The link between school connectedness and alcohol-‐related hospitalization in adulthood
A cohort study
Centre for Health Equity Studies
Master thesis in Public Health (30 credits) Spring 2014
Name: Malin Fransson
Supervisor: Ylva B Almquist
While there is a vast amount of research focusing on the importance of school connectedness for short-term outcomes related to alcohol use, few studies have looked at the longer-term consequences of poor school connectedness. The aim of the current study was therefore to investigate the gender-specific association between school connectedness (1966) and alcohol- related hospitalization in adulthood (1973-2007). Moreover, the role of adjustment problems in adolescence was examined. Logistic regression analysis was based on a cohort of Swedish children, born in 1953 in Stockholm, Sweden, including 6,269 males and 6,106 females. The results revealed a statistically significant relationship between having a lower level of school connectedness in childhood and an increased risk of hospitalization due to alcohol misuse in adulthood, among males. The findings for females were less conclusive. Controlling for socioeconomic background and adjustment problems in adolescence reduced the strength of the association but did not explain it. In sum, this study shows that school connectedness appear to have long-term consequences for alcohol-related diseases, but further research is needed to understand the mechanisms behind this finding.
Key words: School connectedness, adolescents, adjustment problems, alcohol misuse, hospitalizations, gender differences.
Table of contents
1. INTRODUCTION ...1
2. THEORETICAL BACKGROUND...2
2.1 SCHOOL CONNECTEDNESS...3
2.2 ALCOHOL MISUSE...5
2.3 ADJUSTMENT PROBLEMS IN ADOLESCENCE...5
2.4 LIFE-‐COURSE PERSPECTIVE...6
2.5 GENDER DIFFERENCES...8
2.6 POSSIBLE CONFOUNDING FACTORS...8
3. AIM AND RESEARCH QUESTIONS ...9
4. STUDY DESIGN...9
5. DATA AND METHODS... 10
5.1 DATA MATERIAL...10
5.3 ETHICAL ASPECTS...17
6. RESULTS ... 17
6.1 BIVARIATE ASSOCIATIONS...17
7. DISCUSSION ... 19
7.1 STRENGTHS AND LIMITATIONS OF THE STUDY...21
8. POLICY IMPLICATIONS ... 23
ACKNOWLEDGMENTS ... 24
1 1. Introduction
Alcohol consumption is one of the largest risk factors for diseases and disabilities in the world. Alcohol misuse is associated with both chronic diseases like cancers, cardiovascular diseases and neuropsychiatric disorders, and with an increased risk of acute health conditions, such as injuries and traffic accidents (WHO, 2011). In Sweden, the rates of alcohol consumption have increased among both males and females over the last years, and the highest rate is found among young males (Statistics Sweden, 2007). High consumption of alcohol is a major public health problem; the number of alcohol-related deaths in Sweden is about 5,000 per year and to a greater extent affect younger and middle-aged people. In terms of costs for the health care system, hospitalization due to alcohol misuse constitutes between 5 and 7 percent of all the days that patients spend at hospitals (Andréasson, 2002).
Patterns of behaviors and lifestyle choices are established at a young age and affect both the current and the future health. Since young people spend most of their awakening time at school, the school environment can be seen as a central context for the development of health behaviors. The school also constitutes one of the most important environments for implementing universal prevention programs aimed at young people since it generates some of the most important protective factors against alcohol and drug use, such as literacy, intellectual development, social interaction, norms and skills. In addition, schools are easy to mobilize to other sectors and representatives of society as parents, community and cultural groups. Therefore, schools are a highly suitable context for the implementation of alcohol- related health programs (Bohrn, et al. 2008).
The extent to which the children are connected to school – i.e. their level of school connectedness – may constitute a particularly important aspect in terms of creating and shaping behavioral patterns. Engagement in school can work as a buffer against psychological problems. Conversely can lack of engagement in school both be a cause and a consequence of behavioral problems (Steinberg, Brown & Dornbusch, 1996). According to the research done by Bond et al. (2007), there is an association between having a high level of school connectedness at 13-14 years of age, and having the best outcomes 2-4 years later. In contrast, poor school connectedness is associated with an increased risk of anxiety and/or depressive symptoms, regular smoking, drinking and using marijuana in later years. The association
between school connectedness and alcohol use has also been confirmed by Hawkins et al.
Although there have been some studies of school connectedness in relation to health behaviors in childhood and adolescence (Eccles, Early, Fraser, Belansky & McCarthy, 1997;
Resnick, Bearman & Blum, 1997), few have examined how school connectedness may be linked to adult alcohol misuse. Here, a life-course perspective offers a relevant framework for examining these questions. The life-course perspective also provides insight with regard to meditating factors like adjustment problems in adolescence, as well as potentially important background factors that may confound the association between school connectedness and alcohol misuse in adulthood. In other words, studying these relationships within a life-course framework may add more information about some of the mechanisms originating in childhood that may contribute to subsequent alcohol misuse and, hopefully, contribute to the knowledge needed to prevent future poor health outcomes.
2. Theoretical background
School, friends, and activities outside the family have been recognized as increasing in importance as children grow older and mature. Resnick, et al. (1997) found that children and adolescents who were feeling connected to school were less likely to use substances like alcohol and drugs at an early age, were less involved in criminality and gang membership, and were less likely to drop out of school. Eccles, et al. (1997) expressed in their study that high school connectedness was associated with higher levels of emotional well-being.
