child adjustment as a function of chronicity …€¦ · web viewabstract word count ... little is...
TRANSCRIPT
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health
Outcomes
Article
Other: Qualter, P., Brown, S.L., Rotenberg, K.J., Vanhalst, J,* Harris, R.A.*, Goossens, L,
Bangee, M.* & Munn, P. (in press). Trajectories of Loneliness during Childhood and
Adolescence: Predictors and health outcomes. Journal of Adolescence: Special Issue on
Loneliness. iFirst, doi:10.1016/j.adolescence.2013.01.005,
Available at: http://link.springer.com/article/10.1007/s10802-012-9676-x
It is advisable to refer to the publisher’s version if you intend to cite from the work. To link to
this article http://www.sciencedirect.com/science/article/pii/S0140197113000122
1
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
Title: Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health
Outcomes
Authors: Qualter, P.1, Brown, S. L.2, Rotenberg, K. J.3, Vanhalst, J.4, Harris, R. A.1,
Goossens, L.4, Bangee, M.1 & Munn, P.5
Affiliations: 1University of Central Lancashire, Lancashire, England, UK; 2 University of
Liverpool, Liverpool, England, UK; 3 Keele University, Staffordshire, England, UK; 4
Catholic University of Leuven, Leuven, Belgium; 5 University of Strathclyde, Glasgow,
Scotland.
Key words: Loneliness, Latent growth mixture modeling, trajectories, health, temperament,
social acceptance, sociability, self-worth, attribution, trust beliefs.
2
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
Abstract
The present study employed latent growth mixture modeling to discern distinct
trajectories of loneliness using data collected at 2-year intervals from age 7-17 years (N =
586) and examine whether measures taken at age 5 years were good predictors of group
membership. Four loneliness trajectory classes were identified: (1) low stable (37% of the
sample), (2) moderate decliners (23%), (3) moderate increasers (18%), and (4) relatively high
stable (22%). Predictors at age 5 years for the high stable trajectory were low trust beliefs,
low trusting, low peer acceptance, parent reported negative reactivity, an internalizing
attribution style, low self-worth, and passivity during observed play. The model also included
outcome variables. We found that both the high stable and moderate increasing trajectories
were associated with depressive symptoms, a higher frequency of visits to the doctor, and
lower perceived general health at age 17. We discuss implications of findings for future
empirical work.
ABSTRACT WORD COUNT = 150 words
MAIN BODY = 5, 635 (larger word count agreed with editor).
3
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
Introduction
Loneliness is an unpleasant state that arises due to a discrepancy between the
interpersonal relationships people want to have, and those they perceive they currently have
(Peplau & Perlman, 1982). Loneliness is important, not just because it affects the person’s
current emotional state, but it is associated with poor social, behavioural, and health outcomes
in adult, adolescent, and child samples (see Heinrich & Gullone, 2006, for review).
The Course of Loneliness Through Childhood and Adolescence
Little is known about the course of loneliness over the lifespan. Early reviews
concluded that loneliness peaks during early adolescence, drops between young adulthood
and middle age, and then rises slightly in old age (see Heinrich & Gullone, 2006). Prospective
studies support these claims (e.g., Bartels, Cacioppo, Hudziak, & Boomsma, 2008;
Demakakos, Nunn, & Nazroo, 2006; Dykstra, van Tilburg, & de Jong Gierveld, 2005; van
Roekel, Scholte, Verhagen, Goossens, & Engels, 2010).
Such studies yield important information about trends in loneliness from middle
childhood to old age, but they examine only mean loneliness scores, which assumes
homogeneity within the populations under study. This assumption is not realistic because we
know that loneliness is transitory for some people (Hymel, Tarulli, Hayden-Thomson, &
Terrell-Deutsch, 1999; Young, 1982) and prolonged for others (Koenig & Adams, 1999;
Qualter, Brown, Munn, & Rotenberg 2010). Studies that assume homogeneity of loneliness
claim to offer information about the normative developmental patterns of loneliness, but the
mean may be an artifact of averaging across subgroups of individuals who follow different
courses of development. By not considering heterogeneity, we may have inadequate
information regarding processes that occur within distinct subgroups (see Bauer & Curran,
2003). For example, stable loneliness may have a different set of predictors compared to
transient loneliness.
4
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
Only two studies to date have assessed specific patterns of change over a number of
observations in loneliness during middle childhood to early adolescence (Jobe-Shields,
Cohen, & Parra, 2011; Jones, Schinka, van Dulman, Bossarte, & Swahn, 2011). Jobe-Shields
et al. showed that social withdrawal was a risk factor for increases in loneliness across
childhood. Further, both studies showed that outcomes related to differing loneliness
trajectories, with an increasing loneliness trajectory predicting social withdrawal, self-
harming behaviors and suicidal thoughts.
Both these studies identified a distinct group of children who increased in loneliness
from middle childhood into early adolescence; this increase in loneliness was characterized by
poor adjustment. In the current study, we examined loneliness from middle childhood through
to late adolescence and used the growth mixture-modeling framework (Muthén, 2004) to
identify latent classes of loneliness. Thus, we examine loneliness across the full length of the
school journey. We further examined the extent to which membership of these different
profiles of loneliness were predicted by early childhood risk factors, and predicted later health
outcomes.
Correlates of Loneliness: Potential Early Risk Factors
A significant body of research has documented correlates of loneliness, including
temperament, peer rejection, an unhealthy attribution style, low self-worth, and low trust
beliefs in peers. In the current study, we examined these as possible risk factors for the
development of long-term or increasing loneliness across the school years.
Temperament is likely to be important for understanding the development of
loneliness as it influences interpersonal interactions via the individual’s initial response to
new social encounters, their arousal and reactivity during any social situation, and recovery in
response to a social threat (McClowry, 2003). Research shows that lonely people have a
different pattern of social response to non-lonely people: they feel more threatened in social
5
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
encounters (Jones et al., 1981, 1983), are more vigilant to social threats (Cacioppo &
Hawkley’s, 2009; Qualter et al., 2012), and are typically shy, passive and less sociable than
non-lonely people (for review, see Heinrich & Gullone, 2006). Thus, two temperament
characteristics may act as risk factors in predicting chronic, stable and/or increasing
loneliness: (1) negative reactivity, which involves the tendency to experience fear, anger,
sadness, and discomfort, and (2) withdrawal, portrayed by the child’s initial response to new
people and situations. High levels of negative reactivity and withdrawn behaviour pose risks
to children’s social relationships because they are linked directly to behavioral deficiencies
(de Pauw & Mervielde, 2010), which undermine the opportunities to develop and maintain
positive social relationships; thus, they are implicated in the development and/or maintenance
of loneliness. In the current study, we examined whether aspects of temperament, specifically
negative reactivity and approach/withdrawal are important for the development of loneliness.
