Miller-Lewis et al.

Child and Adolescent Psychiatry and Mental Health 2013, 7:6

http://www.capmh.com/content/7/1/6

RESEARCH Open Access

Resource factors for mental health resilience in

early childhood: An analysis with multiple
methodologies
Lauren R Miller-Lewis1,2*, Amelia K Searle1,2,3, Michael G Sawyer1,2, Peter A Baghurst4,5 and Darren Hedley2,6

Abstract
Background: Given that relatively little is known about the development of resilience in early childhood, this
longitudinal study aimed to identify preschool resource factors associated with young children’s mental health
resilience to family adversity.
Methods: A community sample of 474 young Australian children was assessed in preschool (mean age 4.59 years,
49% male), and again two years later after their transition into formal schooling. At each assessment, standard
questionnaires were used to obtain ratings from both parents and teachers about the quality of children’s
relationships with parents and teachers, children’s self-concept and self-control, mental health (Strengths and
Difficulties Questionnaire), and family adversities (including stressful life events and socioeconomic disadvantage).
Results: Greater exposure to cumulative family adversities was associated with both greater teacher- and parent-
reported child mental health difficulties two years later. Multiple methodologies for operationalizing resilience were
used to identify resources associated with resilient mental health outcomes. Higher quality child–parent and child-
teacher relationships, and greater child self-concept and self-control were associated with resilient mental health
outcomes. With the exception of child-teacher relationships, these resources were also prospective antecedents of
subsequent resilient mental health outcomes in children with no pre-existing mental health difficulties. Child–
parent relationships and child self-concept generally had promotive effects, being equally beneficial for children
facing both low- and high-adversity. Child self-control demonstrated a small protective effect on teacher-reported
outcomes, with greater self-control conferring greater protection to children under conditions of high-adversity.
Conclusions: Findings suggest that early intervention and prevention strategies that focus on fostering child-adult
relationship quality, self-concept, and self-control in young children may help build children’s mental health and
their resilience to family adversities.
Keywords: Resilience, Early childhood, Family adversity, Mental health, Child-adult relationships, Self

Significant mental health difficulties such as depressive-, implementing early intervention strategies aimed at
hyperactive- and conduct-disordered symptomatology altering the trajectory of pathways that lead to the emer-
are experienced by about one in eight children [1,2]. gence of these mental health difficulties [4,5]. Research
These problems tend to persist and are associated with indicates it is more effective and economical to inter-
adverse psychosocial, educational, and health outcomes vene early to promote optimal development, as op-
in adolescence and adulthood (e.g., [1,3]). Consequently, posed to intervening after problems become established
early childhood is considered an opportune time for (e.g., [4,6]).
It is well documented that numerous types of family
* Correspondence: lauren.millerlewis@adelaide.edu.au adversity (e.g., socio-economic disadvantage, adolescent
1
Discipline of Paediatrics, School of Paediatrics and Reproductive Health, parenthood, parental separation, parental mental health
University of Adelaide, Adelaide, South Australia 5005, Australia
2
Research and Evaluation Unit, Women’s and Children’s Health Network, 72
problems, stressful family life events) increase the likeli-
King William Road, North Adelaide, South Australia 5006, Australia hood that children will develop mental health difficulties
Full list of author information is available at the end of the article

© 2013 Miller-Lewis et al.; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the
Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use,
distribution, and reproduction in any medium, provided the original work is properly cited.
Miller-Lewis et al. Child and Adolescent Psychiatry and Mental Health 2013, 7:6 Page 2 of 23
http://www.capmh.com/content/7/1/6

(e.g., [2,7-12]). Moreover, such adversities tend to co- resources that are associated with resilience across
occur, and their cumulative effects are associated with various adversities and developmental outcomes. These
the development of childhood mental health difficulties, are grouped in three domains: (a) children’s internal
with evidence suggesting that it is the number rather characteristics and strengths, e.g., self-esteem, self-efficacy,
than a specific type of an individual adversity in isolation self-control; (b) family characteristics and relationships,
that has the greatest impact [11,13-15]. e.g., child–parent closeness, parenting styles; and (c)
However, there is great individual variation in characteristics of children’s social (particularly school)
children’s response to adversity, and many children environment, e.g., student-teacher relationships, school-
exposed to adversity escape relatively unscathed and in- quality [17,22-24].
stead function adequately [13,16]. Resilience refers to Internal child characteristics such as self-concept, in-
this process of positive adaptation despite exposure cluding self-esteem and self-efficacy, have mostly been
to significant adversity [17,18]. Adversity is considered associated with resilience in older children and
‘significant’ when it is commonly associated with poorer adolescents. In longitudinal studies, Werner and Smith
outcomes, and Adaptation is ‘positive’ if functioning [25,26], Masten and colleagues [12], and Elder and
in a developmentally appropriate domain (e.g., mental Conger [27] found that positive self-worth (self-esteem
health) is “better than expected” given the level of and self-efficacy) was longitudinally predictive of psycho-
adversity experienced [17,19]. Because resilience is a socially resilient outcomes in adolescents within the con-
phenomenon that can only be considered within the text of family adversity and stress. In the Rochester child
context of adversity, it is not a fixed or immutable trait resilience project [28,29], self-esteem and perceived self-
that a person ‘has’ – a person may exhibit resilient competence were associated with resilient adjustment
outcomes in one context or domain but not in another for school-aged children experiencing stressful life
[20]. Examining resilient outcomes adds to our know- events, but this was not the case in a similar study
ledge because it involves investigating functioning (or conducted in Australia [30]. However support for a rela-
‘competence’) that is ‘unexpected’, due to the presence of tionship between positive child self-concept and psycho-
adversity. By studying resilient outcomes in children, it social resilience is also provided by other studies
is possible to identify resource factors that enable chil- involving at-risk children and adolescents exposed to
dren to adapt positively to adversity. This is important specific adversities such as socio-economic disadvantage
because adversities are often deep-seated family and so- [31-33], family disintegration [31,34,35], and maternal
cial problems that are difficult to change. A better depression [36].
understanding about why some children are more resili- Children’s self-control or emotional regulation may
ent than others within the context of adversity has the also buffer adversity and promote adaptive outcomes
potential to guide the development of new evidence- by enabling children to respond positively to stressful
based early interventions designed to better prepare chil- circumstances [37,38]. In two cross-sectional studies
dren to cope with current and future adversity [4,18]. of socio-economically deprived preschool children
Understanding factors that promote resilient outcomes attending Head Start, greater emotional regulation
in at-risk children faced with adversity helps to ensure was associated with fewer internalising problems [39],
that children with the odds stacked against them will fewer conduct problems and more pro-social behav-
benefit from prevention programs by targeting resources iour [40]. Longitudinal studies of at-risk young chil-
known to protect at-risk children from poor develop- dren growing up in poverty have found that toddler
mental outcomes [17]. emotional/behavioural regulation and attentiveness/
Given that relatively little is known about the develop- persistence on tasks is predictive of fewer behavioural
ment of resilient outcomes in early childhood [21], this problems 3 to 4 years later [41,42]. Emotional regula-
longitudinal study aimed to identify characteristics of tion (including lower negative emotionality and
4 year-old preschool children, their families, and their greater inhibitory control) has demonstrated both
preschools which predicted ‘better-than-expected’ (i.e., concurrent and longitudinal associations with adaptive
resilient) outcomes on mental health difficulties two mental health outcomes in other high-adversity
years later within the context of cumulative family samples, including children exposed to domestic vio-
adversity. lence, maternal depression, impoverished minority
youth, children experiencing maltreatment and cumu-
Resource factors for mental health resilience lative family adversities, and homeless children
Specific resource factors or assets may have the po- [37,43-51]. Further studies with school-aged children
tential to buffer or ameliorate the detrimental effects found that self-regulation moderated the association
of adversity and lead to resilient outcomes. A consid- between socio-contextual family adversities and men-
erable body of research has identified a core set of tal health outcomes, signifying the potential role of
Miller-Lewis et al. Child and Adolescent Psychiatry and Mental Health 2013, 7:6 Page 3 of 23
http://www.capmh.com/content/7/1/6

self-regulation as a protective factor in the context of In summary, considerable evidence exists for the role
family adversity [43,52]. of each of the groups of child, family, and social/school
Supportive child–parent relationships characterised factors in the development of mental health resilience.
by warmth and closeness have been found to consist- However, some limitations deserve mention. First, the
ently predict mental health resilience in children. For majority of this research has focussed on resilience in
example, in children from the Kauai longitudinal middle childhood and adolescence [21]. In comparison,
study exposed to cumulative family adversities, the re- few studies have investigated resource factors during
silient youth had more supportive relationships and preschool, or resilient outcomes in young children
interactions with capable parents than non-resilient across the preschool to school transition, which is
youth [25,26]. The association between the quality of considered a critical period of rapid developmental
child–parent relationships and positive mental health change [59]. Thus, it is unclear if promoting these
child outcomes is demonstrated in several other lon- factors during preschool will improve mental health
gitudinal studies within different adversity contexts outcomes in young children exposed to family adversity.
including socio-economic disadvantage [27,33,42,53- Second, it is notable that the school environment, and
59], parental death and divorce [35,60,61], stressful particularly the potential role of teachers, has received
life events [12,62], and child maltreatment [31,33,63] far less empirical attention than other resource domains.
some of which focussed on early childhood outcomes As a result, few studies have examined resource factors
[42,53-55,57-59,63]. In longitudinal studies of young from all three child, family, and social/school domains in
children examining family relationships as a moder- the same study (notable exceptions include [25,26] and
ator of the association between adversity exposure [42]). Without knowing what their unique contributions
and child mental health symptoms, O’Grady and Metz are, it is unclear whether one resource may be more im-
[64] found that greater family support provided by portant than another. This is an important omission,
parents for children buffered the effect of stressful life given that evidence already exists of the influence of
event exposure on children’s emotional and behav- resources from all three domains. Third, a considerable
ioural problems, Malmberg and Flouri [65] found that proportion of studies examine single adversity factors in
mother-child relationship quality buffered the effect isolation (e.g., maltreatment, poverty). Comparatively
of socio-economic disadvantage on children’s emo- fewer studies [15,25-27,66,71] have examined cumulative
tional symptoms, and Maughan and colleagues [44] family adversities including combinations of socio-
found that maternal negative parenting moderated the economic factors, stressful life events, parental mental
effect of maternal depression on young children’s health, and parental separation. This is considered
perceptions of social acceptance. In direct contrast, problematic because “focusing on a single risk factor
Calkins and colleagues [66] found that a more re- does not address the reality of most children’s lives”
sponsive relationship between parent and toddler was (p.367) [71].
associated with more externalising and internalising Finally, the vast majority of research on child resilience
behaviour in 5 year old children exposed to high fam- has been conducted in the US and UK. Conducting re-
ily adversity. search in other countries such as Australia is important
There is also a small amount of evidence that close because resource factors relevant to resilience may be
supportive relationships with teachers are associated context and culture specific [72-74]. It is not known
with resilience in children faced with adversity. For ex- whether Australian children may demonstrate unique
ample, the resilient adolescents from the Kauai longitu- developmental patterns and responses to adversity.
dinal study frequently had a favourite teacher who While these countries are all English-speaking multicul-
became a role model for them [25,26]. In a longitudinal tural western societies, the different distributions of
study of children aged 4 to 8 years, Peisner-Feinberg and socio-economic disadvantage, greater income mobility,
colleagues [67] found teacher-child closeness was more less spatial concentration of public housing, and the na-
strongly related to lower levels of behaviour problems tionwide universal provision of free preschool for all 4–
among children identified at-risk due to low maternal 5 year old children in Australia make it difficult to know
education, compared with their low-risk peers. Similar how directly applicable findings from the US and UK
findings have been obtained from cross-sectional studies would be to Australian children [72,75]. Only a handful
of preschool children at-risk due to socio-economic of studies have investigated mental health resilience in
deprivation [40,68]. In Qualitative studies of Australian Australian children (e.g., [30,36,51,69,74-79]), with the
and South-African children experiencing adversity, those evidence for young Australian children limited to studies
identified as ‘resilient’ by their teachers frequently made finding support for positive child–parent relationships
positive comments about special caring teachers who and home environments as correlates of mental health
had a positive impact on their wellbeing [69,70]. in the context of family disadvantage and stress
Miller-Lewis et al. Child and Adolescent Psychiatry and Mental Health 2013, 7:6 Page 4 of 23
http://www.capmh.com/content/7/1/6

