Adolescent risk factors for excessive alcohol use at age 32 years. A 16-year prospective follow-up study
To examine which socioeconomic, family, personal and lifestyle risk factors in adolescence were the strongest independent predictors of excessive alcohol use in adulthood.
In a prospective longitudinal study, all 16-year-olds of one Finnish city completed questionnaires at school, and were followed up by postal questionnaires at 32 years of age [n = 1,471, (females n = 805, males n = 666); response rate 70.3%). The alcohol use disorders identification test (AUDIT) was used to assess alcohol use in adulthood. AUDIT scores of 8 or more for females and 10 or more for males were classified as excessive alcohol use. Adolescent risk factors examined were parental social class, school performance, depressive symptoms, self-esteem, impulsiveness, parental divorce, relationships with parents, parental trust, health behaviour, leisure-time spent with friends, dating, and problems with the law.
All the socioeconomic, family, personal, and lifestyle variables in adolescence, except parental social class in both genders and self-esteem among females, showed significant univariate associations with excessive alcohol use at age 32 years. Multivariate logistic regression analysis showed that among adolescent males, parental divorce, moderate and high level of depressive symptoms, leisure-time spent daily among friends and moderate and drunkenness-orientated drinking were the strongest predictors of excessive alcohol use in adulthood. Among females, the strongest adolescent predictors of excessive alcohol use in adulthood were drunkenness-orientated drinking and frequent smoking.
Early interventions for adolescent substance use and a set of specific psychosocial risk factors should be tailored and evaluated as methods for identifying those at high risk of and preventing excessive alcohol use in adulthood.
KeywordsExcessive alcohol use Risk factors Adolescence Adulthood Longitudinal studies
Excessive consumption of alcohol ranks among the world’s major public health problems. In Finland, alcohol-related diseases and poisonings have become the leading cause of death among working-age men and women . Rising alcohol consumption associates with increasing health problems and other negative phenomena, such as risk of accidents and violence. Although excessive alcohol use is recognised as an important public health issue, there is relatively little information on factors associated with the development of excessive drinking through a life course.
In the fields of developmental psychopathology and addiction there is a fundamental assumption that children and adolescence lay a critical foundation for functioning and adjustment in adulthood . Adolescence is a time when many consequential life decisions are made. It is also a time when risky behaviours become temporarily more normative than they are at other times in the life-span . Based on a broad-based developmental contextual perspective on drug use etiology [20, 42, 57], which is consistent with problem behaviour theory , a common set of predictors in childhood and adolescence have been identified across longitudinal studies: social factors (e.g. parental socioeconomic status); family factors (e.g. family relations, parental psychopathology and substance use); peer relations and social competence; academic functioning and future aspirations; externalising and internalising problems; and substance use .
Evidence is mounting regarding genetic and family influences on, and externalizing and internalizing pathways from childhood and adolescence to alcohol and other drug use disorders through early adulthood. However, knowledge remains relatively limited about how the full range of adolescent characteristics and experiences relate to adult alcohol use and abuse, particularly beyond young adulthood .
Studies which have covered the period from adolescence to adulthood have shown that family problems, e.g. parental separation, less harmonious family relationships and support, maltreatment, and paternal substance use and psychopathology [17, 21, 33, 37, 47], school-related problems [37, 39, 47], substance use [17, 25, 38, 63, 68], more social maladjustment, and various forms of externalising problem behaviour [11, 33, 37, 47] are important predictors of alcohol problems in adulthood. Findings on the role of childhood and adolescent internalising and negative affect symptomatology in alcoholism and problem drinking in early to middle adulthood have been less consistent, with some studies demonstrating links between higher internalising symptomatology and greater/lower alcohol involvement [9, 19, 37, 41] and others not [11, 13].
Differences in alcohol use patterns between males and females [40, 43] indicate the importance of examining gender differences in the associations between earlier risk factors and later alcohol outcomes. In some previous longitudinal studies either non-existent or small gender differences in predictive relationships have been observed [e.g. 11, 17, 37], while in others the relationships have varied depending upon type and developmental stage of the risk variable and the outcome variable [13, 33, 49, 51]. Further, in some studies associations between risk factors and drinking outcomes have tended to be somewhat stronger among males than females .
