Você está na página 1de 8

p u b l i c h e a l t h 1 2 6 ( 2 0 1 2 ) 5 6 6 e5 7 3

Available online at www.sciencedirect.com

Public Health

journal homepage: www.elsevier.com/puhe

Original Research

Individual and environmental predictors of health risk


behaviours among Dutch adolescents: The HBSC study

Z. Harakeh a,*, M.E. de Looze a, C.T.M. Schrijvers c,d, S.A.F.M. van Dorsselaer b,
W.A.M. Vollebergh a
a
Child and Adolescent Studies, Faculty of Social and Behavioural Sciences, Utrecht University, Utrecht, The Netherlands
b
Trimbos Institute (Netherlands Institute of Mental Health and Addiction), Utrecht, The Netherlands
c
Centre for Public Health Forecasting, National Institute for Public Health and the Environment, Bilthoven, The Netherlands
d
IVO, Addiction Research Institute, Rotterdam, The Netherlands

article info summary

Article history: Objective: To examine unique and common predictors of tobacco smoking, binge drinking,
Received 19 October 2010 cannabis smoking, early sexual intercourse and multiple health risk behaviours.
Received in revised form Study design: Cross-sectional survey study.
31 January 2012 Methods: The Dutch Health Behaviour in School-aged Children (HBSC) study was used to
Accepted 9 April 2012 provide data on 1742 adolescents aged 15 and 16 years of age. This study focused on
Available online 16 May 2012 a variety of individual and environmental predictors of health risk behaviours, tapping into
four domains (mental health, family, peers and school), retrieved by adolescent self-
Keywords: reports and corrected for sociodemographic variables. Logistic and linear regression
Substance use analyses were performed.
Early sexual debut Results: Unique predictors (i.e. gender, low and very low education level, general health,
Adolescents hyperactivity problems, conduct problems, incomplete family, religion, knowledge of
Multiple risk behaviours mother, parental rules on alcohol drinking, time spent with friends, number of friends,
Individual factors perceived tobacco use of classmates, truancy) were identified. In addition, common
Environmental factors predictors (i.e. permissive rules on alcohol drinking and much time spent with friends)
were also identified, explaining an increase in engagement in all investigated health risk
behaviours in adolescence, including multiple risk behaviours.
Conclusions: A prevention strategy targeting restrictive parenting and time spent with
friends may be effective to reduce/discourage engagement in health risk behaviours.
ª 2012 The Royal Society for Public Health. Published by Elsevier Ltd. All rights reserved.

Introduction Jessor’s problem behaviour theory,1 risk behaviours (e.g.


alcohol drinking, early sexual involvement, illicit drug use,
Adolescence is a period in which young people are likely to delinquent behaviour) often co-occur in adolescents. More
start experimenting and engaging in risk behaviours that may recently, a study by Willoughby et al. showed that risk
have adverse health consequences. According to Jessor and behaviours can be subdivided into three latent factors: (1)

* Corresponding author. Interdisciplinary Social Science, Utrecht University, PO Box 80.140, 3508 TC Utrecht, The Netherlands. Tel.: þ31
30 2539084; fax: þ31 30 2534733.
E-mail address: z.harakeh@uu.nl (Z. Harakeh).
0033-3506/$ e see front matter ª 2012 The Royal Society for Public Health. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.puhe.2012.04.006
p u b l i c h e a l t h 1 2 6 ( 2 0 1 2 ) 5 6 6 e5 7 3 567