[51,76,78,79]. There is much more knowledge to be with a ‘multiple-groups’ approach, where main-effects
gained in this context. regression analyses predicting resilience residual scores
are run separately for low- and high-adversity groups
Multiple methodologies for measuring resilience [89-92]. Subsequent effect sizes for each group can then
Resilience is a concept that is inferred on the basis of be compared to examine the specificity of processes (i.e.,
associations between the levels of (a) exposure to ad- whether a resource is a general ‘promotive factor’
versity and (b) positive adaptation or positive adjust- associated with good outcomes in both low- and high-
ment outcomes, and therefore it cannot by directly adversity children, or a specific ‘protective factor’ with
measured [18,24,80]. There is no ‘gold standard’ for unique benefits only for high-adversity children) while
operationalising the concept of resilience, and several avoiding the statistical problems related to statistical
different approaches are currently used to combine interaction terms [17].
adversity and adjustment levels to measure resilient In contrast to variable-centred approaches, person-
outcomes. When this occurs it can be difficult to centred approaches involve identifying a group of resilient
compare results from different studies of resilience as children (who experience high adversity but exhibit ad-
it is possible they may not actually be measuring the equate adjustment), and comparing their characteristics
same concept or phenomenon [24,80,81]. with other groups of children showing different patterns
Broadly, methods of measuring resilience can be classed of adversity and adjustment, in order to identify resource
as variable-centred or person-centred approaches. Variable- factors associated with resilience (e.g., [12,25,90]). Using
centred approaches examine statistical associations between Masten and colleagues [12] taxonomy as an example, if
measures of adversity, hypothesised resource factors, and four groups of children with divergent outcomes are
developmentally-relevant functioning, using regression- identified - two high-adversity groups identified as either
based analyses. If a factor modifies (i.e., reduces) the nega- ‘resilient’ (good adjustment) or ‘maladaptive’ (poor adjust-
tive effects of adversity on functioning, then it is labelled ment), and two low-adversity groups classified as ‘compe-
‘protective’, and it is implicated in resilience among the chil- tent’ (good adjustment) or ‘highly vulnerable’ (poor
dren for whom the risk and protective factors co-occur adjustment) - it is possible to determine if a resource is
[82,83]. Researchers typically test such modifying effects truly protective rather than generally promotive by exam-
using a statistical interaction term between the adversity ining if resource levels differ between ‘resilient’ and ‘mal-
and hypothesised protective variables. The ‘statistical inter- adaptive’ children, but not between ‘competent’ and ‘highly
action’ approach draws on the statistical power of the whole vulnerable’ children. A key advantage of the person-
sample. However, the children who meet the criteria for re- centred approach is that it better reflects resilience as it
silience are never explicitly identified, and thus which chil- actually occurs naturally within the whole child, rather
dren are deemed resilient remains unknown [84]. than through associations between variables. Due to this,
Additionally, statistical interaction terms within regression manifestly resilient children can actually be identified
can lack adequate statistical power to fully and reliably de- [17,18,23,36]. However, reducing the vast individual
tect real interactions, leading some researchers to caution differences present in early childhood development into
against relying on statistical interaction terms [16,82,84]. broad dichotomous categories may be problematic, as
Two other variable-centred approaches, used in com- valuable detail becomes lost, particularly if the sample size
bination, can address these two main limitations. First, is substantially reduced by selecting more extreme
the ‘residuals’ approach can identify resilient children subgroups only [84,93]. Furthermore, if cut-points are
who, in a statistical sense, are ‘doing better than somewhat arbitrarily defined (particularly a median-split)
expected’, while also keeping all data as continuous. without a solid reason to suspect different effects between
With this approach, when regressing adjustment on ad- the groups created, then effects that occur within rather
versity, the difference between a child’s actual adjust- than between groups may be obscured [84].
ment score and his/her adjustment score predicted by Despite the considerable methodological variation in re-
adversity (i.e., the standardised residual scores) can be silience studies, the fact that a common set of child, family,
utilised as a continuous vulnerability-to-resilience score. and social resources have been consistently recognised in
Children with positive residual scores (i.e., falling above resilience suggests that these resources are all implicated in
the regression line fitted) show ‘better than expected’ the same underlying phenomenon, and support the validity
adaptation than predicted by their exposure to adversity, of resilience as a construct [18,23,80]. Given their seemingly
and are considered resilient (with the size of the residual universal importance, these particular resource factors
indicating their level of resilience). This residuals meth- could be quite useful for further systematic exploration of
odology is a relatively innovative approach [17] and the resilience construct, and critical examination of its
variants of it have been used in several resilience studies measurement. However, researchers have rarely addressed
[27,85-88]. Second, the ‘residuals’ approach can be used whether similar variables emerge as significant resources
Miller-Lewis et al. Child and Adolescent Psychiatry and Mental Health 2013, 7:6 Page 5 of 23
http://www.capmh.com/content/7/1/6

while employing multiple resilience methodologies within four different methodological approaches for operation-
the same sample. Inferences have needed to be made across ally defining resilient outcomes (as described above).
studies, when many other factors could not be accounted This strategy allowed the investigation of whether simi-
for, such as sample characteristics. Given the relative lar resource factors emerged as predictive of resilient
strengths and weaknesses of both variable- and person- outcomes in young children when different methodo-
focussed resilience methodologies, it seems sensible to use logical techniques were used. To our knowledge, this is
both types of methods in combination in the same study the first study to analyse results from directly compar-
(e.g., [12,84]). able techniques for operationally defining resilient
As different resilience measurement approaches are outcomes.
rarely used in a single study, little information exists There are several unique aspects to this study. We
regarding how different methodologies may affect add to the relatively small body of literature on resilient
results (whilst holding constant the sample and vari- outcomes in young children, and to the limited infor-
able measures). Masten and colleagues [12] conducted mation regarding the various potential resources in the
both variable-centred analyses (examining whether child, family, and school domains that children experi-
resource variables buffered the negative impact of ad- ence during the preschool year [21]. This may inform
versity using regression interactions), and person-centred early intervention efforts designed to maximise positive
analyses (examining whether the same resource factors development in young children and intervene before
distinguished between ‘Resilient’, ‘Maladaptive’ and mental health difficulties become entrenched [4,5]. The
‘Competent’ groups of children in MANOVAs). How- present study also builds upon previous research by
ever, the fourth ‘Highly Vulnerable’ group (low adver- longitudinally investigating resource factors associated
sity + poor adjustment) was omitted because it was an with resilient outcomes in a contemporary cohort of
‘empty cell’, so the possibility that associations between young children. Finally, the present study represents
resources and positive adjustment differed between one of the first investigations of mental health resilience
high-adversity and low-adversity children could not in the context of cumulative family adversities in
be examined. Thus, although complementary, their Australian children. These aspects are important given
variable- and person-centred approaches were not that resilience is considered a contextually and cultur-
directly comparable (see also [46,49,94-96]). To our ally embedded phenomenon, and a multiply-determined
knowledge, only one study has assessed interactive and mutable developmental process [20,74].
effects within both variable- and person-centred ana-
lyses. Lengua [43] examined whether resource levels Method
discriminated not only between two high-adversity Participants
groups (e.g., ‘resilient’ vs. ‘maladaptive’), but also Participants were the families of 485 children
between two low-adversity groups (e.g., ‘competent’ vs. attending the 27 government-funded preschools in
‘highly vulnerable’), using logistic regressions. Findings one South Australian government schooling district
were then compared with those from linear regression (at Time 1, mean age = 4.59 years, SD = 0.33, age
interaction terms. However, these methodologies were range = 3 to 5 years, 49% male). This district is quite
not fully comparable because they used a different diverse, encompassing suburban, rural and remote
adjustment variable – the adjustment variables were areas, with some of these ranked at the highest levels
examined separately within variable-centred analyses, of socio-economic disadvantage in Australia. The
but were combined into a composite adjustment vari- demographic characteristics of this district overall re-
able for person-centred analyses. semble those for South Australia as a whole [97].
In 2006, participation was sought from all families
of children attending preschool a within the district.
The present study At baseline, both a parent survey and a teacher sur-
The aim of the present study was to investigate child, vey were completed for 601 children (representing
family, and preschool resource factors associated with 62% of all district preschoolers). Based on school dis-
the development of resilient mental health outcomes trict records, the 62% of children recruited were of
during the early childhood years. We hypothesised that similar age and gender distribution to the preschool
(a) children’s characteristics (higher self-esteem, self- children in the whole district, but the percentage of
efficacy, and self-control), (b) better quality child–parent children of Aboriginal/Torres Strait Islander (ATSI)
relationships, and (c) better quality child-teacher descent was somewhat lower in the participating sam-
relationships during preschool, would be associated with ple than in the school district population (1.4% versus
greater mental health resilience in children two years 3.9%). This suggests the study findings may not be as
later once at school. To achieve this aim, we utilised the generaliziable to ATSI children. Children were assessed
Miller-Lewis et al. Child and Adolescent Psychiatry and Mental Health 2013, 7:6 Page 6 of 23
http://www.capmh.com/content/7/1/6

two years later after they had commenced formal that were unemployed (24.1% vs. 10.4%), a large number of
schooling. Both parent and teacher surveys were siblings (20.7% vs. 8.7%), and younger fathers (30.8 years vs.
completed for 485 of these children (retention rate = 32.4 years). Those children lost to attrition also had signifi-
81%). At both assessments, the parent-reported cantly greater levels of parent-reported (9.97 vs. 8.47) and
surveys were completed by mothers for the majority of teacher-reported (6.66 vs. 5.28) mental health difficulties at
the sample (92% at both assessments). Eleven children the initial assessment. Hence, those lost from the sample
were missing data for at least one of the study tended to be families with greater exposure to family adver-
variables, so analyses were conducted using the sities and children with greater mental health difficulties.
remaining 474 children will full data. Table 1 provides
demographic information about these 474 participating
children. Measures
The children lost from the sample between assessments Children’s primary care-giving parent and their current
(n= 116) were significantly (i.e., p < .05) more likely to be teacher completed the following standardised ques-
living in a single parent family (28.7% vs. 13.6%) that was tionnaires at the baseline and follow-up assessments.
receiving a means-tested government pension/benefit The internal consistencies of the continuous-measured
(60.9% vs. 40.9%), had experienced more stressful life events scale variables used in the present study were adequate,
in the past 12 months (1.4 vs. 0.9), had mothers who had with Cronbach’s alphas ranging from .78 to .95 (see
not completed high school (37.8% vs. 25.1%) and fathers Table 2).

