EARLY-LIFE PREDICTORS OF INTERNALIZING
SYMPTOM TRAJECTORIES IN CHILDREN
Murray Weeks, Ph.D. ,1 John Cairney, Ph.D. ,2,3,4 T. Cameron Wild, Ph.D. ,5 George B. Ploubidis, Ph.D. ,6
Kiyuri Naicker, MSc.,1 and Ian Colman, Ph.D.1,5*
1 Department of Epidemiology and Community Medicine, University of Ottawa, Ottawa, Canada; 2 Department of
Family Medicine, McMaster University, Hamilton, Canada; 3 Department of Psychiatry & Behavioural
Neurosciences, McMaster University, Hamilton, Canada; 4 Department of Clinical Epidemiology and Biostatistics,
McMaster University, Hamilton, Canada; 5 School of Public Health, University of Alberta, Edmonton, Canada;
6 Department of Population Health, Faculty of Epidemiology and Population Health, London School of Hygiene and
Tropical Medicine, London, UK.
Short title: Predictors of internalizing trajectories
Keywords: depression; anxiety; risk factors; development; epidemiology
Word count: Text: 3702; abstract: 249; Tables: 3; Figures: 1.
Conflict of Interest: All authors declare to have no conflict of interest.
* Address for correspondence: I. Colman, PhD, Department of Epidemiology and Community Medicine, University of Ottawa, 451 Smyth Road, RGN 3230C, Ottawa, ON, K1H 8M5, Canada. Tel: + 1 613 562 5800 ext. 8715; Fax: +1 613 562 5465; Email: [email protected]
1
ABSTRACT
Background: Previous research examining the development of anxious and depressive symptoms (i.e., internalizing
symptoms) from childhood to adolescence has often assumed that trajectories of these symptoms do not vary across
individuals. The purpose of this study was to identify distinct trajectories of internalizing symptoms from childhood
to adolescence, and to identify risk factors for membership in these trajectory groups. In particular, we sought to
identify risk factors associated with early appearing (i.e., child onset) symptoms versus symptoms that increase in
adolescence (i.e., adolescent onset).
Method: Drawing on longitudinal data from the National Longitudinal Survey of Children and Youth, latent class
growth modeling (LCGM) was used to identify distinct trajectories of internalizing symptoms for 6337 individuals,
from age 4-5 to 14-15. Multinomial regression was used to examine potential early-life risk factors for membership
in a particular trajectory group.
Results: Five trajectories were identified: ‘low stable’ (68%; reference group), ‘adolescent onset’ (10%), ‘moderate
stable’ (12%), ‘high childhood’ (6%), and ‘high stable’ (4%). Membership in the ‘adolescent onset’ group was
predicted by child gender (greater odds for girls), stressful life events, hostile parenting, aggression, and
hyperactivity. Membership in the ‘high stable’ and ‘high childhood’ trajectory groups (i.e., child-onset) was
additionally predicted by maternal depression, family dysfunction, and difficult temperament. Also, several
significant gender interactions were observed.
Conclusions: Causal mechanisms for child and adolescent depression and anxiety may differ according to time of
onset, as well as child gender. Some early factors may put girls at greater risk for internalizing problems than boys.
2
INTRODUCTION
Depression and anxiety are among the most common psychiatric problems among adolescents [1]. Often
collectively described as internalizing disorders, these psychiatric problems not only decrease young people’s
quality of life, but also represent a significant social and economic cost to society [2, 3]. Although most individuals
will not meet DSM criteria for internalizing disorders, sub-threshold internalizing symptoms are and have been
identified as risk factors for the development of subsequent internalizing symptoms and disorders [4, 5].
A number of studies have examined typical developmental patterns of child and adolescent internalizing
symptoms using growth curve modeling (i.e., latent curve analysis) [6] to estimate the overall rate of change of such
symptoms over time [7, 8]. However, such methods assume a common trajectory that represents the average rate of
change and thus are unsuitable for identifying different trajectories categories in a population where not all
individuals follow a common developmental course [9]. In contrast, latent class growth modeling (LCGM) is a
semi-parametric approach that is able to identify homogeneous subgroups of individuals showing distinct patterns of
change in internalizing symptoms over time [9]. Importantly, studies using LCGM have shown that some
individuals’ internalizing symptoms increase or decrease over time, whereas others exhibit steadily high or low
symptoms [10-16].
Although such studies have highlighted the potential value of examining psychiatric symptoms across the
life-course, as well as the apparent heterogeneity in the development of internalizing symptoms, there are
nevertheless gaps in our understanding of the development of such symptoms. For example, only a few studies have
identified trajectories of internalizing symptoms [10], whereas a number of studies have examined latent classes of
either depressive or anxious symptoms alone [15, 17], neglecting the considerable overlap often reported between
these types of symptoms. In fact, epidemiological studies suggest rates of comorbidity between 30% and 75% [18,
19] for mood and anxiety disorders. Also, recent studies have typically focused on the development of internalizing
symptoms either in childhood or in adolescence [8, 11, 12, 20], rather than examining the changes in these
symptoms across these developmental periods. As a result, relatively little is known about the factors early in life
that may influence the time of onset of these symptoms. In particular, it is unclear to what extent early-life factors
may differentially predict the emergence of depressive and anxious symptoms in childhood versus during
adolescence.
3
Based on previously identified risk factors for internalizing symptoms, it is possible that early-life factors
could help predict the timing of child versus adolescent internalizing symptom onset. For instance, internalizing
symptoms in childhood have been predicted by difficult child temperament [12, 21], maternal depression [11, 12,
17], family dysfunction and stress [8, 12], as well as negative parenting practices [11]. In contrast, internalizing
symptoms in adolescence have been predicted by delayed motor development [10], being female, lower SES [14],
and negative life events [22].
