Access to this full-text is provided by Springer Nature.
Content available from BMC Public Health
This content is subject to copyright. Terms and conditions apply.
R E S E A R C H A R T I C L E Open Access
Is adolescent multiple risk behaviour
associated with reduced socioeconomic
status in young adulthood and do those
with low socioeconomic backgrounds
experience greater negative impact?
Findings from two UK birth cohort studies
Laura Tinner
*
, Caroline Wright, Jon Heron, Deborah Caldwell, Rona Campbell and Matthew Hickman
Abstract
Background: Adolescent multiple risk behaviour (MRB) is associated with negative outcomes such as police arrests,
unemployment and premature mortality and morbidity. What is unknown is whether MRB is associated with
socioeconomic status (SES) in adulthood. We test whether adolescent MRB is associated with socioeconomic status
(SES) in young adulthood and whether it is moderated by early life SES variables.
Methods: Prospective cohort studies; British Cohort Study 1970 (BCS70) and Avon Longitudinal Study of Parents
and Children (ALSPAC), born in 1991–1992, were used and two comparable MRB variables were derived. Logistic
regression was used to determine the association between MRB and young adult SES. The moderating effect of
three early life SES variables was assessed using logistic regression models with and without interaction parameters.
Evidence to support the presence of moderation was determined by likelihood ratio tests ≤p= 0.05. Multiple
imputation was used to account for missing data.
Results: Adolescents had a median of two risk behaviours in BCS70 and three in ALSPAC. Adolescent MRB was
negatively associated with young adult SES (university degree attainment) in BCS70 (OR 0.81, 95% CI: 0.76, 0.86) and
ALSPAC (OR 0.85, 95% CI: 0.82, 0.88). There was a dose response relationship, with each additional risk behaviour
resulting in reduced odds of university degree attainment. MRB was associated occupational status at age 34 in
BCS70 (OR 0.86 95% CI: 0.82, 0.90). In BCS70, there was evidence that maternal education (p= 0.03), parental
occupational status (p= 0.009) and household income (p= 0.03) moderated the effect of adolescent MRB on young
adult SES in that the negative effect of MRB is stronger for those with low socioeconomic backgrounds. No
evidence of moderation was found in the ALSPAC cohort.
© The Author(s). 2021, corrected publication 2021. Open Access This article is licensed under a Creative Commons Attribution
4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as
long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence,
and indicate if changes were made. The images or other third party material in this article are included in the article's Creative
Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative
Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need
to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/
licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.
0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
* Correspondence: laura.tinner@bristol.ac.uk
Population Health Sciences, Bristol Medical School, University of Bristol, BG3
Oakfield House, Bristol BS8 2BN, UK
Tinner et al. BMC Public Health (2021) 21:1614
https://doi.org/10.1186/s12889-021-11638-3
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Conclusions: Adolescence appears to be a critical time in the life course to address risk behaviours, due to the
likelihood that behaviours established here may have effects in adulthood. Intervening on adolescent MRB could
improve later SES outcomes and thus affect health outcomes later in life. Evidence for a moderation effect in the
BCS70 but not ALSPAC suggests that more detailed measures should be investigated to capture the nuance of
contemporary young adult SES.
Keywords: Adolescence, Socioeconomic factors, Cohort studies, Inequalities, Multiple risk behaviour
Background
Risk behaviours such as excessive alcohol consumption,
tobacco use, risky sexual behaviour and low physical ac-
tivity commonly start in adolescence. Such behaviours
can be detrimental to health and are costly to society
[1]. There is evidence of the co-occurrence of risk be-
haviours, with individuals engaging in one behaviour be-
ing likely to adopt others [2]. Multiple risk behaviour
(MRB) refers to the occurrence of two or more risk be-
haviours. MRB in adolescence has been shown to be as-
sociated with adverse health and social outcomes, such
as unemployment [3], low educational attainment at
GCSE level [4], getting in trouble with the police [5] and
premature mortality and morbidity [1,6]. Therefore,
adolescence is increasingly acknowledged as a critical
time in the life course to address health risk behaviours
due to the likelihood that behaviours established here
may have deleterious effects in adulthood.
Adolescence and young adulthood are also crucial in
addressing health and social inequalities, as it is at this
time that individuals can improve their life chances
through education [7] and entry into the labour market
[8]. Few studies predict from adolescence, with many
studies concerned with exposures in early years, how-
ever, health risk behaviours initiated here are likely to be
‘major contributors to links between deprivation and in-
equality in later life’[9]. Socioeconomic status (SES) is a
construct used to assess inequalities, with commonly
used measures including education, income or occupa-
tion [10]. Previous research has found early life or ‘ori-
gin’SES to strongly predict SES in adulthood [11],
however, examining SES as an outcome has seldom been
done in longitudinal health research. Addressing this
gap will increase our knowledge around how health risk
behaviours contribute to inequalities later in life,
through disturbing social circumstances in young adult-
hood [3,12].
Given the lifestyle perspective of health inequalities,
whereby it is hypothesised that less affluent individ-
uals suffer a heavier burden of health issues due to
the adverse behaviours they engage in [13], it is
plausible to expect risk behaviours and socioeconomic
status to interact across the life course. Risk behav-
iours may be more dangerous for some than for
others due to differential social resources that may
‘buffer some of the costs of adolescent risk behaviour’
[14]. Therefore, accounting for potentially moderating
factors in longitudinal data analyses on consequences
is essential, although so far has been lacking [15].
Shackleton et al. [16]foundtheretobesociallypat-
terned changes in adolescent health behaviours such
as tobacco smoking and obesity when examining two
birth cohorts 30 years apart. This is consistent with
the evidence that inequalities in the general popula-
tion have increased for many health outcomes [17].
In this study, we used two prospective cohort studies,
born 20 years apart to: (1) test the association between
adolescent MRB and young adult SES as measured by
university degree attainment in mid-twenties; and (2) as-
sess whether early life SES moderates this association
[18]. In doing so we aimed to illuminate the nature of
the relationship between adolescent MRB on later life
chances and determine whether this relationship was dif-
ferent for young people from differing socioeconomic
backgrounds. Using two birth cohorts allowed us to test
the use of the MRB variable in two populations and ex-
plore the associations in two different historical
contexts.
Methods
Participants
1970 British Cohort Study
The BCS70 sampled all births in Great Britain within 1
week in 1970 (n= 17,196) with 16,568 participants
followed beyond birth. There have been nine sweeps of
data collection up to age 46 years [19]. Although the ini-
tial cohort was representative of the UK at the time,
those who dropped out of data collection sweeps are
more likely to be male, from low socioeconomic back-
grounds and have a single parent at birth [20]. The sam-
ple does not reflect the ethnic diversity of today’s
population given the difficulty in recruiting immigrants
to the sample at subsequent waves [19].
The Avon Longitudinal Study of Parents and Children (ALSP
AC)
All children in the ALSPAC cohort were born to
mothers residing in Avon in southwest England between
April 1991 and December 1992 [21]. There were 14,541
pregnant mothers recruited, 14,062 live births and
Tinner et al. BMC Public Health (2021) 21:1614 Page 2 of 13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
13,988 (52% males and 48% females) singletons/twins
still alive at 12 months old [22]. Although the original
ALSPAC cohort was representative of the UK 1991 cen-
sus, participants are more likely to be of white ethnicity
(OR 3.85, 95% CI: 3.50, 4.24) and less likely to be eligible
for free school meals (OR 0.46, 95% CI: 0.43, 0.50) than
the national average [22].