Table 1 Child and family demographic background (n = 474)

Time 1 variables M (SD) or %
Child age 4.59 (0.33)
Child Gender Female 51%
Family receives means-tested government pension/benefit 40.3%
Mother was adolescent (<21) when child born 7.2%
Father was adolescent (<21) when child born 2.0%
Child currently lives in a single parent family 12.9%
Child has 3 or more siblings living in home 8.9%
Number of stressful family life events in the last year 0.90 (1.17)
Parental psychological distress (GHQ) score 1.41 (2.49)
Above GHQ clinical cut-off 39.1%
Number of years child lived separate from one (or both) natural parents a 0.50 (1.28)
Mother’s education level b 2.51 (1.07)
Completed university qualifications 20.3%
Technical, trade, or further education (TAFE) certificate 33.4%
Completed high school Year 12 or equivalent 21.6%
Completed some years of high school or less 24.7%
Father’s education level b 2.65 (1.03)
Completed university qualifications 11.2%
Technical, trade, or TAFE certificate or some university 42.7%
Completed high school Year 12 or equivalent 15.9%
Completed some years of high school or less 30.3%
b
Family paid employment level 1.74 (0.99)
Two parents employed full-time 6.3%
One parent full-time and one part-time 38.0%
One parent full-time or 2 parents part-time 39.9%
One parent part-time 7.2%
Both parents unemployed or an unemployed single parent 8.6%
a. b.
Note. i.e., Duration of parental separation Higher scores on these variables indicate less education and employment, respectively.
Miller-Lewis et al. Child and Adolescent Psychiatry and Mental Health 2013, 7:6 Page 7 of 23
http://www.capmh.com/content/7/1/6

Table 2 Bivariate correlations between continuous variables (n = 474)

Variable 1 2 3 4 5 6 7 8 9 10 11 12 Mean SD Alpha
1) P self-esteem 1 55.47 6.10 .80
2) T self-esteem .18*** 1 57.28 8.48 .92
3) P self-efficacy .64*** .17*** 1 28.92 3.90 .78
4) T self-efficacy .16*** .86*** .21*** 1 30.01 5.29 .93
5) P self-concepta .91*** .19*** .90*** .20*** 1 0.00 0.91 .86
a
6) T self-concept .18*** .96*** .20*** .96*** .21*** 1 0.00 0.96 .95
7) P self-control .45*** .13** .47*** .15*** .51*** .15*** 1 20.67 4.05 .86
8) T self-control .09# .64*** .14** .61*** .12** .65*** .14** 1 23.75 5.57 .93
9) P child–parent relationship .52*** .05 .41*** .06 .51*** .06 .49*** .05 1 65.54 6.38 .78
10) T child-teacher relationship .07 .63*** .11* .60*** .10* .64*** .15*** .61*** .06 1 68.19 7.71 .89
11) P cumulative adversity -.09 -.16*** -.05 -.19*** -.08 -.18*** -.06 -.13** -.07 -.10* 1 0.00 3.06 n/a
index
12) P SDQ difficulties (Time 2) -.43*** -.17*** -.37*** -.24*** -.44*** -.21*** -.38*** -.20*** -.39*** -.16*** .26*** 1 8.24 5.43 .83
13) T SDQ difficulties (Time 2) -.15*** -.26*** -.18*** -.34*** -.18*** -.31*** -.17*** -.34*** -.14** -.33*** .13** .42*** 6.94 6.21 .87
Note. P= parent-reported; T= teacher-reported.
a.
A child self-concept composite score for each informant was created by averaging together standardised scores for child self-esteem and child self-efficacy.
* p < .05. ** p < .01. *** p < .001.

Child’s mental health difficulties to create a total score ranging from 9 to 36, with higher
Parents and teachers completed the Strengths and Diffi- scores representing higher self-efficacy. The scale has
culties Questionnaire (SDQ) [98], a screening question- good reliability and factorial validity, and exhibits
naire designed to assess children’s behaviour and expected correlations with Conners’ Teacher Rating
emotions. It consists of 25 items divided between five Scale [102].
subscales: Emotional Symptoms; Conduct Problems;
Hyperactivity; Peer Problems; and Prosocial Behaviour.
Behavioural self-esteem The child’s behavioural self-
Respondents provide answers on the basis of the child’s
esteem was measured with the 14-item Behavior Rating
behaviour (e.g., “generally disobedient”) over the previ-
Form – Revised [103], which measures young children’s
ous six months or the current school year, using a three-
self-esteem as inferred or perceived by a parent or
point response format of “not true” to “certainly true”.
teacher. Each item (e.g., “this child refers to himself/her-
Scores on each subscale can range from 0 to 10.
self in generally negative terms”) is rated using a 5-point
An overall emotional-behavioural difficulties score is
Likert scale from “never” to “always”. Total scores are
generated by summing the subscale scores, with the ex-
derived by summing items and can range from 14 to 70,
ception of the Prosocial subscale. Scores on Total Diffi-
with higher scores indicating higher levels of inferred
culties can range from 0 to 40, with higher scores
self-esteem. This measure has been found to have high
indicating greater mental health difficulties. Total diffi-
internal consistency [103].
culties scores above 14 on parent-reports and above 12
on teacher-reports are considered ‘of concern’ in the ab-
normal/clinical range. The SDQ has well-established Emotional self-control Parents and teachers completed
psychometric properties, including strong relationships the self-control subscale of the Devereux Early Child-
with diagnostic interviews [99-101]. hood Assessment (DECA) [38]. The DECA is a
standardised, norm-referenced behaviour rating scale for
Child’s internal strengths children aged 2 to 5 years. The 8-item self-control
subscale measures children’s ability to experience a
Behavioural self-efficacy The child’s level of self- range of feelings and express them using appropriate
efficacy as perceived by the parent and teacher was actions and words. Respondents rate the frequency of
measured with the Self-Efficacy Scale-Teacher Version behaviours exhibited by the child over the last four
[102]. This scale consists of 9 items reflecting self- weeks (e.g., “calm himself/herself down when upset”) on
efficacious behaviours. Items (e.g., “when presented with a five-point Likert scale ranging from “never” to “very
a new task, the child believes he or she can do it”) are frequently”. Items are summed so that higher scores in-
rated with a four-point Likert scale ranging from “not at dicate greater emotional self-control. The DECA has
all like the child” to “like the child”. Items are summed strong psychometric properties, demonstrating both
Miller-Lewis et al. Child and Adolescent Psychiatry and Mental Health 2013, 7:6 Page 8 of 23
http://www.capmh.com/content/7/1/6

high internal consistency and reliability, and discrimin- had always lived with two natural parents or not, and if
ant validity [38,104,105]. not, the length of time the child had lived with their
mother alone, their father alone, or neither natural par-
Child’s external relationship context ent. This information allowed us to calculate the length
of time each child had spent during their lifetime living
Quality of Child’s relationships with parents and without their two natural parents. Scores could range
teachers Parents and preschool teachers described the from 0 “none, always lived with two parents” through to
quality of their relationships with the children using the 6 “more than five years”, with higher scores indicating a
short forms [106] of the Child–Parent Relationship Scale longer period of parental separation.
(CPRS) [107] and the Student-Teacher Relationship
Scale (STRS) [108], respectively. The two questionnaires Early parenthood Parents reported the age of the child’s
contain 15 identical items on which parents and teachers mother and father. Children’s current age was subtracted
rate their perceptions of their relationship with the child from each parent’s age to calculate the mother’s and
using a 5-point Likert scale (“definitely does not apply” father’s age when the child was born. Each parent was
to “definitely applies”). Items assess the level of close- then categorised as either an adolescent parent (defined
ness, warmth, and conflict in the relationship, and are as ≤ 20 years at the time of the child’s birth), or not (≥
based on behaviours relevant to attachment theory (e.g., 21 years) [10].
“If upset, this child will seek comfort from me”). Total
scores are created by summing the 15 items, with higher Parental psychological distress Parental psychological
scores indicating better quality relationships. Both distress and impairment was assessed using the 12-item
measures have good psychometric properties, including version of the General Health Questionnaire (GHQ-12)
moderate correlations with behavioural ratings of adult- [111]. The GHQ-12 is a widely used screening instru-
child interaction [108-110]. ment designed to detect psychological problems in the
general population. Respondents indicate their state of
Child’s exposure to familial adversity general health over the last four weeks relative to their
Parent-reports at the baseline assessment were used to usual state. For example, the respondent is asked
measure 11 family adversities within five key groups that whether they have lost much sleep over worry over the
are consistently associated with higher rates of mental last four weeks. There are four possible responses ran-
health difficulties among children. ging from ‘not at all’ to ‘much more than usual’ (specific
responses vary depending on the item). In the present
Family socio-economic status Family Socio-Economic study, the standard binary scoring method was used
Status (SES) was measured with five parent-reported [112], for which items are scored as 0-0-1-1. Total scores
variables. The first two variables were mothers’ and can range from 0 to 12, with higher scores indicating
fathers’ level of completed educational qualifications. greater parental psychological distress, and a total score
Third was the family’s paid employment status. To pro- of 1 or more classified as indicating a clinical level of
vide a clearer reflection of economic adversity, the em- psychological distress [113]. The GHQ-12 has well-
ployment status of both the mother and father (where established psychometric properties, including high sen-
present) were combined to reflect the level of full-time sitivity and specificity in detecting psychiatric cases
-equivalent employment within the family, ranging from [112,113].
both parents being employed full-time through to both
parents being unemployed (or where a single parent- Stressful life events Stressful life events occurring
family, that parent was unemployed). Fourth was family within the child’s family were assessed using a modified
receipt of any means-tested government welfare benefits version of the List of Threatening Experiences Question-
for lower-income families. The fifth variable was an indi- naire (LTE-Q) [114,115] that was utilised in the Longitu-
cator of potential economic strain and overcrowding: the dinal Study of Australian Children [116]. The LTE-Q
number of dependent children living with the study asks respondents to report experiencing 12 categories of
child was dichotomised to indicate if the child lived with common negative life events involving moderate or
three or more siblings [66]. marked long-term threat, such as the death of a family
member or friend, or a major financial crisis. The
Parental separation Parents reported the child’s past current study slightly adapted the wording of the LTE-Q
and present living arrangements. Responses on which to identify events occurring within the child’s family unit
parental figures were currently living with the child were rather than for an individual. Wording changes were
dichotomised to reflect a single-parent versus two- based on the Family Inventory of Life Events question-
parent family. Parents also reported whether the child naire [117,118]. For example, one item was changed
Miller-Lewis et al. Child and Adolescent Psychiatry and Mental Health 2013, 7:6 Page 9 of 23
http://www.capmh.com/content/7/1/6

from “you were seeking work unsuccessfully for more tionnaires about these children, in order to allow time to
than one month” to “a parent was seeking work unsuc- get to know the child. Thus teachers had interacted with
cessfully for more than one month”. Parents indicated children for a minimum of 5 weeks before their ratings
whether or not each life event had occurred in the fam- about the child were provided. The average number of
ily over the past 12 months, which was then tallied to months children had been at preschool interacting with the
create a total score ranging from 0 to 12, with higher teacher was 8.08 months (SD = 3.54). The study method-
scores indicating a greater number of stressful life events ology was approved by the Research Ethics Committees at
experienced within the family. The LTE-Q has the Women’s and Children’s Hospital Adelaide, and the
demonstrated good reliability and high sensitivity and South Australian Department of Education and Children’s
specificity to independently rated adversity [114]. Services.

