Alcohol Screening Questionnaires in Women:  A Critical Review FREE

ALTHOUGH less than 5% of women in the US general population develop alcohol abuse and dependence,¹ rates of lifetime alcohol abuse or dependence among women in primary care settings have ranged from 23% to 25%.^2- 3 Nine percent of women seeking primary care have had alcohol-related problems during the past year.⁴ Twelve percent of asymptomatic gynecology patients met interview criteria for current alcohol abuse and dependence in 1 setting,⁵ whereas higher rates have been observed in symptomatic gynecology patients.^6- 9 Of women seeking care in an emergency department, 37% of current drinkers met criteria for alcohol dependence or had at least 1 adverse consequence of drinking.¹⁰

Women with alcohol-related problems are half as likely as men to have received any alcohol-related treatment, and women who seek specialized treatment for alcohol problems have more severe alcohol problems than men, suggesting delayed treatment.⁴ In primary care settings, women's alcohol-related problems are less likely than men's to be recognized and addressed by health care providers (24% compared with 67% in 1 study¹¹). Routine screening programs that identify patients with alcohol abuse or dependence for their providers increase the proportion of women with alcohol-related problems who are counseled.¹²

Women also appear to be more susceptible than men to the medical complications of drinking. A self-report of more than 2 drinks daily on average is associated with increased mortality, cirrhosis, and breast cancer in women.^13- 15 Because brief primary care interventions can decrease heavy drinking and alcohol-related morbidity,^16- 17 screening programs for identifying women who drink heavily are needed.

Screening questionnaires are superior to laboratory tests, including hepatic enzymes, for detecting heavy or problem drinking in unselected populations.^18- 19 Unfortunately, studies validating alcohol screening questionnaires have often not included women or not presented sex-specific analyses. Alcohol screening questionnaires may perform differently in women than men for several reasons. The increased stigma associated with heavy drinking by women²⁰ might lead women to underreport alcohol consumption and related problems more often than men. Also, women are less likely than men to experience overt social consequences of heavy drinking such as employment, economic, or legal difficulties^21- 22 and may therefore be missed by screening questionnaires that target these experiences. Finally, women suffer from adverse consequences of drinking at lower levels of consumption than men. Questions about alcohol consumption that identify men with alcohol abuse or dependence consequently might be less sensitive in women.²³

The purpose of this review was to summarize the published, peer-reviewed literature regarding the performance of screening questionnaires for heavy drinking and/or alcohol abuse or dependence in general clinical populations of women in the United States. Sensitivities, specificities, and areas under receiver operating characteristic curves (AUROCs) for alcohol screening questionnaires were compared for men and women, as well as for women from different ethnic and racial groups.

We limited this review to alcohol screening questionnaires with 10 or fewer items. Longer questionnaires are impractical in many settings because of the need to screen for multiple health-related behaviors (eg, smoking, sexual practices). Six brief screening questionnaires recommended for use with adult women (CAGE, Brief Michigan Alcoholism Screening Test [BMAST], T-ACE, TWEAK, NET, and Alcohol Use Disorders Identification Test [AUDIT]) were identified from a recent National Institute of Alcohol Abuse and Alcoholism review.²⁴ Three additional brief screening questionnaires were known to us.^25- 27

We searched MEDLINE from 1966 to July 1997 for alcoholism or alcohol-drinking and CAGE, BMAST, T-ACE, TACE, TWEAK, MAST, SMAST, SAAST or AUDIT, limited to English. We selected this search strategy, using the names of specific screening questionnaires, because it identified more relevant studies than using broader strategies. We included the Michigan Alcoholism Screening Test (MAST), Short Michigan Alcoholism Screening Test (SMAST), and Self-Administered Alcohol Screening Test (SAAST) questionnaires despite their lengths of more than 10 items because they were the earliest alcohol screening tests and we expected brief screening questionnaires to have been compared with these longer screening tests. Based on this MEDLINE search we identified 432 unique articles.