Adolescent development is a holistic and intertwined process in the context of different aspects of adolescent life. In this study, we prospectively examined which socioeconomic, family, personal and lifestyle risk factors in adolescence were the strongest independent predictors of excessive alcohol use in early 30s, when most individuals have reached a stable life stage after major developmental transitions, using longitudinal data over a 16-year period. Given the aim of this study, we included a wide range of risk factors capturing numerous aspects of adolescents’ lives including individual characteristics, family factors, peer involvement, problem behaviour, substance use and psychological well-being, along with social context. There are few comprehensive analyses examining the effects of multiple domains of adolescent risk factors on problem alcohol use in adulthood. The literature and finding from earlier studies suggest that alcohol use patterns differ between the sexes and some risk variables predicting alcohol use may vary by gender. Therefore, separate analyses were carried out for males and females in order to shed light on whether the risk factors of interest differently predict female versus male excessive alcohol use. Understanding the development of alcohol problems in adulthood is important so that appropriate public health interventions can be tailored to identified risk factors in the individual’s development.
The original study population included all ninth-grade pupils (N = 2,269) attending secondary schools in early 1983 in Tampere, an industrial and university city in southern Finland with 166,000 inhabitants. All the pupils shared the same language and ethnic background, and all had 9 years of education. At baseline, in 1983, 2,194 pupils (96.7%) aged 16 years (mean 15.9, SD 0.3) completed a self-administered questionnaire during school hours. In 1999, postal questionnaires were mailed to the same study population when they were 32 years old. Of the original respondents, 22 had died, five were institutionalised, 14 were unidentifiable because of incomplete or missing identification numbers, and the addresses of 62 subjects were unobtainable. The follow-up thus included 2,091 persons (92%), and the response rate was 70.3% (n = 1471, females n = 805 (54.7%), males n = 666 (45.3%). Of those who participated in the follow-up at 32 years of age, 35% worked in manual jobs and 65% worked in non-manual jobs, and 9% were unemployed. About 48% of the adults were married, 21% were single, 5% were separated or divorced, and 58% had children. Participants represented the same age group of the general population in Finland regarding marital status and unemployment rate, as shown by comparison with the data derived from Statistics Finland [61, 62].The study protocol was approved by the Ethics Committee of the National Public Health Institute.
The analysis of attrition showed that male and poor school performance were the most important predictors of non-response. In addition, non-respondent females were more often from divorced families and characterised by problems with the law, low sense of parental trust and high scores for impulsiveness. No differences between respondents and non-respondents were found in parental social class, psychological health, and drinking and smoking habits. The predictive effect of gender, family background and socioeconomic factors, health and health behaviour on non-response and on the estimation of depression prevalence in this longitudinal survey was studied in detail in Eerola et al. . No indication of informative drop-out was found in that analysis, so missing-at random (MAR) assumption was regarded plausible.
Predictor variables in adolescence
Parental social class was based on adolescents’ open-ended description of their father’s occupation, or, if this was missing (in case of absence of a father or missing information), their mother’s. If neither’s occupation was available, the assignment of socioeconomic status was based on the parents’ education . The division of parental social class into ‘upper non-manual’, ‘lower non-manual’, and ‘manual’ was based on the standard classification of occupations . The pupils’ self-reported grades average from the previous school report was used to measure school performance (range 4 (fail)–10 (excellent)).
The measure of depressive symptoms was constructed using seven items indicative of depression (lack of energy; sleeping difficulties; nightmares; fatigue; irritability; loss of appetite; and nervousness/anxiety) on a four-point scale of 17 distress symptoms [24, 46]. Four of the selected items (lack of energy; sleeping difficulties; fatigue; loss of appetite) are regarded as symptoms of clinical depression in DSM-IV. The theoretical range of the scale was 0–21, with good internal reliability (Cronbach’s alpha = 0.73). The self-esteem measure used a five-point scale developed for Finnish students  and consisted of seven statements of self-worth resembling those used in Rosenberg’s measure : I believe in myself and in my possibilities; I wish I were different from what I am; I suffer from feelings of inferiority; I think I have many good qualities; I feel I lack self-confidence; I am capable of doing the same as others; I am often dissatisfied with myself. Impulsiveness was measured with three items (unpredictable; excitable; restless) rated on a five-point scale from a semantic differential scale of self-image developed by Rauste-von Wright  among Finnish adolescents. The theoretical score range of the self-esteem scale (alpha = 0.81) was 7–35 and of the impulsiveness scale (alpha = 0.84) 3–15, higher scores indicating lower self-esteem and greater impulsiveness.