To our knowledge, the only studies examining the development of internalizing symptoms (i.e., trajectories
of combined anxious and depressive symptoms) from childhood into adolescence have examined average growth
curves (single trajectory), rather than identifying latent classes of symptom trajectories [23]. Therefore, it is unclear
whether risk factors can differentially predict child- versus adolescent-onset internalizing symptoms. Due to the
ability of LCGM to identify homogeneous subgroups represented by distinct trajectories of symptoms, this method
is particularly well suited to the identification of risk factors associated with the emergence of internalizing
symptoms in childhood versus adolescence. Importantly, if risk factors could differentiate between child-onset
versus adolescent-onset internalizing symptoms, this information could play a role in the development of programs
designed to reduce the occurrence and severity of these symptoms.
Aims of the Study
The goals of this study were to identify distinct groups of internalizing symptoms from childhood into
adolescence, and to identify early-life risk factors that may predict group membership in child- versus adolescent-
onset trajectories, in a large prospective cohort of Canadian children.
METHODS
This study was approved by the Research Ethics Board of the Ottawa Hospital. Statistics Canada obtained
written informed consent from parents of survey members (see below).
Sample
The National Longitudinal Study of Children and Youth (NLSCY) is a nationally representative
prospective cohort study of Canadians that is managed by Statistics Canada [24]. Data collection for the longitudinal
sample began in 1994/1995 and continued every two years until 2008/2009. In order to maximize sample size, our
sample was drawn from 6908 individuals who were age 2-3 years in Cycle 1 (1994/1995) or Cycle 2 (1996/1997) of
4
the NLSCY. Our final sample included N=6337 individuals who had information on internalizing symptoms for at
least one age group from age 4-5 to age 14-15. Overall response rates for Cycles 1-8 were 91.5%, 88.7%, 78.7%,
74.1%, 67.9%, 66.9%, and 62.0% respectively [25]. Statistics Canada has acknowledged a number of factors as
important for understanding NLSCY non-response, including privacy concerns, respondent burden, experience and
training of interviewers, and incentives to respondents [26]. In order to deal with survey non-response bias, all
survey responses were weighted using the appropriate sampling weights provided by Statistics Canada. These
weights were based on the inverse of each child’s probability of selection and are adjusted for non-response at each
cycle of the survey [25].
Measures
Internalizing symptoms. Internalizing symptoms were assessed using a questionnaire that included 7 items
from the Ontario Child Health Study (OCHS-R) [27], assessing symptoms of anxiety and depression. The items
were derived from the Child Behavior Checklist (CBCL) [28] because they appeared to operationalize DSM-III
criteria for emotional disorder [29]. The CBCL subscales are derived from DSM-III and DSM-IV criteria [30] and
have been extensively evaluated in terms of their psychometric properties [27, 31]. From ages 4 to 11, parents
reported on their child’s symptoms. The statements were, (my child) “seems to be unhappy, sad or depressed,” “is
not as happy as other children,” “is too fearful or anxious,” “is worried,” “cries a lot,” “is nervous, highstrung, or
tense,” and “has trouble enjoying him/herself”. From ages 12 to 15, these same items are self-reported, albeit with
slightly different wording (e.g., “I am unhappy or sad,” “I am too fearful or nervous”). Response options were:
“never or not true,” “sometimes or somewhat true,” or “often or very true.” Cronbach’s alphas of 0.79 and 0.81 have
been reported for the child and parent reports [25, 32]. In the current study, total scores were grouped into 4
categories representing degrees of severity. This was done in order to account for floor effects and variation in
severity of symptoms (i.e., positively skewed distributions), following previously reported procedures [33, 34]: no
symptoms (scores below the 50th percentile), low symptoms (scores between the 51st and 75th percentile), moderate
symptoms (scores between the 76th and 90th percentile), and severe symptoms (scores above the 90th percentile).
Early-life risk factors. A number of variables were examined as potential predictors of subsequent group
membership in internalizing trajectories from age 4-5 to age 14-15. Therefore, these variables were measured when
5
children were at or below age 4-5 (i.e., at baseline), with survey responses of the person most knowledgeable about
the child (PMK). This person was typically the mother.
Socioeconomic status. Socioeconomic status (SES) is operationalized in the NLSCY using household
income controlled for the ‘low-income cut off’ (LICO) score, which takes into account an individual’s income
relative to the community in which they live, as well as the size of their family [35]. For the current study, SES was
measured at age 2-3 and participants were considered to be ‘low SES’ if this value was in the bottom 10th percentile.
Physical, social and cognitive development. Children’s birth weight and gestational age were reported by
mothers when children were 0-3 years of age [25]. Using sex-specific Canadian population standards of birth weight
for gestational age [36], we classified individuals whose birth weight was in the bottom 10% of these standards as
being small for gestational age (SGA). At age 2-3, the PMK completed the Motor and Social Development (MSD)
Scale as a measure of ‘developmental milestones’. The MSD is a set of 15 ‘yes/no’ questions measuring aspects of
early motor, social, and cognitive development of young children (e.g., “has ____ ever spoken a partial sentence of 3
words or more?”). In the NLSCY, scores are based on the number of ‘yes’ answers, and MSD scores are then
standardized. For the current study, individuals were considered to have ‘low developmental milestones’ if their
score was in the bottom 10th percentile.
Sources of stress. Mothers were asked whether they had ever smoked or drank alcohol during pregnancy.