Composite cumulative family adversities index Scores Statistical analyses

on these 11 family adversities were combined to create a We employed four different statistical methodological
cumulative family adversities index. First, a multiple re- approaches to operationally define mental health resili-
gression containing all 11 adversity variables indicated ence in the context of cumulative family adversity, and
that there was no evidence of multicollinearity within analyse correlates thereof. The same hypothesised set of
these adversities, so all 11 adversities were included in child, family, and preschool resource factors for resilient
the cumulative adversity composite score. Bivariate outcomes was examined within each approach, to enable
correlations between each adversity variable and SDQ direct comparison of results between methods. Effect sizes
outcomes indicated that the effect sizes were all in the were interpreted based on standard recommendations
small to trivial range. Given that there was no large vari- [121-123]. Each set of analyses utilised the composite pre-
ation in effect sizes for the adversities, it was deemed school family adversity index to represent family adversity.
reasonable for the composite adversity score to assume In multivariate analyses, child gender was treated as a
equal weights for each of the 11 adversity factors. This covariate. For each different resilience methodology, we
also meant that the composite adversity score used was examined outcomes at age 6 on: (1) parent-reported child
identical for examining parent-reported and teacher- mental health difficulties, and (2) teacher-reported child
reported SDQ outcomes, respectively. To ensure that mental health difficulties. The use of two separate
each of the five groups of family adversities (SES, paren- informants to describe children’s behaviour is important
tal separation, early parenthood, parental psychological because it allows for the consideration of the context
distress, and stressful life events) was equally weighed in within which the behaviour is observed (i.e., home and
the composite index derived, a standardised average school) or the different attributions and perspectives held
score was computed for the groups of family adversities by the informants, and it can provide a test of convergent
that had more than one indicator. Then the scores for validity for hypotheses. Discrepancies among informants
these five groups of family adversities were standardised regarding child psychopathology have been widely
and summed to create a total composite family adversity documented, and in the absence of a gold-standard assess-
index score (see [52,56,119,120]). ment, one cannot determine if parents or teachers are
more accurate reporters [100,124]. As these informants
Procedure both provide distinct information about children’s
Data collection for the baseline assessment was coordinated problems in different contexts, their reports are not inter-
by each Preschool Director, with assistance from the re- changeable. Researchers who have tested several strategies
search team. At the first assessment, preschool teachers of combining informant reports ultimately recommend
gave consenting parents questionnaires, which were keeping the data as separate, as the unique pattern of
returned to the preschools in sealed envelopes once context-dependent features becomes lost, and questions
completed. Teachers completed questionnaires only after regarding context cannot be addressed [124].
parent consent was received. At the second assessment, The first set of analyses followed the ‘statistical inter-
parents were mailed questionnaires, and returned them to action’ approach. Specifically, hierarchical multiple re-
researchers in stamped self-addressed envelopes. Distribu- gression was used to determine whether any of the
tion and collection of teacher surveys was facilitated by a proposed resource factors moderated the relationship
nominated liaison person at each school b. At both between children’s cumulative adversity exposure scores
assessments, parent questionnaires took approximately 30 and their mental health difficulties at age 6. A significant
minutes to complete, and teacher questionnaires took ap- interaction effect whereby higher resource levels reduced
proximately 10 minutes to complete per child. Teachers the effect of adversity on mental health difficulties would
who were allocated newly commencing students were indicate a protective effect. Covariates were entered at
instructed to wait five weeks before completing ques- Step 1. The cumulative family adversity index was
Miller-Lewis et al. Child and Adolescent Psychiatry and Mental Health 2013, 7:6 Page 10 of 23
http://www.capmh.com/content/7/1/6

entered at Step 2, and all resource factors were entered were divided into thirds, as used in other person-centred
simultaneously at Step 3. The interaction terms between resilience investigations (e.g., see [126]). Low and high
family adversity and each resource factor were then adversity were defined as the bottom and top thirds on
entered at Step 4. All continuous variables were centred the cumulative adversity variable, and poor and good ad-
prior to computing interaction terms, and any significant justment were defined as the top and bottom thirds of
interaction terms were then explored further within each SDQ mental health difficulties variable (as this was
plots using Aiken and West’s [125] methods. negatively scored). This tertile approach balances the
The second set of analyses used the ‘residuals’ ap- dual needs of retaining an adequate sample size for each
proach. Linear regression was first used to compute a group, and ensuring the ‘low’ and ‘high’ groups are con-
‘resilience residuals’ variable, which subsequently be- ceptually distinct (which arguably cannot be achieved
came the outcome variable in a second regression with with a median split). Next, comparisons of groups
all resource factors entered as predictor variables. In the showing extreme functioning were planned, to deter-
first regression, the standardised residual ‘resilience’ mine if resource variables showed different effects in
scores (predicted-obtained discrepancies) were generated low- versus high-adversity conditions [12,90]. For this
by regressing children’s level of mental health difficulties reason, children in the middle third on either or both
on their cumulative family adversity score. In the present variables were not retained in further analyses c. Thus
study, in linear regressions the composite family adver- for the fourth set of analyses, the sample was reduced to
sity index accounted for 7.0% (R2 = 0.070, p < .001) and 237 children for parent-reported outcomes and 234 chil-
1.7% (R2 = 0.017, p < .01) of the variance in children’s dren for teacher-reported outcomes. The resulting four
Time 2 parent-reported and teacher-reported SDQ groups were labelled ‘Resilient’, ‘Maladaptive’, ‘Competent’,
difficulties scores, respectively, with greater family adver- and ‘Highly Vulnerable’, using Masten and colleagues
sity associated with greater SDQ difficulties. The [12] taxonomy, and were then compared on their levels
standardised residual scores generated from these two of resources using MANOVA. When the assumption of
regressions were then reverse-coded so that higher homogeneity of variances was violated, the non-
scores indicated greater mental health resilience on a parametric Kruskal-Wallis (for between-groups effects)
continuum from vulnerability through to resilience [87]. and Mann–Whitney Test (for planned comparisons)
As a result, children with positive residual scores (i.e., were used to analyse group differences for those re-
falling above the regression line fitted) showed ‘better source variables.
than expected’ mental health than predicted by their ex-
posure to adversity, and were considered resilient, to at Results
least some degree. Conversely, children whose mental Preliminary analyses
health was ‘worse than expected’ (i.e., a negative re- Table 1 displays the family adversity background
sidual) were considered more vulnerable. The size of the characteristics of participating children. One in 8 chil-
residual (i.e., the distance from the regression line fitted) dren were living in a single parent family, but typically
provided an indication of their level of resilience or vul- children had spent less than a year living separate from
nerability. These residual scores signify the variance in one (or both) of their natural parents in their home.
adjustment that is not explained by adversity, therefore Families had experienced an average of one stressful life
representing the ‘unexplained variance’ inherent within event in the past year. A total of 39.1% primary caregiv-
resilient outcomes [24]. In the subsequent regressions, ing parents scored above the clinical cut-off on GHQ
these resilience residual scores were treated as the out- psychological distress, which is somewhat higher than
come variable and were regressed on the covariates, all the percentage found in national surveys of Australian
entered at Step 1, and the proposed resource factors, all adults [113]. Approximately a quarter of mothers and a
entered at Step 2. third of fathers had not completed high school. How-
The third set of analyses also utilised these resilience ever, most families (84.2%) were supported on at least
residual scores within multiple-group analyses, which one full-time equivalent job. Nonetheless, 40% of fam-
were conducted to determine whether the effect sizes of ilies were receiving a means-tested government welfare
the resource factors on children’s mental health resili- benefit. Overall, the sample did not differ appreciably on
ence residuals differed between children who were demographic characteristics from other children in the
exposed to low- versus high levels of cumulative family general Australian population, with similar rates of wel-
adversity. fare receipt, employment, and single-parent families
The fourth set of analyses used the ‘person-centred’ [97]. All children had experienced at least some degree
approach. Initially, four key groups of children were of adversity (i.e., no child obtained the lowest possible
identified through a two-step process. First, scores on score on all 11 adversity variables), which is important
adversity and the two mental health difficulties variables given that by definition, the presence of adversity is
Miller-Lewis et al. Child and Adolescent Psychiatry and Mental Health 2013, 7:6 Page 11 of 23
http://www.capmh.com/content/7/1/6