Unfortunately, the attachment to school seems to deteriorate as children grow older (McNeely, Nonnemaker, & Blum, 2002), something which in turn implies that investigation and resources to the school are important at an early age. It has been shown that school connectedness is higher in smaller schools and school classes as well as in schools with a positive classroom management climate and tolerant disciplinary. McNeely, et al. (2002) also found that those who were participating in extracurricular activities had higher school connectedness than those who did not. Overall, these studies show that school connectedness is a beneficial offset for most dangerous behavior, but also a factor for positive emotional health outcomes.
3 2.1 School connectedness
In a literature review made by Libbey (2004) it was made clear that there is variety of definitions of school connectedness used in different studies: commonly used terms were school engagement, school bonding, school attachment, school involvement, school climate and teacher support. The definition of school connectedness is therefore rather unclear, but Libbey (2004) has identified some consistent themes; "A student's sense of belonging and being a part of school, whether or not students like school, level of teacher supportiveness and caring, presence of good friends in school, engagement in current and future academic progress, fair and effective discipline, and participation in extracurricular activities can be traced across several measures of school connection." (Libbey, 2004, p.281). In general, the concept of school connectedness has been discussed from three main theories/models:
attachment theory, control theory, and the social development model (Catalano, Haggerty, Oesterle, Fleming & Hawkins, 2004). These will be discussed further below.
2.1.1 Attachment theory
The foundation of attachment theory is described as an interaction between parents and their infant. This interaction is establishing at an early stage and is important for social connections with others later in life. The bonding between parents and the child is the key to developing motivated behaviors in early life and good attachment to the parents may have a positive effect later on in life. A broader approach to attachment theory is that attachment to other adults than the parents, like teachers or other adults in school, has positive effects on the child’s resilience to adversity. The bonding has been shown to increase positive developmental experiences and buffer the risk of adverse behaviors for the child. School bonding appears to promote healthy development and to prevent problem behaviors both in current and in future contexts (Catalano, et al. 2004).
2.1.2 Control theory
Hirschi’s social bond/social control theory constitutes a paradigm in criminological research into delinquency. Hirschi (1969) found that the bonds that people have in relation to prosocial values, prosocial people, and prosocial institutions, would end up controlling behaviors when people were tempted to engage in criminal or deviant acts. These bonds come in four different forms: attachment, commitment, involvement and beliefs. Attachment refers to relationships to family, friends, and other affiliated persons, as well as to the school. Commitment refers to
the importance of the social relationships that people value. Involvement relates to the opportunity costs associated with how people spend their time, for example being involved in legitimate school-related activities – academically, socially, or athletically. Beliefs refer to the degree to which one follows the values associated with behaviors that conform to the law.
Together, these factors are of critical importance for the level of social control (Pratt, Gau, &
Franklin, 2011). In Hirschi’s social bond/social control theory, school connectedness is described with the two components attachment, characterized by close affective relationships with those at school, and commitment, characterized by an investment in school and doing well in school. Once the social bond/social control has been firmly established, the social bond exerts an informal control on behavior, inhibiting deviant behavior in particular (Catalano, et al. 2004).
2.1.3 The social development model
Similar to control theory, the social development model also has a concept of bonding as composed of attachment and commitment, but this rather reflects being attached and committed to a socializing unit. Involvement is seen as part of a socialization process that leads to bonding, while the students’ beliefs in the values of the social units are seen as a consequence of bonding. The social development model hypothesizes that children must learn behavioral patterns, both prosocial and antisocial, from their social environment. The social environment can include both family and friends, but also the social environment of the school.
Children are being socialized through four processes: 1) Perceived opportunities for involvement in activities and interactions with others, 2) Actual involvement, 3) Skill for involvement and interaction, and 4) Perceived rewards from involvement and interaction.
Once the social bond is strongly established the behaviors is consistent with the norms, values, and behaviors of the social unit (Catalano, et al. 2004). The social development model hypothesizes that the interaction of different factors during the social development influences the degree to which students develop strong social bonds to school. The factors that affect students’ bonding to school are the degree of opportunity for active involvement available in the classroom, the skills possessed and applied by students during participation in these social groups, and the reinforcements provided to students in response to their behaviors in these groups (Hawkins, Catalano, Kosterman, Abbott & Hill, 1999).
5 2.2 Alcohol misuse
Alcohol misuse reflects excessive drinking (more than the recommended limits of alcohol consumption). Increased-risk drinking or ‘hazardous’ drinking is a regularly intake of 22-50 units1 of alcohol a week (adult men), or 15-35 units a week (adult women). Higher-risk drinking or ‘harmful’ drinking is regularly drinking over 50 units of alcohol a week (adult men) or over 35 units a week (adult women). Dependent drinking is drinking over the recommended line and feeling unable to function without alcohol. Drinking excessively can lead to a number of harmful physical and psychological effects, such as cirrhosis of the liver, alcohol poisoning, cancer, inability to work, and destructive behaviors, such as drunk driving and violent behaviors (NHS Choices, 2013). Alcohol consumption has increased among both males and females in Sweden, and high alcohol consumption is more common among males than among females. For males, high alcohol consumption, or at one time drink a large amount of alcohol (binge drinking) is most common among younger people, singles, those living in urban areas, and among less educated. Among females the pattern is different: highly educated females, females in white-collar occupations (middle or higher), and females in metropolitan areas appear to have the highest alcohol consumption whereas among females in manual occupations it is relatively uncommon with high alcohol consumption behavior (Statistics Sweden, 2007).
