1 Centre for Addiction and Mental Health, Toronto, Canada, 2 Addiction Research Institute, Zurich, Switzerland, 3 Department of Public Health Sciences, University of Toronto, Toronto, Canada and 4 School of Psychoeducation, University of Montreal, Montreal, Canada
* Author to whom correspondence should be addressed at: 33 Russell Street, Room 2035B, Toronto, ON M5S 2S1, Canada. Tel.: +1 416 535 8501 (ext. 4495); Fax: +1 416 260 4156; E-mail: jtrehm{at}aol.com
(Received 20 May 2005; first review notified 13 July 2005; accepted in final revised form 17 August 2005)
![]() |
ABSTRACT |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
![]() |
INTRODUCTION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
It has been shown that relatively stable individual patterns of drinking exist in adolescence (Kerr et al., 2002; Andersen et al., 2003
; Wells et al., 2004
; Duhig et al., 2005
), and that heavy drinking occasions are particularly important for this group and young adults (Wechsler and Issac, 1992
; Milgram, 1993
; Chassin and DeLucia, 1996
; Wechsler et al., 1998
; Gmel et al., 2003
; Kuntsche et al., 2004
). For these age groups, heavy drinking occasions have been linked to the following negative consequences: symptoms of intoxication, such as blackouts or hangovers; school problems, such as missing school classes or getting behind in school work; unplanned and unprotected sexual activities; aggression, ranging from arguments with friends to rape; trouble with authorities at school and outside (e.g. police); injury, including, but not limited to, drunk-driving related consequences (for overviews see Wechsler et al., 1994
; Chassin and DeLucia, 1996
; Gmel et al., 2003
).
The objective of the current paper is to test the influence of volume of drinking and heavy drinking occasions on alcohol-related harm. We hypothesize that both factors independently cause harm, with the following specific hypotheses:
![]() |
METHODS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
For the current analysis, the sample was restricted to 2455 secondary school students (grades 912) from 74 schools, who were asked the Alcohol Use Disorders Identification Test (AUDIT) items; by design, only a random half received these items. The average sample size per school was 32.7 (SD 12.6) and ranged from 5 to 82 students. The sample size was further reduced to 2421 cases due to listwise deletion of missing data for the other covariates used in the model.
Outcome
The main outcome was alcohol-related problems derived from the AUDIT (Saunders et al., 1993). The AUDIT can be segmented into two distinct sections: three questions on consumption (i.e. frequency, quantity per occasion, and heavy drinking occasions) and seven items on alcohol-related problems, either indicators of dependence or harmful use/abuse(i) unable to stop drinking, (ii) failed to do what was expected, (iii) needed a morning drink after a night of heavy drinking, (iv) feeling guilty or remorseful after drinking, (v) unable to remember the previous night due to drinking, (vi) injury of self or others due to drinking, and (vii) a friend, relative, doctor, or other health care workers being concerned about drinking and suggested cutting down (Saunders et al., 1993
). These items are scored based on the frequency of occurrence, mostly ranging from 0 (never) to 4 (almost daily) (Babor et al., 2001
).
Since we were interested in forming one summary scale for alcohol-related problems from these items, a confirmatory factor analysis (Jöreskog and Sörbom, 1979) was conducted to determine whether the items can all be loaded on one latent factor using Mplus 3.01 (Muthén and Muthén, 2004
). Items were entered into the factor analysis as ordinal variables. The analysis revealed that all seven standardized factor loadings were >0.6, with an average loading of 0.74, a Cronbach's alpha of 0.72, and a standardized alpha of 0.76. The model was assessed in terms of model fit and showed that the root mean square error of approximation (Brown and Cudeck, 1993
) is well below the suggested cut point of 0.06, as proposed by Hu and Bentler (1999)
. Thus, we could measure alcohol-related problems with the resulting scale, ranging from 0 to 28 (mean: 1.90, standard deviation: 3.31).
Independent variables
Student level
Conceptually, the independent variables at the student level were volume of alcohol consumption and patterns of drinking. The former was measured in drinks per week and was derived from the OSDUS survey (Adlaf et al., 1999), as the product of frequency of drinking occasions and the quantity consumed per occasion. Heavy drinking, derived from the third AUDIT item, was represented by four dummy variables indicating less than monthly, monthly, weekly, and daily heavy drinking occasions (defined as five or more drinks per occasion), with never having a heavy drinking occasion serving as the reference group. The variables selected on the individual level as potential confounders of the relationship were gender (1 = male, 0 = female), age (in years), SES (from 0 to 10, 10 being the highest class; Currie et al., 1997
), and birthplace (whether the subject was born in Canada or not).