Other potential sources of stress included the presence of any chronic illnesses (e.g., diabetes) currently experienced
by the child at age 2-3, as well as the presence of 15 stressful life events (SLEs; e.g., death of a family member) that
children had experienced in the previous 24 months (from age 2-3 to age 4-5). Stressful life events were
dichotomized into ‘high stress’ (2 or more SLEs) versus ‘low stress’.
Parent and family characteristics. At child age 2-3, maternal depression was measured through self-
reported frequency and severity of symptoms over the past week through nine statements (e.g., “I felt depressed”).
Also at age 2-3, the PMK reported on the overall level of dysfunction in their family (e.g., “there are lots of bad
feelings in our family”), and the degree to which they employed a ‘hostile/ineffective’ parenting style (e.g., “How
often do you get angry when you punish your child?”) and ‘punitive/aversive’ parenting (e.g., “When ____ breaks
the rules or does things that he/she is not supposed to do, how often do you use physical punishment?”). These
variables were all dichotomized, where scores in the top 10th percentile represented ‘high’ scores.
6
Child behavior. At child age 2-3, the PMK rated the overall degree of difficulty their child would present
for the average parent (i.e., difficult temperament). Also at age 2-3, scales derived from the OCHS-R [27], CBCL
[28], and the Montreal Longitudinal Survey [37] were used to measure children’s level of ‘physical aggression –
opposition’ (e.g., “gets into many fights”) and ‘hyperactivity – inattention’ (e.g., “can’t sit still, is restless, or
hyperactive”). These variables were dichotomized using a 90th percentile cut-off, where scores in the top 10th
percentile represented ‘high’ scores.
Statistical Analysis
LCGM was performed using the PROC TRAJ application [38] in SAS 9.3 (SAS Institute Inc, Cary, North
Carolina, USA). Models with three, four, five groups were examined. Our choice of model was based on model
parsimony as measured by the Bayesian Information Criterion (BIC) value and posterior probabilities for group
membership. Missing data were handled in PROC TRAJ using full information maximum likelihood (FIML) under
a missing at random (MAR) assumption.
Multinomial logistic regression in SAS 9.3 was used to examine potential risk factors for membership in a
particular trajectory group versus a reference group (Table 2). Child gender and SES were included as covariates in
regression models.
RESULTS
Trajectories of internalizing symptoms
BIC value for the 3-, 4-, 5-, and 6-group models were -39116.33, -37916.34, -37529.36, and
-37588.45 respectively. Average posterior probabilities for the 5-group model were good (.80 - .90). These were
comparable to those of the 4-group model (.84 - .87). Trajectory plots suggested that one trajectory in the 4-group
model may have been combining two meaningful trajectories in the 5-group model. Also, the inclusion of a 6th group
resulted in the largest group from the 5-group model being split into two identical trajectory groups. Thus, the 5-
group model was chosen as the most parsimonious and best fitting model.
As seen in Figure 1, the identified trajectories are as follows: 1) consistently low levels of symptoms—
‘low stable’ (reference group; 67.5%); 2) low levels from age 4-5 to age 8-9 and then a sharp increase until age 14-
15 — ‘adolescent onset’ (9.8%); 3) high levels at age 4-5 followed by a sharp decline until age 8-9, at which point
7
the trajectory is similar to the ‘low stable’ group— ‘high childhood’ (7.4%); 4) mostly moderate levels from age 4-5
to age 14-15— ‘moderate stable’ (11.2%); and 5) consistently high levels throughout— ‘high stable’ (4.1%).
--------- Insert Figure 1 about here ----------
Risk factors for trajectory group membership
Proportions for the examined early-life risk factors are presented in Table 1. These are shown for the entire
sample, as well as stratified by child gender and by trajectory group. These proportions suggest a number of group
differences among trajectory groups. For instance, 2 or more SLEs, conduct/aggression, and
hyperactivity/inattention appeared to be less frequent in the ‘low stable’ group as compared to the other trajectory
groups.
--------- Insert Table 1 about here ----------
Predictors of group membership are described in Table 2. Several factors were associated with membership
in the ‘high stable’, ‘high childhood’, and ‘moderate stable’ trajectory groups (i.e., ‘child-onset’ trajectories) versus
the reference group. In particular, the odds of membership in the ‘high stable’ group were increased significantly in
the presence of maternal depression, family dysfunction, 2 or more SLEs, hostile/ineffective parenting,
conduct/aggression, hyperactivity/inattention, and difficult temperament. Predictors of the ‘high childhood’
trajectory included the presence of a chronic illness, maternal depression, 2 or more SLEs, hostile/ineffective and
punitive/aversive parenting, conduct/aggression, hyperactivity/inattention, and difficult temperament.
--------- Insert Table 2 about here ----------
Membership in the ‘adolescent onset’ trajectory was predicted by fewer early-life factors than either of the
‘child-onset’ trajectories. Most notably, membership in this trajectory was predicted by child gender. Girls had had
almost three times greater odds of being part of this group than boys. Other factors increasing the odds of
8
membership in this trajectory were 2 or more SLEs, hostile/ineffective parenting, as well as conduct/aggression and
hyperactivity/inattention.
Gender differences
Gender differences were further explored by examining interactions with predictor variables and child
gender. Most notably, there was evidence of interactions in the prediction of ‘adolescent onset’ group membership
(see Table 3). Specifically, presence of a chronic illness, punitive/aversive parenting, hostile/ineffective parenting,
as well as conduct/aggression and hyperactivity/inattention were associated with greater odds of ‘adolescent onset’
group membership for girls than for boys. Also, low birth weight was associated with greater odds of ‘adolescent
onset’ group membership for boys than for girls.
