Comparing the AUDIT and 3 drinking indices as predictors of personal and social drinking problems in freshman first offenders.

Abstract

The current study of 376 college freshman adjudicated the first time for breaking university drinking rules tested the predictive power of four alcohol consumption and problem drinking indices--recent changes in drinking (the Alcohol Change Index: ACI), heavy drinking, binge drinking index, and the Alcohol Use Disorders Identification Test (AUDIT) with two subscales of the College Alcohol Problem Scale (personal and social problems) as criterion measure. In addition, the ACI, heavy drinking, and the binge-drinking index were tested for receiver operating characteristics (ROC) (sensitivity, specificity, false positive rate, positive predictive value and the area under the curve) using the AUDIT as criterion. Results demonstrated that the AUDIT was the best predictor of personal and social problems, and the binge showed the best ROC data with the AUDIT as criteria. Recommendations for use of brief instruments are suggested.

INTRODUCTION

Research on college drinking has catalogued an array of problems related to excessive drinking in college students including negative psychological (e.g., depression, suicide, anxiety), interpersonal (e.g., fights, unplanned and unprotected sex, date rape) and community problems (e.g., driving under the influence, vandalism) problems (Wechsler, Lee, Kuo, and Lee, 2000; Wechsler, Davenport, Dowdall, Moeykens, and Costillow, 1994; O'Hare, 1990; O'Hare and Sherrer, 1999; O'Leary, Goodhart, Sweet Jemmott, Boccher-Lattimore, 1992; Abbey, 2002). A recent review of five national college drinking surveys (O'Malley and Johnston, 2002) summarized major findings and trends over the past 20 years: over two thirds of college students drink alcohol, 40% are considered binge drinkers (i.e., consumed five or more drinks at one sitting within the past two weeks), and rates of alcohol use have not changed substantially since the 1950's. Wechsler, Lee, Kuo and Lee (2000) estimate that 44% of college students are binge drinkers, a number that has remained relatively steady throughout the 1990's. In comparison to non-college cohorts, college students consume more alcohol. Males are about one and one-half as likely to be binge drinkers compared to females, although differences in alcohol consumption have narrowed over the past ten years or so. Whites continue to consume the most alcohol per capita than the other racial groups. Overall, the consistency of the data among the large national samples supports the overall reliability and validity of these estimates (O'Malley and Johnston, 2002).

The role and importance of alcohol and problem drinking measures

Many different types of measures have been employed in surveys of youthful drinking. These include the percentage of students who drink within specific rime intervals, binge drinking, various measures of heavy drinking, self-reported problems resulting from drinking, as well as various scales, and screening devices, many of which have been validated against DSM diagnostic criteria. All of these instruments measure something different, and the meaning of each measure has different implications for public health education, primary prevention policies and early intervention strategies with young persons who drink. Data regarding the percentage of students who have "ever used" or "used in the past month," for example, can be somewhat misleading as an indicator of problem drinking since the fact that a young person drinks does not necessarily indicate a problem. Given that most people at some time use alcoholic beverages, and do so in a relatively non-problematic way, it is likely that during late adolescence and early adulthood, young persons are likely to initiate the use of beer, wine or spirits, many of them, in moderation. Measures such as binge drinking (i.e., the consumption of five or more drinks within the past two weeks), identify those who consume more than a moderate amount at one sitting, but the term also suggests an unrestrained drinking spree and may suggest an exaggeration of the current problem on college campuses. Although the consumption of five (four for women) or more drinks at one sitting may be cause for concern, an estimation of consequences would likely depend considerably on the context of drinking (e.g., whether the person is driving afterwards or not). Binge drinking has, perhaps, become the most widely used indicator of excessive drinking in college students over the last decade due largely to the reports from a large national college drinking survey (Wechsler et al. 1994).

Heavy drinking, usually measured as five drinks at one sitting within the past week, may indicate a more serious problem (given a higher overall rate of consumption), but, again, the actual risks would have to be gauged relative to situational factors. The binge measure, as employed since 1980 in national studies of college drinking, appears to be a more liberal measure of excessive consumption than heavy drinking in that the binge index allows the respondent twice the time interval within which to report an incident of having consumed five or more drinks. O'Hare, Cohen, and Sherrer (1997) compared a standard quantity-frequency indicator of heavy drinking (5 or more drinks a week, usually), a 7-day retrospective diary (five or more drinks within the past 7 days) and the binge drinking index (5 or more drinks for men, 4 or more for women, within the past 2 weeks) against a modified version of the Michigan Alcoholism Screening Test as criterion with a cohort of college first offenders and found that positive predictive values for the QFI were generally higher than for the binge, and the false positive rate for the binge was three times that of the QFI. The binge drinking index, although producing the highest sensitivity ratings overall, grossly over-estimated the number of problem drinkers. Peak drinking (O'Hare, 1991; Sobell, Cellucci, Nirenberg and Sobell, 1982) which has been defined as having had five or more drinks in one sitting within the past seven days, also performed better than the binge drinking index with undergraduates in that it had a higher positive predictive value than the binge for both women and men (O'Hare, Cohen and Sherrer, 1997). However, overall, alcohol consumption measures alone provide rather crude estimates of problem drinking in this population. In order for them to be more useful, they should be directly linked to the frequency and severity of negative psychosocial, health or community consequences or compared to the results of standardized scales that have been validated with the youthful drinking population.