Loneliness during childhood is associated with peer rejection and lonely children are
generally not liked by peers and are the target of peer victimization (Asher, Parkhurst, Hymel,
& Williams, 1990; Boivin, Hymel, & Bukowski, 1995; Ladd, Kochenderfer, & Coleman,
1997). Peer rejection serves as a probable cause of loneliness in children (Boivin, Hymel and
Bukowski, 1995). However, there has been no concerted focus on the question of whether low
social acceptance is associated with long-term loneliness.
Loneliness is also correlated with an unhealthy attribution style and low self-worth in
adults (Perlman & Peplau, 1981; Young, 1982), and children (Qualter & Munn, 2002). In the
current study, we examined whether an unhealthy attribution style in early childhood
predisposes a person to stable feelings of loneliness, as is proposed by Peplau, Russell, and
Heim (1979). Also, we examine whether low self worth in early childhood predicts increases
in loneliness over time and/or helps maintain a high level of loneliness. In line with Snyder,
Stephan & Rosenfield, 1978, we expect that children with low self-worth come to make more
6
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
internal attributions for social failings over time; this continued internal negative talk alters a
person’s self-esteem, especially if social relationships are held to be important, and can lead
to long-term loneliness.
Trust beliefs in peers were also examined in the current study. Three short-term
longitudinal studies (Rotenberg et al., 2010) have shown that low trust beliefs in peers predict
loneliness across time. One limitation with that research is that it did not yield evidence for a
link between trust beliefs and loneliness across childhood to adolescence. The current
research redresses that limitation.
Several researchers have shown that that children’s social behavior in school (i.e.,
social engagement) contributes to loneliness (Qualter & Munn, 2002; Rotenberg et al., 2010)
and we examined this in the current study in relation to long-term loneliness by observing
actual social interactions. We examined the role played by low levels of social interaction
when the child was 5 years and whether this predicted long-term loneliness. Another behavior
that likely contributes to loneliness is low behavior-dependent trust, which according to the
Basis, Target, and Basis interpersonal trust framework (BDT: Rotenberg, 2010), comprises
the extent to which children depend on others to fulfill their promises (reliability trust), keep
secrets (emotional trust), and tell the truth (honesty trust). Low behavior-dependent trust
would result in social disengagement, thus contributing to loneliness. In support of that
hypothesis, Rotenberg, MacDonald and King (2004) found that children’s loneliness was
negatively associated with their behavior-dependent trust as assessed by fulfilling a promise
to classmates who had promised to co-operate. Unfortunately, these findings are limited
because the behavior-dependent trust comprised children’s responses in a mixed motive
game, which was described by an experimenter. Consequently, the relation between
children’s loneliness and behavior-dependent trust in naturalistic interactions remains to be
examined. Children’s naturally occurring behavior-dependent trust was assessed in the
7
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
current investigation by others’ (parents, teachers, and peers) observations of the extent to
which individuals are trusting. It is important to note that parents, teachers and peers have
different experiences of the children’s trusting because, according to the BDT framework,
children vary their trusting behavior as a function of the target of trust.
The studies dealing with the correlates of loneliness have generally been cross-
sectional or predicted loneliness at a specific time point only; predictors have not been
evaluated against trajectories of loneliness established over a number of observations.
Consequently, our capacity to develop appropriate intervention strategies for loneliness is
limited. Thus, in the current study, we used latent class analyses to model the course of
loneliness over a 10-year period, while factoring baseline indicators (at age 5 years) into the
analyses as predictors of loneliness class membership via multinomial logit regression.
Correlates of Loneliness: Adolescent Health Behavior and Well-being
Prospective studies show that peer-related loneliness during childhood (the early and
middle school years) predicts poorer emotional health outcomes in adolescence (Qualter, et
al., 2010; Schinka et al., 2011; Vanhalst et al., 2012). The most commonly advanced reasons
for this are that experiencing loneliness consistently causes psychological and physiological
strains that dispose people toward emotional disorders. Cacioppo and Hawkley’s (2009)
model suggests that lonely individuals experience more stress in everyday life, which can
cause extended periods of hypothalamo-pituitary-adrenal activations leading to both
depression and a range of other somatic and psychological symptoms. Non-physiological
mediators of strain include negative self- and other-related cognitions and attributional states,
associated with loneliness (Heinrich & Gullone, 2006) and implicated in the development of
depression and somatic symptoms (Morley & Moran, 2011). Further, lonely people do not
have access to the supportive social relationships that can moderate the link between stress
and poorer mental health (Cohen, 2004). Depression, frequent use of doctors’ services, and
8
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
poorer self-reported health are consequences of stressful experiences (Axelsson & Ejlertsson,
2002), and we would expect them to be affected by consistent loneliness. Whether trajectories
of loneliness from childhood to late adolescence predict depression has yet to be investigated.
Prospective and cross-sectional research also links loneliness to health compromising
behaviors such as alcohol misuse, smoking and drug use (Hawkley & Cacioppo, 2010). This
can be linked to emotional dysfunction described above, as lonely adolescents attempt to self-
medicate feelings of distress (Carrigan & Randall, 2003). Further, dysfunctional social
relationships appear to have a more direct role in undermining self-regulation of health
behaviors. For example, Baumeister et al. (2005) suggest that dysfunctions in social
relationships create needs for individuals to engage personal resources that might otherwise
be used for executive tasks necessary to successfully self-regulate behavior. Another
possibility is that lonely adolescents’ greater integration needs may lead them to engage with
peers in and out of school whom they otherwise may not, and who influence them to practice
undesirable health behaviors (DeWall & Pond, 2011).
However, the evidence linking loneliness to these health behaviors is inconsistent
(Lauder et al., 2006), possibly because such studies cannot discriminate between groups
characterized by transient versus chronic loneliness, which differentially predict outcomes
(Jones et al., 2011). The longitudinal design of the current study allowed us to assess how
developmental trajectories of peer-related loneliness are associated with health outcomes.
The Current Study
This study aims to (1) describe the number and nature of latent classes of peer-related
loneliness during childhood and adolescence using growth mixture modeling (Muthén, 2004),
(2) examine the hypothesis that temperament, social withdrawal, social preference, self-worth,
attribution style, and trust variables at age 5-years will predict trajectory membership at ages
7-17 years, and (3) examine the hypothesis that trajectory memberships, particularly high
9
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
stable or increasing trajectories, will report greater emotional and physical health problems
and engagement in risky health behaviors. The study covers all years of formal schooling in
the UK offering an examination of loneliness trajectories across the school years.
Method
Participants and Procedure
The study is a prospective study of children from the North West of England, UK. All
participants were enrolled in the state education system, were primarily Caucasian, and
parental permission for participation in the study at each time period was obtained. The
sampling frame was developed to ensure that children were chosen from schools in
Lancashire that were reasonably representative of schools in different areas of the UK as
determined by the UK Government Index of Multiple Deprivation. This index combines
indicators over a range of economic, social and housing issues, into a single deprivation score
for local areas in England, and is commonly used in British educational and health research.