Measures
We identified SES and MRB measures from the two co-
horts which were broadly comparable. All measures
were self-reported. Some measures were more readily
comparable than others and further details on the har-
monisation process and derivation of the variables are
available in Supplementary File 1.
Exposure: adolescent MRB
Table 1contains the derivation of the health risk behav-
iors in each birth cohort. For BCS70, information on risk
behaviour engagement was captured through a single
questionnaire distributed by schools at age 16 years. For
ALSPAC, information related to engagement in risk be-
haviour during adolescence was derived from a
computer-based session during a clinic attended at age
15 and a later postal questionnaire at age 16.
Eight comparable MRBs were used in both cohorts:
physical inactivity; car risk; criminal/anti-social behav-
iour; hazardous alcohol consumption; regular tobacco
smoking; cannabis use; illicit drug/solvent use; unpro-
tected sex. The ALSPAC cohort had five additional
MRBs: TV viewing; penetrative sex before age 16;
scooter risk; cycle helmet risk and self-harm.
For our analyses, a total number of risk behaviours
from 0 to 8 (BCS70) and 0 to 13 (ALSPAC) was derived
for each participant. These risk behaviours are not ex-
amined individually but as a composite MRB score.
The use of the MRB measure was informed by previ-
ous research on the ALSPAC cohort that use the same
configuration of health risk behaviors, which has been
shown to be associated with a number of outcomes [1,
4,5]. The individual risk behaviours were chosen based
on discussions with adolescents through a young per-
son’s research advisory group [1,4]. The individual risk
behaviours were modelled to test whether there was one
behaviour that could explain the association (Supple-
mentary File 1). We also conducted sensitivity analyses
using alternative classifications of MRB, using the health
risk behaviours that were individually associated with
the outcomes (0–6 in BCS70, 0–7 in ALSPAC) (Supple-
mentary File 1). These analyses, as well as previous la-
tent class analyses using the ALSPAC cohort, did not
find compelling enough evidence for an alternative clas-
sification of MRB [23].
Outcome: young adult SES
The outcome variable was young adult SES, which was
measured by self-reported university degree attainment.
Education has been found to be the most important and
stable discriminator among young adults [24]. Thus,
while other measures may tell us different elements of a
young person’s SES, education remains a strong indica-
tor of SES and was the primary outcome in this study.
A single question asked for highest educational attain-
ment was asked of BCS70 participants at age 26. A series
of questions were asked of ALSPAC participants be-
tween age 21 and age 24. The variable created from re-
sponses to these was binary with ‘university degree
attainment’denoting ‘high SES’.
Occupational status at age 34 was examined as a sec-
ondary outcome in the BCS70, however, these data were
unavailable for ALSPAC. Occupational status was mea-
sured using the Registrar General’s Social Class classifi-
cation and dichotomised, taking the lower tiers,
capturing mostly manual occupations, as the ‘low SES’
reference category.
Early life SES variables
Parental SES was analysed using three distinct variables:
maternal education, parent occupational status and house-
hold equalised income. Each variable was dichotomised
into ‘high SES’and ‘low SES’for the purposes of compar-
ing the association between two groups, with ‘high SES’as
the reference group as we were specifically concerned with
the effect of MRB on university degree attainment for
those with low SES backgrounds. Maternal education was
dichotomised differently in the two birth cohorts. For
BCS70, high SES denoted O-levels or higher whereas for
ALSPAC, A-levels and higher were coded as high SES as
it was decided this was the most relevant SES milestones
within the context of the time period [25]. Few mothers in
either birth cohort recorded having a degree at the time of
pregnancy (2.57% (BCS70) 12.89% (ALSPAC)). A table
showing the derivation of the moderator variables in both
cohorts is in Supplementary File 1.
Confounders
The analyses were adjusted for known confounders ‘sex’
and ‘conduct problems at age 10’in both cohorts. ALSP
AC had additional covariates, including ‘season of birth’
which has been shown to be an important predictor of
educational attainment [4] as well as ‘IQ score at age 8
years’and ‘key stage 2 educational attainment’to reduce
the likelihood of reverse causality between early educa-
tional performance and MRB engagement. Early life SES
has been treated as a confounder in similar ALSPAC
analyses, for instance between MRB and GCSE attain-
ment [4]. However, a variable may be ‘labelled in one
study as a confounder and in another study of the same
Tinner et al. BMC Public Health (2021) 21:1614 Page 3 of 13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
outcome in the same population as a mediator or mod-
erator, depending on which factor is the focus of each
investigation’[26]. Whereas previous studies have been
concerned with adjusting for early life SES in order to
remove variables that may be masking the true effect of
the exposure and outcome, we are specifically interested
in determining if there is a different effect between the
early life SES groups.
Statistical analysis
All analyses were done in Stata V.15. Data analysis was
conducted in stages starting with descriptive statistics of
the samples. Logistic regression models tested the asso-
ciation between adolescent MRB and young adult SES.
We tested for evidence of a linear relationship between
MRB and SES in young adulthood, with OR denoting re-
duction in odds of attaining a university degree with the
Table 1 Derivation of MRB variables in both cohorts
Health risk
behaviour
How variable was derived in BCS70 How variable was derived in ALSPAC
Physical
inactivity
Young person (YP) has typically over the past year done sport in or out
of school less than once a week (from an extensive list of sports). If
young person said they took exercise last Saturday, they were included
in the non-risk behaviour group.
Young person (YP) has typically over the past year
exercised < 5 times per week.
TV viewing –YP spent 3 or more hours watching television on
average per day across the week.
Car passenger
risk
Young person has drunk and driven once or more AND/OR young
person wears a seatbelt never, a few times or most times.
YP had been in a car passenger at least once in their
lifetime where the driver (a) had consumed alcohol or
(b) did not have a valid licence or (c) the YP chose not
to wear a seat belt last time travelled in a car, van or
taxi.
Cycle helmet
use
If the YP reported that they had last ridden a bicycle
within the previous four weeks and they had not worn a
helmet on the most recent occasion.
Scooter risk YP has driven a motorbike/ scooter off road or without a
licence on a public road at least once.
Criminal/
Antisocial
behaviour
Young person reported that since the age of 10 at least one of the
following offenses: Been questioned by police; let off with a warning;
been arrested; formally cautioned; found guilty in court OR in the past
year reported at least one of the following offenses: broken a window/
smashed others property; stolen item from a shop; used physical force
to get money; broken into a house to steal something; stolen a bike;
broke into a cash dispenser; swore at a teacher; driven a car on the
road underage; sold something shop lifted.
YP reported that at least once in the past year they had
undertaken at least one of the following 7 offences:-
carried a weapon; physically hurt someone on purpose;
stolen something; sold illicit substances to another
person; damaged property belonging to someone else
either by using graffiti, setting fire to it or destroying or
damaging it in another fashion; subjected someone to
verbal or physical racial abuse; or been rude/rowdy in a
public place.
Hazardous
alcohol
consumption
Young person has had four drinks or more in a row in the past two
weeks AND has been really drunk in the last year.
In the past year had scored 8 or more out of 40 on the
Alcohol Use Disorders Identification Test (AUDIT)
indicating hazardous alcohol consumption.
10 questions based on drinking behaviour, scoring
points for never (0 points), less than monthly (1 point),
monthly (2 points), weekly (3 points) and daily/almost
daily (4 points).
Regular
tobacco
smoking
Young person indicates smoking at least one cigarette per week. Has ever smoked and is regularly smoking by currently
smoking at least one cigarette per week.
Cannabis use Young person reports having smoked cannabis 10+ times in the past
year.
Those who reported using cannabis “sometimes but less
often than once a week”or more regular use were
classified as occasional users.