required to be able to exhibit resilience (e.g., [16]). construct, the informant discrepancy indicates that the
Scores on the composite family adversity index variable children’s behaviour reported at home and at school may
ranged from −3.67 to +14.44 (M = 0.00, SD = 3.06). be context or informant specific [100,124].
Means and standard deviations for continuous variables
are also shown in Table 2. Children’s mean scores tended Approach I: Statistical Interaction Resilience Methodology
towards the moderate to upper range, suggesting generally Bivariate correlates of Children’s mental health difficulties
healthy child strengths and relationships. Children’s mean The bottom of Table 2 displays the bivariate associations
SDQ total difficulties scores were within the normal range between the Time 1 cumulative adversity index and
[98]. At Time 2 (age 6), a total of 8.2% and 10.1% of proposed resources, and children’s mental health diffi-
children scored above the clinical cut-off on the parent- culties as reported by parents and teachers two years
reported and teacher-reported SDQ total difficulties, later. Higher levels of cumulative family adversity were
respectively. This compares to 5.3% and 5.9% for parent- positively associated with higher child SDQ mental
and teacher-reported SDQ scores at Time 1 (age 4). These health difficulties. Both parent- and teacher-reported
proportions are fairly similar to those found in national child self-concept and child self-control held significant
cohorts of young Australian children [127]. moderate [121] negative correlations with both parent-
We examined the potential role of child gender, child reported and teacher-reported child SDQ mental health
age, and at Time 2, the number of terms at school, school difficulties. Child–parent relationship quality and child-
year level, and school type (government or private), as teacher relationship quality were each significantly nega-
covariates. Only gender was significantly associated with tively associated with both parent-reported and teacher-
Time 2 SDQ total difficulties (r = −.26, p < .001 for reported child SDQ mental health difficulties. Whilst all
teacher-reported outcomes; not significant for parent- bivariate correlations were significant and in the
reported outcomes). Therefore child gender was treated expected direction, an informant effect was notable –
as a covariate in subsequent multivariate analyses. the size of the correlations was considerably larger when
Table 2 shows the bivariate correlations between the the informant was the same for both the predictor and
continuous variables, and the means, standard deviations, outcome variable.
and Cronbach’s alphas for these variables. As expected,
many of the resource variables showed significant moder- Multivariate correlates of Children’s mental health
ate positive intercorrelations. Possible multicollinearity difficulties
was evident between self-esteem and self-efficacy, which Two hierarchical multiple regressions were conducted
were correlated at .64 when reported by parents, and at assessing Time 1 correlates of parent- and teacher-
.86 when reported by teachers. Collinearity diagnostics reported SDQ mental health difficulties at Time 2, re-
within multiple regression showed the presence of spectively. In these regressions, child gender was entered
multicollinearity, particularly for teacher-reported self- as a covariate at Step 1, followed by the cumulative fam-
esteem and self-efficacy [128]. Thus in this sample it was ily adversity index score at Step 2, the main effects of
difficult to distinguish between these two constructs. From the set of proposed resource variables at Step 3, and fi-
a theoretical perspective these constructs originate within nally the interaction terms between family adversity and
a global higher-order self-concept construct, and thus are each of the proposed resource variables at Step 4. d
expected to be highly conceptually interrelated [129]. Results for the regression model predicting parent-
Thus teacher-reported self-esteem and self-efficacy were reported child mental health SDQ difficulties are shown
combined by averaging their standardised scores to create in Table 3. Greater exposure to cumulative family adver-
a composite variable, and the same approach was used for sity at Time 1 was associated with greater parent-
parent-reported self-esteem and self-efficacy. These com- reported SDQ difficulties two years later. At Step 3, all
posite self-concept variables were utilised in subsequent four resource factors held significant main effects as
analyses. These parent- and teacher-reported composite correlates of subsequent parent-reported child SDQ dif-
self-concept variables had means of zero and SDs of 0.91 ficulties: greater child self-concept and self-control, and
and 0.96, respectively. Their bivariate correlations with the better quality relationships with parents and preschool
study variables are shown in Table 2. Also shown in teachers, were each associated with fewer SDQ difficul-
Table 2, the bivariate correlations between parent- ties. The main effects model at Step 3 accounted for
reported and teacher-reported scores on child self- 30.4% of the variance in parent-reported SDQ difficulties
concept, child self-control, and SDQ total difficulties re- scores. The addition of the four interaction terms at Step
spectively were each small but significant and positive. 4 of the model only increased the variance explained by
Parent and teacher reports on the SDQ total difficulties 0.7%. The interaction between family adversity and
score were correlated at .42. While the size of this correl- child-teacher relationship quality approached statistical
ation suggests they may be measuring a similar underlying significance (p = .06) in the parent-reported SDQ
Miller-Lewis et al. Child and Adolescent Psychiatry and Mental Health 2013, 7:6 Page 12 of 23
http://www.capmh.com/content/7/1/6

Table 3 Multiple regressions predicting time 2 child SDQ mental health difficulties (n=474)
Parent-reported child mental health SDQ Teacher-reported child mental health SDQ
difficulties difficulties
Time 1 predictor variables β R2 (ΔR2) ΔF β R2 (ΔR2) ΔF
Step 1: .003 (.003) 1.32 .065 (.065) 32.79***
Gender (female) -.05 -.26***
Step 2: .073 (.070) 35.77*** .083 (.018) 9.16***
Gender (female) -.06 -.26***
Cumulative Adversity Index .27*** .13**
Step 3: .304 (.230) 38.61*** .197 (.114) 16.61***
Gender (female) -.02 -.19***
Cumulative Adversity Index .22*** .08*
a
Self-concept -.26*** -.03
Self-controla -.14** -.18**
Child–parent Relationship Quality -.17*** -.12**
Child-Teacher Relationship Quality -.08* -.15**
Step 4: .311 (.007) 1.22 .209 (.012) 1.80
Gender (female) -.01 -.19***
Cumulative Adversity Index .23*** .07
Self-concepta -.26*** -.02
Self-controla -.15** -.18**
Child–parent Relationship Quality -.16*** -.11**
Child-Teacher Relationship Quality -.09* -.16**
Adversity x self-concept .04 -.06
Adversity x self-control -.02 -.08
Adversity x child–parent relationship -.03 .05
Adversity x child-teacher relationship .08# .06
Note. a Models predicting parent-reported child mental health difficulties included parent-reported self-concept and self-control as predictor variables, and models
predicting teacher-reported child mental health difficulties included teacher-reported self-concept and self-control as predictor variables.
# p < .10; * p < .05; ** p < .01; *** p < .001.

difficulties model. Because this interaction term did quality child-teacher relationships were each associated
reach statistical significance in post-hoc analyses when with lower teacher-reported SDQ difficulties two years
child-teacher relationship was assessed separately from later. There was no significant main effect for self-
the other resource variables (β = .09, p < .05), we concept. The main effects model at Step 3 accounted for
examined the interaction plot (not shown here), which 19.7% of the variance in teacher-reported SDQ difficul-
indicated that while the size of this moderated effect was ties scores. The addition of the four interaction terms at
very small, it suggested a “protective-reactive” effect Step 4 of the model increased the variance explained by
[80,130]; that is, having better child-teacher relationships only 1.2%, and none of these interaction terms were sig-
conferred advantages for better mental health outcomes, nificantly associated with teacher-reported SDQ difficul-
but less so under conditions of high family adversity. ties. e
The regression model predicting teacher-reported Overall, the findings from these interaction model ana-
child SDQ difficulties is also shown in Table 3. lyses suggest that the resource factors examined tended
Regarding main effects, it can be seen that girls had sig- to have a more general promotive capacity for all chil-
nificantly lower teacher-reported child SDQ difficulties dren regardless of their level of exposure to adversity, ra-
than boys. Greater exposure to cumulative family adver- ther than a specific protective effect present only for
sity at Time 1 was associated with greater teacher- children who have faced significant adversity.
reported child SDQ difficulties two years later (Step 2
and 3). The addition of the four proposed resource Approach II: Residuals Resilience Methodology
factors at Step 3 indicated that greater child self-control, Scores on parent-reported child mental health resilience
better quality child–parent relationships, and better generated from the regression residuals ranged from −4.28
Miller-Lewis et al. Child and Adolescent Psychiatry and Mental Health 2013, 7:6 Page 13 of 23
http://www.capmh.com/content/7/1/6

to +2.31, and from −3.67 to +1.64 for teacher-reported gender accounted for 24.9% of the variance in children’s
child mental health resilience. By definition, the stan- parent-reported mental health resilience residuals at age 6.
dardised residuals had mean scores of zero and SDs of 1. Standardised regression coefficients (β) for each predictor
On parent-reported SDQ, 57% of children had better variable in the final model are shown in Table 5. Higher
outcomes than expected (i.e., a positive residual) and thus parent-reported child self-concept and self-control, higher
were considered to be exhibiting at least some degree of quality child–parent relationship, and higher quality child-
mental health resilience. For teacher-reported SDQ, this teacher relationship were significantly associated with
percentage was 60.5%. Scores on parent- and teacher- greater parent-reported child mental health resilience
reported child mental health resilience showed a significant residuals two years later.
moderate positive correlation (r = .40, p < .001). The model predicting teacher-reported child mental
health resilience residuals is also shown for the whole
Bivariate correlates of Children’s mental health resilience sample in Table 5, with the final model accounting
residuals for 18.1% of the variance in scores. Three Time 1
Table 4 shows the bivariate correlations among both predictor variables were significantly associated with
measures of mental health resilience residuals, and Time 2 teacher-reported child mental health resili-
the set of proposed resources for the whole sample ence residuals, in addition to gender. Higher quality
(n= 474). Both parent- and teacher-reported child self- relationship between the child and their parent and
concept and self-control held significant moderate posi- the child and their preschool teacher, and higher
tive correlations with both parent- and teacher-reported teacher-reported child self-control were significantly
mental health resilience residuals. Child–parent relation- associated with greater teacher-reported child mental
ship quality and child-teacher relationship quality were health resilience residuals. f
each significantly positively associated with both parent-
and teacher-reported child mental health resilience Approach III: Multiple-Groups Residuals Resilience
residuals. Again, an informant effect was notable, with Methodology
correlations larger when the informant was the same for The above bivariate correlations and multiple regressions
both variables. predicting resilience residuals were then run separately for
children facing low-adversity (with scores in the bottom
Multivariate correlates of Children’s mental health resilience third on the family adversity composite index) and high-
residuals adversity (with scores in the top third) g , allowing the effect
Two separate hierarchical multiple regressions were sizes for each group to be compared. The bivariate
conducted assessing Time 1 correlates of each of the two correlations for these two sub-groups are shown in Table 4.
measures of mental health resilience residuals: that The effect sizes for each resource variable were generally
reported by parents and that reported by teachers. The small to moderate, and were in most cases very similar
two-step multiple regression model predicting parent- across the groups. The largest difference between the low-
reported child mental health resilience residuals is shown adversity and high-adversity groups were seen on child-
in Table 5 under the sub-heading ‘whole sample’. The final teacher relationship quality – for parent-reported resilience
model including all 4 predictor variables and the covariate residuals, the effect of positive child-teacher relationship

Table 4 Bivariate correlations between resource factors and child mental health resilience residuals – for the whole
sample, and for low and high adversity groups
Parent-reported child mental health resilience residuals Teacher-reported child mental health resilience residuals
Time 1 predictor Whole sample r Low adversity r High adversity r Whole sample r Low adversity r High adversity r
variables (n=474) (n=158) (n=159) (n=474) (n=158) (n=159)
Gender (female) .06 .06 .01 .26*** .26*** .25**
P self-concept .44*** .45*** .42*** .17*** .23** .19*
T self-concept .17*** .21* .19* .29*** .28*** .38***
P self-control .38*** .39*** .37*** .17*** .13# .17*
T self-control .17*** .25** .15# .33*** .29*** .38***
P Child–parent .38*** .43*** .41*** .13** .19* .08
relationship quality
T Child-teacher .14** .31*** .01 .32*** .36*** .26***
relationship quality
Note. P = parent-reported variable; T = teacher-reported variable.
# p < .10; * p < .05; ** p < .01; *** p < .001.
Miller-Lewis et al. Child and Adolescent Psychiatry and Mental Health 2013, 7:6 Page 14 of 23
http://www.capmh.com/content/7/1/6