substance use and sexual activity; (2) aggression; and (3) Europe. The Dutch HBSC database includes data for 5422
delinquency.2 Our present study focused on the risk behav- students who are representative for the Dutch youth in
iours included in the first factor (i.e. substance use and sexual secondary education aged 12e16 years (mean age 13.83 years,
activity). These behaviours are distinguished from other risk standard deviation 1.27 years). The sample was stratified
behaviours for the following reasons. First, in contrast to the according to urbanicity, and the schools were selected at
behaviours in the two other latent factors, these behaviours random. Schools were excluded when not all required grades
have negative health consequences such as lower quality of were present, or when they provided special education. In
life, anal cancer, cervical cancer, impotence, birth complica- total, 137 schools were approached from all schools in the
tions, etc.3 Second, they are often referred to as adult-like Netherlands, and 64 were willing to participate. Students from
behaviours, and adolescents have been hypothesized to four classes of each school participated: one class was
engage in them ‘to seek acceptance in adult society’.4,5 Third, selected at random for each grade, for each of the first four
they all have seductive, addictive qualities.3 The dopami- grades. Parents were informed of the aim of the study by
nergic system, which is related to pleasure and facilitates means of a letter. If parents had any objections to their child
dependence, is involved in these behaviours.3 Fourth, these participating in this study, they could inform the research
behaviours in adolescence may be a function of behavioural assistant or teacher. This was the case for three students. Data
disinhibition, and reflect low self-control, high impulsivity were not collected for students who were absent on the day of
and sensation seeking.3 Finally, there is no intention to data collection due to truancy (1%) or illness and other reasons
directly harm other people or the property of other people by (7%). Data were collected from October to November 2005
engaging in these behaviours.3 according to the protocol of the international HBSC study.13,14
In line with these reasons, previous studies have suggested All data were collected by means of questionnaires, which
that these behaviours may have an underlying basis and were distributed in classes and administered by research
similar predictors. If so, the development of a prevention assistants during a lesson (usual duration 50 min). The
strategy targeting multiple health risk behaviours would be anonymity of the students was stressed by the research
useful. To date, most interventions targeting adolescent assistants, and was guaranteed by enclosure of each
health risk behaviours have been behaviour specific.6 If completed questionnaire in a sealed envelope.
common predictors of the selected health risk behaviours For the purpose of this study, only the data from adoles-
could be identified, broader interventions could be developed, cents aged 15 or 16 years were used (n ¼ 1742, 51.3% male). The
targeting multiple health risk behaviours simultaneously; this education level was distributed as follows: 29.2% very low
would be more cost-effective. (preparatory middle-level applied education, profession-
According to Bronfenbrenner,7 factors from different oriented learning path), 32.2% low (preparatory middle-level
domains influence human behaviour. Besides individual applied education, theoretic learning path), 24.2% middle
factors (e.g. mental health factors), Bronfenbrenner’s ecology of (higher general continued education) and 14.5% high (prepa-
human development theory7 suggests the importance of envi- ratory scientific education). Furthermore, 74.9% of adolescents
ronmental factors such as the microsystem (e.g. family, peers in the sample had a Dutch origin (native).
and school) and larger social contexts (e.g. socio-economic
status, ethnic background).8e12 The aim of this study was to Measures
examine the associations between individual (mental health
problems) and environmental factors (school, family and peer All variables were assessed through self-reports of the
factors) on the one hand, and health risk behaviours (daily adolescents. Due to limited space, not all variables depicted in
tobacco smoking, binge drinking, monthly cannabis use and Tables 1e3 are described below. More information on these
early sexual intercourse) on the other hand. In addition, this measures can be retrieved from the protocol of the interna-
study also focused on multiple health risk behaviours by the tional HBSC study and/or the HBSC website.13,14
construct of a latent factor, ‘multiple risk behaviours’, and Tobacco smoking, binge drinking, cannabis smoking and
examined whether this can be predicted by a similar set of early sexual intercourse were measured with single questions.
predictors. To the authors’ knowledge, no studies to date have Answer categories were recoded as ‘currently not smoking
examined the predictors of all four risk behaviours and multiple daily’ vs ‘currently smoking daily’; ‘did not drink or occa-
risk behaviours extensively in one design. The advantage of this sionally drank less than five glasses in the previous month’ vs
approach is that, when examining these various health risk ‘drank five or more glasses on one or more occasions in the
behaviours in one design and correcting for sociodemographic previous month’15; ‘not used cannabis the previous month’ vs
variables, the most important predictor(s) can be identified for ‘did use cannabis in the previous month’; and ‘no early sexual
each domain and across domains. intercourse’ vs ‘early sexual intercourse (<17 years old)’.
General health was measured with a single item: ‘What do
you think of your health?’ Response categories were recoded
Methods as ‘not healthy’ vs ‘healthy’.16
Life satisfaction was measured with a single item: ‘How do
Participants and procedure you currently feel about your life?’, and responses were
recoded as ‘happy’ (7e10) vs ‘not happy’ (0e6).17
This study is part of the cross-sectional ‘Health Behaviour in Psychosomatic complaints were measured with eight
School-aged Children’ (HBSC) study, conducted in collabora- items (e.g. headache, stomach ache, feeling dizzy, feeling
tion with the World Health Organization Regional Office for nervous).18e20 Response categories were dichotomized as
568 p u b l i c h e a l t h 1 2 6 ( 2 0 1 2 ) 5 6 6 e5 7 3

Table 1 e aeMultivariate logistic and linear regression analyses regarding sociodemographic predictors of health risk
behaviours (i.e. tobacco smoking, binge drinking, cannabis smoking, early sexual intercourse, and multiple health risk
behaviours), with inclusion of the variables depicted in Tables 2 and 3.
Predictor Tobacco smoking Binge drinking Cannabis smoking Early sexual Multiple risk
(n ¼ 1670) (n ¼ 1651) (n ¼ 1622) intercourse (n ¼ 1642) behaviours (n ¼ 1632)

OR (95%CI) OR (95%CI) OR (95%CI) OR (95%CI) b (coef; SE)