Illicit drug/
solvent use
Young person has used any of the following drugs two or more times
in the past year: cocaine; solvents; LSD; downers; uppers; heroin.
In the year since their 15th birthday, YP had either been
a regular user (i.e. used five or more times) of one or
more illicit drugs (excluding cannabis) including
amphetamines, ecstasy, LSD, cocaine, ketamine or
inhalants including aerosols, gas, solvents and poppers.
Self-harm –Young people who said they had purposely hurt
themselves in some way in their lifetime.
Penetrative sex
before age 16
–YP reported having had penetrative sex in the preceding
year and that they were under 16 at the time.
Unprotected
sex
Young person reports method of contraception is “none”,“boy
withdraws”,“use safe period”or “trust to luck”.
Penetrative sex without the use of contraception on the
last occasion they had had sex in the past year
Tinner et al. BMC Public Health (2021) 21:1614 Page 4 of 13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
increase of one health risk behaviour. Non-linearity was
explored by adding quadratic terms in the MRB variable.
There was no evidence that this had greater predictive
power than a linear effect of the exposure on young
adult SES (likelihood ratio test yielded p= 0.44 in BCS70
and p= 0.29 in ALSPAC) so the linear assumption was
upheld and the imputation models developed on this
basis.
Separate interactions were fitted to assess whether
early life SES moderated the association between MRB
(continuous) and SES in young adulthood (binary),
therefore allowing the linear relationship between MRB
and the log-odds for young adult SES to differ by early
life SES. We hypothesised a negative effect of MRB on
young adult SES. A p-value from the likelihood ratio test
of ≤0.05 was taken as evidence of a moderation effect.
This analysis was undertaken for each hypothesised
moderator variable (household income, parent occupa-
tional status, maternal education) in both cohorts. We
hypothesised a negative moderation effect and therefore
expected an interaction term < 1. The findings from the
moderation analyses are presented in tables and ‘simple
slopes’interaction plots to display the nature of the rela-
tionship interaction plot, using the margins and margin-
splot commands in Stata. Moderation models are best
understood using graphs or plots so that the nature of
the relationship can be visualised [27]. These graphs plot
the predicted values of the outcome (young adult univer-
sity degree) at different levels of the exposure (adoles-
cent MRB) using two separate lines for levels of the
hypothesised moderator (early life SES). The solid line
represents the predicted values of the outcome with the
shaded area denoting the 95% confidence intervals
around these predicted values. If the lines on the plot
were parallel, this indicated there was no evidence of a
moderation effect and the association between adoles-
cent MRB and young adult degree attainment was the
same for both levels of early life SES (high/low).
Missing data
The analysis variables were presumed missing at random
(MAR) as systematic differences between the observed
and missing data could be explained by associations with
the observed data [28]. Therefore, multiple imputation
was adopted to reduce bias even though the proportion
of missing data was large [29] as the MAR assumption is
not determined by the amount of missingness, but rather
the nature of the missingness [30]. Both birth cohorts
benefited from a rich variety of auxiliary variables that
were associated with missingness or the incomplete vari-
ables (BCS70: n= 41 ALAPC: n= 52). These variables
were used to reduce bias, improve the precision of the
imputation model and improve efficiency [4,28]. Please
see Supplementary File 1for information on the
proportion of missingness for each variable that was im-
puted to create our analysis sample (Supplementary Fig-
ures 1and 2). Multiple imputation by chained equations
was implemented using the ice command in Stata.
As the hypothesised moderator variables were categor-
ical, we opted to stratify the dataset into high SES/low
SES strata for each parental SES variable and impute
separately in each stratum. This approach was motivated
by the derivation of the MRB exposure variable. In order
to confidently impute the exposure variable, each indi-
vidual risk behaviour needed to be imputed separately
and then combined (not only imputing the derived MRB
score). The individual risk behaviours have been im-
puted separately in previous work in ALSPAC [1,4].
Therefore, including the interaction term in the imput-
ation model was computationally challenging with the
potential for error. Undertaking the multiple imputation
models in each separate stratum provided an opportun-
ity to preserve any potential moderation effect between
early life SES and the association and impute the MRB
variable at the level of individual risk behaviours.
The imputation sample was those who had complete
early life SES data, not the original cohorts, in order to
conduct the stratified imputation approach described.
For BCS70 this was n= 9691 (56.4%) and for ALSPAC
9001 (64.5%). Incomplete MRB (exposure), young adult
SES (outcome) and incomplete confounder data were
imputed up to the imputation samples.
Monte Carlo errors were used to compare the results
when imputing 25, 100, 250 and 500 data sets [4] using
the mcerror command in Stata. When calculated for the
models for each of the three datasets separated by early
life SES, based on 100 datasets for BSC70 and 500 data-
sets for ALSPAC, Monte Carlo errors were less than
10% of its standard error, with small pvalues (less than
0.01) and tstatistic Monte Carlo errors were less than or
equal to 0.1. As these three rules of thumb were upheld,
reproducibility of the imputed data using 100 and 500
datasets was considered satisfactory [31]. Parameters
were estimated using logistic regression in each imputed
dataset. These estimates were combined by averaging
across the multiple imputed datasets using Rubin’s com-
bination rules [32]. Standard errors were calculated to
show the uncertainty of the missing values, through ac-
counting for ‘between- and within-imputation’variation
in the parameter estimates [33]. A large number of im-
putations was expected, given the large proportion of
missing data and that a previous study on MRB in ALSP
AC also required 500 imputations [4]. Complete case
and imputed data were compared to assess to what ex-
tent missing data were impacting on the results. Un-
paired t-tests were undertaken to compare those in the
complete case samples to those with incomplete analysis
variables (Supplementary File 1).
Tinner et al. BMC Public Health (2021) 21:1614 Page 5 of 13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Results
Descriptive statistics
Of the 17,196 subjects enrolled in the birth cohort, 9691
had complete data for all three early life SES (maternal
education, household income and parental occupational
class) variables and therefore became the imputation
sample, with all other analysis variables imputed up to
this number. Of the 17,196 individuals, 8134 had all
MRB (exposure) data available and 8926 had the out-
come variable ‘young adult degree attainment.’The
complete case sample for the BCS70 cohort was 1358
(14.01% of the 9691 in the imputation sample and 7.90%
of the birth cohort sample).
Of the 13,952 subjects included in the birth cohort
(enrolled cohort, singletons and twins alive at 1 year),
9001 had all three early life SES variables (maternal
education, household income and parental occupa-
tional class) recorded and this made up our imput-
ation sample. Of the 13,952 individuals, 2656 had all
MRB (exposure) data available and 5276 had the out-
come variable ‘young adult degree attainment’.The
complete case sample was 1360 participants (15.11%
of the 9001 in the imputation sample and 9.75% of
the birth cohort sample). Supplementary File 1con-
tains the flow diagrams describing the derivation of
theimputationsampleandthecompletecasesample
for both cohorts (Supplementary Figures 3and 4).
Table 2contains the descriptive statistics of both sam-
ples. The descriptive statistics for the enrolled cohort
samples can be found in Supplementary File 1. The me-
dian and mean number of health risk behaviours en-
gaged in during adolescence by the complete case
sample is approximately two behaviours in the BCS70
cohort (mean:1.74 (SD 1.35), median: 2 (IQR 1–3)) and
three behaviours (mean: 2.91 (SD 1.89), median: 3 (IQR
1–4)) in the ALSPAC cohort. The sensitivity analysis
that restricted the MRB variable to eight risk behaviours
in ALSPAC revealed a lower average of behaviours com-
parable to the BCS70 data (mean 2.00 (SD 1.41), median:
2 (IQR 1–3).