The subject’s relationships with mother and father were measured with three statements scored on a five-point scale (“I feel my mother/father is close to me”; “I often quarrel with my mother/father”; “I often feel my mother/father does not understand me”). The positive item was reverse coded. The theoretical range of the scales was 3–15, higher scores indicating poorer relationship. The relationship with mother scale (Cronbach alpha = 0.70) and with father scale (alpha = 0.61) showed good reliabilities. The scale of parental trust consisted of five self-assertions (alpha 0.71). The items were: “My parents trust me”; “My parents let me make my own decisions”; “I often feel that I disappoint my parent’s expectations”; “I often feel that my parents are disappointed with my academic achievement”; “My parents don’t accept some of my friends”. The positive items were reverse coded. The theoretical range of the scale was 5–25, higher scores indicating lower sense of parental trust.
Aspects of lifestyle were measured by questions on drinking and smoking habits, spending leisure-time daily among friends, dating experiences and problems with the law. Participants were asked to rate how often they used alcohol, and how many times during the school term they had used alcohol to the stage of becoming drunk. A drinking habit indicator was formed using these two indicators. Those reporting no drinking and no drunkenness experience were categorised as ‘non-drinkers’ and those reporting drunkenness at least three times during the school term were deemed ‘drunkenness-orientated drinkers’. The rest were categorised as ‘moderate drinkers’. Based on frequency of smoking, respondents were categorized as non-smokers, smoking less than once a week, and smoking at least once a week. Adolescents who reported breaking the law with consequences during the last 12 months were classified as having problems with the law.
Excessive alcohol use in adulthood
The alcohol use disorders identification test (AUDIT) was used to assess alcohol use at 32 years . The AUDIT is a screening instrument for excessive alcohol use consisting of ten items scored 0–4 points. The items cover three domains: recent alcohol use (hazardous alcohol use), alcohol dependence symptoms, and alcohol-related problems (harmful alcohol use). At the recommended cut-off score of 8, most studies have found very favourable sensitivity and usually lower, but still acceptable, specificity, for current ICD-10 alcohol use disorders and the risk of future harm. Nevertheless, improvements in detection have been achieved in some cases by lowering or raising the cut-off score by one or two points, depending on the population and individual cultures . In Finnish culture, higher cut-off scores for screening purposes have been proposed [29, 64]. In this study, we followed this suggestion and AUDIT scores of 8 or more for females and 10 or more for males were classified as excessive alcohol consumption . Using cut-off scores of 10 or more for males and 8 or more for females have resulted good sensitivity (90%) and clearly greater specificity (94%) than using a cut-off score of 8 or more for both genders for identifying excessive alcohol use .
The associations between adolescent risk variables and excessive alcohol use at 32 years of age were analysed first by comparing means using Student’s t-test and proportions using the Chi-squared test. After that, all the adolescent risk variables were included in the multivariate logistic regression model simultaneously in order to determine the strongest independent predictors of adult excessive alcohol use. In logistic regression analysis, the risk variables with sum scores were classified into three categories using the 25th and 75th percentiles as cut-off points to provide comparability of risk estimates from different variables. A P-value of <0.05 was interpreted as significant and for odds ratios (OR) 95% confidence intervals (95% CI) were computed. The magnitude of multicollinearity among independent variables in the multivariate models was analysed using variance inflation factors (VIF). As all VIF values of independents in the multivariate models were below 2, we considered multicollinearity not to be a problem. The analyses were carried out with SPSS 15.0 software separately for females and males. The selection of variables representing putative risk factors was based on literature, choosing a common set of predictors in childhood and adolescence which have been identified across earlier longitudinal studies.