2.3 Adjustment problems in adolescence
Adjustment problems in adolescence, such as early alcohol debut, behavior problems, criminal acts, and other forms of deviant behaviors, could constitute important mechanisms in the link between school connectedness and alcohol misuse in adulthood. Hawkins et al.
(1997) found that alcohol initiation in younger age was strongly related to a higher level of alcohol misuse in late adolescent and forward into adulthood. Early and persistent behavioral problems, including aggressive behaviors among boys and other conduct problems, are risk factors for alcohol and drug abuse in adolescents (Hawkins, Catalano & Miller, 1992).
Past research has pointed to the importance of parental and school factors for avoiding adjustment problems in adolescence. For example, Lamborn, Mounts, Steinberg and
1 A unit of alcohol is 10ml of pure alcohol, which is about half a pint of normal strength lager or a single
measure (25ml) of spirits. A small glass (125ml) of wine contains about one-and-a-half units of alcohol.
Dombusch (1991) found a relationship between parenting style and adolescent adjustment that indicated a clear advantage for adolescents raised in authoritative homes, and a clear disadvantage for those adolescents raised in neglectful homes. Another study found that students who skipped school experienced more criminal behaviors than those students who were not skipping school; many of them also used alcohol and drugs, were more committed to violations at school, more often had incomplete grades, and were socializing more frequently with antisocial peers (Karlberg & Sundell, 2004). Stafström, Östergren and Larsson (2005) found that psychological, psychosocial and socioeconomic elements were independent risk factors for frequent high alcohol consumption among adolescents in Sweden. Adolescents who had been physically and/or sexually assaulted, who had witnessed violence, or who had family members with alcohol or drug use problems had an increased risk for current substance abuse (Kilpatrick, et al. 2000). The two strongest risk factors for alcohol and drug abuse in adolescents are laws and norms favorable to alcohol and other drugs and social influences to drug use (Hawkins, et al. 1992). Lack of connectedness to school and the family can lead to later alcohol abuse because adults are overlooking the behavioral indicators that students send out (Andersen, Holstein & Due, 2006). When overlooking these indicators, like adjustment problems, the ability to change the drinking patterns from adolescence to adulthood is more difficult. In sum, these studies strengthen the notion of adjustment problems in adolescence as a mediating factor between school connectedness and alcohol related problems in adulthood.
2.4 Life-course perspective
Different events, both good and bad, can be seen as milestones in people’s life. What is happening between birth and death can be looked at from a life-course perspective, including events that will affect (physical and psychological) health or the risk of diseases, both directly and in the future. The life-course perspective is analyzing people’s lives within different structural, social, and cultural contexts. Different kinds of exposures are targeted within this paradigm, present at various stages of life, such as gestation, childhood, adolescence, young adulthood, adult life, and older age. Various biological and social factors that have been experienced at earlier life-course stages can independently, cumulatively, and interactively influence health and disease in adult life (Kuh, Ben-Shlomo, Lynch, Hallqvist & Power, 2003).
Two theoretical models that are commonly used within life-course research are the accumulation of risk model and the chain of risk model (see Figure 1). The accumulation of
risk model explains the health outcome by exposures that gradually accumulate through episodes of illness and injury, poor environment, and health behaviors. The accumulation can consist of different types of exposures such as environmental, socio-economic, and behavioral, and may cause long-term damage with either independent exposures (Model A) or clustered exposures (Model B). Model B is known as an accumulation model with risk clustering (Kuh, et al. 2003). The chain of risk model refers to several exposures linked together that raise the risk of diseases and poor health outcomes. The linked exposures tend to lead to another and then forward to and other. Different types of chains can cause increased or decreased risk. Social, biological, and psychological chains of risk may include ‘‘mediating factors’’ and often also ‘‘modifying factors’’. It is common to talk about two different types of chains of risk. The first model suggests that each exposure not only increases the risk of the subsequent exposure but also has an independent effect on disease risk irrespective of the later exposure (Model C). The second model states that earlier exposures have no effect on poor health outcomes without the final link in the chain that precipitates disease onset (Model D) (Kuh, et al. 2003). In the current study the combination of Model B and Model C may be seen as the most applicable.
FIGURE 1. Life-‐course causal models. Adapted from Kuh, et al. (2003)
8 2.5 Gender differences
Females drink less alcohol and have fewer alcohol-related problems than males. A Taiwanese study by Yeh (2006) with self-reported data, showed that alcohol consumption and problem drinking was 2.22–2.71-fold greater in male than in female adolescents. Alcohol is the leading risk factor for death in males ages 15–59, mainly due to injuries, violence, and cardiovascular diseases. Women’s lower risk, compared to men, is partly due to the fact that women have much higher rates of abstinence (WHO, 2011). Other explanations are that females perceive greater social sanctions for drinking, and are less likely to have characteristics associated with excessive drinking, including aggressiveness, drinking to reduce distress, behavioral under- control and anti-social behaviors (Nolen-Hoeksema, 2004). Hawkins, et al., (1997) found similar gender differences in alcohol misuse among adolescents. Other factors like family management practices, parental drinking, bonding to school, and best friends’ drinking were all associated with alcohol use in adolescence within both genders.
School connectedness also seems to capture partly different aspects for males and females.