--------- Insert Table 3 about here ----------
Additional interactions by child gender emerged with regard to the remaining trajectories. Family
dysfunction was associated with greater odds of group membership in the ‘high childhood’ and ‘high stable’
trajectories for girls than for boys, and conduct/aggression increased the odds of ‘high childhood’ trajectory
membership to a greater extent for girls than for boys. Also, mother drinking and smoking during pregnancy both
were associated with greater odds of ‘moderate stable’ trajectory group membership for girls than for boys.
DISCUSSION
This large nationally representative study identified five distinct trajectories of internalizing symptoms
from childhood into adolescence. This supports the growing number of studies [10-16] identifying several distinct
classes of anxious and depressive symptom trajectories in young people. It is increasingly clear that growth models
that assume that a single trajectory of change adequately represents the population are often inappropriate, in that
they ignore the inherent heterogeneity in the development of a number of phenomena, including internalizing
symptoms. Our study also identified child- versus adolescent-onset symptom trajectories. We were thus able to
identify risk factors associated with symptoms in these two age periods, adding to the understanding of the early
developmental precursors of qualitatively different pathways of internalizing symptoms.
9
In general, early-life factors we examined were more predictive of child-onset than adolescent-onset
internalizing trajectories, suggesting that the early childhood environment may not be as important for predicting
adolescent anxiety and depression. However, previous studies have shown the importance of both early-life factors
[34, 39], as well as factors during late childhood and adolescence [22, 40], in predicting adolescent internalizing
symptoms. It is not clear why these previous results are at odds with those of the current study. One possible reason
is that these studies were methodologically dissimilar to the current study, in that they either did not include analysis
of symptom trajectories [34, 39, 40] or did not analyze trajectories from childhood to adolescence [22], and thus do
not provide directly comparable results.
Our results showed that a smaller number of factors in early childhood were nevertheless predictive of
adolescent-onset internalizing symptoms. For instance, hostile parenting increased the odds of the adolescent-onset
trajectory. This finding is consistent with previous research showing that parent rejection is important for the
development of adolescent depressive symptoms [41]. Early physical aggression and hyperactivity were also
predictive of the adolescent-onset trajectory, which supports findings from studies linking adolescent internalizing
outcomes with childhood aggression [42, 43] and hyperactivity [44, 45]. Also, 2 or more SLEs was also predictive
of this trajectory. Thus, the importance of early childhood for the development of internalizing symptoms in
adolescence may be limited to a relatively small number of factors.
In contrast, factors in late childhood and early adolescence may be of much greater importance in terms of
the development of adolescents’ internalizing symptoms. Indeed, some research suggests that, although early
childhood adversity can increase the risk for depression in adolescence, this effect may be mediated through
adolescent stress [46]. Other studies have shown the importance of late childhood stressors such parent
psychopathology [47] and negative life events [22] on depressive symptom trajectories in adolescence and young
adulthood. Thus, factors in late childhood and adolescence can be seen as leading toward dramatic changes in
internalizing symptoms for certain children, with early-life factors playing a less prominent role in this shift in
symptom severity.
Our results also showed that membership in the ‘adolescent-onset’ trajectory was strongly predicted by
child gender, with girls having almost 3 times higher odds of group membership than boys. This was not surprising,
considering studies showing not only that rates of depression and anxiety are higher among girls [48, 49], but that
these gender differences tend to emerge in adolescence [50, 51]. Other research has shown that girls are more likely
10
to be in depressive symptom trajectories in adolescence [13, 14, 47]. We also found no effect of gender for the ‘high
childhood’ or ‘high stable’ trajectories, consistent with some previous research [12, 17]. However, other studies
suggest such gender differences may indeed emerge in childhood [15, 21].
The underlying reasons for this gender difference in the adolescent-onset trajectory are unclear. However,
our further analysis of gender differences suggests that some early risk factors might be particularly important for
the development of adolescent internalizing symptoms in girls. Punitive/aversive and hostile/ineffective parenting
were associated with membership in the ‘adolescent-onset’ trajectory for girls, but not boys. These findings are
consistent with previous studies suggesting that parenting style might have a stronger influence on depression for
college-age females [52] and that coercive parenting is associated with increased risk of depression in adolescent
girls, but not boys [53]. Conduct/aggression and hyperactivity/inattention also were associated with membership in
the ‘adolescent-onset’ trajectory for girls, but not boys. These results are consistent with studies suggesting that
physical aggression is positively related to depression among adolescent girls, but negatively related to depression
among adolescent boys [54], conduct problems in preschool lead to increased internalizing symptoms in adolescence
for girls only [55], and some subtypes of ADHD are associated with comorbid internalizing symptoms among girls,
but not boys [56].
Some researchers have proposed that, although there may be gender differences in early-life risk factors for
depression, gender differences in internalizing symptoms may only occur during adolescence because such risk
factors may trigger stronger reactions to stress for girls at this age [57]. Indeed, there is some evidence that
adolescent girls are more likely than boys to exhibit symptoms of anxiety and depression in response to various
types of stress such as peer and relationship stress [58]. However, our results also showed that being small for
gestational age increased the odds of membership in the ‘adolescent onset’ trajectory for boys only, suggesting that
some sources of stress may also be particularly problematic for boys. This finding is consistent with a recent study
examining the role of birth weight and stress on adolescent internalizing symptoms [33].
Gender interactions in the current study were also not limited to group membership in the ‘adolescent
onset’ group. In fact, family dysfunction increased the odds of group membership in the ‘high stable’ and ‘high
childhood’ groups for girls, but not boys. This supports a previous study of infants of depressed mothers, which
found that early exposure to marital conflict was associated with a subsequent increase in childhood internalizing
symptoms for girls, but not boys [59]. Also, conduct/aggression increased the odds of membership in the ‘high
11
childhood’ group to a greater extent for girls than boys. Indeed, among children with conduct disorder, girls may be
at greater risk for comorbid disorders such as depression [60].