Associations between young adult SES and adolescent
multiple risk behaviour and parental SES
BCS70
There was evidence of a negative association between
MRB and university degree by mid-twenties (Table 3).
In the BCS70, an odds ratio of 0.81 (0.76, 0.86) p< 0.001
shows that with each individual risk behaviour, young
people have 19% reduction in odds of attaining a univer-
sity degree. There was also evidence of a negative associ-
ation between adolescent MRB and occupational status
at age 34 (Adjusted OR 0.86 95% CI: 0.82, 0.90). There
was little change in the association between adolescent
MRB and young adult SES when including early life SES
variables as confounders (Table 3).
All three parental SES variables were negatively associ-
ated with both young adult SES variables. Parents having
low SES was associated with offspring not attaining a
university degree. For instance, low maternal education
reduced the odds of young people attaining a university
degree by 62% (OR 0.38 95% CI: 0.34, 0.43). All associa-
tions in the complete case sample were comparable to
those in the imputed data (Supplementary File 1).
ALSPAC
The direction of the association was the same in the
ALSPAC cohort (Table 3), with similar odds ratios to
BCS70 (Unadjusted OR: 0.83, 95% CI: 0.81, 0.86, Ad-
justed OR: 0.85 95% CI: 0.82, 0.88). In ALSPAC, MRB
was made up of more risk behaviours (13 behaviours),
however, the sensitivity analyses which modelled the as-
sociation with only eight behaviours yielded similar re-
sults (Adjusted OR: 0.84, 95% CI: 0.80, 0.89).
In the ALSPAC cohort, all three early life SES vari-
ables were associated with young adult SES (university
degree attainment). Parents having low SES was associ-
ated with offspring not attaining a university degree.
Low maternal education reduced the odds of attaining a
university degree by 51% (Adjusted OR 0.49, 95% CI:
0.42, 0.56). There was little change in the association be-
tween adolescent MRB and young adult SES when in-
cluding early life SES variables as confounders (Table 3).
All associations in the complete case were comparable
to those in the imputed data (Supplementary File 1).
Moderation analysis
BCS70
Figure 1and Table 4presents estimated associations be-
tween adult SES and adolescent MRB by maternal edu-
cation. With increasing MRB scores, both early life SES
groups had decreasing predictions of university degree
attainment, however, for the low SES group the impact
was greater as illustrated by the steeper gradient. Table 4
confirms that in the BCS70 sample there was evidence
that maternal education (p= 0.03), parental occupational
status (p= 0.009) and household income (p= 0.03) mod-
erated the linear association between adolescent MRB
and young adult SES (degree attainment at age 26). This
means the negative effect of MRB on young adult degree
attainment is stronger for those with low socioeconomic
backgrounds. For the complete case sample there was
evidence of a moderation effect for parental occupation
only (Supplementary File 1). For the secondary outcome,
there was evidence that parental occupational status
moderated the relationship between adolescent MRB
and occupational status at age 34 (p= 0.01) (Supplemen-
tary File 1).
Tinner et al. BMC Public Health (2021) 21:1614 Page 6 of 13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Table 2 Descriptive statistics for imputed and complete case samples
BCS70 ALSPAC
n Complete case (n=
1358) (%)
Imputed sample (n =
9691) (SE)
n Complete case (n=
1360) n (%)
Imputed sample (n=
9001) % (SE)
Young Adult SES Degree
attainment (outcome)
6193 4425
High SES 499 (36.8%) 25.5% (0.6%) 871 (64.0%) 51.8% (0.8%)
Low SES 859 (63.3%) 74.8% (0.6%) 489 (36.0%) 48.2% (0.8%)
Young Adult SES occupation
status (outcome)
5158
High SES 742 (54.6%) 60.5% (0.5%) –3.20 (0.03)
Low SES 616 (45.4%) 39.5% (0.5%) ––
MRB total (exposure) 5805 2345
Mean (SD) 1.74 (1.4) 1.88 (0.02) 2.91 (1.9) 3.20 (0.03)
Median (IQR) 2 (1–3) –3(1–4) –
Maternal Education SES
(moderator)
9691 9001
High SES 567 (41.8%) 29.3% (0.5%) 697 (51.3%) 39.8% (0.5%)
Low SES 791 (58.3%) 70.7% (0.5%) 663 (48.8%) 60.2% (0.5%)
Household Equivalised Income
(moderator)
9691 9001
High SES (higher income
brackets)
209 (15.4%) 22.5% (0.4%) 694 (51.0%) 42.1% (0.5%)
Low SES (lower income
brackets)
1149 (84.6%) 77.5% (0.4%) 666 (49.0%) 57.9% (0.5%)
Parental social class (moderator) 9691 9001
High SES 338 (24.9%) 25.3% (0.4%) 908 (66.8%) 57.4% (0.5%)
Low SES 1020 (75.1%) 74.7% (0.4%) 452 (33.2%) 42.6% (0.5%)
Gender 9691 9001
Female 779 (57.4%) 51.7% (0.5%) 840 (61.8%) 48.8% (0.5%)
Male 579 (42.6%) 48.3% (0.5%) 520 (38.2%) 51.2% (0.5%)
Season of birth 9001
Autumn –– – 456 (33.5%) 33.0% (0.5%)
Winter –– – 186 (13.7%) 14.1% (0.5%)
Spring –– – 329 (24.2%) 23.2% (0.5%)
Summer –– – 389 (28.6%) 29.7% (0.5%)
Previous educational
attainment/ability
IQ at age 8 Mean (SD) –– – 5841 109.69 (15.0) 103.5 (0.2)
KS2 educational attainment
Mean (SD)
–– – 6533 913.84 (154.6) 800.6 (2.6)
Conduct problems score (age
10 years)
9083 6007
0 1144 (84.2%) 80.0% (0.4%) 983 (72.3%) 67.4 (0.6%)
1 178 (13.1%) 15.2% (0.4%) 316 (23.2%) 24.9 (0.6%)
2 36 (2.7%) 4.8% (0.2%) 61 (4.5%) 7.7% (0.5%)
Tinner et al. BMC Public Health (2021) 21:1614 Page 7 of 13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
ALSPAC
These same models applied to the imputed data in the
ALSPAC cohort revealed no evidence of a moderation
effect for maternal education (p= 0.83), parental occupa-
tion (p= 0.72) or household income (p= 0.31) (Fig. 2
and Table 4). There was no evidence of a moderation ef-
fect in the complete case (Supplementary File 1).
Discussion
Main findings
Adolescent MRB was negatively associated with young
adult degree attainment in mid-twenties. The association
was present in both cohorts. With an incremental in-
crease of each additional risk behaviour, young people
had reduced odds of attaining a degree in young adult-
hood. MRB was also strongly associated with occupa-
tional status at age 34 in the BCS70 cohort. There was
evidence of a moderation effect in all three models for
the primary outcome in the BCS70 cohort but no
evidence of a moderation effect in the ALSPAC cohort.
Therefore, for BCS70 participants the negative effect of
adolescent MRB upon university degree attainment was
stronger for young people from low socioeconomic
backgrounds. For ALSPAC participants, however, the as-
sociation between MRB and university degree attain-
ment was the same regardless of early life SES.
Strengths and limitations
This is the first longitudinal study to examine adolescent
MRB and young adult SES in a UK. We also compared
two birth cohorts 20 years apart and were able to derive
a comparable adolescent MRB variable containing a wide
range of health risk behaviours. There are several limita-
tions. First, both cohort studies suffer from high levels of
attrition. Further, non-participation and loss to follow-
up is usually more pronounced among less advantaged
and less healthy groups, potentially leading to bias and
underestimation of inequalities [34]. To account for so-
cially patterned attrition, we used imputed datasets de-
rived from data collected since recruitment to maximise
the chances that the assumption that observations are
missing at random is satisfied. However, the method for
multiple imputation may have impacted on the results.