Sociodemographic characteristics
Age 1.47 (1.05e2.06) 0.97 (0.71e1.31) 1.33 (0.95e1.87) 1.48 (1.09e2.00) 0.06 (0.07; 0.03)
Gender (ref. girls) 0.81 (0.59e1.12) 1.87 (1.42e2.45) 2.03 (1.43e2.87) 0.96 (0.69e1.34) 0.07 (0.09; 0.03)
Education level
High (ref.) 1.00 1.00 1.00 1.00 ref.
Middle 1.51 (0.76e2.98) 1.49 (0.84e2.63) 0.99 (0.52e1.87) 1.91 (1.14e3.19) 0.06 (0.09; 0.04)
Low 3.13 (1.56e6.28) 1.65 (0.98e2.76) 1.25 (0.63e2.49) 2.17 (1.32e3.56) 0.11 (0.16; 0.04)
Very low 4.55 (2.26e9.18) 1.64 (0.95e2.80) 1.07 (0.57e2.04) 3.16 (1.90e5.28) 0.16 (0.21e0.05)
Ethnicity (ref. Dutch origin) 1.21 (0.68e2.14) 0.65 (0.45e0.93) 1.43 (0.86e2.36) 1.32 (0.89e1.97) 0.01 (0.01; 0.04)

Ref., reference group; CI, confidence interval; SE, standard error.


Significant odds ratio (OR) in are shown in bold (P < 0.01). Empty cells with ‘e‘ represent non-significance in the first model. The results in Tables
1e3 were derived from one analysis but, to enhance visual clarity, the tables have been split up into three. However, the findings in these tables
should not be interpreted separately.

‘none’ vs ‘at least one’ psychosomatic complaint more than Quality of communication with mother was assessed with
once a week in the previous 6 months. a single item: ‘How easy is it for you to talk to your mother
Emotional problems, hyperactivity problems, conduct about things that really bother you?’ Responses were recoded
problems and problems with peers were assessed with the as ‘good’ vs ‘not good’.
Strength and Difficulties Questionnaire (SDQ).21 The Dutch Knowledge of mother was defined as how much the
translation of the SDQ was used,22 and each of the four adolescents’ mothers knew about their friends, whereabouts
subscales consisted of five items. The responses were recoded and activities. It was assessed with five items, and recoded as
as ‘normal’ vs ‘borderline/abnormal’ (i.e. the highest 15e20%) ‘mother knows much’ vs ‘mother knows little/nothing’.26
in the previous 6 months. Cronbach’s alpha was 0.77.
Prosperity refers to the current material goods in the family Parental rules on alcohol drinking refers to the tolerance
and is developed as a proxy indicator for socio-economic of parents with regard to the adolescent drinking alcohol at
position. It was measured with the four items of the Family home or at parties.27 An abbreviated version was used with
Affluence Scale (FAS),23 inquiring about the number of cars, four items (i.e. ‘I am allowed to have one glass of alcohol
computers and family holidays, and the adolescent’s disposal when my parents are at home’, ‘I am allowed to drink several
of a private and single bedroom. Previous studies have glasses of alcohol when my parents are not at home’, ‘I am
assessed the psychometric properties of this instrument and allowed to drink alcohol at a party with my friends’ and ‘I am
indicated good (criterion) validity.24,25 allowed to drink alcohol on the weekend’). Response

Table 2 e Multivariate logistic and linear regression analyses regarding individual predictors of health risk behaviours (i.e.
tobacco smoking, binge drinking, cannabis smoking, early sexual intercourse, and multiple health risk behaviours), with
inclusion of the variables depicted in Tables 1 and 3.
Predictor Tobacco smoking Binge drinking Cannabis smoking Early sexual Multiple health
(n ¼ 1670) (n ¼ 1651) (n ¼ 1622) intercourse risk behaviours
(n ¼ 1642) (n ¼ 1632)

OR (95%CI) OR (95%CI) OR (95%CI) OR (95%CI) b (coef; SE)

Individual factors
Mental health
General health (ref. not healthy) 0.36 (0.23e0.56) e 0.58 (0.38e0.88) e L0.08(-0.12e0.04)
Life satisfaction (ref. happy) e e e e e
Any psychosomatic problem (ref. none) e e e e e
Emotional problems (ref. normal) e e e
Hyperactivity problems (ref. normal) 1.52 (1.04e2.20) 1.48 (1.06e2.07) 1.33 (0.93e1.88) e 0.07 (0.11e0.03)
Conduct problems (ref. normal) 1.51 (0.99e3.31) 1.65 (1.14e2.40) 2.02 (1.42e2.89) 2.58 (1.88e3.56) 0.12 (0.21; 0.04)
Problem with peers (ref. normal) e e e e e

Ref., reference group; CI, confidence interval; SE, standard error.