The current study has several limitations that should be mentioned. For instance, parents reported on their
children’s internalizing symptoms up until age 10-11, after which children began to self-report on these symptoms.
Therefore, it is unclear whether changes in internalizing trajectories were due partly to the change in respondent,
rather than entirely due to changes in symptom levels over time. However, this is a problem that is inherent to the
assessment of symptoms via questionnaires versus clinician ratings or other methods. Young children are often not
the best informants, and thus parents must often provide information on their child’s behalf. Older children are
capable of providing self-reports, and are arguably the most reliable source of information regarding their own
emotional state. Although trajectory groups were based on a large sample of children, the relatively small sample
sizes in some trajectory groups may have resulted in a lack of power to detect significant effects, including gender
interactions. Our study utilized survey data and, although well-known confounders of the association between the
risk factors and trajectories were included in the models, bias due to unmeasured confounders cannot be ruled out.
Furthermore, we only examined predictors of trajectory groups at baseline. It is important to note that there may be
time-varying factors influencing trajectories through important cross-sectional effects.
Our study also had a number of strengths. We used data from a large, nationally representative sample and
included incomplete cases with a principled approach to missing data. We were able to identify distinct trajectories
representing child- versus adolescent-onset of internalizing symptoms, as well as early-life risk factors for
membership in these trajectories. We also identified some important gender differences in the role played by various
risk factors, and thus added to the existing knowledge of early-life predictors of the course of internalizing
symptoms in young people.
12
Acknowledgments
This research was supported by a grant from the SickKids Foundation and the Canadian Institutes of Health
Research (grant number SKF 116328), as well as funding from the Canada Research Chairs program for Dr.
Colman. The authors thank Dr. Zacharie Tsala Dimbuene and Dr. Jean-Michel Billette of Statistics Canada for their
assistance with data access and use. The research and analysis are based on data from Statistics Canada and the
opinions expressed do not represent the views of Statistics Canada.
13
REFERENCES
1. Costello EJ, Egger H, Angold A. 10-year research update review: the epidemiology of child and
adolescent psychiatric disorders: I. Methods and public health burden. Journal of the American Academy
of Child and Adolescent Psychiatry. 2005;44(10):972-86. Epub 2005/09/22.
2. Greenberg PE, Kessler RC, Birnbaum HG, Leong SA, Lowe SW, Berglund PA, et al. The
economic burden of depression in the United States: how did it change between 1990 and 2000? Journal
of Clinical Psychiatry. 2003;64(12):1465-75. Epub 2004/01/20.
3. Greenberg PE, Sisitsky T, Kessler RC, Finkelstein SN, Berndt ER, Davidson JR, et al. The
economic burden of anxiety disorders in the 1990s. Journal of Clinical Psychiatry. 1999;60(7):427-35.
Epub 1999/08/24.
4. Pine DS, Cohen E, Cohen P, Brook J. Adolescent depressive symptoms as predictors of adult
depression: Moodiness or mood disorder? American Journal of Psychiatry. 1999;156(1):133-5.
5. Ashford J, Smit F, van Lier PAC, Cuijpers P, Koot HM. Early risk indicators of internalizing
problems in late childhood: a 9-year longitudinal study. Journal of Child Psychology and Psychiatry.
2008;49(7):774-80.
6. Mcardle JJ, Epstein D. Latent Growth-Curves within Developmental Structural Equation Models.
Child Development. 1987;58(1):110-33.
7. Burstein M, Ginsburg GS, Petras H, Ialongo N. Parent psychopathology and youth internalizing
symptoms in an urban community: a latent growth model analysis. Child Psychiatry and Human
Development. 2010;41(1):61-87. Epub 2009/08/12.
8. Mathiesen KS, Sanson A, Stoolmiller M, Karevold E. The nature and predictors of undercontrolled
and internalizing problem trajectories across early childhood. Journal of Abnormal Child Psychology.
2009;37(2):209-22. Epub 2008/09/04.
9. Nagin DS. Analyzing developmental trajectories: A semiparametric, group-based approach.
Psychological Methods. 1999;4(2):139-57.
10. Colman I, Ploubidis GB, Wadsworth MEJ, Jones PB, Croudace TJ. A longitudinal typology of
symptoms of depression and anxiety over the life course. Biological Psychiatry. 2007;62(11):1265-71.
14
11. Feng X, Shaw DS, Silk JS. Developmental trajectories of anxiety symptoms among boys across
early and middle childhood. Journal of Abnormal Psychology. 2008;117(1):32-47. Epub 2008/02/13.
12. Cote SM, Boivin M, Liu XC, Nagin DS, Zoccolillo M, Tremblay RE. Depression and anxiety
symptoms: onset, developmental course and risk factors during early childhood. Journal of Child
Psychology and Psychiatry. 2009;50(10):1201-8.
13. Brendgen M, Wanner B, Morin AJ, Vitaro F. Relations with parents and with peers, temperament,
and trajectories of depressed mood during early adolescence. Journal of Abnormal Child Psychology.
2005;33(5):579-94. Epub 2005/10/01.
14. Costello DM, Swendsen J, Rose JS, Dierker LC. Risk and protective factors associated with
trajectories of depressed mood from adolescence to early adulthood. Journal of Consulting and Clinical
Psychology. 2008;76(2):173-83. Epub 2008/04/02.
15. Dekker MC, Ferdinand RF, van Lang ND, Bongers IL, van der Ende J, Verhulst FC.
Developmental trajectories of depressive symptoms from early childhood to late adolescence: gender
differences and adult outcome. Journal of Child Psychology and Psychiatry and Allied Disciplines.