Different domains of adolescent risk factors by excessive alcohol use at age 32 years among males and females
No excessive alcohol use (n = 444–460)a
Excessive alcohol use (n = 198–205)a
t or X2
No excessive alcohol use (n = 667–677)a
Excessive alcohol use (n = 122–128)a
t or X2
Socioeconomic factors at age 16
Parental social class %
School performance mean (SD)b
Family factors at age 16
Parental divorce %
Relationship with mother mean (SD)c
Relationship with father mean (SD)c
Parental trust mean (SD)c
Personal factors at age 16
Depressive symptoms mean (SD)c
Self-esteem mean (SD)c
Impulsiveness mean (SD)c
Lifestyle at age 16
Spent leisure-time daily among friends %
Drinking habit %
Smoking habit %
Smoking less than once a week
Smoking at least once a week
Problems with the law %
Multivariate logistic regression analyses of the impact of all adolescent risk factors on excessive alcohol use at 32 years among males and females
Males (n = 613)
Females (n = 774)
Parental social class
Relationship with mother
Relationship with father
Spent leisure-time daily among friends
Smoking less than once a week
Smoking at least once a week
Problems with the law
This study prospectively examined which risk factors in adolescence, across socioeconomic, family, personal, and lifestyle domains, were the strongest independent predictors of excessive alcohol use in adulthood till the age of 32 among an urban Finnish community cohort. The results showed that among males there were several domains in adolescence influencing excessive alcohol use in adulthood, while for females the most important domain was substance use.
Consistent with previous longitudinal studies [17, 38, 47, 51] we found that alcohol use, especially drunkenness-orientated drinking, during adolescence is a risk factor for excessive alcohol use in adulthood among both genders. Young peoples’ experimentation with alcohol is often seen as modelling adult behaviour, attempting to achieve peer acceptance, and as an effort to overcome age-typical challenges in psychological growth . However, it is also known that alcohol use, especially to the stage of drunkenness, can be a sign of adolescent developmental problems predicting current and future problems with alcohol [45, 56]. Because alcohol use may interfere with normal development [10, 35] and cause not only short-term risks but also many negative long-term consequences, identifying problematic drinking behaviour during adolescence is vital. Particular attention should focus on early initiators, who are at high risk of heavier consumption and problem drinking in adulthood [48, 65]. In Finland, already at 16 years of age, some adolescents have had extensive experience with alcohol , especially in urban areas [32, 53], though alcohol dependence symptoms are quite rare at age 14 .
Alcohol use is associated frequently with the use of other addictive and psychoactive substances, such as smoking cigarettes . Findings on the predictive role of smoking in adolescence on future substance use and substance-use related problems are not entirely consistent, with some studies showing links between smoking and alcohol use [35, 37, 51], while others not . We found a stronger role of adolescent smoking for excessive alcohol use in adulthood among females when other adolescent risk factors were taken into account, which is consistent with the cross-sectional and longitudinal findings of some other previous studies [27, 34]. Cigarette smoking among girls may be more indicative of psychological problems, particularly in societies and times where male smoking has been more socially acceptable than that of women. Studies of gender differences of the associations of smoking with future risk of alcohol problems are needed in those current societies where smoking is becoming rarer and socially unacceptable.
The results of existing longitudinal studies on the relationship between depression and alcohol use are somewhat conflicting and vary with measurements, sampling frames, age groups, gender, time intervals and adjustments for confounding factors, such as genetic liability and parental psychopathology. We found a significant effect of adolescent depressive symptoms on excessive alcohol use in adulthood among males only, when all other risk factors were taken into account.
It has been suggested that young females may be more sensitive to identifying their depressive symptoms and those feeling depressed may be more readily to turn to other people and health care for help . The non-significant relation between adolescent depressive symptoms and subsequent excessive drinking among females in our study may thus partly be explained by depressed girls more often than boys seeking help from their family, friends, other important persons or professionals [5, 59]. The relationship between depressive symptoms and excessive alcohol use among boys especially may reflect a self-medication process. There is some evidence that boys and males are more likely to drink alcohol to escape from or cope with distress and relieve depression [28, 43]. Boys may engage in activities, including drinking, designed to distract themselves or to alleviate their depression .