Two of our targeted screening questionnaires did not have acronyms. Instead of using MEDLINE to identify studies of these questionnaires, we performed searches of Social Science and Science Citations Indexes for the original articles describing these screening tests.^25- 26 In addition, we used Citations Indexes to search for 2 other recently described screening questionnaires, the PRIME-MD and the NET.^27- 28 An additional 206 articles were identified by these searches.

To identify recent studies of screening questionnaires described since publication of the National Institute of Alcohol Abuse and Alcoholism's 1995 review,²⁴ we searched MEDLINE from 1996 to July 1997 for alcoholism or alcohol-drinking and screening , also limited to English. This search identified 91 new articles. Thirty additional articles evaluating alcohol screening questionnaires were identified from the references of articles identified by these searches.

Studies were included in this review if they compared a brief alcohol screening questionnaire with an appropriate criterion standard for heavy drinking or alcohol abuse or dependence in a generalizable, clinical population of US women.

Titles and abstracts were reviewed initially to exclude studies of inappropriate populations. We limited our review to US studies because cultural norms and drinking habits vary markedly in different countries and the performance of screening questionnaires may therefore vary as well. We excluded studies of nonclinical populations because alcohol-related questions likely perform differently when administered in circumstances such as entry into college or after driving while intoxicated. We excluded special clinical populations because of their lack of generalizability. Studies comparing patients in alcohol or drug treatment with other drinkers were excluded because of spectrum bias.

The remaining articles or their abstracts were reviewed to determine whether they compared a brief alcohol screening questionnaire with a validated criterion standard. Studies without an adequate comparison, as defined herein, were excluded.

Alcohol Abuse or Dependence.Alcohol dependence refers to a repetitive pattern of excessive alcohol use with serious adverse consequences, often including lack of control, tolerance, and withdrawal. Alcohol abuse refers to continued drinking despite adverse consequences (in the absence of dependence). Acceptable criterion standards for alcohol abuse or dependence were based on Diagnostic and Statistical Manual of Mental Disorders (DSM) or International Classification of Diseases (ICD) criteria.^29- 30 Standardized interviews based on these criteria include the alcohol module of the Diagnostic Interview Schedule (DIS),³¹the substance abuse and dependence sections of the Composite International Diagnostic Interview,³² and the Alcohol Use Disorder and Associated Disabilities Interview Schedule.³³ Several studies used self-administered versions of the DIS.^3,34 In general,DSM and ICD criteria show good to excellent agreement for alcohol dependence, but not between ICD harmful use and DSM alcohol abuse.³⁵ Therefore, we excluded data regarding screening for ICD harmful use.

Heavy Drinking. Acceptable criterion standards for heavy drinking were based on in-depth interviews and included the Alcohol Use Disorder and Associated Disabilities Interview Schedule³³ and a timeline follow-back procedure.³⁶ We did not review studies using self-administered questions asking for patients' estimates of typical quantity and frequency of drinking as a criterion standard,³⁷ because such self-report measures are less accurate. Studies have varied slightly in their definitions of heavy drinking, but in general, heavy drinking for women is defined as having an average of 2 or more standard-size alcoholic drinks daily.

Data regarding sensitivity, specificity, and AUROCs were extracted from reviewed studies by the first author, and the accuracy of all extracted data was checked by a second author (J.B.W. or M.L.B.). Sensitivity refers to the true-positive rate (number who meet diagnostic criteria and screen positive/total number who meet diagnostic criteria), and specificity refers to the true-negative rate (number who do not meet diagnostic criteria and screen negative/total number who do not meet diagnostic criteria). The ROC curves plot sensitivity vs (1−specificity) at each cut point, and the AUROC reflects the overall performance of a screening test, with 1.0 being a perfect test and 0.5 representing a test that provides no information.³⁸ We did not pool data for meta-analytic purposes because studies meeting inclusion criteria were heterogeneous, with important population and methodological differences that appeared to affect questionnaire performance. Methodological limitations of the reviewed studies were divided into 7 categories(Table 1). Two authors (K.A.B. and M.L.B.) independently coded each reviewed study for the presence or absence of each of 7 limitations.