Table 5 Multiple regressions predicting child mental health resilience residuals – for the whole sample, and for low
and high adversity groups
Parent-reported mental health resilience residuals Teacher-reported mental health resilience residuals
Time 1 predictor variables Whole sample Low adversity High adversity Whole sample Low adversity High adversity
β (n = 474) β (n = 158) β (n = 159) β (n = 474) β (n = 158) β (n = 159)
Step 2:
Gender (female) .02 -.11 .08 .19*** .17* .21**
Self-concept .27*** .27*** .22* .02 .04 .19#
# #
Self-control .14** .14 .16 .18** .07 .24*
#
Child–parent relationship quality .17*** .22** .25** .12** .13 .08
Child-teacher relationship quality .08* .25*** -.04 .15** .24* -.01
R2 .25 .34 .26 .18 .19 .21
F 31.00*** 15.43*** 10.47*** 20.71*** 7.14*** 8.35***
ΔR2 .25 .34 .26 .11 .12 .15
ΔF 38.20*** 19.12*** 13.08*** 16.28*** 5.79*** 7.41***
Note. Models predicting parent-reported child mental health resilience residuals included parent-reported self-concept and self-control as predictor variables, and
models predicting teacher-reported child mental health resilience residuals included teacher-reported self-concept and self-control as predictor variables. Variable
coefficients are standardised regression coefficients (Betas). Step 1 of the model adjusted for the covariate child gender, with the resource variables entered at
Step 2.
# p < .10; * p < .05; ** p < .01; *** p < .001.

was .30 larger for children with low-adversity than for those adversity groups. Among children facing low-adversity
facing high-adversity, indicating a small “reactive” effect (i.e., (but not those facing high-adversity), child-teacher rela-
conferring greater benefit for children with low-adversity). tionship quality was significantly associated with
The multiple regression results for the low-adversity teacher-reported resilience residuals, showing a small
and high-adversity groups are displayed in Table 5. For positive effect. Child-teacher relationship quality re-
the model predicting parent-reported mental health resili- sembled a small “reactive” effect (conferring greater
ence residual scores, there were only small differences in benefit among low-adversity children), with a between-
the effects of resource variables at low and high levels of group beta difference of .25. Among children facing
adversity. Among children experiencing low-adversity, sig- high-adversity (but not those facing low-adversity), self-
nificant Time 1 correlates of parent-reported resilience control was significantly associated with resilience
residuals were self-concept, child–parent relationship residuals, showing a small positive effect. This beta dif-
quality, and child-teacher relationship quality, all showing ference of .17 resembled a small “protective” effect, with
small positive effects [121]. However, significant Time 1 greater self-control conferring greater protection under
correlates among children facing high-adversity were self- conditions of high-adversity.
concept and child–parent relationship quality (with small
positive effects). For child-teacher relationship quality, ef- Approach IV: Person-Centred Resilience Methodology
fect sizes were smaller among children facing high-adversity Formation of adaptation groups
, with a beta difference between the adversity groups of .29. For both parent-reported and teacher-reported outcomes,
Whilst this difference was small, it resembled a “reactive” children were classified into four key groups based on
effect, conferring higher benefits among the group exposed their combinations of scores on the adversity (lowest/
to low-adversity: child-teacher relationship quality did not highest tertile) and mental health difficulties (lowest/
show any benefits for children experiencing high-adversity. highest tertile) variables. When using parent-reported
There was little difference in betas between low- and high- mental health difficulties, this classification yielded 39 Re-
adversity groups on self-concept, self-control, and child– silient (high adversity, low mental health difficulties), 79
parent relationship quality, indicating they could be Maladaptive (high adversity, high mental health difficul-
considered to have small but generally promotive effects. ties), 72 Competent (low adversity, low mental health diffi-
For the model predicting teacher-reported mental culties), and 47 Highly Vulnerable (low adversity, high
health resilience residual scores, there was some small mental health difficulties) children. When using teacher-
differences in the effects of resource variables between reported mental health difficulties, classification yielded
the low- and high-adversity groups. Self-concept and 50 Resilient, 71 Maladaptive, 65 Competent, and 48
child–parent relationship did not reach significance as Highly Vulnerable children. Chi-square tests for inde-
correlates of resilience residuals in low- or high- pendence indicated that the mental health difficulties were
Miller-Lewis et al. Child and Adolescent Psychiatry and Mental Health 2013, 7:6 Page 15 of 23
http://www.capmh.com/content/7/1/6

not evenly distributed across the adversity groups, when (see partial η2 values in Table 6). Also shown in Table 6,
using both parent-reported mental health difficulties χ2 (4) ‘teacher-reported’ adaptation group was significantly
= 18.63 (p < .001), and teacher-reported mental health dif- associated with teacher-reported self-concept, self-control,
ficulties, χ2 (4) = 7.40 (p = .10). For instance, 46% of the and child–parent relationship quality, with small to
children scoring in the lowest adversity group showed low medium effect sizes (see F and partial η2 values in Table 6).
parent-reported mental health difficulties (the ‘Competent’ Child-teacher relationship quality was not significantly
children), whereas only 25% of children scoring in the related to ‘teacher-reported’ adaptation group.
highest adversity group showed low parent-reported child Next, a series of three planned comparisons were made
mental health difficulties (the ‘Resilient’ children), z = 4.02, for each of the univariately significant resource variables,
p < .001. Although less apparent, this effect was present to determine if differences existed between (1) the Resili-
when using teacher-reported mental health difficulties ent and the Maladaptive group, (2) the Resilient and the
(41% of low-adversity children showed low mental health Competent group, and (3) the Competent and Highly
difficulties, compared with 31% of high-adversity children, Vulnerable group. These are first reported for the ‘parent-
z = 1.88, p = .06). Thus, consistent with the definition of reported’ adaptation groups, and then for the ‘teacher-
resilience as ‘unexpected’ positive adaptation, children ex- reported’ adaptation groups. Results are presented in
periencing high levels of adversity were significantly less Table 6.
likely to show low levels of mental health difficulties
compared with their low-adversity counterparts. Parent-reported planned comparisons For parent–
Additionally, MANOVAs were conducted on adversity child relationship quality, Resilient children were rated
and mental health difficulties scores to ensure the cre- significantly higher than Maladaptive children (large effect
ation of groups had worked as intended. As a direct re- size, d = .89). Additionally, Competent children showed
sult of the cut-off method, the Resilient and Competent significantly higher levels than did the Highly Vulnerable
children did not differ on their levels of mental health children (large effect size, d = .97). The Resilient and
difficulties, and the Resilient and Maladaptive children Competent children did not differ. For self-concept, Resili-
did not differ on their levels of adversity. ent children were rated significantly higher than Maladap-
tive children (large effect size, d = .99). Additionally,
Comparison of groups Competent children showed significantly higher levels
Next, MANCOVAs were conducted to determine than Highly Vulnerable children (large effect size,
whether the resource variables were associated with d = 1.07). The Resilient and Competent children did not
parent-reported and teacher-reported adaptation group differ. For self-control, Resilient children were rated sig-
classification. All resource variables held good internal nificantly higher than Maladaptive children (moderate
consistencies when examined for each group separately effect size, d = .76), Competent children showed signifi-
(Cronbach’s Alphas from .75 to .96). Within MANCOVA cantly higher levels than Highly Vulnerable children
(after adjusting for child gender), there was a statistically (moderate effect size, d = .77), and the Resilient and Com-
significant difference on the combined resource variables petent children did not differ. None of the three planned
between (i) the four ‘parent-reported’ adaptation groups comparisons were significant for child-teacher relation-
(Wilks’ λ = .69, F (15, 632) = 6.05, p < .001), and (ii) the ship quality. The overall pattern of group differences
four ‘teacher-reported’ adaptation groups (Wilks’ λ = .76, suggested that all three variables functioned as promotive
F (15, 625) = 4.45, p < .001). Both demonstrated a factors; that is, higher levels of child–parent relationship
medium effect size (partial η2’s of .12 and .09, quality, self-concept and self-control were associated with
respectively). lower levels of mental health difficulties, regardless of the
Mean scores for the parent-reported and teacher- level of adversity experienced.
reported adaptation groups on each resource variable are
displayed in Table 6. The F values in Table 6 represent the Teacher-reported planned comparisons For self-
univariate between-subjects tests for each resource vari- control, the Resilient children were rated as significantly
able, which were adjusted using the Bonferroni procedure. higher than Maladaptive children (moderate effect size,
The effect of the covariate gender was also partialed-out d = .76). However, there was no significant difference be-
of these results (not shown in Table 6 for ease of presenta- tween the Competent children and the Highly Vulnerable
tion). Results indicated that ‘parent-reported’ adaptation children. Furthermore, the Resilient and Competent chil-
group was significantly associated with parent-reported dren did not differ. Altogether, these effects suggested that
self-concept, self-control, child–parent relationship qual- self-control worked as a protective factor, where higher
ity, and child-teacher relationship quality (see F values in self-control levels led to lower levels of mental health diffi-
Table 6). With the exception of child-teacher relationship culties, specifically under conditions of high-adversity ex-
quality, these resource variables showed large effect sizes posure. For self-concept, the Resilient children were rated
Miller-Lewis et al. Child and Adolescent Psychiatry and Mental Health 2013, 7:6 Page 16 of 23
http://www.capmh.com/content/7/1/6

Table 6 MANCOVA results: resource variable means (and SDs) for the four adaptation groups, between-subjects
effects, and planned contrasts
Resource variable 1. 2. 3. 4. Highly- Between-subjects effects Planned
Resilient Maladaptive Competent vulnerable contrasts
F Partial η2
(high adv+ (high adv+ (low adv+ (low adv+
low SDQ) high SDQ) low SDQ) high SDQ)
Parent-reported adaptation groups (n = 237)
(n = 39) (n = 79) (n = 72) (n = 47) df = 3, 233
a
P Child–parent relationship 68.23 (5.80) 62.70 (6.61) 68.84 (4.67) 63.40 (6.37) 45.42*** .19 1 > 2; 1 = 3; 3 > 4
T Child-teacher relationshipa 68.36 (7.14) 66.53 (8.64) 70.03 (5.51) 66.38 (9.41) 7.48* .04 1 = 2; 1 = 3; 3 = 4
P Self-concept 0.39 (0.76) −0.46 (0.94) 0.44 (0.71) −0.39 (0.83) 21.31*** .22 1 > 2; 1 = 3; 3 > 4
P Self-control 22.23 (4.08) 19.09 (4.15) 22.19 (3.52) 19.13 (4.37) 11.84*** .13 1 > 2; 1 = 3; 3 > 4
Teacher-reported adaptation groups (n = 234)
(n = 50) (n = 71) (n = 65) (n = 48) df = 3, 230
P Child–parent relationship 65.16 (7.05) 64.44 (7.04) 67.61 (5.55) 64.78 (6.95) 3.04* .04 1 = 2; 1 = 3; 3 = 4
T Child-teacher relationshipa 68.10 (5.79) 65.70 (9.80) 69.65 (5.30) 66.13 (10.21) 4.45 .04 n/a
T Self-concept a
−0.01 (0.87) −0.60 (1.09) 0.27 (0.66) −0.07 (1.07) 23.51*** .12 1 > 2; 1 = 3; 3 = 4
T Self-controla 24.77 (4.62) 20.90 (5.56) 24.85 (4.71) 22.63 (6.85) 20.43*** .09 1 > 2; 1 = 3; 3 = 4
Note. These results adjust for the covariate gender (not shown for ease of presentation). For the three planned contrasts, all significant group differences found
were at p < .05. Numbers within the planned contrasts column refer to adaptation group shown in the mean scores column headings. P = parent-reported
variable; T = teacher-reported variable; adv = adversity; SDQ = total SDQ mental health difficulties score; n/a = planned contrasts not conducted as no significant
univariate group differences.
a
Non-parametric tests used (Kruskal-Wallis Test for between-subjects, Mann–Whitney Test for paired comparisons) due to unequal variances across groups. In
these cases, a χ2 value is reported instead of an F value.