School level
Culture of drinking at the school level was indicated by the mean volume of alcohol consumed in a week by all students and the proportion of students that engage in heavy drinking occasions weekly or more frequently. The actual volume of alcohol consumed weekly was group-centred (calculated by subtracting the corresponding school mean alcohol volume from the individual's alcohol volume consumption) and modelled as a random slope i.e. its slope on the score for alcohol-related problems was allowed to be estimated differently by each school (Raudenbush and Bryk, 2002).
Statistical analysis
Our hierarchical model was developed using HLM 5.05 (Raudenbush et al., 2000), with students grouped by schools. Thus, both individual-level (level 1) and school-level (level 2) effects for a student's score on the alcohol-related problems can be estimated by
![]() |
The random intercept ß0j, which varies by school, is further modelled by
![]() |
The random slope coefficient ß1j indicating the impact of alcohol volume is further determined by
![]() |
The most significant advantage of constructing this model is that we can ascertain not only which individual factors influence the score on the alcohol-related problems scale, but also whether a school culture of drinking, measured by either the volume consumed (per week) or instances of heavy drinking, can influence an individual's level of alcohol-related problems. For sensitivity analyses, weighted regressions were conducted, taking into account the sampling probabilities related to grade and region in the province. In addition, we tested the model without group-centring (for different interpretations of group-centred vs uncentred models see Kreft et al., 1995).
![]() |
RESULTS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
|
Hierarchical linear model analysis
Table 2 shows the results of the analysis for individual-level and school-level effects. Variables at both the student level and school level were used to model the effects of student and school on an alcohol-related problems scale for a student. The model was first developed as a random effects ANOVA (a null model) in order to compare the intraclass correlation coefficient (ICC) with the full model. The ICC comparison of the null vs full model showed that the full model decreased the variation in problems across schools (3.71.9%) when adjusted for influencing factors at the student and school level. However, for both models, variation between schools was highly significant, meaning that explanatory variables did not fully account for between-school variation.
|
The average volume of alcohol consumption was significantly associated with alcohol-related problems (ß = 0.08, t = 5.52, P < 0.001). This slope was school specific and later analyses found that there were significant differences between schools (see Table 3, and below). Of particular interest is that the likelihood of students reporting greater alcohol-related problems increased cumulatively as the frequency of heavy drinking occasions increased. Specifically, having such occasions less than monthly led to, on an average, a score that was 1.2 points higher (on the 28 point scale) compared with never experiencing such occasions. Having monthly heavy drinking occasions resulted in problem scores 2.2 points higher, having weekly occasions resulted in scores
4.1 points higher, and having daily occasions resulted in scores
6.6 points higher. These effects were found after adjusting for all other variables, including average volume of consumption.
|
Sensitivity analyses Weighted analyses revealed practically identical results, where all significant effects remained significant with about the same effect size, but the influence of gender at the student level changed from significance to marginal significance (ß = 0.23, t = 1.86, P = 0.062).
The models using uncentred volume of alcohol consumption did not yield any substantively changed results i.e. all the significant effects stayed significant, with similar effect sizes, and all the non-significant remained non-significant (details not shown).
Variance explained at level 2, and reliability of coefficients
Table 3 presents the estimated variances of the random effects and the test of the hypothesis that these variances were null. Specifically, the estimated variance for the random effect on the alcohol-related problem scores between schools (the random intercept) was 0.12, which was significant (
; P < 0.01) after adjusting for the level 2 predictors. This provides evidence against the null hypothesis and indicates that schools do vary significantly amongst each other in terms of mean alcohol-related problems. Also, it can be seen that the mean alcohol volume per week consumed also varies significantly by school at the P < 0.001 level, with a variance estimate of 0.01.
To look at the effect sizes of the different variance components, the procedures provided by Raudenbush and Bryk (2002) were applied. Overall, in the basic model with random intercept, schools accounted for 3.7% of the overall variance. This effect is highly significant, but small in size. Adding the school-level drinking as the only explanatory variable to this model, we can see that 89.0% of the variance between schools was explained by mean school drinking.
The reliability estimate of the random intercept and the alcohol volume consumed per week random slope were both adequate at 0.38 and 0.60, respectively (Raudenbush and Bryk, 2002).