Previous studies have shown that indicators of poor school connectedness among females included feelings of their parents not engaging in their schoolwork, low satisfaction with the school, and poor affiliation to the school. Among males, poor school connectedness reflected not feeling safe at school, feelings of their parents not being ready to help with school problems, and feelings of being treated unfairly at school (Andersen, et al. 2006). Moreover, studies have demonstrated that female students feel less connected to school than male students (Bonny, et al., 2000; McNeely, et al., 2002). Given the gender differences both in terms of school connectedness and alcohol misuse, it is highly relevant to examine the role of gender also in the current study.
2.6 Possible confounding factors
Population health can be seen as made up by a range of social, material and psychosocial factors like work, unemployment, early life, addiction, food, transport, stress, social exclusion, social support, and the social gradient. Poor social and economic circumstances affect health throughout the life course. People further down the social ladder usually have twice the risk of illness and premature death compared to those further up in the ladder. Good health involves improving housing standards, reducing insecurity, reducing unemployment, and reducing levels of educational failure (Wilkinson & Marmot, 2003).
Alcohol-related problems often affect people in low socioeconomic groups due to cumulative disadvantages, which can be explained by socioeconomic disadvantages occurring early in life that multiply over the life course, contributing to adverse health outcomes. Education, occupational class and income are some factors that have showed statistically significant results as predictors of alcohol-related mortality after adjusting for other socioeconomic factors (WHO, 2010). In turn, parental alcohol use may affect the relation children have to alcohol. A study by Dube, et al. (2001) found that persons who grew up with parental alcohol abuse had approximately 2 to 13 times higher risk of adverse childhood experiences. Another study found that the risk of developing adolescent problem drinking was fourfold among students whose fathers had habits of drinking (Yeh, 2006). Parental support, specific parental guidelines for adolescent behaviors, and parental attitudes toward drinking have been found to be significant predictors of alcohol abuse in adolescents (Barnes & Windle, 1987). Parents’
drinking patterns are a determinant influencing alcohol consumption among their children. A Danish study, made by data from a cohort born in 1953, found that low socioeconomic status in childhood has been shown to be associated with increased risk of alcohol or drug dependence in adulthood (Osler, Nordentoft & Nybo Andersen, 2006). In sum, it is thus reasonable to assume that parental socioeconomic status may confound the association between poor school connectedness and alcohol-related hospitalization.
3. Aim and research questions
The overall aim of this thesis was to analyze the association between school connectedness in childhood and alcohol misuse in adulthood. The specific research questions were:
1. Is school connectedness in childhood associated with alcohol misuse in adulthood, in terms of being hospitalized in 1973-2007 due to alcohol misuse?
2. If so, is this association explained by adjustment problems in adolescence?
3. Are there any differences between the genders?
4. Study design
The study has a longitudinal design with data from the Stockholm Birth Cohort (SBC). The life course perspective in this study starts with the family background of the students, which includes the children’s socioeconomic conditions and parental alcohol abuse, followed by
school connectedness during middle childhood, adjustment problems in adolescence, and ends with alcohol misuse in adulthood (see Figure 2).
FIGURE 2. Study design
5. Data and methods
5.1 Data material
This thesis is based on data from the Stockholm Birth Cohort study, which was created by a probability matching of two anonymous datasets. The first part is the Project Metropolitan that covers the period between 1953 and 1985, and consists of all children born in 1953 who were living in the Stockholm metropolitan area in 1963 (n=15,117). The second part is the Swedish Work and Mortality Database which was matched to the Stockholm Metropolitan Study in 2004/2005. The probability matching rendered a sample of 14,294 individuals (7,305 males and 6,989 females). Below, the variables used in the present thesis are described.
5.2.1 Main independent variable
Information about school connectedness was derived from the “School Study” of 1966. The School Study was a part of Project Metropolitan, and was based on a questionnaire that contained a cognitive test, orientation of interests, attitudes towards school, and questions on leisure and educational plans. In May 1966 the students in the sixth grade of the schools in the Stockholm Metropolitan area filled out the questionnaires.
The current study created a measure of school connectedness largely based on Hirschi’s social bond/social control theory. An index of school connectedness was created from five separate questions about school. The first variable “The pupils’ feeling of security at school” originally consisted of ten items with yes or no answers. The no answers reflected high security whereas yes answers reflecting low security at school. In the current thesis, this variable was dichotomized so that 0-2 no answers were coded as low school connectedness (1) and 3-10 no answers were coded as high school connectedness (0). The second variable “Interest in school work” also consisted of ten items with yes or no answers. This category was dichotomized in the same way as for the previous variable. The third variable was based on the question “If school were completely voluntary and you could quit tomorrow or stay if you wanted to, what would you do if you could decide yourself?” and had four response alternatives: leave school at once; leave school after this grade; leave after elementary school; and leave later on. The first two response alternatives were coded as low school connectedness (1) and the two latter as high school connectedness (0). The fourth variable was based on the question “Have you ever been absent from school even though you were not ill?” and had three response alternatives: no; yes, one or twice; and yes, several times. The first two responses were coded as high school connectedness (0) whereas the third response alternative was coded as low school connectedness (1). The fifth variable was based on the question “Have you ever been told to leave the classroom during a lesson because of something you did?” It had the same response alternatives as the previous question and was recoded in the same way. The distribution of all five variables can be seen in Table A-C in the Appendix.
Based on the five dichotomous variables reflecting school connectedness, an index was created. The index ranged from 0 to 5, where higher scores represented lower school connectedness. Because of the skewed distribution of the index, it was collapsed into four categories: 0 problems; 1 problem; 2 problems and 3-5 problems. A ‘problem’ reflects that the individual has low school connectedness for any of the five variables.