as significantly higher than Maladaptive children (moder- and above their synchronous association with baseline
ate effect size, d = .60). However, there was no significant mental health difficulties.
difference between the Competent children and the Smaller effect sizes and lower proportion of variance
Highly Vulnerable children. Furthermore, the Resilient explained were evident throughout the sensitivity ana-
and Competent children did not differ. Again, this pattern lyses using all four resilience methodologies, possibly
suggested that self-concept worked as a protective factor, suggesting that the associations are mediated through
where higher self-concept levels led to lower levels of preschool mental health difficulties. The largest differ-
mental health difficulties, specifically under conditions of ence from previously reported results was that the small
high-adversity exposure. For child–parent relationship effects (both protective and promotive) of child-teacher
quality, none of the three planned comparisons were relationship on parent- and teacher-reported mental
significant. health outcomes did not persist when children with pre-
existing mental health difficulties were excluded. Fur-
thermore, the small promotive effect of child–parent
Sensitivity Analyses: Prospective Longitudinal Antecedents relationship quality on teacher-reported mental health
In order to determine the sensitivity of the resource outcomes present on all resilience methodologies in the
factors as predictors of the onset of new mental health former analyses was no longer apparent in the prospect-
difficulties in addition to correlates of the subsequent ive analyses excluding children with clinical-level mental
absolute level of mental health difficulties, we replicated health difficulties at baseline. Nonetheless, several re-
the statistical analyses described above for a reduced source factors were prospectively predictive of subse-
sample of 425 children for whom there was no evidence quent mental health outcomes in the context of
of mental health difficulties at the Time 1 baseline as- adversity. Greater child self-concept, self-control, and
sessment. The 49 children who scored above the clinical child–parent relationship quality continued to demon-
cut-off on either parent- or teacher-reported SDQ diffi- strate small promotive effects on parent-reported child
culties at baseline were excluded. Following the mental health outcomes in all four resilience methodolo-
guidelines of Kraemer and colleagues [131], this sensitiv- gies. On teacher-reported mental health outcomes, self-
ity analysis allowed the examination of whether the control continued to indicate a small protective effect
proposed resource variables could prospectively predict according to the person-centred and the multiple-groups
(as temporal antecedents) the onset and escalation of methodologies and a small promotive effect when exam-
mental health difficulties between age 4 and age 6, over ining statistical interactions. Self-concept continued to
Miller-Lewis et al. Child and Adolescent Psychiatry and Mental Health 2013, 7:6 Page 17 of 23
http://www.capmh.com/content/7/1/6

have little apparent effect on teacher-reported mental adversity [5,18]. The results also highlight the import-
health outcomes. Statistical tables showing these results ance of promoting factors from several systems, includ-
are available upon request from the corresponding ing the family, school, and the child, to achieve the
author. largest benefit [4]. Additionally, in most cases these
resources were found to have predominately promotive
Discussion effects, generally being beneficial for children experien-
This study aimed to identify resource factors during pre- cing both low- and high- levels of adversity. Thus, these
school that were associated with early childhood mental resource factors may be well-suited for use in universal
health resilience in the context of cumulative family adver- prevention strategies, given that promoting high quality
sity. As is conditional within resilience research, we found child-adult relationships, positive child self-concept and
a small positive association between cumulative family ad- good self-control may benefit the mental health of
versity experienced during preschool and the level of all children, regardless of whether they have yet ex-
childhood mental health difficulties reported by parents perienced significant family adversity.
and teachers two years later, consistent with previous re- Consistent with previous research (e.g., [12]), person-
search [12,90]. Inherent to the phenomenon of resilient centred analyses suggested that the Resilient children
outcomes as ‘unexpected adaptation’, the individual were similar to the Competent children (who had lower
differences in how children responded to adversity expos- levels of adversity) on every resource variable. This
ure was demonstrated by a notable proportion of the vari- finding highlights the ‘self-righting tendencies’ within
ance in their mental health difficulties being left human development [26], where children generally
unexplained by family adversity. We found that several re- achieve good outcomes if certain resources are available,
source factors were bivariately associated with these even in the presence of adversity.
‘unexpected’ outcomes, such that higher levels of parent– A unique aspect of the present study was the ability to
child relationship quality, teacher-child relationship qual- directly compare the results of several different resili-
ity, self-concept, and self-control were positively related to ence methodologies. It is noteworthy that the four differ-
resilient outcomes, or ‘better than expected’ mental health ent methodologies utilised led us to fairly similar
outcomes in the context of children’s levels of family ad- conclusions. Even though these methods approached the
versity. Furthermore, greater child–parent relationship operationalisation of resilience in slightly different ways,
quality, greater self-concept, and greater self-control dur- there were only some variations in results. First, the re-
ing preschool were prospectively found to be antecedent source variables showed similar effect sizes on mental
predictors of subsequent mental health difficulties within health outcomes within all methods. Second, the notable
the context of adversity. These variables have consistently reduction in the effect of child-teacher relationship on
been identified as associated with better mental health mental health outcomes in the prospective longitudinal
outcomes for at-risk children (e.g.,[17,26,28,37]), although analyses predicting new mental health difficulties was
rarely examined during the preschool and early school apparent within all resilience methodologies used. Third,
years. Together, these results indicate that the correlates significant resources tended to show predominately pro-
and antecedents of mental health in the context of family motive rather than protective (or interactive) effects on
adversity in young Australian children appear concordant the different methodologies. This convergent evidence
with those found in older children and children in other suggests that results are not necessarily an artefact of
western countries. the type of analysis used. It also provides validation for
While the effect sizes of these resource factors were our operationalisation of resilience, and for the construct
generally small to moderate, they were fairly robust of resilience more broadly.
correlates of resilience, being related to good mental We did find some evidence of protective effects in our
health outcomes in the presence of adversity bivariately, analyses. In variable-centred interactions and multiple-
but also uniquely, when all other resource variables were group methods, child-teacher relationship quality showed
adjusted for (the main exception to this was self-concept a very small protective-reactive effect, conferring greater
in regards to teacher-reported outcomes). Furthermore, benefits for children facing low-adversity than high-
our sensitivity analyses indicated that greater child self- adversity. However this small effect disappeared in sensi-
concept, self-control, and child–parent relationship tivity analyses prospectively predicting the onset of mental
quality were prospective antecedents of subsequent health difficulties. The most robust finding regarding pro-
parent-reported resilient mental health outcomes. These tective effects was for teacher-reported self-control, which
results suggest that more can be gained within interven- demonstrated a protective effect on teacher-reported
tion programmes with every additional resource that is mental health outcomes in the multiple-groups and
promoted. This aligns with the contention that ‘cumula- person-centred methodologies, but not in the potentially
tive protection’ is needed to counteract cumulative lower-powered statistical interaction model. Greater self-
Miller-Lewis et al. Child and Adolescent Psychiatry and Mental Health 2013, 7:6 Page 18 of 23
http://www.capmh.com/content/7/1/6

control during preschool conferred greater protection to possible causal sequences. Examining the potential re-
children under conditions of high-adversity. This effect ciprocal or transactional processes between child-adult
was evident in both the analyses of absolute mental health relationships, self-concept, self-control, and mental
outcomes at age 6, and the sensitivity analyses prospect- health outcomes within family adversity was beyond the
ively predicting age 6 mental health outcomes in children scope of this study.
with no pre-existing mental health difficulties in Third, a limitation of key pertinence for studies of resili-
preschool. ence is that within our community-based sample, few chil-
It must be noted that an ‘informant’ effect was present dren had experienced very high levels of family adversity.
within our findings: although both parent-reported and pre- While our cumulative family adversity index showed
school teacher-reported resource factors were bivariately sufficient variability, and the association between adversi-
related to children’s mental health difficulties as reported by ties and mental health difficulties was similar in magni-
either parents or school teachers, the effect sizes were lar- tude to those found in other studies (e.g., [12,90]), overall
ger when the informant-type was the same, and some of this association was relatively weak, particularly in com-
the associations between variables assessed with different parison to those between some resources and mental
informants diminished to non-significance in multivariate health. The ability to detect the degree of moderation by
analyses. Overall, the strongest effects were detected when resource factors on the association between adversity and
assessing associations between parent-reported resources mental health may have been compromised by the modest
and subsequent mental health reported by the parent. Con- level of cumulative adversity in the sample [66]. Further
sequently, the associations found may be partly due to exacerbating the underrepresentation of children facing
shared method variance for parent-reported mental health higher levels of adversity in this study was the higher rate
outcomes. However, this informant effect may also be a re- of attrition for children facing greater adversity. Therefore,
sult of children’s self-concept, self-control, and mental our results may be less generalisable to children facing
health being context specific. Although parent and teacher- great adversity, and it is unclear whether the resources we
reported resources were not strongly associated with each identified as promotive would maintain their beneficial
other, it seems they were related to mental health outcomes effects at extremely high levels of adversity. It is also not
in a similar manner. In sum, although children’s self- known if our findings would generalise to children in
concept and self-control may manifest or be perceived dif- other regions, or whether within-preschool clustering
ferently at home and at preschool, the manners in which effects influenced our results.
they influence their mental health appear to be similar. Given that only a minority of children experienced both
The findings of this study should be interpreted within high adversity and low mental health difficulties within
the context of the following limitations. First, whilst a person-centred analyses illustrates that although a number
strength of this study was the inclusion of reports from of children showed resilient functioning, they were
two informants, our sole reliance on survey methodology fighting against the odds. This highlights the need for fur-
poses limitations on the interpretations of our findings. It ther research to determine how such children manage to
is possible that the use of direct observations, child transcend their family circumstances, when many others
interviews, or diagnostic interviews may have changed the become engulfed by them. It would be worthwhile to de-
pattern of results. These more objective assessment termine whether preschool-age relationships, self-concept,
methods were unfortunately beyond the scope of this and self-control are equally important for different subcat-
study. The reliance on parent and teacher informant egories of mental health difficulties in young children (e.g.,
reports also meant parents and teachers had to infer internalising and externalising problems). More research
children’s internal self-beliefs and emotions from behav- on the preschool-age correlates of resilience within other
ioural manifestations associated with these internal child developmentally relevant domains, such as social and aca-
constructs [16]. It is unclear how accurately parents’ and demic functioning, would provide a more complete pic-
teachers’ perceptual judgements regarding these internal ture of resilience during early childhood. Because the
characteristics would correspond to children’s own self- resource factors included in this study did not explain the
assessments (although our ability to obtain accurate infor- majority of the variance in mental health, it is clear that
mation from children themselves is limited by the young other resource variables are involved in the development
age of the sample). of mental health resilience. It would be worthwhile to
Second, even though this study included analyses that examine the role of relationships with other important
examined longitudinal preschool correlates of absolute adults (e.g., grandparents, regular carers), and the role of
levels of mental health outcomes at school, and pro- other potential internal characteristics (e.g., optimism) in
spectively antecedent preschool predictors of the onset understanding early childhood resilience to mental health
or escalation of mental health difficulties once at school, difficulties. Furthermore, the role of biological processes is
our results can still only suggest but not confirm a recently burgeoning field within resilience research, and
Miller-Lewis et al. Child and Adolescent Psychiatry and Mental Health 2013, 7:6 Page 19 of 23
http://www.capmh.com/content/7/1/6