![]() |
DISCUSSION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Before discussing these implications further, we would like to point out the limitations of the study. The main limitation is the cross-sectional nature of the study. We cannot be sure that the level and patterns of drinking measured at the time of the survey are a good operationalization of the drinking style that caused the alcohol-related problems. We assume that they are, and we find significant relations, but we cannot exclude measurement error or other systematic relationships between the variables, e.g. when drinking led to problems and was consequently reduced. However, the latter relationship would have worked against our hypotheses. Secondly, the operationalization of school drinking culture certainly could be improved. Future research should try to capture indicators for school drinking culture that are recorded independently from the students' responses. While this limitation needs to be acknowledged, it could not alternatively explain the main results.
Another limitation of our results is self-reports. Although we used validated questions from the AUDIT for outcome and validated questions for alcohol consumption for the main independent variable, we cannot exclude that for some respondents the co-variation between them reflects attitudes or preconceptions about alcohol rather than a true relationship (Rehm et al., 1999). However, it is very unlikely that this explanation could explain all of our results. Another limitation of our analysis is the absence of a 3-level model, we cannot be sure that the school effects identified are not neighbourhood effects. Also, as the AUDIT was originally designed with an adult population in mind, there has been some concern of its applicability to an adolescent population. However, previous research has shown that the AUDIT can be used among adolescents (Chung et al., 2000
; Kelly et al., 2002
; Knight et al., 2003
).
With respect to theory, it becomes more evident that average volume or volume alone is not sufficient to predict social or medical harms. This statement seems to be true for adults as well as adolescents. Thus, given the strong effects of frequency of heavy drinking occasions on alcohol-related problems, future epidemiological research, including social epidemiological research, should always include at least one indicator for patterns of drinking (Dawson and Room, 2000). However, simply adding a variable for patterns of drinking may be insufficient. To substantially predict alcohol-related problems and harm, indicators for the environment and interaction terms between consumption indicators and key environmental variables must be included (Rehm et al., 2004
) in future research in this area. In our models there were substantial differences between schools in terms of alcohol-related problems, even after adjusting for individual-level variables and environmental (school level) variables. This finding is an important addition to the existing body of work, especially to that of Kairouz and Adlaf (2003)
, who found that both school-level and student-level variables are important determinants of heavy drinking behaviour. In addition, these findings add weight to the argument that the school setting is a fixed attribute of adolescent alcohol use and abuse (Rountree and Clayton, 1999
; Kairouz and Adlaf, 2003
), regardless of individual-level variables.
Future research in this area may focus on the contextual make-up of schools. In this study, drinking culture within a school was measured from aggregated individual-level attributes, but determining those school specific features that contribute to its unique culture are important in uncovering real school-level effects in adolescent substance use.
The fact that the culture in schools is important in influencing alcohol-related harm, over and above individual drinking levels, offers important avenues for prevention. On the other hand, student-level educational programmes often show none or very modest effects in changing long-term behaviour (Babor et al., 2003; Foxcroft et al., 2003
), measures to change the drinking culture or drinking environment in schools may be more effective. This may include policies for school events and parties, as well as strict enforcement of rules with respect to alcohol in schools. The drinking environment may also be influenced by the availability of alcohol for students around schools (Maes and Lievens, 2003
), e.g. whether age limits are enforced in public sales places and bars.
![]() |
REFERENCES |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Aquilino, W. S. (1997) Privacy effects on self-reported drug use: interactions with survey mode and respondent characteristics. In The Validity of Self-Reported Drug Use: Improving the Accuracy of Survey Estimates, Harrison, L and Hughes, A. eds, pp. 383415. National Institute on Drug Abuse, Rockville.
Adlaf, E. M. and Paglia, A. (2003) Drug Use among Ontario Students: OSDUS Highlights, 19772001: (CAMH Research Document No. 14). Centre for Addiction and Mental Health, Toronto.
Adlaf, E. M., Paglia, A. and Ivis, F. J. (1999) Drug Use among Ontario Students, 19771999: Findings from the OSDUS (CAMH Research Document No. 5). Centre for Addiction and Mental Health, Toronto.
Alexander, C., Piazza, M., Meckos, D. et al. (2001) Peers, schools, and adolescent cigarette smoking. Journal of Adolescent Health 29, 222230.[CrossRef]
Babor, T. F., Higgins-Biddle, J. C., Saunders, J. B. et al. (2001) AUDITThe Alcohol Use Disorders Identification Test: Guidelines for use in primary care, 2nd edn. World Health Organization, Geneva.
Babor, T., Caetano, R., Casswell, S., et al. (2003) Alcohol: no ordinary commodity. Research and public policy. Oxford University Press, Oxford, London.