Imputing the data by early life SES stratum was neces-
sary in order to impute each individual risk behaviour
that made up the MRB variable while also preserving a
potential moderation effect of early life SES on the asso-
ciation between MRB and young adult SES. As the early
life SES variables by which the data were split were also
strong predictors of the outcome variable, which had a
large proportion missing, the binary outcome variable
(young adult university degree) was imputed using other
auxiliary variables that may not have been as
informative.
A further limitation is the risk behaviours that made
up the MRB variables were self-reported and reduced to
binary variables, with engagement was based on cut-offs
informed by the literature, which may have influenced
whether an association was found [34]. The derivation of
the MRB variable was based on literature and advice
Table 3 Associations between adolescent unit increase in MRB score and SES variables in young adulthood
BCS70 (N= 9691) ALSPAC (N= 9001)
Unadjusted OR (95% CI)
p-value
Adjusted
a
OR (95% CI)
p-value
Unadjusted OR (95% CI)
p-value
Adjusted OR
a
(95% CI)
p-value
Outcome variables
b
Young adult degree attainment in
mid-twenties
0.80 (0.75, 0.84) p< 0.001 0.81 (0.76, 0.86) p< 0.001 0.83 (0.81, 0.86) p< 0.001 0.85 (0.82, 0.88) p< 0.001
Occupational status at age 34 years 0.84 (0.81, 0.88) p< 0.001 0.86 (0.82, 0.90) p< 0.001 ––
a
Models in BCS70 were adjusted for sex and conduct score at age 10 and models in ALSPAC were adjusted for sex, IQ score age 8, conduct score at age 10, Key
Stage 2 score and season born
b
OR are presented indicating the odds of the outcome for each incremental single behaviour out of a possible eight behaviours for BCS70 and thirteen
behaviours for ALSPAC
Degree
No degree
0 1 2 3 4 5 6 7 8
MRB score
High maternal education Low maternal education
Fig. 1 Predicted values of young adult education at each level of
MRB, stratified by maternal education (n=9,691). Legend: Each line
represents the association predicted values of the outcome (young
adult degree attainment) at each level of MRB, with blue denoting
the high maternal education group and red the low maternal
education group. The shaded area around the lines represents the
95% confidence intervals around the predicted values
Tinner et al. BMC Public Health (2021) 21:1614 Page 8 of 13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
from young people’s groups, informing a body of MRB
research in ALSPAC. There might also be differences in
how the variables were reported between each cohort
due to changes in data collection methods, language in
questions and socially acceptable responses [16]. MRB
treats all risk behaviours as equally weighted, yet out-
comes may differ across the behaviours and this might
be historically dependent. Reducing the number of risk
behaviours in ALSPAC to the same eight behaviours in
BCS70 had little effect on the regression results but did
result in a lower mean and median average of risk be-
haviours engaged in among the ALSPAC participants.
The young adult SES variable was also self-reported as
higher education data linkage was unavailable at the
time of this study. It is possible that individuals aged 21
to 24 years may not have finished their education, but
most will have started it given that the vast majority of
individuals applying to UK universities are under age 21
[35]. Participants that were ‘currently studying’for an
undergraduate degree or for a higher degree were in-
cluded in the ‘high SES’category, given the non-
continuation rate of university students by year 1 is low
at 6.3% [36] and the social patterning of university appli-
cations [37]. However, this non-continuation rate is so-
cially patterned, with a greater proportion of young
people from low income backgrounds not continuing
their degree beyond the first year (8.3%) [36], therefore,
including these observations should be noted as a limita-
tion of the dataset. Further, the early life SES variables
were also self-reported. We chose to examine the mod-
erators in separate analyses, but using a composite meas-
ure of SES as the moderator may have produced
different results. Future research may consider this
approach.
Comparability between the cohorts more generally is
also limited given that BCS70 is representative of the
UK, whereas ALSPAC is confined to southwest England.
As the ALSPAC participants were between 21 and 24
years-old when the SES outcome variable was measured,
it is possible that these young people do not yet have a
settled measure of SES.
Table 4 Logistic regression of young adult SES on MRB score, stratified by early life SES variables
BCS (n=9691)
a
ALSPAC (n=9001)
a
All
participants
High origin
SES
Low origin
SES
Pvalue for
moderation
c
All
participants
High origin
SES
Low origin
SES
Pvalue for
moderation
c
Maternal education SES
b
MRB unit increase
(0-8) and (0-13)
1 (REF) 1 (REF) 1 (REF) 1 (REF) 1 (REF) 1 (REF)
0.81 (0.77,
0.86) p<0.001
0.87 (0.80,
0.94) p=0.001
0.77 (0.71,
0.84) p<0.001
0.03 0.85 (0.82,
0.89) p<0.001
0.85 (0.80,
0.90) p<0.001
0.84 (0.81,
0.89) p<0.001
0.83
Parent occupation SES
b
MRB unit increase
(0-8) and (0-13)
1 (REF) 1 (REF) 1 (REF) 1 (REF) 1 (REF) 1 (REF)
0.81 (0.77,
0.86) p<0.001
0.91 (0.83,
1.01) p=0.06
0.78 (0.73,
0.83) p<0.001
0.009 0.85 (0.82,
0.89) p<0.001
0.85 (0.82,
0.90) p<0.001
0.84 (0.79,
0.90) p<0.001
0.72
Household income SES
b
MRB unit increase
(0-8) and (0-13)
1 (REF) 1 (REF) 1 (REF) 1 (REF) 1 (REF) 1 (REF)
0.81 (0.77,
0.86) p<0.001
0.87 (0.81,
0.94) p=0.001
0.77 (0.72,
0.84) p<0.001
0.03 0.85 (0.82,
0.22) p<0.001
0.87 (0.82,
0.92) p<0.001
0.85 (0.80,
0.89) p<0.001
0.31
a
Models in BCS70 were adjusted for sex and conduct score at age 10 and sex, IQ score age 8, conduct score at age 10, Key Stage 2 score and season born
for ALSPAC
b
Early life SES variables are binary variables, with high SES as the reference category
c
Likelihood ratio test p-values are presented, with p≤0.05 taken as evidence of difference between the groups and thus a moderation effect. OR are presented
indicating the odds of the outcome for each incremental single behaviour out of a possible eight behaviours for BCS70 and thirteen behaviours for ALSPAC,
stratified by origin SES
Degree
No degree
0 1 2 3 4 5 6 7 8 9 10 11 12 13
MRB score
High maternal education Low maternal education
Fig. 2 Predicted values of young adult education at each level of
MRB, stratified by maternal education (n=9,001). Legend: Each line
represents the association predicted values of the outcome (young
adult degree attainment) at each level of MRB, with blue denoting
the high maternal education group and red the low maternal
education group. The shaded area around the lines represents the
95% confidence intervals around the predicted values
Tinner et al. BMC Public Health (2021) 21:1614 Page 9 of 13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
The potential impact of confounding bias should be
noted as a limitation. There are a number of covariates
we could have included as confounders, such as adverse
childhood experiences (ACEs). However, when our study
was conceptualised, we set out to include confounders
consistent with a similar study in ALSPAC using MRB
and an education outcome [4]. There were also limited
comparable ACEs variables between the cohorts, with
ALSPAC having developed an ACEs score and BCS70
using socioeconomic variables such as income and over-
crowded housing as indicators of ACEs. In future studies
we will also include parental risk behaviours as potential
confounders, given that recent epidemiological research
in ALSPAC has included these [5]. Another limitation
and one potential reason for the differential results be-
tween the cohorts is the presence of residual confound-
ing [38], which is a limitation inherent in observational
epidemiology. Even large scale studies with ‘substantial
adjustment may still be affected by residual confounding’
[39]. Of particular relevance to our study, the confound-
ing effects of unmeasured genetic factors may have im-
pacted on the results. Future research could use
Mendelian Randomisation or generate polygenic scores,
which has already been done in ALSPAC to investigate
mental health, individual traits and substance use [40].