Significant odds ratio (OR) in are shown in bold (P < 0.01). Empty cells with ‘e‘ represent non-significance in the first model. The results in Tables
1e3 were derived from one analysis but, to enhance visual clarity, the tables have been split up into three. However, the findings in these tables
should not be interpreted separately.
p u b l i c h e a l t h 1 2 6 ( 2 0 1 2 ) 5 6 6 e5 7 3 569

Table 3 e Multivariate logistic and linear regression analyses regarding environmental predictors of health risk behaviours
(i.e. tobacco smoking, binge drinking, cannabis smoking, early sexual intercourse, and multiple health risk behaviours),
with inclusion of the variables depicted in Tables 1 and 2.
Predictor Tobacco Binge Cannabis Early sexual Multiple health
smoking drinking smoking intercourse risk behaviours
(n ¼ 1670) (n ¼ 1651) (n ¼ 1622) (n ¼ 1642) (n ¼ 1632)

OR (95%CI) OR (95%CI) OR (95%CI) OR (95%CI) b (coef; SE)

Environmental factors
Family
Incomplete family (ref. e 1.56 (1.11e2.20) 1.54 (1.09e2.17) 1.54 (1.14e2.09) 0.07 (0.11; 0.03)
complete family)
Prosperity: e e e e e
High (ref.)
Middle
Low
Religious (ref. not religious) e e 0.55 (0.37e0.82) 0.64 (0.49e0.84) 0.06 (0.07; 0.02)
Quality of communication e e e e e
with mother (ref. good)
Knowledge of mother (ref. 2.20 (1.30e3.72) e 3.41 (2.00e5.82) 1.44 (0.82e2.52) 0.11 (0.24; 0.07)
knows much)
Rules on alcohol drinking:
Strict (ref.) 1.00 1.00 1.00 1.00 ref.
Moderate 2.13 (1.04e4.36) 4.87 (3.09e7.66) 2.57 (1.24e5.32) 1.34 (0.77e2.34) 0.18 (0.26; 0.05)
Permissive 4.62 (2.30e9.25) 15.28 (9.81e23.82) 4.06 (1.94e8.52) 2.72 (1.66e4.44) 0.40 (0.49; 0.05)
Peers
Time with friends:
Little (ref.) 1.00 1.00 1.00 1.00 ref.
Moderate (33%) 2.48 (1.34e4.61) 1.99 (1.39e2.85) 3.28 (1.91e5.66) 1.49 (0.93e2.39) 0.12 (0.15; 0.03)
Much (33%) 6.37 (3.53e11.49) 3.19 (2.25e4.52) 5.84 (3.27e10.43) 3.04 (2.15e4.30) 0.30 (0.37; 0.03)
Number of friends 1.25 (1.12e1.38) e e e
Quality of communication e e e e e
with friends (ref. difficult)
Perceived tobacco use of 1.64 (1.21e2.24) 1.64 (1.18e2.28) 1.51 (0.96e2.40) e 0.10 (0.13; 0.04)
classmates (ref. none/little)
Perceived alcohol use of e 1.40 (1.02e1.92) e e e
classmates (ref. none/little)
Perceived cannabis use of e e 1.31 (0.89e1.92) e e
classmates (ref. none/little)
School
Satisfaction with school 0.88 (0.65e1.19) e e e 0.01(0.01; 0.04)
(ref. not fun)
School-related stress e e e e e
(ref. no pressure)
Truancy (ref. no) 1.58 (1.04e2.40) 1.52 (1.08e2.13) 2.05 (1.44e2.90) e 0.08 (0.12; 0.04)
Academic achievement 0.85 (0.60e1.21) e e e 0.02(0.02; 0.03)
(ref. not good)
Classmate support e e e
Bullying (ref. never bullied) e 1.10 (0.80e1.52) 1.37 (1.05e1.78) 1.16 (0.90e1.51) 0.02 (0.03; 0.03)
Been bullied (ref. never e e e e d
been bullied)

Ref., reference group; CI, confidence interval; SE, standard error.


Significant odds ratio (OR) in are shown in bold (P < 0.01). Empty cells with ‘e‘ represent non-significance in the first model. The results in Tables
1e3 were derived from one analysis but, to enhance visual clarity, the tables have been split up into three. However, the findings in these tables
should not be interpreted separately.

categories ranged from 1 ¼ completely applicable to 5 ¼ not classmates smoke cigarettes?’; ‘How many of your classmates
applicable at all. Cronbach’s alpha was 0.92. drink alcohol weekly?’; ‘How many of your classmates use
Time with friends refers to how much the adolescent has cannabis?’), and responses were coded as ‘no/few classmates’
contact with their peers directly after school, in the evenings vs ‘half/most/many classmates’.
or via electronic-media communication (phone, e-mail, MSN Satisfaction with school was assessed with a single item:
or SMS). ‘How do you feel about school at present?’ Responses were
Number of friends at present was assessed with two items. recoded as ‘not experiencing school as fun’ vs ‘experiencing
Responses were combined from 0 ¼ none to 6 ¼ six or more. school as fun’.
Perceived tobacco, alcohol or cannabis use of classmates School-related stress was assessed with a single item:
were each assessed with single items (‘How many of your ‘How pressured do you feel by the schoolwork you have to do?’
570 p u b l i c h e a l t h 1 2 6 ( 2 0 1 2 ) 5 6 6 e5 7 3