Thirty-six articles met our inclusion criteria with regard to population and criterion standards, but 3 studies included only men, and 20 studies did not specify sex or presented pooled data for women and men. The 13 articles meeting inclusion criteria described 9 studies and evaluated 8 brief screening questionnaires (Table 2).^{2- 3,5,10,28,34,36,39- 44} (Articles that met study criteria but did not present sex-specific data, alcohol screening questionnaires that have not been compared with an adequate criterion standard in a US clinical population, and numbers of articles identified by each search and reasons for exclusion are available from the authors on request.)

Table 3 summarizes sex-specific data on sensitivity, specificity, and AUROCs for screening questionnaires compared with either heavy drinking or alcohol abuse and dependence. Two studies were omitted from Table 3, one because reported sensitivities appear to be based on only 6 women with alcohol dependence,⁴⁴ the other because it did not present sensitivity, specificity, or the AUROC.⁵ Methodological limitations of each study are noted in Table 3. Agreement between 2 authors' coding of study limitations was 60 (95%) of 63, with incongruities resolved by consensus after re-review of the relevant studies.^2,41,45

Alcohol Abuse and Dependence. The CAGE, AUDIT, and TWEAK questionnaires were the optimal tests for identification of alcohol dependence in women (Table 3). Although an emergency department study suggested that sensitivities are higher for the TWEAK and CAGE questionnaires than for the AUDIT questionnaire,⁴² the sensitivity of the AUDIT using cut points less than 7 was not reported for women. The high specificity of the AUDIT at a cut point of 7 (95%) suggests that using lower cut points might result in the optimal balance of sensitivity and specificity.

The sensitivity of screening questionnaires for alcohol dependence may be affected by the race or ethnicity of the screened population (Table 3). Although the CAGE questionnaire has performed adequately in predominantly black populations of women, it has had a sensitivity of only 0.50 for past-year alcohol abuse and dependence in white, female emergency department patients using the traditional cut point.¹⁰ The sensitivity of the CAGE questionnaire for a lifetime diagnosis of alcohol abuse or dependence was 0.38 in the only primary care study of a predominantly white population of women.³⁴ There are fewer data available for the AUDIT or TWEAK questionnaires in predominantly black or white populations. In emergency department patients, however, the AUDIT and TWEAK questionnaires trended toward increased sensitivity for alcohol dependence in black women.¹⁰

Heavy Drinking. The only study to evaluate a screening questionnaire for detecting heavy drinking in nonpregnant women found the AUDIT questionnaire to be effective (Table 3). In studies of black obstetric populations, the CAGE questionnaire had a relatively low sensitivity for periconceptional heavy drinking, whereas questionnaires asking about tolerance, a diminished effect of alcohol with continued use, were more sensitive (TWEAK, T-ACE, and NET).

The AUCs for alcohol screening questionnaires do not appear to be significantly different for men and women (Table 3).^39,41,43 However, equal AUROCs in women and men do not rule out systematic differences between women and men in the sensitivity and specificity of questionnaires at each cut point. Most studies of alcohol screening questionnaires have observed trends toward lower sensitivities in women than men at equivalent cut points (Table 3). Sex differences in the sensitivity of alcohol screening questionnaires, particularly the CAGE questionnaire, may be greater in white populations than in black populations.^{2- 3,10,34,39,46}

Several studies of alcohol screening questionnaires found higher specificities in women than men,^10,42,46 although this finding is not consistent (Table 3). In 1 study, the differences in specificities between men and women were far greater than the differences in sensitivities, resulting in much higher positive likelihood ratios for women than for men.¹⁰

Several trends emerged from our review, despite the diverse settings and methods of the included studies. First, the performance of alcohol screening questionnaires in women may vary by race and ethnicity. The CAGE, AUDIT, and TWEAK questionnaires were more sensitive for alcohol abuse and dependence in studies of black women than of white women. In addition, the TWEAK questionnaire may perform better than the CAGE or AUDIT questionnaires in white women. Using traditional cut points, the CAGE or AUDIT questionnaires have missed 41% to 62% of white women with alcohol dependence.^10,34 Second, brief alcohol screening questionnaires may be less sensitive for alcohol abuse or dependence among women than among men, particularly screening questionnaires asking about alcohol consumption. Therefore, it may be necessary to use different cut points in women than in men.