Of those primary schools approached, a representative group of 30 schools agreed to
take part in the study. All children between 58 and 62 months who attended the targeted
schools were possible participants. A total of 842 children were selected, of which 640 (76%)
participated in the study at Time 1 when they were aged 5 years. 50.16% of this sample was
female, 4.84% were from ethnic minority groups that all originated from areas of Southern
Asia, including Bangladesh, India and Pakistan; 97.5% of the sample were born in the UK.
Drop out was relatively high with 586 (92%) children participating at T2 (age 7 years), 400
(62%) at T3 (age 9 years), 389 (61%) at T4 (age 11 years) 296 (46%) at T5 (age 13 years),
364 (57%) at T6 (age 15 years), and 361 (56%) at T7 (age 17 years). Analyses revealed that
children retained at each time point were no different to those that had dropped out at that
particular time point on peer-related loneliness (t < .23, p > .05), or predictor variables (t
< .69, p > .05). There was one exception: children who did not take part at T3 (age 9 years)
10
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
were more likely to have been observed ‘alone’ at Time 1 than those children who provided
data at Time 3 (t = 2.66, p = .009).
All questionnaires for the child at Time 1 were delivered in an interview with the first
author; several trained coders collected observational data of playground behaviour; parent
and teacher measures were given to the parent/teacher respectively to complete at home.
Loneliness measures up to and including age 15 years were completed in school during an
interview with one of several trained researchers, while the loneliness and health
questionnaires at Time 7 were sent to the adolescent’s home and returned via post.
Measures
Loneliness. The peer-related loneliness subscale from the Loneliness and Aloneness
Scale for Children and Adolescents (LACA: Marcoen, Goossens, & Caes, 1987) was used.
Participants rate each statement as it applies to them using a rating of 1 (never) to 4 (often).
Higher scores indicate higher loneliness. An example item is ‘I feel cut-off from other people’.
The questionnaire was developed for school children and early-adolescents, so we adapted the
measure for our participants when they were aged 13 years and above; items were changed so
that they were more appropriate for that age group. For example, ‘ I feel sad because nobody
wants to play with me’ was changed to ‘I feel sad because nobody wants to join in with me’.
This scale demonstrated acceptable internal consistency at each time point (α
= .82, .75, .83, .89, .82, and .78 at ages 7, 9, 11, 13, 15, and 17 years respectively).
Predictors of trajectory class membership. Parent reports, teacher reports,
behavioural observations, and self-reports were used as measures of child risk factors at age 5
years.
Parent reported temperament. Negative Reactivity and Approach/Withdrawal
subscales of the School-aged Temperament Inventory (SATI: McClowry, 1995; McClowry,
Halverson, & Sanson, 2003) were used. Higher scores indicate that the child is high in
11
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
negative reactivity and has a tendency to withdraw in new situations, respectively. Both
subscales have adequate internal consistency (McClowry, 1995, 2002; McClowry et al.,
2003). In the current sample, α = .89 and .84 for negative reactivity and approach/withdrawal
respectively.
Social preference. Using the standard protocol for administering sociometric measures
to young children (Cassidy & Asher, 1992), children were asked to name three children with
whom they most like to play and three children they do not like to play with very much. A
proportion score was computed for each child (i.e., the number of nominations received was
divided by the number of classmates contributing sociometric data) to permit comparison of
scores across classrooms varying in size; these proportion scores were standardized within
each classroom. These standardized most-liked (ML) and disliked (DL) scores were used to
calculate a social preference score following the methods outlined in Coie and Dodge (1983).
A social preference (SP) score was computed as the ML score minus the DL score and is an
indication of likability and acceptance by peers.
Attribution style. The interview instrument designed by Qualter and Munn (2002) was
used to assess children’s habitual style of explaining social events. There are 10 social
outcomes (i.e., 5 for positive-relational outcomes, 5 for negative relational outcomes) and
children are asked to provide reasons for that event. Responses were transcribed and coded as
attributing outcome either to external or internal mechanisms. Total scores used in the current
study are (1) number of positive events that were attributed to external causes, and (2) number
of negative social events that were attributed to internal causes. The inter-rater agreement for
the coding of the attributions was good (kappa co-efficient = .92).
Global self-worth. The global self-worth subscale of the Self-Perception Profile for
Children (Harter, 1985) constitutes a global judgment of one’s worth as a person. The
subscale includes six items. The young child is presented with two pictures and is asked to
12
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
decide which of these children is most like them. They are then asked whether this is sort of
true or really true of them. Each item was scored according to Harter (1985), with a total
mean score of 4 indicating the highest level of perceived self-worth. Reliability was good (α =
.85).
Observed sociability. Children were observed for a 5-minute period on four separate
occasions whilst they were at recess. The protocol is described in detail elsewhere (Qualter
and Munn, 2002). Behaviors of interest for this study are (1) amount of time playing alone,
and (2) aloneness (child is alone, non-involvement in play or interaction with peers or
watches other children). Greater frequencies of these behavior denoted higher social
withdrawal.
Trust beliefs in peers. Trust beliefs in peers were measured using the peer subscale
from Imber's Children's Trust Scale. This scale comprises 10 items with a forced-choice, two-
alternative-answer format (Imber, 1971, 1973). The possible scores ranged from 10 to 20. The
items pertain to the belief in the sincerity of words or behavior of a person or a group of
people; the items in the scale depict issues regarding keeping promises by generic peers and
peer friends. This scale demonstrated adequate internal consistency (α = .75). Scores on the
T-CARS were reversed so that higher scores were indicative of higher trust on all measures.
Trusting (Behavior-Dependent Trust). Teacher and parent reports of children’s
trusting were measured using the single item ‘doesn’t trust other people’ from The Teacher-
Classroom Adjustment Rating Scale (T-CARS: Connors, 1969). Peers reported on trust using
the item ‘trusts other kids’ during sociometric testing. Children’s peer ratings as trusting were
standardized within classes.
Loneliness at 5 years. In order to determine whether early reports of loneliness could
also predict loneliness trajectories, we collected loneliness data from all children at age 5
years. Here we were interested to determine whether all of our other risk factors were able to
13
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
predict membership of loneliness trajectory once a self-report measure of loneliness taken at
age 5 years was included in the model. The LACA has been adapted for use with 5-year-old
children (Qualter & Munn, 2002) and the peer-subscale was used in the current sample.
Internal consistency for the revised measure was good (α = .72).
Health Outcome Measures. Participants completed a series of self-report measures
of health at age 17.