Problem drinking screens such as the MAST or the AUDIT were developed to provide brief valid detection of alcohol abuse or alcoholism. Often this distinction is ambiguous and studies may differ on exactly what these scales measure. Alcohol Use Disorders Identification Test (AUDIT) (Saunders, Aasland, Babor, de la Fluente, and Grant, 1993), although developed with outpatient clinical samples, has been shown to be reliable and valid with undergraduates as well (Fleming, Barry & MacDonald, 1991; O'Hare & Sherrer, 1999). Given that it is considered a valid indicator of problem drinking based on DSM criteria, it should serve as a reasonably valid criteria against which to measure the sensitivity and specificity of selected alcohol consumption indicators.

Lastly, there are a host of other simple indicators that may be useful. For example, asking a young person about recent changes in drinking behavior may have some value. In one study of college students (O'Hare and Tran, 1997) the alcohol change index (ACI) was demonstrated to be one of the best predictors of overall problems related to alcohol use. This item simply queries whether the respondent has increased, decreased their consumption or kept it about the same over the past three months. Logistic regression analysis based on that sample showed the ACI to be a better predictor of a total score of self-reported alcohol related problems than the CAGE questionnaire, the quantity-frequency index, subjective problem self-ratings and other selected descriptive factors (O'Hare and Tran, 1997).

Although these various approaches to measuring alcohol problems in young persons have their advantages, researchers, prevention and intervention professionals need to be clear about what these indicators actually measure. Although problem drinking in youth may have some long-term prognostic value for future alcoholism (Bennett, McCrady, Johnson, Pandina, 1999), and up to almost one third may meet the DSM criteria for alcohol abuse (Knight, Wechsler, Kou, Seibering, Weitzman, and Schuckit, 2002), it is well established that most college students who drink and even those who report serious psychosocial consequences are likely to attenuate their use during the college years with fewer problems resulting over time (O'Neill, Para and Sher, 2001). Indeed, Because of the transitory nature of most youthful problem drinking (Fillmore, 1988), substance abuse prevention specialists, practitioners, parents, and school administrators may be most concerned about the acute personal and community consequences of excessive alcohol use rather than the development of long-term addiction, which, even with youthful problem drinkers, is very hard to predict.

Nevertheless, alcohol problem inventories and scales have inherent problems. Evidence has consistently shown that, although college students are willing to disclose prodigious amounts of drinking, they do not necessarily link their drinking with negative consequences, or consider the consequences to be serious despite being well aware of the potential dangers of alcohol abuse (Wechsler et al., 1994; O'Hare & Tran, 1997; Perkins, 2002). Even when students do attribute a problem to their drinking, such an inference must be considered somewhat of a guess. A student who reports depression, interpersonal problems or poor grades "as a result of drinking" may not be aware of other factors related to their complaint. Such cause-effect inferences about the effects of drinking must be made cautiously. In addition, many alcohol inventories, although ostensibly useful, measure problems dichotomously (e.g., Have you ever gotten sick from drinking?) rather than measuring frequency or severity on a continuum, and some of these instruments have been subjected to little reliability or validity testing. Thus, little is often known about how these instruments perform relative to other measures of substance abuse such as the AUDIT. Measures such as consumption level, problem indicators and diagnostic tools are all useful, but their relative significance should be determined by both point of view of the person reporting the data (e.g., the student or researcher), the focus of concern (i.e., consumption level, psychosocial problems, health complaint or diagnosis), documented track record of reliability and validity, as well as the purpose of the measure (e.g., epidemiological research, clinical assessment, or outcome goals for primary prevention and early intervention programs).