Other evidence
While it has long been understood that adolescent
health risk behaviours have potentially a short and long
term negative impact on health (e.g. unprotected sex
putting young people at risk of sexually transmitted in-
fections, or in the long term, physical inactivity contrib-
uting to obesity later in life), there has been less research
on social indicators of adult functioning such as educa-
tion and employment [3]. Our finding that adolescent
MRB is associated with lower educational attainment in
young adulthood builds upon some of this research that
has determined adolescent MRB to have negative socio-
economic outcomes later in life [3,5,15]. This study
also mirrors findings from ALSPAC, which show a de-
crease in GCSE points with each additional health risk
behaviour [4]. One reason this negative association
might occur could be due to engagement in risk behav-
iours leading to disengagement in school [3]or‘getting
side-tracked’, a phrase used by one of the ALSPAC par-
ticipants in the qualitative study undertaken simultan-
eously to this research [41]. While some engagement in
health risk behaviours during adolescence may expected
or even beneficial [15], our research further highlights
that adolescent MRB is a public health problem that has
the potential for negative consequences for young people
as they enter adulthood.
The lack of evidence for a moderation effect in ALSP
AC suggests that the effect of MRB is the same for
young people in the cohort regardless of their SES back-
ground. This result is, however, is challenging to the
wealth of evidence that people from low SES back-
grounds suffer a greater burden of poor outcomes [17],
as is shown in the BCS70 cohort. Furthermore, previous
research has found SES to moderate the relationship be-
tween obesity and depressive symptoms in young people
[42], health cognition and health behaviour [43], as well
parenting practices and adolescent drinking behaviour
[44]. It is also an unexpected finding given that inequal-
ities are thought to have increased since 1970s [17].
However, the results show that early life SES is associ-
ated with young adult degree attainment, meaning if
parents have low SES their children’s odds of attaining a
university degree are lower. Therefore, even though
there was no evidence to support the hypothesis that
MRB has a greater impact on those from low SES back-
grounds, it appears that there is intergenerational trans-
mission of socioeconomic inequalities within this cohort.
Given that more young people are now attending uni-
versity, one explanation for the divergent result between
the cohorts is that the outcome measure of university
degree attainment is not nuanced enough to explain so-
cial inequalities in contemporary young people. There-
fore, while the moderating impact is less apparent for
the younger cohort, there are still socioeconomic in-
equalities operating in terms of attendance at higher sta-
tus Russell Group universities, for instance, which in
turn comes with the increased likelihood of achieving a
high status occupation and high income [45]. Therefore,
we would be cautious of suggesting that society is more
equal today than when the BCS70 participants were
young adults. Instead, we propose further research to
develop a measure that captures the range and com-
plexity of young adult SES is needed to address this
issue. This young adult socioeconomic measure could
be a composite incorporating different income, em-
ployment and educational markers as well as subject-
ive social status, which has been developed in the
past for adolescents [10].
We know adult SES is strongly associated with a range
of morbidities and mortality [17], therefore determining
the impact of adolescent MRB upon young adult SES is
instructive. It provides further rationale for intervening
on adolescent MRB, with the aim of being more cost-
effective and efficient than single behaviour interven-
tions. Recent work has shown the success of universal
school-based interventions in addressing adolescent
MRB [46], thus, there is already a suite of interventions
that could improve health and social outcomes.
Conclusion
Adolescent multiple risk behaviour was negatively asso-
ciated with young adult SES, measured by university
Tinner et al. BMC Public Health (2021) 21:1614 Page 10 of 13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
degree attainment in mid-twenties and occupational sta-
tus age 34. The association was similar in two birth co-
horts 20 years apart. Adolescence appears to be a critical
time in the life course to address risk behaviours, due to
the likelihood that behaviours established here may have
effects in adulthood. Each additional health risk behav-
iour was associated with a reduction in odds of achieving
a university degree, consistent with previous work that
highlighted the importance of intervening on each and
every risk behaviour [4]. Evidence for a moderation ef-
fect in the BCS70 but not ALSPAC suggests that other
measures should be investigated to capture the complex-
ity of contemporary young adult SES.
Abbreviations
MRB: Multiple Risk Behaviour; SES: Socioeconomic Status; BCS70: British
Cohort Study 1970; ALSPAC: The Avon Longitudinal Study of Parents and
Children; GCSE: General Certificate of Secondary Education; N: Number;
CI: Confidence Interval; OR: Odds ratio
Supplementary Information
The online version contains supplementary material available at https://doi.
org/10.1186/s12889-021-11638-3.
Additional file 1.
Acknowledgments
This study is sponsored by the University of Bristol. We are extremely grateful
to all the families who took part in this study, the midwives for their help in
recruiting them, and the whole ALSPAC team, which includes interviewers,
computer and laboratory technicians, clerical workers, research scientists,
volunteers, managers, receptionists and nurses. We thank the Centre for the
Development and Evaluation of Complex Interventions for Public Health
Improvement (DECIPHer) Advice leading to Public Health Advancement
(ALPHA) young person’s research advisory group for informing the
development of the MRB variable.
Authors’contributions
LT, RC, MH and DC developed the research project. LT conducted the data
analysis with the support of JH, CW, DC and MH. LT wrote the first draft of
the manuscript. All authors read and commented on the manuscript before
submission. LT had full access to the data and final responsibility for the
decision to submit the publication. The authors read and approved the final
manuscript.
Funding
LT is a PhD student funded by the Medical Research Council within the
Centre for the Development and Evaluation of Complex Public Health
Interventions (DECIPHer). CW is funded by a Cancer Research UK Population
Research Postdoctoral Fellowship (C60153/A23895). The work was
undertaken with the support of The Centre for the Development and
Evaluation of Complex Interventions for Public Health Improvement
(DECIPHer), a UKCRC Public Health Research Centre of Excellence. Joint
funding (MR/KO232331/1) from the British Heart Foundation, Cancer
Research UK, Economic and Social Research Council, Medical Research
Council, the Welsh Government and the Wellcome Trust, under the auspices
of the UK Clinical Research Collaboration, is gratefully acknowledged. The
funding sources had no involvement with the study design, collection,
analysis, writing the report or the decision to submit for publication. The UK
Medical Research Council and Wellcome (Grant ref.: 102215/2/13/2) and the
University of Bristol provide core support for ALSPAC. This publication is the
work of the authors and we will serve as guarantors for the contents of this
paper.
Availability of data and materials
BCS70 Cohort data comply with ESRC data sharing policies, readers can
access data via the UK Data Archive (www.data-archive.ac.uk), through a
formal request. This data set is open to the public.
ALSPAC Data used for this submission will be made available on request to
the Executive (alspac-exec@bristol.ac.uk). The ALSPAC data management
plan (available here: http://www.bristol.ac.uk/alspac/researchers/data-access/
documents/alspac-data-management-plan.pdf) describes in detail the policy
regarding data sharing, which is through a system of managed open access.