Health Status Questionnaire (HSQ). Three items from the Roy (1994) HSQ were used
in the current study: (1) how would you describe your health generally? (rated ‘excellent’ to
‘very bad’), (2) When did you last consult your GP about your own health, other than for a
check-up required for work or insurance, or for a vaccination? (rated ‘in the last week’ to ‘so
long ago I can’t remember’), and (3) If you had a consultation with your doctor or with a
specialist within the last year, how many consultations did you have? (rated ‘more than 4 per
month’ to ‘less that 4 per year’). General health was reverse scored so that a higher score
indicated better health. The possible ranges of scores were general health (1-5), last visit to
the doctors (1-7), and frequency of visits to doctors (1-4).
Behavioral Risk Factor Surveillance System (BRFSS). Questions from this measure
(Center for Disease Control and Prevention [CDC], 2006) were used to assess health-
impacting behaviors carried out by participants. Responses to each question are treated as
separate measures. Smoking information was assessed by the question ‘do you smoke
cigarettes everyday, some days, or not at all?’. Alcohol intake was assessed with two
questions: ‘on the days you drank in the last 30 days about how many units did you drink on
average per session?’ (units are explained and examples given) and ‘How many times during
the past 30 days did you have 5 units or more on one occasion?’. Drug taking was assessed
using the question ‘During the past 30 days on approximately how many days have you taken
non prescribed drugs or used substances?’.
14
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
Depressive symptoms. Depressive feelings and behavior were assessed using the 19
items from the CES-D (Radloff, 1977). The item 'I felt lonely’ was removed from the scale.
This scale includes items such as “I felt depressed”. Participants were asked to rate how often
they felt the way described during the past week on a scale ranging from 0 (rarely or none of
the time) to 3 (most or all of the time). Items were summed, with higher scores indicating
higher levels of depressive symptoms. Cronbach’s alpha was .85.
Overview of Data Analyses
First, Latent Growth Curve Modeling (LGCM) was used to examine mean level
changes in loneliness. Second, we employed Latent Growth Mixture Modeling (LGMM) to
identify discrete growth classes of loneliness, to test predictors of class membership, and
establish health outcomes (Muthén, 2004). All analyses were conducted in Mplus version 5.1
(Muthén & Muthén, 1998-2008), and missing values were estimated with full information
maximum likelihood estimation (FIML), which is widely endorsed (Schafer & Graham,
2002).
The LGMM proceeded in three stages. First, a series of models, with varying numbers
of classes, were fit to the loneliness data to determine the optimal number of latent classes that
underlie the data and the form of changes over time in loneliness within each class. We
compared 2-, 3- 4- and 5-group trajectory models. The Bayesian information criteria (BIC),
Akaike Information Criteria (AIC), Adjusted BIC, entropy, and the Lo-Mendell-Rubin
likelihood ratio test (LMR) were used to inform the decision of number of classes
(McLachlan & Peel, 2000). Lower AIC, BIC, and adjusted BIC values indicate a more
parsimonious model. Entropy is a measure of classification accuracy, with values closer to 1
indexing greater precision (range: 0-1); the LMR test provides a k−1 likelihood ratio-based
method for determining the ideal number of trajectories, with low p value indicating a better
fit to the data.
15
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
In the second stage of the LGMM, covariates taken when the child was 5 years were
explored as predictors of membership of the loneliness classes using logistic regression.
Third, once the latent loneliness classes were identified and described, within the final model,
we examined the relation between loneliness classes and health outcomes at 17 years of age
using Wald tests of mean equality (Morin et al., 2011).
Results
Identifying Different Classes of Loneliness
Examination of the mean scores for the full sample showed that loneliness was
relatively stable up until age 13 at which point it peaked, but then dropped quickly (see Table
1). LGCM revealed that all estimates of variance related to the intercept and slope were
significant (intercept = 28.05, slope = -1.26, p < .001), which justified an examination of
interindividual differences in loneliness over time using LGMM. Table 2 summarizes the fit
statistics for the 2-5 models from the LGMM. Based on the fit statistics and interpretation of
the later predictor variable analyses, four latent classes of loneliness were identified1. Figure 1
presents the four–class solution. Descriptive statistics on loneliness by age and latent class are
presented in Table 1. We found that most of the children (37%) followed a low stable
trajectory of loneliness, whilst another 23% followed a moderate declining trajectory of
loneliness. Furthermore, 22% followed a high stable trajectory, whilst another 18% followed
an increasing loneliness trajectory that started at a moderate start point on the loneliness scale.
The predictor variables were incorporated into our LGMM model (Table 3). Six
loneliness contrasts were examined: high stable lonely vs. moderate decreasers, high stable
lonely vs. low stable loneliness, high stable lonely vs. moderate increasers, moderate
increasers vs. moderate decreasers, increasers vs. low stable lonely, and low stable vs.
1 The means and percentages reported here are from the model with all covariates included. The class membership changed when the covariates were added to the model, but, for reasons of parsimony, the changes in class membership are not noted; they are available from the corresponding author upon request.
16
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
moderate decreasers. Table 3 shows that seven risk factors differentiated the high stable
lonely adolescents from the other three latent classes in the LGMM model: high loneliness
scores at age 5, low social preference, high negative reactivity, greater frequency of playing
alone, externalization of positive events, low self worth, and trust beliefs. Further, the
moderate increasers could be differentiated from the two non-lonely classes by two risk
factors: (1) low social acceptance by peers, and (2) low trusting scores as reported by teachers
and trust beliefs in peers. These increasers also reported higher self-worth than all other
groups at age 5 years. Approach/Withdrawal was the only variable that did not predict
membership of classes.
Age 17 Health Outcomes as a Function of Loneliness Class
In the model, we included the predictor variables at age 5 years, but also health
outcome variables at age 17 years. Using Wald’s tests of mean equality, we examined
whether membership of a given loneliness class predicted health behavior and outcome. We
found that the high stable lonely and moderate increasers could be differentiated from the
non-lonely groups by their higher reports of depression (W >1.03, p > .05, ORs > 1.29 [1.14-
1.89]), frequency of visits to the doctor (W >5.71, p > .02, ORs > 2.30 [1.18-4.72]), and
general health (W >2.13, p > .03, ORs < .63 [.35-.93]). The high stable loneliness class was
also different to the non-lonely classes in the amount of alcohol units they consumed (W
>4.69, p > .03, ORs > 1.19 [1.02-1.39]) and in the number of days they consumed over 5 units
of alcohol in the last month (W >1.27, p > .05, ORs > 1.27 [1.19-1.46]). For completeness, we
tested for differences between low stable lonely and moderate, decreasing groups: they
showed no differences on any of the outcome variables. Descriptive statistics on health
outcome measures can be found in Table 4.
Discussion
This study is one of the first to prospectively describe the developmental trajectories
17
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
of loneliness from middle childhood to late adolescence. We examined loneliness from 7-17
years, covering the compulsory years of schooling in the UK, and explored whether early risk
factors were associated with different profiles of loneliness, and investigated health outcomes
measured in late adolescence. First, four trajectory profiles were identified. Results showed
that most children followed a consistently low level of loneliness throughout the 12 years.