Freshmen are an important group to study because they can be considered an at-risk group within an at-risk group. College students drink more than their non-college cohorts in the community (Gfoerer, Greenblatt & Wright, 1997), college freshmen increase their drinking from high school to college (Yu and Schacket, 2001; Wechsler, Isaac, Grodstein and Sellers, 1994), and apparently experience more drinking related problems than upper classmen (O'Neill, Para, and Sher, 2001). In addition, moving from high school to college marks an important developmental transition, a time during which drinking may become a coping mechanism to deal with new stressors associated with academic achievement, becoming socially integrated, as well as becoming sexually more active (Schulenberg and Maggs, 2002). The current study of these college freshman adjudicated the first time for breaking university drinking rules tests the predictive power of four important and common alcohol consumption and problem drinking indices--recent changes in drinking (the Alcohol Change Index), heavy drinking (in the current study, 6 or more drinks at one setting on at least a weekly basis, slightly higher than the standard QF; Straus and Bacon, 1953; O'Hare, 1991), the binge drinking index (5 or more drinks at one sitting within the past 2 weeks), and the AUDIT at the conventional cutoff score of 8. Two subscales of the College Alcohol Problem Scale (i.e., personal and social problems) using recently improved factoring (O'Hare, 1997; Maddock, LeForge, Rossi, and O'Hare, 2001) will be employed as a criteria to judge the predictive power of these drinking measures in logistic regression models. In addition, the ACI, peak drinking and the binge drinking index will be tested for sensitivity, specificity, false positive rate, positive predictive value and the area under the curve using the AUDIT as criterion variable given its concurrent validity with DSM criteria for alcohol abuse. (1) Receiver operating characteristics (i.e., sensitivity, specificity, positive predictor value, and area under the curve) (Griner, Mayewski, Mushlin, & Greenland, 1981; O'Hare, Cohen & Sherrer, 1997) represent the most straightforward way of comparing a test scale against a criterion scale by combining "sensitivity and specificity in a manner that allows examination of the ability of tests to discriminate between two populations regardless of the cutoff level or weighting selected" (Wolford et al. 1999, p.320). ROC data are more compelling than correlational analysis since those who deny substance abuse on both the drinking indicator and the criterion variable (in this case, the AUDIT) would artificially inflate the correlation and, thus, erroneously suggest greater concurrent validity (Wolford et al. 1999).

METHODS

Sample and procedure

Three hundred and eighty-nine college freshmen (age M = 18.24, Mdn = 18, SD = .532) at the University of Rhode Island responded consecutively to an anonymous questionnaire during the mid to late 1990's as part of an adjudication process for having been cited by campus authorities (campus police or residence hall assistants) for violating university rules concerning under-aged drinking or the use of illicit drugs. When reporting to the Office of Student Life to pay a required fine, students were provided with a private office to complete the anonymous questionnaire in order to provide an "accurate picture of the range of experiences [relevant to substance use] among the students referred to us." The self-report questionnaire included background questions (gender, age, etc ...), measures of alcohol consumption, related problems and other questions relevant to youthful drinking. Instructions to the respondent were generally self-explanatory, but office staff consisting of the project director (MSW), other trained counselors and graduate students in related fields were available to answer questions. Anonymity was assured by not requesting names or other unique identifiers on the questionnaire. After completing the questionnaire, students were invited to return for further consultation if they had any concerns about their own substance use. The university human subjects review committee approved the instrument and procedures for administering it. In the instructions included with the self-report questionnaire, students were informed that a "drink" was equivalent roughly to a 1.5 ounce shot of liquor (86 proof), one 12 ounce beer, or one 5 ounce glass of wine, all of which approximate .60 ounces of absolute alcohol). These equivalents are very similar to those used by other prominent alcohol researchers (Wechsler, Davenport et al., 1994; Sanchez-Craig, Wilkinson, Davila, 1995).

Study participants are described as follows (in frequencies, percents): gender (female, n = 143, 36.8%/male, n = 246, 63.2%). Almost all of the respondents (98.7%) lived on campus, and almost all respondents were white (95.6%). Family income was self-reported as poor (7, 1.8%), lower middle class (23, 5.9%), middle class (197, 50.6%), upper middle class (143, 36.8%) and wealthy (14, 3.6%) with five respondents (1.3%) not reporting.