Bjarnason, T., Andersson, B., Choquet, M. et al. (2003) Alcohol culture, family structure and adolescent alcohol use: Multilevel modeling of frequency of heavy drinking among 1516 year old students in 11 European countries. Journal of Studies on Alcohol 64, 200208.[ISI][Medline]
Brown, M. W. and Cudeck, R. (1993) Alternative ways of assessing model fit. In Testing Structural Equation Models. Bollen, K and Long, J. S. eds, pp. 136162. Sage Publications, Newbury Park.
Bruun, K., Edwards, G., Lumio, M. et al. (1975) Alcohol control policies in public health perspective. Finnish Foundation for Alcohol Studies, Helsinki, Finland.
Burton, R. P. D., Johnson, R. J., Ritter, C. et al. (1996) The effects of role socialization on the initiation of cocaine use: an event history analysis from adolescence into middle adulthood. Journal of Health and Social Behaviours 37, 7590.
Chassin, L. and DeLucia, C. (1996) Drinking during adolescence. Alcohol Health and Research World 20, 175180.
Chung, T., Colby, S. M., Barnett, N. P. et al. (2000) Screening adolescents for problem drinking: performance of brief screens against DSM-IV alcohol diagnoses. Journal of Studies on Alcohol 61, 579587.[ISI][Medline]
Currie, C. E., Elton, R. A., Todd, J. et al. (1997) Indicators of socioeconomics status for adolescents: the WHO Health Behaviour in School-aged Children Survey. Health Education Research: Theory and Practice 12, 385397.
Dawson, D. A. and Room, R. (2000) Towards agreement on ways to measure and report drinking patterns and alcohol-related problems in adult general population surveys: Skarpo Conference Overview. Journal of Substance Abuse 12, 122.[CrossRef][ISI][Medline]
Duhig, A. M., Cavallo, D. A., McKee, S. A. et al. (2005) Daily patterns of alcohol, cigarette, and marijuana use in adolescent smokers and nonsmokers. Addictive Behaviours 30, 271283.[CrossRef][ISI][Medline]
Duncan, S. C., Duncan, T. E. and Strycker, L. A. (2002) A multilevel analysis of neighbourhood context and youth alcohol and drug problems. Prevention Science 3, 125133.[CrossRef][Medline]
Edwards, E., Anderson, P., Babor, T. F. et al. (1994) Alcohol policy and the public good. Oxford University Press, Oxford.
Ellickson, P. L., Bird, C. E., Orlando, M. et al. (2003) Social context and adolescent health behavior: Does school-level smoking prevalence affect students' subsequent smoking behavior? Journal of Health and Social Behavior 44, 525535.[ISI][Medline]
Foxcroft, D. R., Ireland, D., Lister-Sharp, D. J. et al. (2003) Longer-term primary prevention for alcohol misuse in young people: A systematic review. Addiction 98, 397411.[CrossRef][ISI][Medline]
Gfroerer, J. (1985) Influences of privacy on self-reported drug use by youths. In Self-Report Methods of Estimating Drug Use. NIDA Research Monograph 57. Rouse, B. A., Kozel, N. J. and Richards, L. G. eds, pp. 2230. Department of Health and Human Services, Rockville.
Gmel, G., Rehm, J. and Kuntsche, E. (2003) Binge drinking in Europe: definitions, epidemiology, and consequences. Sucht 49, 105116. Available at: http://www.camh.net/research/population_life_course.html (accessed 30 September, 2004).
Hu, L. T. and Bentler, P. M. (1999) Cutoff criteria for fit indices in covariance structure analysis: conventional criteria versus new alternatives. Structural Equation Modeling 6, 155.
Jöreskog, K. G. and Sörbom, D. (1979) Advances in factor analysis and structural equation models. Abt Books, Cambridge, MA.
Kairouz, S. and Adlaf, E. F. (2003) Schools, students and heavy drinking: a multilevel analysis. Addiction Research and Theory 11, 427439.[CrossRef]
Kelly, T. M., Donovan, J. E., Kinnane, J. M. et al. (2002) A comparison of alcohol screening instruments among under-aged drinkers treated in emergency departments. Alcohol and Alcoholism 37, 444450.
Kerr, W. C., Fillmore, K. M. and Bostrom, A. (2002) Stability of alcohol consumption over time: evidence from three longitudinal surveys from the United States. Journal of Studies on Alcohol 63, 325333.[ISI][Medline]
Klingemann, H. and Gmel, G. (2001) Mapping social consequences of alcohol consumption. Kluwer, Boston.