5.2.2 Dependent variable
The dependent variable of this study was alcohol misuse. Information about alcohol misuse in adulthood was derived from The Swedish Hospital Discharge Register 1973-2007. Individuals who had been admitted to the hospital and underlying diagnosis with “Mental and behavioral disorders due to use of alcohol” (code F10 in ICD 10 as well as corresponding diagnose codes
for ICD 8 and ICD 9) at any point during the period of 1973-2007 were coded into one category, whereas the remaining individuals were coded into another category.
5.2.3 Mediating variables
Information about adjustment problems in adolescence was derived from the Social Register.
The mediating variables were based on the criminal acts and other forms of deviant behaviors being the causes of the Child Welfare Committee’s (CWC) actions during each year between 1967 and 1972. Individuals could have been registered one or several times for each action taken.
Four indicators of adjustment problems in adolescence were included in the present study:
drunk driving; drunkenness; use of alcohol; and adaption problems. Each indicator was dichotomized. For the variable drunk driving, 0 represented not driving drunk, and 1 represented driving drunk one or more times during the period of 1967 -1972. For the variable drunkenness, 0 represented not being drunk, and 1 represented being drunk one or more times during the period of 1967 -1972. For the variable use of alcohol, 0 represented not using alcohol and 1 represented using alcohol one or more times during the period of 1967 -1972.
Finally, for the variable adaption problems, 0 represented not having adaption problems, and 1 represented having adaption problems one or more times during the period of 1967 -1972.
5.2.4 Confounding variables
Three variables reflecting the socio-economic background of the individuals were included as possible confounding factors: parental social class, parental education, and parental income.
Parental social class (1963) was categorized into: upper and upper middle class; officials;
entrepreneurs; skilled workers; unskilled workers and other. Parental education (1960) was specified by the number of persons in the household with at least a high school diploma, and was categorized into: at least one graduate and no graduates/no information. Parental income (1963) was based on the combination of two variables: the father’s income and the mother’s income. These two variables were combined divided by two to get the parental average income. Parental income was subsequently categorized into: at or above the mean; below mean and no registered income. An additional control variable reflecting parental alcohol use (1953-1965) was also included, for which the information was derived from the Social Register. It was divided into two categories: no (not having a parent with alcohol misuse) and
13 yes (at least one of the parents having alcohol misuse).
5.2.5 Characteristics of the analytic sample
Table 1 presents the gender-specific prevalence of alcohol misuse between the years 1973- 2007 as well as the distribution of school connectedness, adjustment problems in adolescence, and the confounding variables.
Among males, 4.7 percent experienced one or more alcohol-related hospitalizations during the years of 1973-2007. Among females, 2.1 percent had been hospitalized due to alcohol misuse during the same time period. The parental socioeconomic background variables were more or less in the same level for both of the genders. A total of 41.5 percent of the males and 31.7 percent of the females had one or more problems related to school connectedness. About 2.2 percent of the males and only 0.1 percent of the females have been driving drunk. A similar connection between genders can been seen with drunkenness with 4.8 percent of the males and 1.1 percent of the females. Among the males, 1.4 percent had been taken action against by the Child Welfare Committee because of alcohol use whereas the corresponding percentage among females was 0.5. Moreover, 3.9 percent of the males and 3.3 percent of the females had adjustment problems in adolescence (see Table 1).
TABLE 1. Distribution (%) of all study variables (n=12,375)
Males (n=6,269) Females (n=6,106)
Hospitalization No 95.3 97.9
Yes 4.7 2.1
Parental social class Upper and upper middle 14.4 14.2
Officials 30.0 30.9
Entrepreneurs 6.4 7.5
Skilled workers 19.7 20.1
Unskilled workers 14.0 14.1
Others 15.5 13.1
Parental education At least one graduate 76.8 76.4
No graduate/ no
information 23.2 23.6
Parental income At or above mean 29.9 31.9
Below mean 47.3 49.4
No registered income 22.8 18.7
Parental alcohol misuse No 96.8 96.8
Yes 3.2 3.2
School connectedness 0 problems 58.6 68.5
1 problem 25.9 21.1
2 problems 10.3 7.6
3-‐5 problems 5.3 2.9
Drunk driving No 97.8 99.9
Yes 2.2 0.1
Drunkenness No 95.2 98.9
Yes 4.8 1.1
Use of alcohol No 98.6 99.5
Yes 1.4 0.5
Adaption problems No 96.1 96.7
Yes 3.9 3.3
5.2.6 Statistical analysis
Of the SBC cohort (n=14,294) a total of 12,821 individuals had participated in The School Study. The external attrition was primarily due to emigration out of the Stockholm Metropolitan area as well as absence on the day of the data collection. Of these 12,821 individuals, 12,375 had full information on the variables reflecting school connectedness (6,269 males and 6,106 females). None of the other variables contributed with any atttition (see Figure 3). The analytical sample was slightly possitively selected in terms of e.g.
FIGURE 3. Sources of attrition
Figure 4 illustrates the tested relationships in the current study. The big arrow illustrates the main association investigated: the relationship between school connectedness in childhood and alcohol misuse in adulthood. The small arrows illustrate the confounding and mediating factors that have been included in the regression analysis. Parental class, parental education, parental income, and parental alcohol misuse were controlled for as confounders in the statistical analyses. Adjustment problems in adolescents were included for as mediating factors, and consisted of drunk driving, drunkenness, use of alcohol and adaptation problems.