Instrument

The College Alcohol Problem Scale

The College Alcohol Problem Scale (CAPS) was developed and replicated with previous cohorts of the current sample (O'Hare, 1997; O'Hare, 1998). The initial item pool was based on an array of instruments used in prominent college drinking studies conducted over the past twenty years (e.g., O'Hare, 1990; Wechsler, Davenport, Dowdall et al., 1994). Chronbach alphas in both samples were comparable (socio-emotional .88, .89; community .79, .76), and the CAPS demonstrated good concurrent validity with the quantity-frequency index, a version of the MAST, and peak drinking index from the AUDIT (O'Hare, 1997; O'Hare, 1998). Using the same 20 items as in the original CAPS, Maddock, LaForge, Rossi, and O'Hare (2001) employed confirmatory factor analysis with university, undergraduates to refine the original CAPS. The study resulted in an 8-item version of the CAPS employing two subscales similar to the original, but labeled personal problems: 1) feeling sad, blue, depressed; 2) nervousness, irritability;; 3) caused to feel bad about oneself; and 4) problems with appetite or sleeping) and social problems (5) engaged in unplanned sexual activity; 6) drove under the influence; 7) did not use protection when engaging in sex; 8) engaged in illegal activity associated with drug use. Internal consistency reliabilities were comparable to those of the original scale (i.e., personal problems alpha = .79, social problems alpha = .75). The current version employs the new items (drawn from the original 20) but retains the original response format: How often have you had any of the following problems over the past year as a result of drinking too much alcohol? (Responses were scored 4 to 0 corresponding with "very often, often, moderate degree, seldom, and never/almost never," respectively). Because the subscale distributions are very positively skewed, both the personal and social problem subscales were recoded into dichotomous measures (experienced one or more incidents of a problem or did not), and these measures will be used as dependent variables in logistic regression models.

The AUDIT

The ten-question AUDIT includes three questions on alcohol consumption (how much, how often, and a question for heavy or peak drinking--"six or more drinks"), four questions on dependence (unable to stop, failed responsibilities, eye opener, guilt), and three on consequences (memory loss, injury to others, advised to cut down). Items are rated (0-4) on a frequency continuum ("never" to "daily"), request measures over the past year, and the scale is scored by summing the 10 items with a possible range of 0-41. A respondent who obtains a cut-off score of [greater than or equal to] 8 is considered a problem drinker in this analysis. The AUDIT has been shown to have an overall sensitivity, of 80% and specificity, of 98% for detecting problem drinking, with good to excellent subscale reliabilities in healthcare patients (Saunders, Aasland, Babor et al. 1993), and good sensitivity (84%) and internal consistency (.80) with undergraduates (Fleming, Barry & MacDonald, 1991). O'Hare & Sherrer (1999) collected data with a previous cohort (n = 312) to evaluate the factorial validity and internal consistency of the ten item AUDIT. Two factors accounted for 59.9% of the variance (alcohol consumption and drinking problems), and were composed of three and seven items, respectively. Chronbach's alphas were calculated at (.78) for consumption, and (.84) for drinking problems, and these factors correlated significantly with the subscales from the original 22-item DCS and the CAPS.

Alcohol Change Index

The Alcohol Change Index (ACI) (O'Hare and Tran, 1997) is a single item question: "In the past three months, how would you best describe your pattern of alcohol use?" Respondents are instructed to choose one of the following answers: 1 "reduced [your alcohol use] alot", 2 reduced ... somewhat, 3 "kept ... about the same", 4 "increased ... somewhat", 5 "increased ... alot". The current study employs a bivariate version of the ACI: increased drinking or not.

Binge drinking

Binge drinking measures five or more (four or more for women) drinks consumed at one sitting within the past two weeks. No adjustment was made in this study for gender differentials given that there are no such norms set for the other measures in this study.

Heavy drinking

Item #3 on the AUDIT requests the respondent to estimate "how often do you have six or more drinks on one occasion?" Since the AUDIT directions imply "usual" drinking practices, this item is close in definition to "heavy drinking" as defined by the standard QF index (Straus and Bacon, 1953; O'Hare, 1991), which is 5 or more (usually) per week.

RESULTS

Major univariate measures

Univariate data for all study variables for total respondents (n = 389) and by gender reveal that almost two-thirds of these adjudicated freshmen met the criteria for problem drinking on the AUDIT and slightly fewer met the criteria for binge drinking. More than one-third met the criteria for heavy drinking (a more conservative measure than the binge index), and almost one-fifth said they had increased their consumption within the past three months. Over one-third reported personal problems and about one-half reported social problems that they attributed to their own alcohol consumption. Males scored significantly higher in every, consumption category than did women, and men reported significantly more social problems related to drinking than did their female counterparts. These results can be further examined in Table 1.

Correlation of major factors

Correlations among the various alcohol indices and the subscales of the CAPS were all statistically significant, but ranged from low to moderate overall. Higher correlations were shown among the binge index, heavy drinking, and the AUDIT. Correlations among the CAPS subscales and the consumption indices were significant but low overall with the exception of the AUDIT, which revealed more moderate correlations with problems related to drinking. A correlation matrix can be examined in Table 2.