Responses were recoded as ‘not feeling pressure of school- sociodemographic variables) with a P-value <0.01 in the first
work’ vs ‘feeling much pressure of schoolwork’. model were included in the final model. A P-value <0.01 was
Truancy was measured with a single item: ‘In the last 4 taken to indicate significance in the final model due to the
weeks, how many classes did you skip?’ Responses were large sample size and multiple testing in order to minimize
recoded as ‘not cutting classes’ vs ‘cutting classes’. random associations. Only the results of the final model are
Academic achievement refers to the adolescent’s percep- shown in this paper (see Tables 1e3; the results of the first step
tion of what his/her class teacher thinks about his/her school are available on request from the corresponding author).
performance compared with his/her classmates. Responses
were recoded as ‘not good (i.e. average or below average)
school performance’ vs ‘good (i.e. good or very good) school Results
performance’.
Classmate support was assessed with three items [e.g. In this sample, 16.7% of the adolescents smoked daily, 57.8%
‘Most of the students in my class(es) are kind and helpful.’]. had drunk five or more glasses of alcohol on one or more
Response categories varied from 1 ¼ completely agree to occasions in the previous month, 14.4% had used cannabis in
5 ¼ completely disagree. Cronbach’s alpha was 0.71. the previous month, and 29.7% had their sexual debut before
Both bullying and being bullied at school were assessed 17 years of age (early sexual intercourse). In addition, 5.7%
with single items,28 and responses were recoded as ‘never were engaged in all four health risk behaviours, 9.3% in three
(being) bullied’ vs ‘(being) bullied once or more’ in the past health risk behaviours, 17% in two health risk behaviours,
couple of months. 33.5% in one health risk behaviour; and 34.6% in none of these
health risk behaviours.
Statistical analyses The question of whether tobacco smoking, alcohol
drinking, cannabis smoking and early sexual intercourse
Multivariate logistic regression analyses were conducted to formed one underlying latent construct ‘multiple health risk
examine which determinants predicted tobacco smoking, behaviours’ was tested. The model showed a good fit:
binge drinking, cannabis smoking and early sexual intercourse. c2(2) ¼ 2.84, P ¼ 0.242, CFI ¼ 0.999, TLI ¼ 0.998, RMSEA ¼ 0.016.
Next, the question of whether tobacco smoking, binge drinking, The standardized coefficients for tobacco smoking, alcohol
cannabis smoking and early sexual intercourse formed one drinking, cannabis smoking and early sexual intercourse were
underlying latent construct ‘multiple health risk behaviours’ 0.86, 0.63, 0.88 and 0.68 respectively (P ¼ 0.000).
was tested using confirmatory factor analysis (CFA) with the The associations between sociodemographic, individual
software package Mplus (Muthén & Muthén, Los Angeles).29 (mental health problems) and environmental factors (school,
A good model fit is when the Comparative Fit Index (CFI) and family and peer factors) were examined on the one hand, and
Tucker Lewis Index (TLI) are both >0.95 and Root Mean Square the four health risk behaviours and multiple risk behaviours
Error of Approximation (RMSEA) >0.05.30 Furthermore, the on the other hand. The results of the final model for all five
weighted least square method was used, with adjusted mean outcome variables are depicted in Tables 1e3. The predictors
and variance chi-square statistic (WLSMV) estimator in explained 39.9% of the variance of engaging in multiple health
Mplus,29 and the derived factor scores for this underlying latent risk behaviours, 25.6% for tobacco smoking, 29.3% for alcohol
construct were transported to Stata version 11 (Stata Statistical drinking, 24.7% for cannabis smoking, and 15% for early
Software, release 11.1. Stata Corporation, College Station, TX)40. sexual intercourse.
Multivariate linear regression analyses were conducted to test
which determinants predicted multiple risk behaviours. Sociodemographic characteristics
Data were weighted for school level, grade, gender and
urbanicity. In addition, all analyses were corrected for Low and very low education were found to be a risk factor
possible cluster effects of pupils within the same schools by compared with high education: adolescents with low and very
using the first-order Taylor series linearization method. low education were more likely to engage in tobacco smoking,
Analyses were conducted in Stata Version 9.0 as this enables early sexual intercourse and multiple health risk behaviours.
correction for cluster effects using weighted data, and In particular, adolescents with a very low education were at
controls for multicollinearity by excluding variables that are greatest risk of engaging in these behaviours. However,
highly correlated from the analyses. education level was not related to binge drinking and
For each domain, two models for each of the five outcome cannabis smoking. With respect to gender, males were more
variables were tested in order to identify which predictors likely to binge drink, smoke cannabis, and engage in multiple
were most influential for the pertinent outcome. In the first health risk behaviours.
model, the predictors were tested separately for each domain
(mental health problems, family, peers and school), control- Mental health
ling for sociodemographic factors. For example, all variables
for mental health problems were tested in one model Conduct problems were found to be a risk factor for engage-
controlling for all the sociodemographic variables. The same ment in all health risk behaviours except for tobacco smoking.
approach was used to test family factors, peer factors and Feeling healthy was only a protective factor for tobacco
school factors. In the second and final model, all the significant smoking and engagement in multiple health risk behaviours.
variables from the first model were included, again correcting Hyperactivity problems were only found to be a risk factor for
for all the sociodemographic variables. Variables (except the multiple health risk behaviours.
p u b l i c h e a l t h 1 2 6 ( 2 0 1 2 ) 5 6 6 e5 7 3 571