Another group showed a falling trajectory of modest to low loneliness. A third group included
children who started followed a rising trajectory of loneliness from an average starting point;
the fourth group of children had relatively high levels of loneliness throughout the 12 years.
These patterns of results are in line with previous studies (Jobe-Shields et al., 2011; Jones et
al., 2011; others in this issue) that show heterogeneity in the course of loneliness through
childhood and early adolescence.
In the current study, there were considerably more individuals in the high loneliness
group than in other trajectory studies and more than we would expect based on recent survey
statistics that show 11% of British adults report feeling lonely often (The Mental Health
Foundation: Griffin, 2010). This difference probably represents the fact participants in the
high lonely class in the current study were only relatively high on loneliness; they did not
score very high levels on the loneliness measure, even though they always scored higher than
other groups. However, the fact that this pattern of relatively high loneliness was associated
with health outcomes provides validation that this subgroup is ‘at risk’.
Predictive Links Between Early Risk Factors and Loneliness Trajectories
We examined the extent to which different longitudinal profiles of loneliness were
predicted by a number of risk factors. Negative reactivity, low levels of social engagement
and low social preference increased the risk of following the relatively high stable loneliness
trajectory. How these risk factors work together has not been established here, and should be
examined in future research. For example, this pre-existing temperamental vulnerability may
18
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
amplify the effects of environmental stressors on lonely feelings. So, children following the
relatively high stable loneliness trajectory may be emotionally reactive children who are
simply less able to cope than other children with negative peer experiences.
Cognitive style also acted as a significant risk factor. Specifically, those individuals
who externally attributed positive events, had low self worth, reported low generalized trust
beliefs, and were noted by peers and teachers to have low trust in peers were at increased odds
of following the relatively high stable lonely trajectory. This suggests that as young as 5 years
of age, lonely children have already developed a sense of self that is linked to not trusting
peers, not liking themselves, and seeing positive things that happen to them as luck or out of
their control. We might expect to see low self-worth linked to this profile because (1) feelings
of pessimism and hopelessness about the future often lead to an unhealthy attribution style
(Ickes & Layden, 1978), but (2) causal attributions may alter self-worth, especially when
social relationships are important for that individual (Snyder et al., 1978). The current study
shows that it is the combination of these characteristics that predisposes an individual to
relatively high stable loneliness across childhood/adolescence.
High self-worth, low social acceptance, low trust beliefs in others and low trusting in
early childhood increased the odds of following the increasing loneliness trajectory. Others
(Murrey-Close et al., 2010) have noted that inflated self-perceptions often place children at
risk for rejection by peers, so it may not be surprising that overly high self-worth is associated
with low social acceptance in this sample. Further, decreases in positive biases regarding
social and behavioral competence have been shown to be associated with increases in
depressive symptoms over time (Hoza et al., 2010). In line with these findings, we found that
overly positive self-perceptions are associated with the development of loneliness, but we
have not shown that loneliness increases alongside decreases in positive self-perception bias.
Thus, future research should establish (1) whether this positive self-bias is evident for specific
19
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
domains of self-perception for children following this increasing trajectory of loneliness, and
(2) whether an increase in loneliness over time is associated with a decrease in positive bias.
The findings further demonstrate that children’s low trust beliefs in others, as well as
low trusting, predict sustained loneliness across childhood to adolescence. The findings
extend those of Rotenberg et al. (2010) that showed low trust beliefs in others predicted
loneliness at early childhood, middle and young adulthood. Our findings also complement
those obtained by Rotenberg, MacDonald and King (2004) in which children’s loneliness was
negatively associated with behavior-dependent trust. The finding that trusting reported by
teachers and peers predicted loneliness, rather than that reported by parents, may be the result
of two principles. First, teachers and peers are reporting the extent to which children showed
low trusting behavior in peer interactions. Second, the low trusting quality of those peer
interactions contributed to the children’s loneliness.
Loneliness Trajectories and Health Outcomes
Loneliness trajectories between ages 7-17 years predicted depression, more frequent
use of services offered by doctor’s surgeries, poorer self-reported general health and more
units of alcohol consumed per drinking occasion at age 17. Interestingly, patterns of
association differed between the outcomes. High stable loneliness and moderately increasing
loneliness trajectories were associated with greater depression than moderate declining and
low-stable loneliness trajectories. Visits to the doctors and self-reported health status showed
a similar profile of association with the trajectories.
This notion that loneliness in early childhood precedes depression in adolescents is
one we tested previously with data from this same sample (Qualter et al., 2010). The current
analysis shows that this predictive relationship extends into later adolescence. Here, we also
found that loneliness is associated with the frequency of doctors’ visits and reported concerns
20
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
about health. It is not clear whether these indicators reflect objectively compromised physical
health or somatic complaints that are related to emotional distress.
Greater alcohol use was associated with the relatively high stable loneliness group, but
not the moderately increasing loneliness trajectory. This suggests that loneliness may affect
drinking differently to the way in which it influences distress. This is more consistent with the
social processes suggested by Baumeister et al (2005) or de Wall and Pond (2011), than it is
with an emotional dysregulation explanation.
Strengths and Limitations
The present study is one of the first to explore different developmental trajectories of
loneliness during childhood and adolescence, and our findings offer important new insights
regarding the development of loneliness, its predictors, and how it impacts on health in late
adolescence. The prospective design, with early measurement of predictive variables assessed
when the participants were aged 5 years provides a broad picture of the development of
loneliness during childhood and adolescence. In addition, self-reports were combined with
parent-reports, peer-reports and observational data, limiting shared method variance.
Despite these strengths, the present study has some limitations. First, the participants
were primarily Caucasian. A sample with greater ethnic diversity would be more appropriate
to ensure findings are generalizable. Also, there was significant drop-out over the course of
the study, which may have impacted on the overall results. Third, growth mixture modeling is
a sample-specific technique and replication of the present findings in other samples is
recommended.
These limitations not withstanding, we believe these findings provide important new
insights into the development of loneliness during the school years, from young childhood to
late adolescence. Moreover, when considered alongside the findings from the other papers in
this special issue, our findings have implications for prevention and intervention programs
21
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
with children and adolescents at risk of long-term loneliness. Specifically, providing
interventions for young people who have low self-worth, are not socially accepted, and do not
trust others may prevent them from embarking on a stable high or increasing trajectory of
loneliness through middle childhood and adolescence. Emotion regulation and social skills
training may be particularly helpful in this regard too.
22
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
References
Axelsson., L., & Ejlertsson, G. (2002). Self-reported health, self-esteem and social support
among unemployed young people – a population-based study. International Journal of
Social Welfare, 11, 111-119. doi: 10.1111/1468-2397.00205
Bartels, M., Cacioppo, J. T., Hudziak, J. J., & Boomsma, D. I. (2008). Genetic and
environmental contributions to stability in loneliness throughout childhood. American
Journal of Medical Genetics Part B: Neuropsychiatric Genetics, 147B, Issue 3, 385–
391.