Please note that the ALSPAC study website contains details of all the data
that is available through a fully searchable data dictionary and variable
search tool:
http://www.bristol.ac.uk/alspac/researchers/our-data/
Study data were collected and managed using REDCap electronic data
capture tools hosted at University of Bristol [47]. This dataset is closed to the
public, with our permission access statement appearing in the ‘ethics
approval and consent to participant’section.
Declarations
Ethics approval and consent to participate
This study used secondary data analysis and the authors did not collect any
primary data for this work. The ethics statements for the birth cohorts used
in this study are reported here.
BCS70: All procedures were approved by the London Multi-Centre Research
at the 2000 wave (process 98/2/120) and by an internal committee from the
Centre for Longitudinal Studies, Institute of Education, University of London,
for the 2004 wave. Participants provided verbal consent during the inter-
views. Before the interviews, study members were sent an advance letter ad-
vising them about the survey. The letter was accompanied by detailed
information about the survey and the cohort members were free to request
further information, or to opt out of the survey at this point. Also, the cohort
members could request further information or refuse involvement during all
the survey process, including when the interviewer attempted to make an
appointment to visit, when the interviewer visited and at any point during
the administration of any elements of the surveys. The verbal consent was
approved as it was a routine since the beginning of the cohort, in 1970.
ALSPAC: Ethical approval for the study was obtained from the Avon
Longitudinal Study of Parents and Children Ethics and Law Committee and
local Research Ethics Committees: Bristol and Weston Health Authority:
E1808 Children of the Nineties: Avon Longitudinal Study of Pregnancy and
Childhood (ALSPAC). (28th November 1989), Southmead Health Authority:
49/89 Children of the Nineties - “ALSPAC”. (5th April 1990), Frenchay Health
Authority: 90/8 Children of the Nineties. (28th June 1990), 15 Year Clinic:
Central & South Bristol Research Ethics Committee (UBHT): 06/Q2006/53
Avon Longitudinal Study of Parents and Children (ALSPAC), Hands on
Assessments: Teen Focus 3 (Focus 15+). (7th August 2006) (Confirmed 15th
September 2006) and 17 Year clinic: North Somerset & South Bristol Research
Ethics Committee: 08/H0106/9 Avon Longitudinal Study of Parents and
Children (ALSPAC), Hands on Assessments: Teen Focus 4 (Focus 17+) (18th
November 2008) National Research Ethics Service Committee South West –
Frenchay: 14/SW/1173 ALSPAC Focus at 24+ (24th February 2015, confirmed
20th March 2015). Informed consent for the use of data collected via
questionnaires and clinics was obtained from participants following the
recommendations of the Avon Longitudinal Study of Parents and Children
Ethics and Law Committee at the time.
Study participants who complete questionnaires consent to the use of their
data by approved researchers. Up until age 18 an overarching parental
consent was used to indicate parents were happy for their child (the study
participant) to take part in ALSPAC. Consent for data collection and use was
implied via the written completion and return of questionnaires. Study
participants have the right to withdraw their consent for specific elements of
the study, or from the study as a whole, at any time.
Ethical approval for administrative permission to access the data was granted
as part of a larger programme of work on multiple risk behaviour by the
ALSPAC Law and Ethics Committee (ALEC) for study number B1369.
Consent for publication
Not applicable.
Tinner et al. BMC Public Health (2021) 21:1614 Page 11 of 13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Competing interests
The authors declare no competing interests.
Received: 14 August 2020 Accepted: 18 August 2021
References
1. Kipping RR, Smith M, Heron J, Hickman M, Campbell R. Multiple risk
behaviour in adolescence and socio-economic status: findings from a UK
birth cohort. Eur J Public Health. 2015;25(1):44–9. https://doi.org/10.1093/
eurpub/cku078.
2. Brooks FM, Magnusson J, Spencer N, Morgan A. Adolescent multiple risk
behaviour: an asset approach to the role of family, school and community. J
Public Health. 2012;34(suppl 1):i48–56. https://doi.org/10.1093/pubmed/
fds001.
3. Hair EC, Park MJ, Ling TJ, Moore KA. Risky behaviors in late adolescence: co-
occurrence, predictors, and consequences. J Adolesc Health. 2009;45(3):253–
61. https://doi.org/10.1016/j.jadohealth.2009.02.009.
4. Wright C, Kipping R, Hickman M, Campbell R, Heron J. Effect of multiple risk
behaviours in adolescence on educational attainment at age 16 years: a UK
birth cohort study. BMJ Open. 2018;8(7):e020182. https://doi.org/10.1136/
bmjopen-2017-020182.
5. Campbell R, Wright C, Hickman M, Kipping RR, Smith M, Pouliou T, et al.
Multiple risk behaviour in adolescence is associated with substantial adverse
health and social outcomes in early adulthood: findings from a prospective
birth cohort study. Prev Med. 2020;138:106157. https://doi.org/10.1016/j.
ypmed.2020.106157.
6. Biglan A. Helping adolescents at risk: prevention of multiple problem
behaviors. London: Guilford Press; 2004.
7. Karvonen S, Rimpelä AH, Rimpelä MK. Social mobility and health related
behaviours in young people. J Epidemiol Community Health. 1999;53(4):
211–7. https://doi.org/10.1136/jech.53.4.211.
8. Blane D, Smith GD, Bartley M. Social selection: what does it contribute to
social class differences in health? Sociol Health Illness. 1993;15(1):1–15.
https://doi.org/10.1111/1467-9566.ep11343777.
9. Viner R, Hargreaves D, Varnes L, Heys M. The social determinants of young
people’s health: identifying the key issues and assessing how young people
are doing in the 2010s. London: The Health Foundation Working Paper;
2018.
10. Currie C, Molcho M, Boyce W, Holstein B, Torsheim T, Richter M. Researching
health inequalities in adolescents: the development of the health behaviour
in school-aged children (HBSC) family affluence scale. Soc Sci Med. 2008;
66(6):1429–36. https://doi.org/10.1016/j.socscimed.2007.11.024.
11. Hasselberg M, Vaez M, Laflamme L. Socioeconomic aspects of the
circumstances and consequences of car crashes among young adults. Soc
Sci Med. 2005;60(2):287–95. https://doi.org/10.1016/j.socscimed.2004.05.006.
12. Bambra C, Thomson K. Reducing inequities in health across the life-course:
transition to independent living–young adults. In: European Health Equity
Status Report initiative; 2019.
13. Thrane C. Explaining educational-related inequalities in health: mediation
and moderator models. Soc Sci Med. 2006;62(2):467–78. https://doi.org/10.1
016/j.socscimed.2005.06.010.
14. Maggs JL, Frome PM, Eccles JS, Barber BL. Psychosocial resources,
adolescent risk behaviour and young adult adjustment: is risk taking more
dangerous for some than others? J Adolesc. 1997;20(1):103–19. https://doi.
org/10.1006/jado.1996.0067.
15. Viner RJ. Co-occurrence of adolescent health risk behaviors and outcomes
in adult life: findings from a national birth cohort. J Adolesc Health. 2005;
36(2):98–9.
16. Shackleton N, Hale D, Viner RM. Trends and socioeconomic disparities in
preadolescent's health in the UK: evidence from two birth cohorts 32 years
apart. J Epidemiol Community Health. 2016;70(2):140–6. https://doi.org/1
0.1136/jech-2015-205603.
17. Marmot M, Atkinson T, Bell J, Black C, Broadfoot P, Cumberlege J, et al. Fair
society, healthy lives: The marmot review. 2010.