FIGURE 4. Illustration of the variables of the study and their tested relationships
Logistic regression was used to assess the relationship between school connectedness and alcohol misuse in adulthood. The results were presented as odds ratios (OR) with 95%
confidence intervals (CI). In Model 1, the bivariate associations between the independent variables and the dependent variable are examined, separately for males and females. Since the association between school connectedness and alcohol misuse was only statistically significant for males, the proceeding models were not estimated for females. Model 2 is adjusted for socioeconomic background (parental social class, parental education, parental income, and parental alcohol misuse), while Model 3 is also taking into account adjustment problems in adolescence.
In order to ensure that the adjusted odds ratios for school connectedness are not biased due to residual variation, the analyses were repeated by means of linear regression; a strategy which has been recommended by Mood (2010). The results from these analyses do not suggest that the comparisons of adjusted odds ratios are invalid (data not presented). Model diagnostics were also performed by means of Hosmer-Lemeshow likelihood ratio tests and showed satisfactory results.
17 5.3 Ethical aspects
The data of the participants are de-identified, and researchers only have access to small segments of the database. The data are stored in encrypted form, and the master copy is locked up in a safe. The Swedish Research Council ethical committee has approved the creation of the Stockholm Birth Cohort study, and the Regional Ethics Committee in Stockholm has approved the research (dnr 03-629).
6.1 Bivariate associations
The results of the gender-stratified logistic regression analysis for the unadjusted association between school connectedness and alcohol misuse are shown in Table 2. For example, compared to males without any problems related to school connectedness, the risk of alcohol misuse in adulthood is higher among males with one problem (OR 1.59; CI 1.21-2.09) and those with two problems (OR 1.84; CI 2.63-4.76). Males with three to five problems have the highest risk (OR 3.22; CI 2.71-4.76) of alcohol misuse out of all four groups. Among females, the results are less conclusive. Although the estimates are in the expected direction, there is no statistically significant association between school connectedness and alcohol misuse for females.
TABLE 2. School connectedness (1966) and alcohol misuse in adulthood (1973-‐2007). Results from logistic regression, for males and females separately (n=12,375)
Since the association between school connectedness and alcohol misuse was not statistically significant among females, the decision was to show the subsequent models only for males (Table 3). In the male sample, when including indicators of socioeconomic background in the model (Model 2), the association between school connectedness and hospitalization due to alcohol misuse is reduced in strength but, remains statistically significant. Thus, parental socioeconomic background only explains a part of the association between school connectedness and alcohol misuse among males. When also including adjustment problems in
OR (95% CI) Females (n=6,106) OR (95% CI)
0. (ref.) 1.00 1.00
1 problem 1.59 (1.21-‐2.09) 1.34 (0.89-‐2.04) 2 problems 1.84 (2.63-‐4.76) 1.52 (0.84-‐2.76) 3-‐5 problems 3.22 (2.71-‐4.76) 1.57 (0.63-‐3.92)
adolescence (Model 3) the odds ratios of alcohol misuse decrease in all categories of school connectedness. For example, among males with three to five problems the odds ratio decreases from 2.60 (CI 1.74-3.89) to 2.16 (CI 1.43-3.26) compared to the previous model.
Moreover, the estimate for those with two problems is no longer statistically significant.
TABLE 3. School connectedness (1966) and alcohol misuse in adulthood for males (1973-‐2007) by the different models (n=6,269).
OR (95% CI) Model 2
OR (95% CI) Model 3 OR (95% CI)
0 problems (ref.) 1.00 1.00 1.00
1 problem 1.59 (1.21-‐2.10) 1.51 (1.15-‐2.00) 1.39 (1.05-‐1.84) 2 problems 1.84 (1.28-‐2.63) 1.70 (1.18-‐2.44) 1.40 (0.96-‐2.04) 3-‐5 problems 3.22 (2.18-‐4.76) 2.60 (1.74-‐3.89) 2.16 (1.43-‐3.26)
Parental social class
Upper and upper middle (ref.) 1.00 1.00 1.00
Officials 2.77 (1.53-‐5.02) 1.90 (1.00-‐3.61) 1.88 (0.99-‐3.59) Entrepreneurs 2.82 (1.35-‐5.93) 1.54 (0.68-‐3.46) 1.52 (0.68-‐3.44) Skilled workers 3.50 (1.91-‐6.42) 1.92 (0.96-‐3.86) 1.74 (0.86-‐3.52) Unskilled workers 3.43 (1.83-‐6.44) 1.76 (0.85-‐3.61) 1.65 (0.80-‐3.41) Other 7.26 (4.03-‐13.07) 3.01 (1.43-‐6.35) 2.88 (1.36-‐6.10)
At least one graduate (ref.) 1.00 1.00 1.00
No graduate/ no information 2.89 (1.96-‐4.26) 1.47 (0.91-‐2.35) 1.43 (0.89-‐2.29)
At or above the mean (ref.) 1.00 1.00 1.00
Below mean 2.13 (1.50-‐3.01) 1.43 (0.97-‐2.10) 1.39 (0.95-‐2.06) No registered income 3.65 (2.55-‐5.22) 1.79 (1.09-‐2.94) 1.67 (1.01-‐2.77)
Parental alcohol misuse
No (ref.) 1.00
Yes 2.60 (1.65-‐4.12) 1.83 (1.14-‐2.95) 1.54 (0.94-‐2.52)
No (ref.) 1.00 1.00
Yes 3.59 (2.20-‐5.84) 1.61 (0.93-‐2.78)
No (ref.) 1.00 1.00
Yes 4.02 (2.86-‐5.66) 2.31 (1.58-‐3.37)
Use of alcohol
No (ref.) 1.00 1.00
Yes 9.45 (5.87-‐15.22) 4.22 (2.45-‐7.28)
No (ref.) 1.00 1.00
Yes 3.77 (2.59-‐5.48) 1.60 (1.04-‐2.47)
Model 1: Unadjusted
Model 2: Adjusted for socioeconomic status
Model 3: Adjusted for socioeconomic status and adjustment problems
19 7. Discussion
The findings presented in this study suggest that there is an association between school connectedness and alcohol-related hospitalization in males. This association decreased in strength when controlling for socioeconomic background, and was even further reduced when adjustment problems in adolescence were included in the analysis. In other words, adjustment problems in adolescence can be seen as a mediating factor but it does not explain the entire association. The results can be interpreted with a combination of two life-course models; the chain of risk model and the accumulation of risk model; where several exposures linked tougher in sequences raise the risk of poor health outcomes and are, more or less, clustered by a common risk factor. In the current study a low level of school connectedness in childhood can be seen as the central exposure, which is preceded by socioeconomic background and followed by the mediating factor adjustment problems in adolescence. This study shows that low school connectedness both can be a risk factor by itself but also a risk factor clustered and accumulated together with adjustment problems and social background.