Baumeister, R. F., DeWall, C. N., Ciarocco, N. J., & Twenge, J. M. (2005). Social exclusion
impairs self-regulation. Journal of Personality and Social Psychology, 88, 589-604. doi:
10.1037/0022-3514.88.4.589
Boivin, M., Hymel, S., & Bukowski, W. M. (1995). The roles of social withdrawal, peer
rejection, and victimization by peers in predicting loneliness and depressed mood in
childhood. Developmental Psychopathology, 7, 765-785. doi:
10.1017/S0954579400006830
Cacioppo, J. T., & Hawkley, L. C. (2009). Perceived social isolation and cognition. Trends in
Cognitive Science, 13, 447-454. doi: 10.1016/j.tics.2009.06.005
Carrigan., M. H., & Randall, C. L. (2003). Self medication in social phobia: A review of the
alcohol literature. Additive Behavior, 28, 269-284. doi: 10.1016/S0306-4603(01)00235-
0
Cassidy, J., & Asher, S. R. (1992). Loneliness and peer relations in young children. Child
Development, 63, 350-365. doi: 10.1111/j.1467-8624.1992.tb01632.x
Cohen, S. (2004). Social Relationships and Health. American Psychologist, 59, 676-684. doi:
10.1037/0003-066X.59.8.676
Connors, C. (1969). A teacher rating scale for use in drug studies with children. American
23
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
Journal of Psychiatry, 126, 884–888.
Demakakos, P., Nunn, S., & Nazroo, J. (2006). Loneliness, relative deprivation and life
satisfaction. In J. Banks, E. Breeze, C. Lessoff, & J. Nazroo (Eds.), Retirement, health
and relationships of the oler population in England: the 2004 English Longitudinal
Study of Ageing. London: Institute for Fiscal Studies.
DeWall, C. N., & Pond Jr., R. S. (2011). Loneliness and smoking: The costs of the desire to
reconnect. Self and Identity, 10, 375-385. doi: 10.1080/15298868.2010.524404
Dykstra P. A., van Tilburg. T. G, & de Jong Gierveld. J. (2005) Changes in older adult
loneliness: results from a seven-year longitudinal study. Research into Aging 27, 725–
747. doi: 10.1177/0164027505279712
Griffin, J. (2010). The lonely society. London: Mental Health Foundation
Harter, S. (1985). Manual for the Self-Perception Profile for Children and Adolescents –
Revised. University of Denver.
Hawkley, L. C., & Cacioppo, J. T. (2010) Loneliness Matters: A theoretical and empirical
review of consequences and mechanisms. Annals of Behavioral Medicine, 40, 218-227.
doi: 10.1007/s12160-010-9210-8
Hawkley, L. C., Thisted, R. A., Cacioppo, J. T. (2009). Loneliness predicts reduced physical
activity: Cross-sectional and longitudinal analyses. Health Psychology, 28, 354–363.
doi: 10.1037/a0014400
Heinrich, L. M., & Gullone, E. (2006). The clinical significance of loneliness: A literature
review. Clinical Psychology Review, 26, 695–718. doi: 10.1016/j.cpr.2006.04.002
Hoza, B., Murray-Close, D., Arnold, L. E., Hinshaw, S. P., Hechtman, L., & MTA
Cooperative Group (2010). Time-Dependent Changes in Positively Biased Self-
Perceptions of Children with ADHD: A Developmental Psychopathology Perspective.
Developmental Psychopathology, 22, 375-390. doi: 10.1017/S095457941000012X
24
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
Hymel S., Tarulli D., Hayden-Thomson L., & Terrell-Deutsch B. (1999) Loneliness through
the eyes of children. In: Rotenberg K. J. & Hymel S. (Eds.) Loneliness in childhood and
adolescence. Cambridge University Press, Cambridge.
Ickes, W. J., & Layden, M. A. (1978). Attributional styles. In J. H. Harvey, W. J., Ickes, & R.
F. Kidd (Eds.). New Directions in Attribution Research: Vol 2. Hillside: Erlbaum.
Imber, S. C. (1971). The Imber Trust Scale for children and teachers. Unpublished
manuscript.
Imber, S. C. (1973). Relationship of trust to academic performance. Journal of Personality
and Social Psychology, 28, 145–150.
Jobe-Shields, L., Cohen, R., & Parra, G. (2011). Patterns of Change in Children’s Loneliness:
Trajectories from Third Through Fifth Grades. Merrill Palmer Quarterly, 57, 25-47.
Jones, A. C., Schinka K. C., van Dulmen, M. H. M , Bossarte R. M. & Swahn M. H. (2011).
Changes in Loneliness during Middle Childhood Predict Risk for Adolescent
Suicidality Indirectly through Mental Health Problems. Journal of Clinical Child &
Adolescent Psychology, 40, 818-824. doi: 10.1080/15374416.2011.61458
Kochenderfer-Ladd, B., & Wardrop, J. L. (2001). Chronicity and instability of children’s peer
victimization experiences as predictors of loneliness and social satisfaction trajectories.
Child Development, 72, 134-151. doi: 0009-3920/2001/7201-0009
Koenig, L. J., & Abrams, R. F. (1999) Adolescent loneliness and adjustment: a focus on
gender differences. In Rotenberg K. J. & Hymel S. (Eds.) Loneliness in childhood and
adolescence. Cambridge University Press: Cambridge.
Lauder, W., Mummery, K., Jones, M., & Caperchione, C. (2006). A comparison of health
behaviours in lonely and non-lonely populations. Psychology, Health, and Medicine, 11,
233–245. doi: 10.1080/13548500500266607
Marcoen, A., Goossens, L., & Caes, P. (1987). Loneliness in pre through late adolescence:
25
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
exploring the contribution of a multidimensional approach. Journal of Youth and
Adolescence, 16, 561–577.
McClowry, S. G. (1995). The development of the School-Age Temperament Inventory.
Merrill-Palmer Quarterly, 41, 271-285.
McClowry, S. G., Halverson, C. F., & Sanson, A. (2003). A re-examination of the validity
and reliability of the School-age Temperament Inventory. Nursing Research, 52, 176-
182.
McLachlan, G. J., & Peel, D. (2000). Finite Mixture Models. New York: Wiley.