Knight, J. R., Sherritt, L., Harris, S. K. et al. (2003) Validity of brief alcohol screening tests among adolescents: a comparison of the AUDIT, POSIT, CAGE, and CRAFFT. Alcoholism: Clinical and Experimental Research 27, 6773.[ISI][Medline]
Kreft, I. G., de Leeuw, J. and Aikem, L. S. (1995) The effect of different forms of centering in hierarchical linear models. Mulitvariate Behavioural Research 30, 121.
Krohn, M. D., Lanza-Kaduce, L. and Akers, R. L. (1984) Community context and theories of deviant behaviour. An examination of social learning and social bonding theories. Sociology Quarterly 25, 353371.[CrossRef]
Kuntsche, E., Rehm, J. and Gmel, G. (2004) Characteristics of binge drinkers in Europe. Social Science and Medicine 59, 113127.[CrossRef][ISI][Medline]
Maes, L. and Lievens, J. (2003) Can the school make a difference? A multilevel analysis of adolescent risk and health behaviour. Social Science and Medicine 3, 125133.
McPartland, J. M. and Mcdill, E. L. (1977) Violence in Schools: Perspectives, Programs, and Positions. Lexington Books, Lexington, MA.
Milgram, G. (1993) Adolescents, alcohol and aggression. Journal of Studies on Alcohol 11, 5361.
Muthén, L. K. and Muthén, B. O. (2004) Mplus User's Guide. 3rd edn. Muthén and Muthén, Los Angeles, CA.
Norton, E. C., Lindrooth, R. C. and Ennett, S. T. (1998) Controlling for the endogeneity of peer substance use on adolescent alcohol and tobacco use. Health Economics 7, 439453.[CrossRef][ISI][Medline]
Patton, G. C., Carlin, J. B., Coffey, C. et al. (1998) The course of early smoking: a population-based cohort study over three years. Addiction 93, 12511260.[CrossRef][ISI][Medline]
Pink, W. T. (1984) Schools, youth, and justice. Crime and Delinquency 30, 439461.
Raudenbush, S. W. and Bryk, A. S. (2002) Hierarchical linear models: applications and data analysis methods, 2nd edn. Sage Publications, Thousand Oaks, CA.
Raudenbush, S., Bryk, A., Cheong, Y. F. et al. (2000) HLM 5 Hierarchical Linear and Nonlinear Modeling. Scientific Software International, Inc., Lincolnwood, Illinois.
Rehm, J., Ashley, M. J., Room, R. et al. (1996) On the emerging paradigm of drinking patterns and their social and health consequences. Addiction 91, 16151621.[CrossRef][ISI][Medline]
Rehm, J., Fischer, B., Graham, K. et al. (2004) The importance of environmental modifiers of the relationship between substance use and harm. Addiction 99, 663666.[CrossRef][ISI][Medline]
Rehm, J., Frick, U. and Bondy, S. (1999) A reliability and validity analysis of an alcohol-related harm scale for surveys. Journal of Studies on Alcohol 60, 203208.[ISI][Medline]
Rehm, J., Room, R., Graham, K. et al. (2003) The relationship of average volume of alcohol consumption and patterns of drinking to burden of diseaseAn overview. Addiction 98, 12091228.[CrossRef][ISI][Medline]
Rootman, I. and Smart, R. G. (1985) A comparison of alcohol, tobacco and drug use as determined from household and school surveys. Drug and Alcohol Dependence 16, 8994.[CrossRef][ISI][Medline]
Rountree, P.W. and Clayton, R. R. (1999) A contextual model of adolescent alcohol use across the rural-urban continuum. Substance Use and Misuse 34, 495519.
Saunders, J. B., Aasland, O. G., Babor, T. F. et al. (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, 791804.[ISI][Medline]
Wechsler, H., Davenport, A. E., Dowdall, G. W. et al. (1994) Health and behavioural consequences of binge drinking in collegea national survey of students at 140 campuses. Journal of the American Medical Association 272, 16711677.
Wechsler, H., Dowdall, G. W., Maenner, G. et al. (1998) Changes in binge drinking and related problems among American college students between 1993 and 1997. Results of the Harvard School of Public Health College Alcohol Study. Journal of American College Health 47, 5768.[ISI][Medline]
Wechsler, H. and Isaac, N. (1992) Binge drinkers at Massachusetts colleges. Prevalence, drinking style, time trends, and associated problems. Journal of the American Medical Association 267, 29292931.[Abstract]
Wells, J. E., Horwood, L. J. and Fergusson, D. M. (2004) Drinking patterns in mid-adolescence and psychosocial outcomes in late adolescence and early adulthood. Addiction 99, 15291541.[CrossRef][ISI][Medline]
|