To the author’s knowledge there are no other studies focusing on the relationship between school connectedness in childhood and alcohol misuse in adulthood. The findings from previous studies have only looked at this connection in a short-term perspective. For example did Catalano, et al. (2004), McNeely, et al. (2002), and Resnick, et al. (1997) find that students who were feeling connected to school were less likely to use substances such as alcohol. Andersen, et al. (2006) also found that lack of connectedness to school and family, can lead to later alcohol abuse. Early behavioral problems have shown to be a risk factor for alcohol abuse in adolescence, especially among boys (Hawkins, et al. 1992). Another study made by Hawkins, et al. (1997) examined behavioral problems in adolescence, found a connection between low commitment to school and alcohol and drug use in adolescence. This connection can be seen as a link in an accumulative process within a longer life-course perspective. In relation to current findings, alcohol use and adjustment problems in adolescence can be seen as the mediating factor that enhances the connection between poor school connectedness and hospitalization due to alcohol misuse. Hawkins, et al. (1997) also found that alcohol use in younger ages is strongly related to a higher level of alcohol misuse in late adolescent and forward to adulthood, which can be seen as an additional link to the life-course perspective and will strengthens the findings of this study.
Previous studies have found that parental alcohol use may affect children’s alcohol use (Barnes et al., 1987; Dube et al., 2001; Yeh, 2006). The current study have found that parental alcohol misuse (and other parental socioeconomic factors) explains some of the association between school connectedness and hospitalization due to alcohol misuse - but parental alcohol misuse could not explain the association by itself. More general types of parental alcohol behavior were not accounted for and could have an underlying influence of the association.
For example, it is a large difference between growing up in a home with parents who are not drinking and parents who are having a liberal view at alcohol. The two highest risk factors for alcohol abuse in adolescents are laws and norms favorable to alcohol and social influences to alcohol use (Hawkins, et al. 1992). The attitudes towards alcohol have to change, both within the society, for this example in families and schools, but also at the government level with regulations and laws.
With regard to gender differences, the current study showed, firstly, that males had a higher prevalence of both school-connectedness problems (5.3% of the males and 2.9% of the females had 3-5 school connectedness problems) and alcohol misuse in adulthood (4.7% of the males and 2.1% of the females had be hospitalized due to alcohol misuse). Males had also higher prevalence of adjustment problems in all of the five studied variables (see Table 1).
The results from the regression analysis showed that there was a clear and statistically significant association between school connectedness and alcohol misuse among males but not among females. One explanation behind these gender differences could be the higher level of social support females have compared to males (Geckova, Van Dijk, Stewart, Groothoff, Post, 2003). Females in the risk zone more often have friends and family to support them in times of difficulties, where men do not talk about problems and anxiety to the same extent. A possible explanation is that the social support schools give to children is more important for male students than to female students; the social support females got from other contexts than school can buffer from risk factors resulting in alcohol misuse. Harm associated with alcohol use is expected to be associated to bonding at school and best friends’ drinking (Hawkins, et al., 1997). A Swedish study found an association between high alcohol consumption and psychosocial factors differed by gender. Peer pressure is higher among males, and can be a risk factor of drinking alcohol. Norms regarding the acceptability of alcohol misuse may differ for males and females (Hawkins, et al., 1997), where females often have the pressure of being the “good girl”, and therefore not drinking the same amount of alcohol and as often as males. Parental acceptance of alcohol consumption has been found to be more important for
females then for males (Stafström, Östergren & Larsson, 2005). These previous studies examining alcohol use among males and females all indicate difference in attitude between genders. Females generally consider that norms and parental acceptance are more important for drinking/ not drinking alcohol then males. If the norms change toward a more liberal view the females may start to drink larger amount of alcohol. In sum, this could be interpreted as males and females responding to school connectedness in different ways, or that females not get hospitalized due to alcohol misuse under the same premises that males.