Morin A. J. S., Maïano C., Nagengast B., Marsh H. W., Morizot J., Janosz M. (2011). General
growth mixture analysis of adolescents’ developmental trajectories of anxiety: the
impact of untested invariance assumptions on substantive interpretations. Structural
Equation Modeling, 18, 613–48. doi: 10.1080/10705511.2011.607714
Morley, T. E., & Moran, G. (2011). The origins of cognitive vulnerability in early childhood:
Mechanisms linking early attachment to later depression. Clinical Psychology Review,
31, 1071-1082. doi: 10.1016/j.cpr.2011.06.006
Murray-Close, D., Hoza, B., Hinshaw, S. P., Arnold, L. E., Swanson, J., Jensen, P. S.,
Hechtman, L., & Wells, K. (2010). Developmental Processes in Peer Problems of
Children with ADHD in the MTA Study1: Developmental Cascades and Vicious
Cycles. Developmental Psychopathology, 22, 785–802. doi:
10.1017/S0954579410000465
Muthén, B. O. (2004). Latent variable analysis: Growth mixture modeling and related
techniques for longitudinal data. In D. Kaplan (Ed.), Handbook of quantitative
methodology for the social sciences (pp. 345-368). London: Sage Publications.
Muthén, L.K., & Muthén, B.O. (1998-2008) Mplus Users Guide (5th Edition). Muthén &
Muthén, Los Angeles, CA.
26
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
Page, R. M., & Hammermeister, J. (1995). Shyness and loneliness: Relationship to the
exercise frequency of college students. Psychological Reports, 76, 395-398. doi:
10.2466/pr0.1995.76.2.395
Pauw, S. S. W., & Mervielde, I. (2010). Temperament, personality, and developmental
psychopathology: A review based on the conceptual dimensions underlying childhood
traits. Child Psychiatry and Human Development, 41, 313-329. doi: 10.1007/s10578-
009-0171-8
Peplau, L., & Perlman, D. (Eds.). Loneliness: A Sourcebook of Current Theory, Research and
Therapy. New York: John Wiley and Sons.
Radloff, L. S. (1977). The CES-D scale: A self-report depression scale for research in the general
population. Applied Psychological Measurement 1: 385-401.
Rotenberg, K. J. (2010). The conceptualization of interpersonal trust: A basis, domain, and
target framework. In Ken J. Rotenberg (Ed). Interpersonal trust during childhood and
adolescence (pp. 8-27). New York: Cambridge University Press.
Rotenberg, K. J., Addis, N., Betts, L. R., Fox, C., Hobson, Z., Rennison, S., Trueman, M. &
Boulton, M. J. (2010). The relation between trust beliefs and loneliness during early
childhood, middle childhood and adulthood. Personality and Social Psychology
Bulletin, 36, 1086-1100.
Rotenberg, K. J., MacDonald, K. J., & King, E. V. (2004). The relation between loneliness
and interpersonal trust during middle childhood. Journal of Genetic Psychology, 165,
233-249.
Roy, M. P. (1994). Psychobiological reactivity and responses to stress: A laboratory and field study
in firefighters. Thesis. St. George's Hospital Medical School, University of London.
Schafer., J. L., & Graham, J. W. (2002). Missing data: Our view of the state of the art. Psychological
Methods, 7, 147-177. doi: 10.1037/1082-989X.7.2.147
27
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
Snyder, M. L., Stephan, W. G., & Rosenfield, D. (1978). Attributional egoticism. In J. H.
Harvey, W. J., Ickes, & R. F. Kidd (Eds.). New Directions in Attribution Research: Vol
2. Hillside: Erlbaum.
Qualter, P., Brown, S. L., Munn, P., & Rotenberg, K. J. (2010). Childhood loneliness as a predictor
of adolescent depressive symptoms: an 8-year longitudinal study. European Child &
Adolescent Psychiatry, 19, 493-501. doi: 10.1007/s00787-009-0059-y
Qualter, P., & Munn, P. (2002). The separateness of social and emotional loneliness in childhood.
Journal of Child Psychology and Psychiatry, 43, 233-244. doi: 10.1111/1469-7610.00016
Qualter, P., & Munn, P. (2005). The friendships and play partners of lonely children. Journal of
Social and Personal Relationships, 22, 379-397. doi: 10.1177/0265407505052442
Vanhalst, J., Klimstra, T. A., Luyckx, K., Scholte, R. H. J., Engels, R. C. M. E., & Goossens, L.
(2012). The interplay of loneliness and depressive symptoms across adolescence: Exploring the
role of personality traits. Journal of Youth and Adolescence, 41, 776-787. doi: 10.1007/s10964-
011-9726-7.
van Roekel, E., Scholte, R. H. J., Verhagen, M., Goossens, L., & Engels, R. C. M. E. (2010).
Loneliness in adolescence: Gene x environment interactions involving the serotonin transporter
gene. The Journal of Child Psychology and Psychiatry, 51, 747 - 754. 10.1111/j.1469-
7610.2010.02225.x
Young J. E. (1982). Loneliness, depression, and cognitive therapy: theory and application. In: Peplau
L. A. & Perlman D. (Eds.) Loneliness: a sourcebook of current theory, research, and therapy.
Wiley: London
28
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
Table 1. Adjusted Loneliness Scores Over Time by Peer Loneliness Class
_____________________________________________________________________
Loneliness class
_______________________________________
Participant’s Relatively Moderate Moderate Low/Stable Whole
age in years High Increasing Decreasing Sample
7 32.87 (4.82) 26.72 (5.67) 28.55 (5.29) 22.13 (4.05) 27.00 (6.34)
9 33.68 (5.23) 27.46 (4.66) 25.45 (3.07) 24.83 (3.39) 27.55 (5.36)
11 32.84 (7.98) 30.34 (8.66) 24.57 (7.23) 18.86 (5.75) 26.05 (9.44)
13 37.87 (4.82) 31.72 (5.67) 23.08 (5.30) 21.19 (5.23) 31.20 (5.83)
15 36.03 (6.96) 32.80 (6.81) 21.31 (7.26) 18.36 (5.46) 26.18 (9.93)
17 30.41 (6.63) 32.41 (5.72) 22.49 (4.24) 21.81 (3.85) 24.46 (6.11)
___________________________________________________________________________
Note. Scores are adjusted for missing data using full information maximum likelihood
estimation (FIML) within Mplus version 5.1. Scores for trajectories are also adjusted for the
following covariates shown to distinguish class membership: social preference, negative
reactivity, self-worth, playing alone, externalization of positive events, and trust beliefs.
Scores on peer loneliness range from 12-48.
29
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
Table 2. Conditional Growth Mixture Modeling for Peer Loneliness: Global Fit Statistics
________________________________________________________________________
Class AIC BIC Adjusted BIC Entropy LRT p value
________________________________________________________________________
1 22526.93 22575.04 22540.12 - -
2 22449.43 22510.66 22466.22 .85 .0807
3 22371.37 22445.72 22391.75 .87 .0001
4 22347.53 22434.99 22371.51 .95 .0001
5 22353.53 22464.12 22391.10 .91 .50
_________________________________________________________________________
Notes: AIC = Akaike Information Criteria, BIC = Bayesian information criteria, LRT = Lo-
Mendell-Rubin test. AIC, BIC, Adjusted BIC = lower values indicate a more parsimonious
model; Entropy = values closer to 1 index greater precision (range: 0-1). The LRT = a low p
value indicates a better fit to the data. The final model selection of a four-class solution was
based on the model fit information detailed above, in combination with the predictor and
outcome effects noted in the remainder of this paper.