Only 7 studies of acceptable quality have reported data on the validity of alcohol screening questionnaires for female patients in the United States, and only emergency department and obstetric studies have compared multiple alcohol screening questionnaires in a single population of women. The data are especially limited with regard to identification of heavy drinking among nonpregnant women. The sole study to compare a screening questionnaire with an adequate criterion standard for heavy drinking in nonpregnant women evaluated only the AUDIT questionnaire.⁴³ However, 3 studies of obstetric patients used a criterion standard of heavy drinking before patients recognized they were pregnant,^28,36,40 and therefore may suggest effective screening questionnaires for identification of heavy drinking in nonpregnant women.

Inconsistent findings among studies may in part reflect methodological differences. In studies comparing multiple screening questionnaires or using interview criterion standards for both alcohol consumption and related problems, the order of screening questions and interview components may affect responses.⁴⁷ Self-administered screening questionnaires may perform less well than interviewer-administered screening questionnaires.⁴⁸

Publication bias may limit the generalizability of this review. Studies finding no effect of sex may have been more likely to report aggregate data and, thus, would be excluded from our review. Additional biases may result from the populations studied. Emergency department patients with injuries or women obtaining prenatal care, especially those with alcohol abuse or dependence,⁴⁹ may underreport alcohol consumption and related problems. Geographic variation may also bias our findings. Studies comparing questionnaire performance in black and white patients and describing predominantly black primary care patients have been conducted in the South, and attitudes toward drinking may be different in the Northeast and West. Finally, several studies have presented data only for current drinkers or women who ever drank alcohol, which might lower the reported performance of screening questionnaires.⁴³ Therefore, pending additional data, our conclusions should be viewed as tentative.

Additional research is needed to assess the performance of alcohol screening questionnaires in diverse clinical populations of women. Several screening questionnaires have not been evaluated in US clinical populations of women or compared with appropriate criterion standards. In addition, a sex-specific modification of the AUDIT questionnaire should be evaluated. Because women report increased psychosocial problems at lower levels of consumption than men,⁵⁰ the AUDIT question that asks about the frequency of heavy drinking could ask women about the frequency of drinking "4 or more" instead of "6 or more" drinks. Finally, the performance of the AUDIT, CAGE, and TWEAK questionnaires in women should be further evaluated, using lower cut points than have been recommended for men. Future studies should strive to have adequate numbers of women to stratify analyses by race, ethnicity, and age.

Based on the data reviewed, the 5-item TWEAK questionnaire appears to be the optimal screening questionnaire for identifying women with heavy drinking or alcohol abuse and dependence in racially mixed populations. Although the "hold" version of the TWEAK questionnaire has been validated more extensively, many women will have never passed out from alcohol use. Therefore, the "high" tolerance question may be more appropriate for screening some populations. If space on self-administered instruments permits, the 10-item AUDIT questionnaire is a reasonable alternative, with the advantage of providing specific information regarding patients' alcohol consumption and symptoms of dependence. The CAGE questionnaire is a reasonable choice for identification of past year or lifetime alcohol dependence (but probably not heavy drinking) in predominantly black female populations. Whichever questionnaire is used, lower thresholds for a positive screening result should be used for women than for men to identify equal proportions of women and men who have alcohol-related problems.⁴² For women, reasonable cut points are 2 points or more for the TWEAK questionnaire, 4 points or more for the AUDIT questionnaire, and 1 point or more for the CAGE questionnaire.

Screening is only the first step in the process of assessment of alcohol problems. For women with positive results on screening questionnaires, current drinking practices, adverse consequences of drinking, symptoms of dependence, and motivation to change should be assessed. When possible, use of a standardized instrument for diagnosing alcohol abuse or dependence^27,32 will improve diagnostic accuracy. Depending on the severity of problem drinking, women should be offered referral to an alcohol treatment program and/or brief feedback and advice.¹⁷ In referring women to alcohol treatment, the approach should be empathetic and nonconfrontational,⁵¹ and care should be taken to address specific barriers, including child care. Referring women to all-female alcohol treatment programs may improve dropout rates and outcomes.⁵²