2007;48(7):657-66. Epub 2007/06/27.
16. Duchesne S, Vitaro F, Larose S, Tremblay RE. Trajectories of anxiety during elementary-school
years and the prediction of high school noncompletion. Journal of Youth and Adolescence.
2008;37(9):1134-46.
17. Sterba SK, Prinstein MJ, Cox MJ. Trajectories of internalizing problems across childhood:
heterogeneity, external validity, and gender differences. Development and Psychopathology.
2007;19(2):345-66. Epub 2007/04/27.
18. Angold A, Costello EJ. Depressive comorbidity in children and adolescents: empirical, theoretical,
and methodological issues. American Journal of Psychiatry. 1993;150(12):1779-91. Epub 1993/12/01.
19. Kessler RC, Stang P, Wittchen HU, Stein M, Walters EE. Lifetime co-morbidities between social
phobia and mood disorders in the US National Comorbidity Survey. Psychological Medicine.
1999;29(3):555-67. Epub 1999/07/15.
15
20. Wickrama T, Wickrama KA. Heterogeneity in adolescent depressive symptom trajectories:
implications for young adults' risky lifestyle. Journal of Adolescent Health. 2010;47(4):407-13. Epub
2010/09/25.
21. Fanti KA, Henrich CC. Trajectories of pure and co-occurring internalizing and externalizing
problems from age 2 to age 12: findings from the National Institute of Child Health and Human
Development Study of Early Child Care. Developmental Psychology. 2010;46(5):1159-75. Epub
2010/09/09.
22. Stoolmiller M, Kim HK, Capaldi DM. The course of depressive symptoms in men from early
adolescence to young adulthood: identifying latent trajectories and early predictors. Journal of Abnormal
Psychology. 2005;114(3):331-45. Epub 2005/08/25.
23. Leve LD, Kim HK, Pears KC. Childhood temperament and family environment as predictors of
internalizing and externalizing trajectories from ages 5 to 17. Journal of Abnormal Child Psychology.
2005;33(5):505-20. Epub 2005/10/01.
24. Statistics Canada and Human Resources and Skills Development Canada. 2009. National
Longitudinal Survey of Children and Youth: Survey Overview for the 2008/2009 Data Collection, Cycle 8.
Ottawa, ON: Statistics Canada.
25. Statistics Canada and Human Resources and Skills Development Canada. 2008. National
Longitudinal Survey of Children and Youth (NLSCY), Cycle 8 - Microdata User Guide. Ottawa, ON:
Statistics Canada.
26. Swain L, Dolson D, editors. Nonresponse in survey research. Eighth International Workshop on
Household Survey Nonresponse; 1997 24-26 September; Manheim, Germany: ZUMA; 1998.
27. Boyle MH, Offord DR, Hofmann HG, Catlin GP, Byles JA, Cadman DT, et al. Ontario Child Health
Study .1. Methodology. Archives of General Psychiatry. 1987;44(9):826-31.
28. Achenbach TM, Edelbrock CS. Behavioral-Problems and Competencies Reported by Parents of
Normal and Disturbed-Children Aged 4 through 16. Monographs of the Society for Research in Child
Development. 1981;46(1):1-82.
29. Statistics Canada and Human Resources and Skills Development Canada. 1998. National
Longitudinal Survey of Children and Youth - Cycle 1 Data Dictionary. Ottawa, ON: Statistics Canada.
16
30. Boyle MH, Offord DR, Racine Y, Sanford M, Szatmari P, Fleming JE. Evaluation of the Original
Ontario-Child-Health-Study Scales. Canadian Journal of Psychiatry-Revue Canadienne De Psychiatrie.
1993;38(6):397-405.
31. Boyle MH, Offord DR, Racine Y, Fleming JE, Szatmari P, Sanford M. Evaluation of the revised
Ontario Child Health Study scales. Journal of Child Psychology and Psychiatry and Allied Disciplines.
1993;34(2):189-213. Epub 1993/02/01.
32. Statistics Canada and Human Resources and Skills Development Canada. 2002. National
Longitudinal Survey of Children and Youth (NLSCY), Cycle 3 - Microdata User Guide. Ottawa, ON:
Statistics Canada.
33. Colman I, Ataullahjan A, Naicker K, Van Lieshout RJ. Birth Weight, Stress, and Symptoms of
Depression in Adolescence: Evidence of Fetal Programming in a National Canadian Cohort. Canadian
Journal of Psychiatry-Revue Canadienne De Psychiatrie. 2012;57(7):422-8.
34. Naicker K, Wickham M, Colman I. Timing of first exposure to maternal depression and adolescent
emotional disorder in a national Canadian cohort. PLoS One. 2012;7(3):e33422. Epub 2012/03/31.
35. Cotton C. Developments in the Low Income Cutoffs. Cotton, C. Ottawa, ON: Statistics Canada;
2001.
36. Kramer MS, Platt RW, Wen SW, Joseph KS, Allen A, Abrahamowicz M, et al. A new and
improved population-based Canadian reference for birth weight for gestational age. Pediatrics.
2001;108(2).
37. Tremblay RE, Masse B, Perron D, Leblanc M, Schwartzman AE, Ledingham JE. Early disruptive
behavior, poor school achievement, delinquent behavior, and delinquent personality: longitudinal
analyses. Journal of Consulting and Clinical Psychology. 1992;60(1):64-72. Epub 1992/02/01.
38. Jones BL, Nagin DS, Roeder K. A SAS procedure based on mixture models for estimating
developmental trajectories. Sociological Methods & Research. 2001;29(3):374-93.
