, the sensitivity of the TWEAK was lower in the general population, compared to an emergency room sample, but higher compared to primary-care patients, in identifying alcohol dependence (Cherpitel, 1998). In a second study, no differences in sensitivity were found between a general population and a primary-care sample in detecting heavy drinking (Chan et al., 1993). Using alcohol dependence as gold standard, differences in sensitivity of the TWEAK between samples depended on two versions of the tolerance item. In summary, there is evidence that screening questionnaires show different psychometric properties in the general population, compared to samples drawn in medical settings. No data with respect to the validity in the general population could be found for three more recently developed instruments: the Alcohol Use Disorders Identification Test (AUDIT; Babor et al., 1989b; Saunders et al., 1993), the AUDIT Alcohol Consumption Questions (AUDIT-C; Bush et al., 1998) and the Lbeck Alcohol Dependence and Abuse Screening Test (LAST; Rumpf et al., 1997). The AUDIT has been used in the general population; however, data are restricted to subgroups (unemployed; Claussen and Aasland, 1993) and do not give clear estimates of validity such as sensitivity and specificity based on a gold standard (Holmila, 1995; Fleming, 1996; Medina-Mora et al., 1998). The aims of the present study were: (1) to assess and compare the performance of the AUDIT, the AUDIT-C and the LAST in a general population sample; (2) to examine different cut-off points for the three instruments; (3) to analyse age and gender effects; (4) to test whether sensitivity and internal consistency varied in the subsamples of individuals reporting general hospital admissions or general practice visits in the previous 12 months. H.-J. RUMPF et al. SUBJECTS AND METHODS Population area of study The study was part of the project on Transitions in Alcohol Consumption and Smoking (TACOS; Hapke et al., 1998; Rumpf et al., 1998a). The present data are derived from a general population sample in Lbeck, a northern German city with 217 000 inhabitants, and 46 adjoining communities. Individuals born between 1932 and 1978 were randomly drawn from the official resident registration office files (representing the age group 1864 years in the mid of data gathering). In Germany, residents are bound by law to register within 4 weeks after moving to a new place. Therefore, these files are a valuable source to obtain representative samples. All individuals with German nationality (to avoid language problems) and not living (...truncated)