Logistic regression

The purpose of employing logistic regression was primarily to test the relative predictive power of each drinking index when controlling for the others. Logistic regression was chosen since the dependent variables are dichotomous and the distributions of the independent variables are not normally distributed (Hosmer & Lemeshow, 1989; Tabachnick & Fidell, 2001). As was expected, the models significantly predicted both personal problems (Chi square (df, 4) = 65.62, p [less than or equal to] .001) and social problems (Chi square (df, 4) = 86.12, p [less than or equal to] .001) significantly. The models also predicted 65.5% and 70.5% of respondents with personal and social drinking-related problems, respectively. Of more specific interest, however, are the odds ratios (i.e., Exp(b) in Table 3) for each of the dichotomous indices. Expanded beta's show proportionately how much the independent predictor changes for each increase in the dependent variable. The AUDIT was the only significant predictor of personal problems (by almost a factor of 6), but both the heavy drinking index (i.e., 5 or more drinks at one sitting within the past week) and the AUDIT predicted social problems significantly by roughly a factor of 2 and 5 times, respectively. The other indices were not significantly predictive of alcohol-related problems as measured by the CAPS. These results can be examined in Table 3.

Receiver Operating Characteristics of selected measures with AUDIT (> 8) as criteria

In the current sample, the binge index demonstrated the best overall balance of sensitivity and specificity of all the indicators, as well as the highest rating for area under the curve (.83) compared to .78 and .58 for heavy drinking and the ACI, respectively. Men, however, showed a somewhat higher false positive rate than did women.

DISCUSSION

Overall, the AUDIT questionnaire appears to be the strongest predictor of personal and social problems (as measured by the CAPS) of all the indicators. The heavy drinking measure also was shown to be a significant predictor of community problems, although the odds ratio was less than half the magnitude of that for the AUDIT. However, when the AUDIT was used as a diagnostic criteria to test the receiver operating characteristics of the three consumption indices (i.e., ACI, heavy drinking and binge), the binge clearly demonstrated the best overall balance of sensitivity and specificity, and accounted for the greatest area under the curve.

Relevance and use for these measures in primary and secondary prevention programs

Certainly, the choice of measures for alcohol surveys and prevention programming is a critical one given the impact that research reports can have on decisions regarding prevention and early intervention practices in high schools, colleges and in communities. Perhaps there is no "best" measure. Most have some strength, and all have their weaknesses. However, determining the validity and clinical significance of excessive or problem drinking in young people is important for a number of reasons. These indicators concisely represent the magnitude of youthful problem drinking in research reports that are disseminated to primary and secondary prevention professionals, policy makers, and often make their way into news reports targeted at the general public. Qualifiers like "epidemic" are often associated with such reports that may alternately heighten public concern out of proportion to the real problem. The perception of exaggerated findings may ultimately reduce the credibility of further reports when closer scientific scrutiny is exercised. This "cry wolf" syndrome only serves to further alienate a public already skeptical of reports of drunken youth running rampant in the streets of college communities. More careful analysis of common problem drinking indices and prevention/early intervention outcomes should include validating them against more specific, meaningful indicators including rates of psychological, social, health and community problems.

As with other enjoyable but potentially harmful behaviors, drinking alcohol is not a risk-free activity (Sanchez-Craig et al. 1995), and relevant prevention policies must also focus on realistic goals which are both significant and achievable. These goals should be measured in a way that is accurate and meaningful in terms of assessing the psychosocial wellbeing and health of young people. However, the meaning of alcohol consumption, diagnostic and drinking problem measures also depends on the point of view of the respondent. Young persons in college tend to see drinking as a generally enjoyable activity associated with having fun and facilitating social relations (Carey, 1993), and some evidence suggests that students who use alcohol report better social adjustment and satisfaction (Senchak, Leonard & Greene, 1998). Nevertheless, drinking to excess has also been shown to be associated with psychological distress as some young persons may use alcohol in a dysfunctional attempt to cope with negative emotions (Evans and Dunn, 1995). Few of them would likely argue that the more acute effects from alcohol abuse (e.g., date rape, car accidents, serious vandalism) are a problem, but would likely see these as low frequency events and not likely to "happen to me." However, the point of view shared by health and mental health clinicians, university administrators, prevention experts and local law enforcement officials is that youthful drinking is a significant problem as measured with diagnostic tools, incidents of alcohol poisoning, reports of alcohol-related psychological symptoms, interpersonal problems, fights, acts of vandalism, accidents, and sometimes death. Perhaps the best recommendation to be made from current and previous data would be to continue to use the binge measure as an indicator of overall rates of problem drinking in broad surveys, but to employ multidimensional, brief and valid measures such as the AUDIT or the CAPS to enhance screening and assessment with at-risk groups, such as these adjudicated freshmen, who are specifically targeted for early intervention programs. These tested measures, used in combination, would reliably identify consumption level as well as physiological indicators of more serious drinking problems, personal and social consequences of abuse such as depression and interpersonal problems, and community indicators of problem drinking such as risky sexual behaviors and driving under the influence. In using measures that go beyond consumption level, assessment and intervention goals can be targeted to the individual young person's needs.