39. Voci SC, Beitchman JH, Brownlie EB, Wilson B. Social anxiety in late adolescence: The
importance of early childhood language impairment. Journal of Anxiety Disorders. 2006;20(7):915-30.
17
40. Meng XH, Tao FBA, Wan YH, Hu Y, Wang RX. Coping as a Mechanism Linking Stressful Life
Events and Mental Health Problems in Adolescents. Biomedical and Environmental Sciences.
2011;24(6):649-55.
41. Wickrama KA, Conger RD, Lorenz FO, Jung T. Family antecedents and consequences of
trajectories of depressive symptoms from adolescence to young adulthood: a life course investigation.
Journal of Health and Social Behavior. 2008;49(4):468-83. Epub 2009/02/03.
42. Underwood MK, Beron KJ, Rosen LH. Joint trajectories for social and physical aggression as
predictors of adolescent maladjustment: Internalizing symptoms, rule-breaking behaviors, and borderline
and narcissistic personality features. Development and Psychopathology. 2011;23(2):659-78.
43. Burk LR, Armstrong JM, Park JH, Zahn-Waxler C, Klein MH, Essex MJ. Stability of Early
Identified Aggressive Victim Status in Elementary School and Associations with Later Mental Health
Problems and Functional Impairments. Journal of Abnormal Child Psychology. 2011;39(2):225-38.
44. Hinshaw SP, Owens EB, Sami N, Fargeon S. Prospective follow-up of girls with
attention-deficit/hyperactivity disorder into adolescence: Evidence for continuing cross-domain
impairment. Journal of Consulting and Clinical Psychology. 2006;74(3):489-99.
45. Biederman J, Monuteaux MC, Mick E, Spencer T, Wilens TE, Silva JM, et al. Young adult
outcome of attention deficit hyperactivity disorder: a controlled 10-year follow-up study. Psychological
Medicine. 2006;36(2):167-79.
46. Hazel NA, Hammen C, Brennan PA, Najman J. Early childhood adversity and adolescent
depression: the mediating role of continued stress. Psychological Medicine. 2008;38(4):581-9.
47. Olino TM, Klein DN, Lewinsohn PM, Rohde P, Seeley JR. Latent trajectory classes of depressive
and anxiety disorders from adolescence to adulthood: descriptions of classes and associations with risk
factors. Comprehensive Psychiatry. 2010;51(3):224-35. Epub 2010/04/20.
48. Kessler RC, Mcgonagle KA, Swartz M, Blazer DG, Nelson CB. Sex and Depression in the
National Comorbidity Survey .1. Lifetime Prevalence, Chronicity and Recurrence. Journal of Affective
Disorders. 1993;29(2-3):85-96.
49. Wittchen HU, Jacobi F. Size and burden of mental disorders in Europe - a critical review and
appraisal of 27 studies. European Neuropsychopharmacology. 2005;15(4):357-76.
18
50. Duggal S, Carlson EA, Sroufe LA, Egeland B. Depressive symptomatology in childhood and
adolescence. Development and Psychopathology. 2001;13(1):143-64.
51. Wade TJ, Cairney J, Pevalin DJ. Emergence of gender differences in depression during
adolescence: national panel results from three countries. Journal of the American Academy of Child and
Adolescent Psychiatry. 2002;41(2):190-8. Epub 2002/02/12.
52. Barton AL, Kirtley MS. Gender Differences in the Relationships Among Parenting Styles and
College Student Mental Health. Journal of American College Health. 2012;60(1):21-6.
53. Compton K, Snyder J, Schrepferman L, Bank L, Shortt JW. The contribution of parents and
siblings to antisocial and depressive behavior in adolescents: A double jeopardy coercion model.
Development and Psychopathology. 2003;15(1):163-82.
54. Knox M, Carey M, Kim WJ. Aggression in inpatient adolescents - The effects of gender and
depression. Youth & Society. 2003;35(2):226-42.
55. Zahn-Waxler C, Park JH, Essex M, Slattery M, Cole PM. Relational and overt aggression in
disruptive adolescents: Prediction from early social representations and links with concurrent problems.
Early Education and Development. 2005;16(2):259-82.
56. Levy F, Hay DA, Bennett KS, McStephen M. Gender differences in ADHD subtype comorbidity.
Journal of the American Academy of Child and Adolescent Psychiatry. 2005;44(4):368-76.
57. Nolenhoeksema S, Girgus JS. The Emergence of Gender Differences in Depression during
Adolescence. Psychological Bulletin. 1994;115(3):424-43.
58. Rudolph KD. Gender differences in emotional responses to interpersonal stress during
adolescence. Journal of Adolescent Health. 2002;30(4):3-13.
59. Essex MJ, Klein MH, Cho E, Kraemer HC. Exposure to maternal depression and marital conflict:
Gender differences in children's later mental health symptoms. Journal of the American Academy of Child
and Adolescent Psychiatry. 2003;42(6):728-37.
60. Loeber R, Keenan K. Interaction between Conduct Disorder and Its Comorbid - Effects of Age
and Gender. Clinical Psychology Review. 1994;14(6):497-523.