Family predictors which have a predictive association with all health


risk behaviours may, therefore, be highly effective.
Rules of parents on alcohol drinking were found to be an First, the findings showed that permissive parental rules
important predictor: adolescents perceiving permissive on alcohol drinking are the strongest predictor in the family
parental rules on alcohol drinking were more likely to engage domain, which is in line with previous studies examining
in all five outcome variables. Instead, moderate parental rules alcohol use.31e33 However, the findings showed that this is
on alcohol drinking were only found to be a risk factor for also related to other health risk behaviours among adoles-
binge drinking and engaging in multiple health risk behav- cents (tobacco and cannabis use, early sexual intercourse and
iours. Furthermore, adolescents with mothers who had little multiple health risk behaviours). Abdelrahaman et al. showed
knowledge about their friends, whereabouts and activities that adolescents with parents who set clear rules against
were more likely to engage in health risk behaviours, except alcohol and drug use are also less likely to smoke cigarettes.34
for tobacco smoking and binge drinking. Finally, adolescents To the authors’ knowledge, this is the first study to show that
from incomplete families or adolescents who had not been these alcohol-specific rules are also a protective factor for
brought up in a religious way were more likely to have early other health-comprising behaviours such as sexual inter-
sexual intercourse and engage in multiple health risk behav- course and multiple health risk behaviours. A possible
iours. In addition, adolescents who had not been brought up in explanation for this finding may be that alcohol-specific rules
a religious way were also more likely to smoke cannabis. are an expression of underlying general restrictive parenting
practices (e.g. control and supervision), norms and beliefs35
Peers targeting the adolescent’s risk behaviour.
Second, spending much or moderate time with peers was
Time spent with friends was shown to be an important risk found to be the strongest predictor within the peer domain, as
factor for health risk behaviours: adolescents who spent a lot of it increases the risk of engagement in all four health risk
time with their friends were more likely to engage in all five behaviours as well as in multiple risk behaviours. A possible
outcome variables. Adolescents who spent a moderate amount explanation may be that adolescents who spend a lot of time
of time with their friends were more likely to engage in all health with their peers are also those whose parents undertake little
risk behaviours except for early sexual intercourse. The number efforts to control their child, such as parental control and
of friends was only found to be a risk factor for binge drinking. monitoring.36 Therefore, in future, it would be interesting to
Furthermore, adolescents who perceived much tobacco use examine whether the level of parental control and monitoring
among their classmates were more likely to engage in tobacco is related to the time spent with peers, which may, in turn,
smoking, binge drinking and multiple health risk behaviours. influence engagement in risk behaviours. Another possible
explanation may be that adolescents who are engaged in risk
School behaviours often do so in public settings where they subse-
quently meet others, spend more time with them and estab-
Truancy appeared to be a risk factor for health risk behav- lish new peer and intimate relationships.37 It is likely that
iours: adolescents who cut classes were more likely to engage adolescents who spend more time with their peers are more
in cannabis smoking and multiple risk behaviours. exposed to the health risk behaviours of their peers.

Limitations and future research


Discussion
This study has some limitations which should be mentioned
The objective of this epidemiological study was to identify the and considered in future research. First, self-reports of health
predictors of engagement in single and multiple health risk risk behaviours were used. Self-reports may be biased by
behaviours among a nationally representative sample of inaccuracies arising from cognitive processes such as
Dutch adolescents aged 15 or 16 years. Sociodemographic, comprehension and recall, or from factors related to social
individual (mental health) and environmental (family, peers desirability and interviewing conditions.38 However, in the
and school) factors were examined, and the strongest present study, these biases may have been reduced by
predictors were: gender and education level (sociodemo- guaranteeing anonymity and privacy to the subjects. Second,
graphic factors), conduct problems (mental health factor), the results were derived from cross-sectional data. It is
permissive parental rules on alcohol drinking (family factor) therefore not possible to provide information about cau-
and much time spent with friends (peer factor) and truancy seeconsequence relationships. Finally, this study focused on
(school factor). However, only permissive parental rules on four domains which have been described in previous studies.
alcohol drinking and much time spent with friends were However, these domains may not be all-encompassing.
identified as common predictors of engagement in all health
risk behaviours (tobacco smoking, binge drinking, cannabis
smoking and early sexual intercourse) as well as in the latent Conclusions and implications
construct ‘multiple health risk behaviours’. These two
predictors remained important when considering predictors This study showed that although some predictors are specif-
from other domains, and they explained a major part of the ically related to single health risk behaviours, these behav-
variance of each of these health risk behaviours. A broad iours have common predictors. These tap into the family
prevention programme focusing on these two common (parental rules on alcohol drinking) and peer (time spent with
572 p u b l i c h e a l t h 1 2 6 ( 2 0 1 2 ) 5 6 6 e5 7 3