7.1 Strengths and limitations of the study
A major strength of this study is the large data material and rich information collected over a long period of time, and the longitudinal design enables a life-course perspective to the studied questions. Another advantage of the study is the numbers of covariates used to control for confounding variables to decrease the risk of false correlations, as well as the possibility to include mediating factors in terms of adjustment problems in adolescence. Since the data material contained a lot of different information about circumstances at school collected from the children themselves, a more inclusive measure of school connectedness could be constructed. The socioeconomic status includes parental social class, parental education, and parental income, which are seen as the “classic” socioeconomic confounding factors. The study also controlled for parental alcohol abuse and found that the association between school connectedness and adult alcohol misuse still was significant. The large data material included register information, which contributes to good statistical power and provided more objective measurements and a small amount of attrition. The life-course perspective applied in the current study looks at the impact of historical and social change on human behavior, which other perspectives do not always takes into account. The life-course perspective also acknowledges people’s strengths and capacity for change. Another strength of the life-course perspective is that it gives opportunities to change the risk factors and the possibility to prevent future health outcomes with different kinds of interventions. This study contributes to the existing literature by studying objective health effects on individuals in a life course perspective. This study is showing a directly health disadvantage by looking at the data from the Swedish Hospital Discharge Register, allowing the connection to be more objective than self-rated health.
Nevertheless, a limitation of this study was the use of the hospital discharge register, which only includes the most serious forms of alcohol misuse, and not the outpatient care register with milder cases of alcohol misuse. Moreover, the study did not include information about how many times the participation has been hospitalized due to alcohol misuse. The measurement period for alcohol misuse was restricted to 1973-2007, which means that individuals who had been hospitalized due to alcohol before or after will not be included in the results. It is possible that a slightly different pattern had emerged if alcohol misuse had been measured during a different or a longer period of time.
School connectedness was constructed from five questions, which may not optimally fit into existing theories. Also, since each item was dichotomized, it was not feasible to look at their interrelations by means of factor analysis or correlation analysis. A sensitivity analysis was however done, which proved that all items were related to the outcome in the same way and therefore it was not considered problematic to merge them into an index. The measure of school connectedness can also be the reason why males has a stronger association between school connectedness and alcohol misuse in adulthood; the five indicators that were included measured more extrovert variables like; “If school were completely voluntary and you could quit tomorrow or stay if you wanted to, what would you do if you could decide yourself?”,
“Have you ever been absent from school even though you were not ill?” and “Have you ever been told to leave the classroom during a lesson because of something you did?”. Only two variables measured introvert behaviors; “The pupils’ feeling of security at school” and
“Interest in school work”. No variables included social factors like numbers of friends in school or liking the teacher. If the number of variable were more even distributed between extrovert behaviors, introvert behaviors and social behaviors, the associations may have looked different in terms of gender differences.
Another weakness of the current study is the fact that the major part of the data material was collected in the 1960s, and the level of school connectedness can have been different compared to today. But this is always a problem when looking in a life course perspective; the starting-point of the life course has to be much earlier in order to enable the possibility to look at later outcomes. The life-course perspective’s leading challenge is heterogeneity, because it makes it harder to generalize and find causes and connections that can be transferable to other populations. The study did not include any variables about the students’ school performance, which could be an important factor that explains the relationship between school
connectedness and alcohol misuse. Moreover, there can be other factors affecting the association and pathways that were not examined here.
A problematic point, which should be discussed, is the lack of a relationship between school connectedness and alcohol-related hospitalization among females. There was however a weak pattern that suggested a connection among females as well, but since the association was not statistically significant at the 5 % level, it was not analyzed further. Significance levels are however arbitrary and if another level – for example, the 10 % level – had been used, females could have been included in the further analysis. This is a tough issue that has no clear-cut solutions.
8. Policy implications
The findings of this study suggest that school connectedness is associated with alcohol misuse in adulthood; at least among males, and that adjustment problems in adolescence to some extent can explain this association. In terms of implications for policy, actions for preventing low school connectedness and alcohol misuse should perhaps focus the efforts on the school and the school environment. For example, Centres for Disease Control and Prevention (2009) have put forward four factors that can increase school connectedness: Adult support, Belonging to a positive peer group, Commitment to education, and School environment (see Figure 5). Students who feel supported and cared for by teachers and other adults at school are more likely to be more engaged in school and learning. Students’ health and educational outcomes are also influenced by the characteristics of their peers. Students who feel connected to school are having a lot of friends at school and having friends from several different social groups. School connectedness is higher for students who are interested in schoolwork and believe that a good education is important for reaching their life goals. The students who are committed to their education spend more time on homework and on school activities. School connectedness is also enhanced by a healthy and safe school environment and a supportive psychosocial climate. This includes both a good classroom management and a clean and pleasant environment (Centres for Disease Control and Prevention, 2009). These four factors can increase school connectedness that further on increase positive health outcomes and positive educational outcomes.
FIGURE 5. Promoting School Connectedness. Adapted from Centres for Disease Control and Prevention (2009)
Further studies that also include other possible mechanisms are needed in order to better understand the link between school connectedness and alcohol misuse in adulthood. Since the current study is based on a Swedish cohort, studies looking at the same association in other cultural contexts should be performed.
The creation and maintenance of the Stockholm Birth Cohort Data Base is a collaboration between CHESS and SOFI, financed by the Swedish Research Council. Sten-Åke Stenberg of SOFI prepared the original Metropolitan Data Base, Denny Vågerö of CHESS prepared the follow-up data for 1980-2002, and Reidar Österman of CHESS organized the probability matching of the two data sets.
I would like to give an extra big thank to my supervisor at CHESS, Ylva B Almquist, for supporting and helping me, especially with the statistics and methodology. Ylva has shared a lot of her knowledge and skills, and always been available for questions and comments.