Achieving robust successes with broad-brush primary prevention strategies remains a challenge for prevention programmers (Wechsler, Lee, Kuo, Seibring, Nelson and Lee, 2002), but programs that target at-risk groups with coping skills strategies have demonstrated modest success with adolescents and young adults dealing with a range of health risk behaviors including substance abuse, risky sexual behaviors, and smoking (Botvin, Griffin, Diaz, and I fill-Williams, 2001; Sikkema, Winet and Lombard, 1995; Marlatt et al., 1998). Successful strategies may include some combination of the following ingredients: psychoeducation that includes models of addiction, consequences of alcohol abuse, estimating blood alcohol levels, teaching drinking moderation skills, training in relaxation, nutrition and encouragement to engage in aerobic exercise, monitoring and coping with risky drinking situations more assertively, challenging erroneous alcohol expectancies (e.g., alcohol makes me sexier), and relapse prevention skills to better adhere to one's self-imposed limits and rules concerning moderate drinking. A battery of brief, valid and reliable instruments such as those examined here can provide a cost-effective approach to individual assessment and program evaluation for prevention and early intervention programs.

Author Note

Correspondence concerning this article should be addressed to Tom O'Hare, Boston College, McGuinn Hall, Room 202, Chestnut Hill, MA. 02167-3807. E-mail: [e-mail]

References

Abbey, A. (2002). Alcohol-related sexual assault: A common problem among college students. Journal of Studies on Alcohol, 63, 118-128.

Bennett, M. E., McCrady, B. S., Johnson, V., & Pandina, R. J. (1999). Problem drinking from young adulthood to adulthood: Patterns, predictors and outcomes. Journal of Studies on Alcohol, 60, 605-614.

Botvin, G. J., Griffin, K. W., Diaz, T., & I fill-Williams, M. (2001). Preventing binge drinking during early adolescence: One- and two-year follow-up of a school-based prevention intervention. Psychology of Addictive Behaviors, 15, 360-365.

Carey, K. B. (1993). Situational determinants of heavy drinking among college students. Journal of Counseling Psychology, 40, 217-220.

Evans, D. M., & Dunn, N. J. (1995). Alcohol expectancies, coping responses and self-efficacy judgments: A replication and extension of Cooper et al's 1988 study in a college sample. Journal of Studies on Alcohol, 56, 186-193.

Fillmore, K. M. (1988). Alcohol use across the life course. Toronto, Ontario, Canada: Addiction Research Foundation.

Fleming, M. F., Barry, K. L., & MacDonald, R. (1991). The Alcohol Use Disorders Identification Test (AUDIT) in a college sample. The International Journal of the Addictions, 26, 1173-1185.

Gfoerer, J. C., Greenblatt, J. C., & Wright, D. A. (1997). Substance use in the US college-age population: Differences according to educational status and living arrangement. American Journal of Public Health, 87, 62-65.

Griner, P. F., Mayewski, R. J., Mushlin, A. I., & Greenland, P. (1981). Selection and interpretation of diagnostic tests and procedures. Annals of Internal Medicine, 94, 553-592.

Hosmer, D. W., & Lemeshow, S. (1989). Applied logistic regression. New York: Wiley.

Knight, J. R., Wechsler, H., Kou, M., Seibering, M., Weitzman, E. R., & Schuckit, M. A. (2002). Alcohol abuse and dependence among U.S. college students. Journal of Studies on Alcohol, 63, 263-270.

Maddock, J. E., Laforge, R. G., Rossi, J. S., & O'Hare, T. (2001). The college alcohol problems scale. Addictive Behaviors, 26, 385-398.

Marlatt, G. A., Baer, J. S., Kivlahan, D. R., Dimeff, L. A., Larimer, M. E., Quigley, L. A., Somers, J. M., & Williams, E. (1998). Screening and brief intervention for high-risk college student drinkers: Results from a 2-year follow-up assessment. Journal of Consulting and Clinical Psychology, 66, 604-615.

O'Hare, T. (1990). Drinking in college: Consumption patterns, problems, sex differences and legal drinking age. Journal of Studies on Alcohol, 51, 536-541.

O'Hare, T. (1991). Measuring alcohol consumption: A comparison of the retrospective diary and the quantity frequency methods in a college drinking survey. Journal of Studies on Alcohol, 52, 500-502.

O'Hare, T. (1997). Measuring problem drinking in first time offenders: Development and validation of the College Alcohol Problem Scale. Journal of Substance Abuse Treatment, 14, 383-387.