There is some evidence that poor coping resources in adolescence may raise the risk of affect-related alcohol use, particularly among subjects relying heavily on their peer groups for support and guidance . In our study, we had no direct information on societal influences of peer group modelling. However, spending leisure-time daily with friends in adolescence, earlier found to associate with high level of substance use and susceptibility to peer pressure , was more frequent among males and significantly predicted risky drinking in adulthood.
We found a strong significant independent effect of parental divorce on subsequent excessive alcohol use among males. Some previous studies on the long-term effects of parental divorce in childhood on adult offspring’s alcohol use and alcohol-related problems have found no gender differences [21, 55], while others have found parental divorce to be a greater risk of alcohol problems for either adult females [18, 36, 52] or for adult males .
On the whole, it has been suggested that changes in family structure and living in a single-parent household may afford children more opportunity to experiment with alcohol (at least partly because of less parental control) leading to increased likelihood of substance use later on . Divorce and other circumstances surrounding parental loss, such as decreased social support, greater exposure to social stress, financial difficulties of single parent families, may lead to poor psychosocial adjustment and thereby contribute to the high rates of alcohol consumption in adults from divorced families [4, 67]. Divorce often involves a lengthy sequence of pre-divorce experiences, such as continued parental discord, the effects of which may be even more important for children’s welfare than the separation as such . Divorce is far more likely to occur among couples with personal, social, and economic problems and to be preceded by troubled family relationships and parenting processes . For example, a history of parental alcoholism increases the probability of parental divorce and distress and the risk of an adolescent developing alcohol problems in later life, given the longstanding and strong evidence of the familial nature of alcoholism [7, 30]. Latendresse et al.  found that associations between parents’ and adolescents’ alcohol-related behaviors were partially mediated by adolescents’ perceptions of the parenting that they received. In particular parental monitoring and discipline had unique mediating capabilities independent of the effects of all other parenting behaviors.
Our study showed no effect of parental social class on adult excessive alcohol use. A recent systematic review of longitudinal studies, mostly based on data from the United States, New Zealand and Scandinavia, found little evidence to support an association between socioeconomic status in early life and later use of alcohol, irrespective of whether the studies were grouped according to length of follow-up, type of socioeconomic measures or geographical location . In a Finnish longitudinal study in a single community , higher problems drinking was related to lower parental socioeconomic status, but high parental socioeconomic status related to higher frequency of drinking in adulthood.
As regards education, poor school achievement in adolescence has been related to higher levels of drinking in adulthood [37, 47], but the reverse can also be true, i.e. early onset drinking can affect future scholastic achievement. We also identified the significance of school achievement for problem alcohol use in adulthood. However, in multivariate analyses, the effects of other risk factors including substance use and psychosocial factors on alcohol use were found to be stronger than school success.
Strengths and limitations of the study
The original study population comprised a total age cohort of 16-year-olds in one city, and the same children were followed up to 32 years of age. The data were collected using self-report questionnaires in a classroom survey at 16 years of age, resulting in a minimal dropout rate at baseline. Though the participation rates in each phase were relatively high, it is possible that through a long follow-up sample attrition caused bias in the results. If so, some significant relationships may not have been detected (e.g. school performance), or those that were detected may underestimate the relationships between variables (e.g. problems with the law and impulsiveness among females). Another limitation relates to the fact that only the end point at 32 years of age analysed not the process in between adolescence and that age.
All data were derived using self-report questionnaires, and are thus prone to the general problems of questionnaire studies, such as the validity of data. Further, the self-report method may carry the risk of underestimating, especially in questions related to alcohol use and problems with law. Although the study included several adolescent assessments, we were unable to consider a number of potential factors, such as age of onset of drinking, genetic predisposition, and societal influences such as paternal alcohol consumption and peer group modelling. Some of our instruments in adolescence were developed for Finnish students (self-esteem, impulsiveness) or for this study (e.g. depressive symptoms), and some scales were relatively crude and included only of three items (tailored to the limited space of the questionnaires). However, the scales showed good internal consistency reliabilities, but their validity was not tested against other, more widely used measures. Despite the limitations, however, our study adds to the existing literature as few prospective longitudinal studies have explored the role of multiple domains of adolescent factors on alcohol problems up to adulthood.