peers) domains and should be considered when developing 2. Willoughby T, Chalmers H, Busseri MA. Where is the
intervention programmes targeting multiple risk behaviours syndrome? Examining co-occurrence among multiple
simultaneously. These results imply that adolescents engaged problem behaviors in adolescence. J Consult Clin Psych
2004;72:1022e37.
in health risk behaviours are those whose parents exert less
3. Sussman S. The relations of cigarette smoking with risky
control and supervision, and who have closer bonds and sexual behaviour among teens. Sex Addict Compulsiv
attachments with their peers (time spent with peers). This 2005;12:181e99.
does not automatically mean that increased peer contacts are 4. Moffitt TE. Adolescence-limited and life-course-persistent
detrimental for adolescents’ development; on the contrary, antisocial behaviour: a developmental taxonomy. Psychol Rev
intense peer relationships enable youth to reflect on their own 1993;100:674e701.
ideas and opinions with people of their own age,39 discuss 5. Moffit TE. A review of research on the taxonomy of life-course
persistent versus adolescence-limited antisocial behavior. In:
their worries and learn specific communication skills,37 and in
Flannery DJ, Vazsonyi AT, Waldman ID, editors. The
this way serve an important function in young people’s Cambridge handbook of violent behavior and aggression. New
learning to function in society. Moreover, adopting health risk York: Cambridge University Press; 2007. p. 49e74.
behaviours and/or positive attitudes and beliefs regarding risk 6. Guilamo-Ramos V, Litardo HA, Jaccard J. Prevention programs
behaviours may even be an essential part for the maturation for reducing adolescent problem behaviors: implications of
of adolescents.4,12 Nevertheless, increased peer contact may the co-occurrence of problem behaviors in adolescence. J
Adolesc Health 2005;36:82e6.
become detrimental when the increase in bonding and
7. Bronfenbrenner U. The ecology of human development: history
attachment with peers is accompanied by low parental
and perspectives. Psychologia-Wychowawcza 1976;19:537e49.
control and supervision, especially when adolescents are still 8. Buhi ER, Goodson P. Predictors of adolescent sexual behaviour
young and need supervision to prevent them from excessively and intention: a theory-guided systematic review. J Adolesc
high levels of engagement in multiple health risk behaviours. Health 2007;40:4e21.
9. Donovan JE. Adolescent alcohol initiation: a review of
psychosocial risk factors. J Adolesc Health 2004;35:7e18.
10. Galea S, Nandi A, Vlahov D. The social epidemiology of
Acknowledgements substance use. Epidemiol Rev 2004;26:36e52.
11. Goodson P, Evans A, Edmundson E. Female adolescents and
onset of sexual intercourse: a theory-based review of
The authors are grateful to the study participants. Further- research from 1984 to 1994. J Adolesc Health 1997;21:147e56.
more, the authors would like to thank Dr Quinten A.W. 12. Tyas SL, Pederson LL. Psychosocial factors related to
Raaijmakers for his helpful advice in computing an underlying adolescent smoking: a critical review of the literature. Tob
construct ‘multiple health risk behaviours’. Control 1998;7:409e20.
13. Currie CW, Samdal O, Boyce W, Smith R. Health behaviour in
Ethical approval school-aged children: a World Health Organization cross-national
study. Research protocol for the 2001/02 survey. Edinburgh: HBSC;
2002.
The Advisory Board of the Ethics Committee of the Faculty of
14. http://www.hbsc.org/[last accessed 17.05.10].
Social Sciences at Utrecht University gave their approval. 15. Hibell B, Guttormsson U, Ahlström S, Balakireva O,
Bjarnasson T, Kokkevi A, et al. The 2007 ESPAD Report.
Funding Substance use among students in 35 European countries. Sweden:
Modintryckoffset AB; 2009.
Data collection of 2005 was supported by grants from the 16. Idler EL, Benyamini Y. Self-rated health and mortality:
Netherlands Organization for Health Research and Develop- a review of twenty-seven community studies. J Health Soc
Behav 1997;38:21e37.
ment (ZonMw; 6000.0970), the Netherlands Organization for
17. Cantril H. The pattern of human concern. New Brunswick, NJ:
Scientific Research (NWO; 481-08-002), the Netherlands Insti- Rutgers University Press; 1965.
tute for Social Research, the Ministry of Youth and Family, 18. Ravens-Sieberer U, Erhart M, Torsheim T, Hetland J,
Trimbos Institute and Utrecht University. Z. Harakeh was Freeman J, Danielson M, et al. An international scoring
supported by a grant (VENI) from NWO during the preparation system for self-reported health complaints in adolescents.
of this manuscript. None of these sponsors had any involve- Eur J Public Health 2008;18:294e9.
19. Hagquist C, Andrich D. Measuring subjective health among
ment in the study design; in the collection, analysis and
adolescents in Sweden. Soc Indic Res 2004;68:201e20.
interpretation of the data; in the writing of the manuscript; or
20. Haugland S, Wold B, Stevenson J, Aarø LE, Woynarowska B.
in the decision to submit the manuscript for publication. Subjective health complaints in adolescence e a cross-
national comparison of prevalence and dimensionality. Eur J
Competing interests Public Health 2001;11:4e10.
21. Goodman R, Meltzer H, Bailey V. The strengths and
None declared. difficulties questionnaire: a pilot study on the validity of the
self-report version. Eur Child Adolesc Psychiatry 1998;7:125e30.
22. Van Widenfelt BM, Goedhart AW, Treffers PDA, Goodman R.
Dutch version of the strengths and difficulties questionnaire
references (SDQ). Eur Child Adolesc Psychiatry 2003;12:281e9.
23. Parry-Langdon N, Clements A, Harris L, Roberts C, Kingdon A.
Family affluence: a parentestudent validation study in Wales.
1. Jessor R, Jessor SL. Problem behaviour and psychosocial development: Health Promotion Division/Is-adran Hybu Iechyd, Welsh
a longitudinal study of youth. New York: Academic Press; 1977. Assembly Government; 2005.
p u b l i c h e a l t h 1 2 6 ( 2 0 1 2 ) 5 6 6 e5 7 3 573