19
All(N=6337)
Low Stable (n=4279)
Moderate stable
(n=710)
Adolescent Onset
(n=618)
High Childhood(n=471)
High Stable
(n=259)
Female 48.6 46.7 43.0 70.9 43.3 52.1
Low SES 9.3 9.4 9.4 8.1 8.5 10.0
Low PPVT 9.7 9.2 11.6 7.5 13.1 11.3
Low MSD 9.8 9.8 12.1 7.2 9.9 9.7
Small for gestational age 7.7 7.1 9.9 7.7 9.0 9.0
Presence of chronic illness 25.6 24.5 27.4 22.0 32.2 33.3
Smoking during pregnancy 25.7 24.1 29.8 29.6 27.2 27.7
Drinking during pregnancy 16.8 16.1 20.7 13.6 21.4 18.7
Maternal depression 11.0 9.2 13.9 12.8 17.1 20.4
≥ 2 SLE 2.4 1.6 2.8 2.9 5.9 5.4
Family dysfunction 11.5 10.8 15.4 11.8 16.1 18.7
Hostile/ineffective parenting 12.0 10.7 12.4 15.2 19.7 20.4
Punitive/aversive parenting 14.9 13.7 14.2 17.2 18.7 18.0
Conduct/Aggression 13.1 10.9 15.4 16.4 19.4 24.3
Hyperactivity/Inattention 14.9 11.1 20.5 16.4 20.9 25.5
Difficult temperament 10.3 8.2 12.4 10.3 16.2 20.5
Table 1. Proportions for early-life risk factors
20
Risk FactorAdolescent Onset
(n=618)High Childhood
(n=471)High Stable
(n=259)Moderate stable
(n=710)
Gender (male as reference) *2.84 [2.35, 3.42] 0.90 [0.77, 1.05] *1.45 [1.12, 1.87] 0.90 [0.77, 1.05]
Low SES 0.73 [0.53, 1.01] 0.88 [0.63, 1.23] 0.91 [0.58, 1.42] 0.93 [0.71, 1.22]
Low MSD 0.69 [0.48, 1.00] 0.76 [0.54, 1.07] 0.79 [0.49, 1.29] 1.18 [0.92, 1.51]
Low PPVT 0.90 [0.64, 1.25] 1.07 [0.76, 1.51] 1.07 [0.68, 1.68] *1.42 [1.09, 1.84]
Small for gestational age 1.11 [0.81, 1.52] 1.15 [0.80, 1.66] 1.07 [0.66, 1.73] 1.21 [0.91, 1.62]
Chronic illness 0.86 [0.66, 1.13] *1.81 [1.38, 2.37] 1.37 [0.97, 1.94] 1.30 [1.03, 1.64]
Smoking during pregnancy 1.09 [0.80, 1.48] 0.92 [0.62, 1.37] 1.32 [0.87, 2.00] *1.34 [1.00, 1.78]
Drinking during pregnancy 0.68 [0.47, 1.00] 1.25 [0.84, 1.85] 1.30 [0.83, 2.02] *1.40 [1.03, 1.91]
Maternal depression 1.17 [0.88, 1.56] *2.15 [1.65, 2.80] *2.18 [1.54, 3.08] *1.65 [1.30, 2.09]
≥ 2 SLE *1.88 [1.03, 3.42] *4.39 [2.70, 7.14] *4.08 [2.16, 7.70] *4.71 [3.11, 7.12]
Family dysfunction 0.91 [0.68, 1.21] *1.59 [1.21, 2.08] *1.94 [1.38, 2.73] *1.77 [1.42, 2.21]
Hostile/ineffective parenting *1.65 [1.28, 2.11] *2.19 [1.71, 2.81] *2.03 [1.45, 2.84] 1.04 [0.81, 1.34]
Punitive/aversive parenting 1.25 [0.98, 1.60] *1.34 [1.04, 1.74] 0.95 [0.64, 1.39] 1.08 [0.86, 1.35]
Conduct/Aggression *1.68 [1.31, 2.15] *2.46 [1.94, 3.12] *2.69 [1.97, 3.68] *1.49 [1.18, 1.87]
Hyperactivity/Inattention *1.56 [1.22, 2.01] *1.71 [1.32, 2.22] *2.66 [1.95, 3.63] *2.17 [1.77, 2.67]
Difficult temperament 1.21 [0.89, 1.64] *3.02 [2.36, 3.88] *2.42 [1.71, 3.43] *1.76 [1.38, 2.24]
* Odds ratio significant at p < .05
21
Table 3. Significant gender interactions predicting the odds of trajectory group membership (‘low stable’ group as reference)
Interactions by Gender Trajectorygroup
Interactionp-value
OR (95% CI)for girls
OR (95% CI)for boys
Mother drink while pregnant Moderate stable p<.001 2.41 [1.59, 3.66] 0.78 [0.49, 1.27]
Mother smoked while pregnant Moderate stable p=.006 2.07 [1.37, 3.12] 0.93 [0.63, 1.39]
Small for gestational age Adolescent Onset p=.020 0.89 [0.60, 1.32] 1.93 [1.15, 3.25]
Chronic illness Adolescent Onset p=.014 1.12 [0.81, 1.56] 0.53 [0.33, 0.87]
Family DysfunctionHigh Childhood p=.024 2.24 [1.51, 3.33] 1.20 [0.82, 1.75]
High Stable p=.042 2.62 [1.71, 4.00] 1.24 [0.69, 2.23]
Punitive/aversive Parenting Adolescent Onset p=.007 1.58 [1.18, 2.11] 0.73 [0.45, 1.18]
Hostile ParentingModerate stable p=.031 1.43 [0.99, 2.08] 0.81 [0.57, 1.16]
Adolescent Onset p=.012 2.10 [1.55, 2.85] 1.02 [0.64, 1.64]
Conduct/Aggression Adolescent Onset p=.038 2.09 [1.53, 2.85] 1.17 [0.75, 1.83]
High Childhood p=.020 3.52 [2.41, 5.15] 1.98 [1.46, 2.69]
Hyperactivity/Inattention Adolescent Onset p=.027 1.93 [1.42, 2.63] 1.03 [0.65, 1.64]
22