18. Tinner L, Wright C, Heron J, et al. P32 Is adolescent multiple risk behaviour
associated with reduced socioeconomic status in young adulthood and do
those with low socioeconomic backgrounds experience greater negative
impact? Findings from two UK birth cohorts. J Epidemiol Community
Health. 2020;74:A59.
19. Elliott J, Shepherd P. Cohort profile: 1970 British birth cohort (BCS70). Int J
Epidemiol. 2006;35(4):836–43. https://doi.org/10.1093/ije/dyl174.
20. Mostafa T, Wiggins R. The impact of attrition and non-response in birth
cohort studies: a need to incorporate missingness strategies. Longitud Life
Course Stud. 2015;6(2):131–46.
21. Fraser A, Macdonald-Wallis C, Tilling K, Boyd A, Golding J, Davey Smith G,
et al. Cohort profile: the Avon longitudinal study of parents and children:
ALSPAC mothers cohort. Int J Epidemiol. 2012;42(1):97–110.
22. Boyd A, Golding J, Macleod J, Lawlor DA, Fraser A, Henderson J, et al.
Cohort profile: the ‘children of the 90s’—the index offspring of the Avon
longitudinal study of parents and children. Int J Epidemiol. 2013;42(1):111–
27. https://doi.org/10.1093/ije/dys064.
23. Wright C, Heron J, Campbell R, Hickman M, Kipping RR. Adolescent multiple
risk behaviours cluster by number of risks rather than distinct risk profiles in
the ALSPAC cohort. BMC public health. 2020;20(1):1–8.
24. Rahkonen O, Arber S, Lahelma E. Health inequalities in early adulthood: a
comparison of young men and women in Britain and Finland. Soc Sci Med.
1995;41(2):163–71. https://doi.org/10.1016/0277-9536(94)00320-S.
25. Galobardes B, Shaw M, Lawlor DA, Lynch JW, Smith GD. Indicators of
socioeconomic position (part 1). J Epidemiol Community Health. 2006;60(1):
7–12. https://doi.org/10.1136/jech.2004.023531.
26. Kraemer HC, Stice E, Kazdin A, Offord D, Kupfer D. How do risk factors work
together? Mediators, moderators, and independent, overlapping, and proxy
risk factors. Am J Psychiatr. 2001;158(6):848–56. https://doi.org/10.1176/a
ppi.ajp.158.6.848.
27. Aiken LS, West SG, Reno RR. Multiple regression: testing and interpreting
interactions: sage; 1991.
28. Hughes RA, Heron J, Sterne JAC, Tilling K. Accounting for missing data in
statistical analyses: multiple imputation is not always the answer. Int J
Epidemiol. 2019;48(4):1294–304. https://doi.org/10.1093/ije/dyz032.
29. Madley-Dowd P, Hughes R, Tilling K, Heron J. The proportion of missing
data should not be used to guide decisions on multiple imputation. J Clin
Epidemiol. 2019;110:63–73. https://doi.org/10.1016/j.jclinepi.2019.02.016.
30. Sterne JA, White IR, Carlin JB, Spratt M, Royston P, Kenward MG, et al.
Multiple imputation for missing data in epidemiological and clinical
research: potential and pitfalls. BMJ. 2009;338(jun29 1):b2393. https://doi.
org/10.1136/bmj.b2393.
31. White IR, Royston P, Wood AM. Multiple imputation using chained
equations: issues and guidance for practice. Stat Med. 2011;30(4):377–99.
https://doi.org/10.1002/sim.4067.
32. Rubin DB. Multiple imputation for nonresponse in surveys: Wiley; 2004.
33. Royston P. Multiple imputation of missing values: update of ice. Stata J.
2005;5(4):527–36. https://doi.org/10.1177/1536867X0500500404.
34. Howe LD, Tilling K, Galobardes B, Lawlor DA. Loss to follow-up in cohort
studies: bias in estimates of socioeconomic inequalities. Epidemiology. 2013;
24(1):1.
35. UCAS. UCAS undergraduate sector-level end of cycle data resources 2020.
Explore this section. 2020. Available from: https://www.ucas.com/data-and-a
nalysis/undergraduate-statistics-and-reports/ucas-undergraduate-sector-
level-end-cycle-data-resources-2020.
36. HESA. Non-continuation summary: UK performance indicators. 2019. Available
from: https://www.hesa.ac.uk/news/07-03-2019/non-continuation-summary.
37. Machin S, Vignoles A. Educational inequality: the widening socio-economic
gap. Fisc Stud. 2004;25(2):107–28.
38. Vandenbroucke JP, Von Elm E, Altman DG, Gøtzsche PC, Mulrow CD, Pocock
SJ, et al. Strengthening the reporting of observational studies in epidemiology
(STROBE): explanation and elaboration. Ann Intern Med. 2007;147(8):163.
39. Brion M-JA, Lawlor DA, Matijasevich A, Horta B, Anselmi L, Araújo CL, et al.
What are the causal effects of breastfeeding on IQ, obesity and blood
pressure? Evidence from comparing high-income with middle-income cohorts.
Int J Epidemiol. 2011;40(3):670–80. https://doi.org/10.1093/ije/dyr020.
40. Iob E, Schoeler T, Cecil CM, Walton E, McQuillin A, Pingault JB. Identifying
risk factors involved in the common versus specific liabilities to substance
use: A genetically informed approach. Addict Biol. 2020;26:e12944.
41. Tinner L, Caldwell D, Hickman M, Campbell R. Understanding adolescent
health risk behaviour and socioeconomic position: a grounded theory study
of UK young adults. Sociol Health Illness. 2021;43(2):528–44. https://doi.org/1
0.1111/1467-9566.13240.
42. Beydoun MA, Wang Y. Pathways linking socioeconomic status to obesity
through depression and lifestyle factors among young US adults. J Affect
Disord. 2010;123(1–3):52–63. https://doi.org/10.1016/j.jad.2009.09.021.
Tinner et al. BMC Public Health (2021) 21:1614 Page 12 of 13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
43. Conner M, McEachan R, Jackson C, McMillan B, Woolridge M, Lawton R.
Moderating effect of socioeconomic status on the relationship between
health cognitions and behaviors. Ann Behav Med. 2013;46(1):19–30. https://
doi.org/10.1007/s12160-013-9481-y.
44. Spijkerman R, van den Eijnden RJJM, Huiberts A. Socioeconomic differences
in alcohol-specific parenting practices and adolescents’drinking patterns.
Eur Addict Res. 2008;14(1):26–37. https://doi.org/10.1159/000110408.
45. Boliver V. How fair is access to more prestigious UK universities? Br J Sociol.
2013;64(2):344–64. https://doi.org/10.1111/1468-4446.12021.
46. MacArthur G, Caldwell DM, Redmore J, Watkins SH, Kipping R, White J, et al.
Individual-, family-, and school-level interventions targeting multiple risk
behaviours in young people. Cochrane Database Syst Rev. 2018;10:
CD009927.
47. Harris PA, Taylor R, Thielke R, Payne J, Gonzalez N, Conde JG. Research
electronic data capture (REDCap)—a metadata-driven methodology and
workflow process for providing translational research informatics support. J
Biomed Inform. 2009;42(2):377–81. https://doi.org/10.1016/j.jbi.2008.08.010.
Publisher’sNote
Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
Tinner et al. BMC Public Health (2021) 21:1614 Page 13 of 13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
1.
2.
3.
4.
5.
6.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:
use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
use bots or other automated methods to access the content or redirect messages
override any security feature or exclusionary protocol; or
share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at
onlineservice@springernature.com
Available via license: CC BY 4.0
Content may be subject to copyright.