24. Boyce W, Torsheim T, Currie C, Zambon A. The family 33. Van der Vorst H, Engels RCME, Meeus W, Dekovic  M. Van
affluence scale as a measure of national wealth: validation Leeuwe. The role of alcohol-specific socialization in
of an adolescent self-report measure. Soc Indic Res adolescents’ drinking behaviour. Addiction 2005;100:1464e76.
2006;78:473e87. 34. Abdelrahman AI, Rodriguez G, Ryan JA, French JF,
25. Currie C, Molcho M, Boyce W, Holstein BE, Torsheim T, Weinbaum D. The epidemiology of substance use among
Richter M. Researching health inequalities in adolescents: the middle school students: the impact of school, familial,
development of the health behaviour in school-aged Children community and individual risk factors. J Child Adoles Subst
(HBSC) family affluence scale. Soc Sci Med 2008;66:1429e36. 1998;8:55e75.
26. Rispens J, Hermanns JMA, Meeus WHJ. Opvoeden in Nederland 35. Van Zundert RMP, Van der Vorst H, Vermulst AA,
[Parenting in the Netherlands]. Assen: van Gorcum; 1997. Engels RCME. Pathways to alcohol use among Dutch students
27. Van der Vorst H, Engels RCME, Meeus W, Dekovic  M. The in regular education and education for adolescents with
impact of alcohol-specific rules, parental norms about early behavioral problems: the role of parental alcohol use, general
drinking and parental alcohol use on adolescents’ drinking parenting practices, and alcohol-specific parenting practices.
behavior. J Child Psychol Psyc 2006;47:1299e306. J Fam Psychol 2006;20:456e67.
28. Olweus D. The revised Olweus bully/victim questionnaire. Bergen, 36. Wang J, Simons-Morton B, Farhart T, Luk JW. Socio-
Norway: research Center for health Promotion (HEMIL Center). demographic variability in adolescent substance use:
University of Bergen; 1996. mediation by parents and peers. Prev Sci 2009;10:387e96.
29. Muthén LK, Muthén BO. Mplus user’s guide. 5th ed. Los 37. Engels RCME, Ter Bogt T. Influences of risk behaviors in the
Angeles: Muthén & Muthén; 1998e2007. pp. 55e6. quality of peer relations in adolescence. J Youth Adolesc
30. Kline RB. Principles and practice of structural equation modelling. 2001;30:675e95.
3rd ed. New York: The Guilford Press; 2010. 38. Brener ND, Billy JOG, Grady WR. Assessment of factors
31. Koning IM, Engels RCME, Verdurmen JEE, Vollebergh WAM. affecting the validity of self-reported health-risk behavior
Alcohol-specific socialization practices and alcohol use in among adolescents: evidence from the scientific literature. J
Dutch early adolescents. J Adolesc 2010;33:93e100. Adolesc Health 2003;33:436e57.
32. Spijkerman R, Van den Eijnden RJJM, Huiberts A. 39. Johnson KA, Jennison KM. The drinking-smoking syndrome
Socioeconomic differences in alcohol-specific parenting and social context. Int J Addict 1992;27:749e92.
practices and adolescents’ drinking patterns. Eur Addict Res 40. Stata Corp., 2009. Stata Statistical Software, release 11.1. Stata
2008;14:26e37. Corporation, College Station, TX.