O'Hare, T. (1998). Replicating the College Alcohol Problem Scale with college first offenders. Journal of Alcohol and Drug Education, 43, 75-82.

O'Hare, T., Cohen, E C., & Sherrer, M. V. (1997). Comparing the QFI, the retrospective diary, and binge drinking in college first offenders. Journal of Alcohol and Drug Education, 42, 40-53.

O'Hare, T., & Sherrer, M. (1999). Validating the alcohol use disorder identification test with college first offenders. Journal of Substance Abuse Treatment, 17, 113-119.

O'Hare, T., & Tran, T. V. (1997). Predicting problem drinking in college students: Gender differences and the CAGE questionnaire. Addictive Behaviors, 22, 13-21.

O'Leary, A., Goodhart, E, Sweet Jemmott, L., & Boccher-Lattimore, D. (1992). Predictors of safer sex on the college campus: A social cognitive theory analysis. Journal of American College Health, 40, 254-263.

O'Malley, P. M., & Johnston, L. D. (2002). Epidemiology of alcohol and other drug use among American college students. Journal of Studies on Alcohol, 63, 23-39.

O'Neill, S. E. Parra, G. R., & Sher, K. J. (2001). Clinical relevance of heavy drinking during the college years: Cross-sectional and prospective perspectives. Psychology of Addictive Behaviors, 15, 350-359.

Perkins, W. (2002). Surveying the damage: A review of research on consequences of alcohol misuse in college populations. Journal of Studies on Alcohol, 63, 91-100.

Sanchez-Craig, M., Wilkinson, A., & Davila, R. (1995). Empirically based guidelines for moderate drinking: 1-year results from three studies with problem drinkers. American Journal of Public Health, 85(6), 823-828.

Saunders, J. B., Aasland, O. G., Babor, T. E, de la Fluente, J. R., & Grant, M. (1993). Development of the Alcohol Use Disorders Identification Test (AUDIT): WHO collaborative project on early detection of persons with harmful alcohol consumption-II. Addiction, 88, 791-804.

Schulenberg, J. E., & Maggs, J. L. (2002). A developmental perspective on alcohol use and heavy, drinking during adolescence and the transition to young adulthood. Journal of Studies on Alcohol, 63, 54-70.

Senchak, M., Leonard, K. E., & Greene, B. W. (1998). Alcohol use among college students as a function of their typical social drinking context. Psychology of Addictive Behaviors, 12, 62-70.

Sikkema, K. J., Winett, R. A., & Lombard, D. N. (1995). Development and evaluation of an HIV-risk reduction program for female college students. Aids Education and Prevention, 7, 145-159.

Sobell, L. C., Cellucci, T., Nirenberg, T. D., & Sobell, M. B. (1982). Do quantity-frequency data underestimate drinking-related health risks? American Journal of Public Health, 72(8), 823-828.

Straus, R., & Bacon, S. D. (1953). Drinking in college. New Haven, CT: Yale University Press.

Tabachnick, B. G., & Fidell, L. S. (2001). Using multivariate statistics, 4th edition. Needham Heights, MA: Allyn and Bacon.

Wechsler, H., Davenport, A., Dowdall, G., Moeykens, B., & Costillow, S. (1994). Health and behavioral consequences of binge drinking in college. Journal of the American Medical Association, 272, 1672-1677.

Wechsler, H., Isaac, N. E., Grodstein, E, & Sellers, D. (1994). Continuation and initiation of alcohol use from the first to the second year of college. Journal of Studies on Alcohol, 55, 41-45.

Wechsler, H., Lee, J. E., Kuo, M., & Lee, H. (2000). College binge drinking in the 1990's: A continued problem: Results of the Harvard School of Public Health 1999 college alcohol study. American Journal of College Health, 48, 199-210.

Wechsler, H., Lee, J. E., Kuo, M., Seibring, M., Nelson, T. E, & Lee, H. (2002). Trends in college binge drinking during a period of increased prevention efforts: Findings from 4 Harvard School of Public Health College Alcohol Study Surveys: 1993-2001. Journal of American College Health, 50, 203-217.

Wolford, G. L., Rosenberg, S. D., Drake, R. E., Mueser, K. T., Oxman, T. E., Hoffman, D., Vidaver, R. M., Luckoor, R., & Carrieri, K. L. (1999). Evaluation of methods for detecting substance use disorder in persons with severe mental illness. Psychology of Addictive Behaviors, 13, 313-326.

Yu, J., & Shacket, R. W. (2001) Alcohol use in high school: Predicting students' use and alcohol problems in four-year colleges. American Journal of Drug and Alcohol Abuse, 27, 775-793.