30
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
Table 3. Latent growth mixture modeling (LGMM) of loneliness predicted by risk factors
____________________________________________________________________________________________________________________
Loneliness Class Comparisons
_________________________
Relatively Relatively Relatively
High Stable Lonely vs. High Stable Lonely vs. High Stable Lonely vs.
Moderate Increasers Moderate Decreasers Low Stable Lonely
______________________________________________________________________________________________________________
Variable OR (95% CI) p value OR (95% CI) p value OR (95% CI) p value
______________________________________________________________________________________________________________
Risk factor
Loneliness age 5 1.15 (1.10-1.20) .001*** 1.05 (1.01-1.09) .008** 1.06 (.02-1.09) .001***
socpref .78 (.56-1.09) .15 .55 (.34-1.05) .05* .53 (.37-.76) .001**
Approach/withdrawal 2.06 (.84-5.06) .12 1.05 (.47-2.30) .90 1.40 (.59-3.34) .47
NREACT 3.29 (1.40-7.55) .005*** 3.19 (1.35-7.56) .008** 3.04 (1.23-7.54) .01**
playing alone 1.18 (.95-1.09) .16 1.28 (1.03-1.60) .03* 2.67 (.93-7.65) .05*
31
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
alone 1.02 (.95-1.09) .63 1.10 (.98-1.11) .19 .99 ( .88-1.16) .83
intneg .87 (.72-1.04) . 11 .85 (.66-1.09) .21 1.01 (.83-1.21) .94
expos 1.72 (1.23-2.41) .001*** 2.33 (1.65-3.39) .001*** 2.08 (1.10-3.94) .02*
Lself-worth .24 (.11-.49) .05* .56 (.30-1.04) .05* .47 (.26-.85) .01**
Trusting Teacher .95 (.17-5.31) .95 .44 (.24-.81) .01** .49 (.04-1.92) .01**
Trusting Peers 1.23 (.61-2.46) .57 .47 (.01-1.47) .05* .91 (.51-1.66) .75
Trust Beliefs .95 (.67-1.31) .76 .31 (.10-.92) .03* .53 (.37-.76) .001**
_______________________________________________________________________________________________________________
Loneliness Class Comparisons
___________________________________
Moderate Increasers vs. Moderate Increasers vs. Low, Stable Lonely vs.
Moderate Decreasers Low Stable Lonely Moderate Decreasers
______________________________________________________________________________________________________________
Variable OR (95% CI) p value OR (95% CI) p value OR (95% CI) p value
______________________________________________________________________________________________________________
Risk factor
32
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
Loneliness age 5 .92 (.88-.96) .001*** .92 (.89-.96) .001*** 1.007 (.97-1.04) .70
socpref .57 (.45-.71) .001*** .73 (.50-.99) .04* 1.16 (.98-1.38) .11
Approach/withdrawal .51 (.22-1.21) .13 .68 (.28-1.67) .34 .76 (.32-1.79) .52
NREACT .97 (.44-2.15) .94 .51 (.24-1.08) .11 .98 (.97-1.07) .86
playing alone .92 (.70-1.20) .53 1.19 (.95-1.49) .13 1.00 (.99-1.02) .93
alone 1.06 (.99-1.13) .09 1.01 (.95-1.07) .86 1.05 (.98-1.07) .89
negative int. .97 (.84-1.10) .61 1.01 (.92-1.12) .70 .98 (.89-1.13) .67
intneg .58 (.30-1.13) .11 .88 (.74-1.03) .10 .96 (.89-1.06) .53
expos 1.28 (.51-1.52) .69 1.07 (.61-1.84) .82 1.30 (.98-1.74) .07
Lself-worth 2.33 (1.14-4.76) .02* 1.99 (.99-3.88) .05 1.02 (1.01-1.04) .93
Trusting Teacher .12 (.01-1.02) .05* .42 (.08-1.21) .07 .93 (.29-2.99) .91
Trusting Peers .59 (.27-1.27) .17 .84 (.65-1.91) .69 .98 (.72-1.59) .57
Trust Beliefs .24 (.04-.98) .03* .36 (.06-.75) .01** 1.06 (.82-1.89) .68
___________________________________________________________________________________________________________________
Notes. CI = confidence interval; OR = odds ratio. socpref = social preference score, NREACT = negative reactivity (SATI), Approach =
Approach/withdrawal (SATI), , expos = externalizing positive events (attribution style), intneg = internalizing negative events (attribution style),
33
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
Lself-worth = low global self-worth, TrustingT = Teacher’s reports of child’s trusting, TrustingP – Parent’s reports of child’s trusting, TrustingPe
= Peers reports of child’s trusting, Trust Beliefs = the child’s trust beliefs in peers as measured using the peer subscale of the Imber's Children's
Trust Scale.
34
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
Table 4. Health Outcomes at 17 years by Loneliness Class
___________________________________________________________________________
Loneliness class
_____________________________________________________________________
Relatively
High/Stable Moderate Increasers Moderate Decreasers Low/Stable
Outcome M SE M SE M SE M SE
___________________________________________________________________________
GenH*** 3.09 .15 3.27 .14 3.98 .15 3.66 .13
GPLast 1.93 .15 2.01 .16 1.93 .17 2.19 .14
GP Freq*** 1.69 .09 1.92 .10 1.47 .08 1.48 .06
Smoker 1.60 .09 1.55 .09 1.43 .08 1.52 .07
Alc.units*** 4.36 .39 3.23 .39 2.25 .23 2.30 .21
Alc.5unit*** 5.53 .26 3.11 .22 1.08 .27 1.70 .15
DrugDays 1.29 .26 .90 .06 .76 .03 .88 .03
CES-D*** 8.23 .94 7.78 .99 5.18 .79 5.28 .86
__________________________________________________________________________
Notes. GenH = General Health, Reversed scored so higher score indicates better health,
GPLast = ‘When did you last consult your GP?’ (1-in the last week, to 7 -so long ago I can’t
remember), GP Freq = Frequency of visits to GP for the same problem (1-more than 4 per
month, to 4 -less that 4 per year). Smoker – ‘how often you do smoke?’ (1= everyday, 2=
some days, 3= not at all), Alc.units = ‘when drinking, how many units did you drink?’,
Alc.5units = ‘How many times during the past 30 days did you have 5 units or more on one
35
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
occasion?’, DrugDays = ‘During the past 30 days on approximately how many days have you
taken non prescribed drugs or used substances?’ ***p < .001.
36
Trajectories of Loneliness During Childhood and Adolescence: Predictors and Health Outcomes
Figure 1. Four-Class Solution for Peer Loneliness
37