Early interventions for adolescent substance use and a set of specific psychosocial risk factors should be tailored and evaluated as methods for identifying those at high risk of and preventing excessive alcohol use in adulthood.
This study was supported by a grant from the Juho Vainio Foundation. Jaakko Kaprio is supported by the Academy of Finland Centre of Excellence in Complex Disease Genetics.
- 3.Babor TF, Higgins-Biddle JC, Sauders JB, Monteiro MG (2001) AUDIT the alcohol use disorders identification test. Guidelines for use in primary care, 2nd edn. WHO, Department of Mental Health and Substance Dependence, GenevaGoogle Scholar
- 5.Biddle L, Gunnell D, Sharp D, Donovan JL (2004) Factors influencing help seeking in mentally distress young adults: a cross-sectional survey. Br J Gen Pract 53:248–253Google Scholar
- 6.Central Statistical Office of Finland (1975) Classification of occupations. Central Statistical Office of Finland, HelsinkiGoogle Scholar
- 26.Jessor R, Donovan JE, Costa FB (1991) Beyond adolescence: problem behavior and young adult development. Cambridge University Press, New YorkGoogle Scholar
- 27.Kandel DB, Johnson JG, Bird HR, Canino G, Goodman SH, Lahey BB, Regier DA, Schwab-Stone M (1997) Psychiatric disorders associated with substance use among children and adolescents: findings from the methods for the epidemiology of child and adolescent mental disorders (MECA) study. J Abnorm Child Psychol 25:121–132CrossRefPubMedGoogle Scholar
- 29.Kuokkanen M, Teirilä J (2001) Alkoholin suurkuluttajien mini-interventio työterveysasemalla. Kokeilu S-työterveyspalveluissa Helsingissä. Suom Lääkäril 56:3489–3493Google Scholar
- 40.Mäkelä P, Gmel G, Grittner U, Kuendig H, Kuntsche S, Bloomfield K, Room R (2006) Drinking patterns and their gender differences in Europe. Alcohol Alcohol 41(1 Suppl):8–18Google Scholar
- 45.Pape H, Hammer T (1996) How does young people’s alcohol consumption change during the transition to early adulthood? A longitudinal study of changes at aggregate and individual level. Addiction 9:1345–1358Google Scholar
- 50.Rauste-von Wright M (1975) The image of man among Finnish girls and boys. Report No. 41. University of Turku, Department of Psychology, TurkuGoogle Scholar
- 53.Rose RJ, Kaprio J, Winter T, Koskenvuo M, Viken RJ (1999) Familial and socioregional environmental effects on abstinence from alcohol at age sixteen. J Stud Alcohol 13(Suppl):63–74Google Scholar
- 54.Rosenberg M (1965) Society and the adolescent self-image. Princeton University Press, PrincetonGoogle Scholar
- 57.Schulenberg JE, Maggs JL (2002) A developmental perspective on alcohol use and heavy drinking during adolescence and the transition to young adulthood. J Stud Alcohol 14(Suppl):54–70Google Scholar
- 60.Statistics Finland (2006) Cause of death Statistics. Statistics Finland, HelsinkiGoogle Scholar
- 61.Statistics Finland. Labour force survey. http://tilastokeskus.fi/til/tyti. Accessed 10 March 2009
- 62.Statistics Finland. Population structure. http://tilastokeskus.fi/til/vaerak. Accessed 10 March 2009
- 64.Vuorilehto M, Isometsä E, Ihalainen R, Korhonen A (2003) Terveyskeskuspotilaiden masennus, ahdistuneisuus ja alkoholin riskikäyttö [Mental symptoms among unselected users of primary health care, in Finnish with English abstract]. Suom Lääkäril 58:905–908Google Scholar