Thomas O'Hare, Boston College

(1) The sensitivity (i.e., probability that the test or indicator will be positive when the condition is present), specificity (i.e., probability that the test or indicator will be negative when the condition is not present), positive predictive value (i.e., probability that the condition is present when the test or indicator is positive), and false positive rate (i.e., probability that the consumption measure will be positive when the condition is not present)

Table 1
Univariate data for major factors for total sample (n = 389) and by
women (n = 143) and men (246) Univariate data for major study
variables for total samples (N= 389) and by gender

                                 Total         Female

                                  f     %      f     %

ACI                               73   18.8    23   16.1
Binge drinking                   241   62.0    66   46.2
Heavy drinking                   142   36.5    31   21.7
AUDIT ([greater than or equal
  to] 8)                         257   66.1    73   51.0
Personal problems                145   37.3    47   32.9
Social problems                  191   49.1    55   38.5

                                  Male

                                   f    %    Chi Sq.     Cramer's V

ACI                                50  20.3    .99
Binge drinking                    175  71.1  23.95 **  .25 **
Heavy drinking                    111  45.1  21.44 **  .24 **
AUDIT ([greater than or equal
  to] 8)                          184  74.8  22.75 **          .24 **
Personal problems                  98  39.8   1.88
Social problems                   136  55.3  10.24 **          .16 **

** P [less than or equal to] .01

Table 2
Pearson correlation matrix of major factors (N = 389)

                               ACI Binge Heavy AUDIT ([greater than
                                     or equal to] 8) Personal

Binge                       .24 **
Heavy drinking              .23 **   .45 **
AUDIT ([greater than or
  equal to] 8)              .20 **   .65 **   .54 **
Personal problems           .08      .28 **   .27 **   .38 **
Social problems             .13 **   .33 **   .35 **   .44 **   .46 **

** p [less than or equal to] .01

Table 3
Logistic regression using dichotomous measures of Personal and Social
drinking problems as dependent variables (n=389) examining selected
consumption indices and the AUDIT as predictors.

                                          Dependent variable: College
                                             Alcohol Problem scale:
                                               personal problems

Factor                                      B       S.E.      Wald

ACI                                        -.074    .286       .067
Binge drinking                              .250    .320       .611
Heavy drinking                              .323    .259      1.553
AUDIT ([greater than or equal to] 8)       1.774    .385     21.22
Constant                                  -2.131    .292     53.411

                                          Dependent variable: College
                                             Alcohol Problem scale:
                                               personal problems

Factor                                      df     Signif.   Exp(B)

ACI                                         1       .796       .929
Binge drinking                              1       .434      1.284
Heavy drinking                              1       .213      1.381
AUDIT ([greater than or equal to] 8)        1       .000      5.896
Constant                                    1       .000       .119

                                          Dependent variable: College
                                             Alcohol Problem scale:
                                               personal problems

Factor                                    B       S.E.      Wald

ACI                                       .150      .300     .252
Binge drinking                            .152      .312     .237
Heavy drinking                            .648      .648     .271
AUDIT ([greater than or equal to] 8)     1.655      .346   22.847
Constant                                -1.550      .239   44.986

Factor                                    df     Signif.   Exp(B)

ACI                                       1       .616      1.162
Binge drinking                            1       .626      1.164
Heavy drinking                            1       .017      1.911
AUDIT ([greater than or equal to] 8)      1       .000      5.235
Constant                                  1       .000       .212

Table 4
Receiver Operating Characteristics for study variables
for total sample (N = 389) and by gender (2)

Alcohol Change Index           Total     Women       Men

Sensitivity                     24.61     26.03     24.04
Specificity                     92.13     94.03     90.00
Positive predictive value       86.30     82.61     88.00
False positive rate              7.87      5.97     10.00
Area under the curve              .58       .60       .57

Heavy drinking                 Total     Women       Men

Sensitivity                     55.25     42.47     60.33
Specificity                    100.00    100.00    100.00
Positive predictive value      100.00    100.00    100.00
False positive rate              0.00      0.00      0.00
Area under the curve              .78       .71       .80

Binge drinking                 Total     Women       Men

Sensitivity                     84.44     79.45     86.41
Specificity                     81.82     88.57     74.19
Positive predictive value       90.04     87.88     90.86
False positive rate             18.18     11.43     25.81
Area under the curve              .83       .84       .80

(2) Sensitivity=True Positives/(True Positives + False Negatives);
Specificity=True Negatives/(True Negatives + False Positives); Positive
Predictive Value (PPV)=True Positives/(True Positives + False
Positives); False Positive Rate (FPR)=False Positives/(False
Positives+True Negatives)