future investigations would benefit from the consideration factors should be prioritised within intervention strat-
of such factors alongside other child, family, and wider so- egies [81,132].
cial factors previously implicated in the development of
resilient outcomes [20,66,120]. Researchers should also Conclusion
conduct prospective longitudinal research in order to in- In drawing the findings from this multiple-methodology
vestigate the temporal precedence of preschool-age re- study together, the many internal child and external en-
source factors and subsequent mental health resilience, vironmental resources for child mental health resilience
and whether these resources are able to predict change identified in this study reinforces that early intervention
over time in resilience. Examining the accumulation of strategies developed will need to be multifaceted in
family adversities over several time points rather than just order to address the complexities of the development of
one, and the use of weights on each adversity based on childhood resilience. Boosting positive child-adult
the size of their association with mental health, would also relationships, self-concept and self-control as resources
provide a more complete and realistic picture of the influ- in early childhood may hold promise for helping chil-
ence of family adversity on mental health in young dren establish a firm foundation that will carry them for-
children. ward into healthy futures, regardless of what adverse
There are several other worthwhile avenues for future family circumstances come their way.
research. Given that the methodologies we used in com-
bination provided a holistic view of resilience in young Endnotes
a
children, we urge other researchers to conduct multiple- Preschool is a government-funded programme which
methodology resilience studies. This will help to deter- is available to all four year-old children in the year im-
mine if any results found are likely to be reflecting real mediately prior to commencing formal schooling. In this
effects, or if they are artefacts of the analysis method. 12 month period, 11 to 15 hours per week of preschool
Further evidence of convergent findings across methods education is provided free of charge. While attending
will bolster the construct validity of resilience [80]. Al- preschool in South Australia is not compulsory, most
ternatively, if other studies do not find such conver- children do: approximately 93% of eligible four year olds
gence, the pattern of results may provide researchers attended government-funded state preschools in South
with important information on differences in resilience Australia from 2006–2007 [133].
b
measurement techniques, and perhaps eventually lead Children were tracked regardless of their school des-
the field towards the adoption of a consistent method- tination and were attending 92 different schools across
ology. As there are few multiple-methodology resilience Australia. The majority (69%) were attending govern-
studies from which to draw firm conclusions, much ment schools in the same district that they had attended
more research needs to be conducted in this area. preschool. Schools with less than 5 participating children
In the present study we found evidence of child, fam- were directly sent teacher surveys by mail.
c
ily, and school resources for mental health resilience, as We also conducted analyses including children within
well as indication that these resources were interrelated. the ‘middle’ groups, to enable more direct comparison
Furthermore, in some cases the effects of particular between resilience methods by using the same sample
resources were diminished when several resources were size. However, results were almost identical to those that
considered together in multivariate analyses. Such did not include the ‘middle’ group. For ease of presenta-
findings suggest that future resilience studies should tion, only the results for the 4 extreme groups are
examine the possible mediating processes by which re- reported here.
d
source factors exert their effects on mental health We also tested models including both parent-reported
[81,132]. For example, Luthar and Brown [132] assert and teacher-reported self-concept and self-control as pre-
that “relationships lie at the roots of resilience . . . the dictor variables of SDQ difficulties, in addition to the
presence of support, love, and security fosters resilience informant-specific models presented here. The addition of
in part by reinforcing people’s innate strengths” (p.947). both informants on these predictor variables and their
The resilience residuals methodology in combination interaction terms made little improvement to the variance
with multiple-groups resilience methodology provides a explained in the models (R2 increases of .005 and .005 for
valuable platform for the examination of meditational parent-reported and teacher-reported child SDQ difficulties
pathways and processes at play between the resource respectively), and made very little change to the size of
factors in their prediction of resilient outcomes. Because most β coefficients.
e
resources found to catalyse the development of other In post-hoc analyses for both parent-reported and
resources are most likely to have the greatest benefits teacher-reported SDQ difficulties, we also examined
within interventions, information generated from such each predictor variable separately in a series of hierarch-
studies may provide valuable insight into which resource ical multiple regressions to determine their total main
Miller-Lewis et al. Child and Adolescent Psychiatry and Mental Health 2013, 7:6 Page 20 of 23
http://www.capmh.com/content/7/1/6

effects and interactions with adversity. Whilst all of the Adelaide, South Australia 5005, Australia. 4Public Health Research Unit,
predictor variables showed significant main effects on Women’s and Children’s Health Network, 72 King William Road, North
Adelaide, South Australia 5006, Australia. 5School of Population Health,
both parent-reported and teacher-reported child mental University of Adelaide, Adelaide, South Australia 5005, Australia. 6Child
health difficulties, none of the interaction terms were Development Center, Nationwide Children’s Hospital and Ohio State
statistically significant, with the exception of the inter- University, Columbus, Ohio, USA.

action between adversity and child-teacher relationship Received: 14 September 2012 Accepted: 14 February 2013
in its association with parent-reported SDQ difficulties. Published: 22 February 2013
f
We also tested 7-variable models including both
parent-reported and teacher-reported self-concept and References
self-control as predictor variables of mental health resili- 1. Costello EJ, Egger H, Angold A: 10-year research update review: The
epidemiology of child and adolescent psychiatric disorders: I. Methods
ence residuals, in addition to the 5-variable models and public health burden. J Am Acad Child Adolesc Psychiatry 2005,
presented. The addition of both informants on these pre- 44(10):972–986.
dictor variables made little improvement to the variance 2. Sawyer MG, Arney F, Baghurst P, Clark J, Graetz B, Kosky R, Nurcombe B,
Patton G, Prior M, Raphael B, Rey J, Whaites L, Zubrick S: The Mental Health
explained in the models (R2 increases of .006 and .005 of Young People in Australia. Child and Adolescent Component: National
for parent-reported and teacher-reported child mental Survey of Mental Health and Well-being. Canberra: Mental Health and Special
health resilience respectively), and made very little Programs Branch, Commonwealth Department of Health and Aged Care;
2000.
change to the size of most β coefficients. 3. Caspi A, Moffitt TE, Newman DL, Silva PA: Behavioral observations at age
g
Children classified in the high adversity group 3 years predict adult psychiatric disorders: Longitudinal evidence from a
(highest tertile on the adversity composite index) had in- birth cohort. Arch Gen Psychiatry 1996, 53(11):1033–1039.
4. Luthar SS, Cicchetti D: The construct of resilience: Implications for
deed experienced significantly high levels of family ad- interventions and social policies. Dev Psychopathol 2000, 12(4):857–885.
versity, e,g., 85% of this group had experienced 3 or 5. Masten AS, Gewirtz AH: Resilience in development: The importance of
more adversities, compared to 0% in the low adversity early childhood. In Encyclopedia on Early Childhood Development. Volume
2009. Edited by Tremblay RE, Barr RG, Peters RD. Montreal, Quebec: Centre
group and 8.5% in the moderate adversity group. Of the of Excellence for Early Childhood Development; 2006:1–6. http://www.child-
children in the high adversity group, 88% lived in a fam- encyclopedia.com/Pages/PDF/Masten-GewirtzANGxp.pdf.
ily receiving a government benefit, 42% were living in a 6. Heckman JJ: The real question is to how use the available funds wisely. The
best evidence supports the policy prescription: Invest in the very young.
single-parent family, 51% of mothers and 64% of fathers Chicago: Ounce of Prevention Fund and the University of Chicago Harris
had not completed high school, 54% had a parent scor- School of Public Policy Analysis; 2000.
ing above the clinical cut-off on the GHQ, and 43% had 7. Amato PR: Children of divorce in the 1990s: An update of the Amato and
Keith (1991) meta-analysis. J Fam Psychol 2001, 15(3):355–370.
experienced 2 or more stressful life events in the past 8. Beardslee WR, Versage EM, Gladstone TR: Children of affectively ill parents:
12 months. A review of the past 10 years. J Am Acad Child Adolesc Psychiatry 1998,
37(11):1134–1141.
Competing interests 9. Bradley RH, Corwyn RF: Socioeconomic status and child development.
The authors declare that they have no competing interests. Annu Rev Psychol 2002, 53:371–399.
10. Moffitt TE: Teen-aged mothers in contemporary Britain. J Child Psychol
Authors’ contributions Psychiatry 2002, 43:727–742.
L.M-L. contributed to the conception and design of the study, data 11. Fergusson DM, Horwood JL: The christchurch health and development
collection, data analysis and interpretation, and drafted the manuscript. A.S. study: review of findings on child and adolescent mental health.
assisted in the design, data collection, data analysis and interpretation for the Aust N Z J Psychiatry 2001, 35(3):287–296.
study, and helped draft the manuscript. M.S. participated in the design of the 12. Masten AS, Hubbard JJ, Gest SD, Tellegen A, Garmezy N, Ramirez M:
study, interpretation of data, and drafting and revising the manuscript. P.B. Competence in the context of adversity: Pathways to resilience and
participated in the design of the study, provided statistical advice and data maladaptation from childhood to late adolescence. Dev Psychopathol
interpretation, and critically revised the manuscript. D.H. assisted with data 1999, 11:143–169.
analysis and helped to draft and revise the manuscript. All the authors have 13. Mistry RS, Benner AD, Biesanz J, Clark S, Howes C: Family and social risk,
read and approved the final manuscript. and parental investments during the early childhood years as predictors
of low-income children’s school readiness outcomes. Early Child Res Q
Acknowledgements 2010, 25:432–449.
We would like to acknowledge the contributions of Mike Hudson, Debra Kay, 14. Shaw DS, Vondra JI, Hommerding KD, Keenan K, Dunn M: Chronic family
Richard Costi, and other members of our Project Advisory Team from the adversity and early child behavior problems: A longitudinal study of low
South Australian Department of Education and Children’s Services regarding income families. J Child Psychol Psychiatry 1994, 35:1109–1122.
the design and conduct of this research. We would like to thank the 15. Sameroff AJ, Rosenblum KL: Psychosocial constraints on the development
teachers and families who participated in this research. We are also grateful of resilience. Annals New York Acad Sci 2006, 1094:116–124.
to Nancy Briggs for statistical advice. This research was supported by funding 16. Rutter M: Resilience reconsidered: Conceptual considerations, empirical
from the National Health and Medical Research Council, Channel 7 Children’s findings and policy implications. In Handbook of Early Childhood
Research Foundation South Australia, Australian Rotary Health, and University Intervention. 2nd edition. Edited by Shonkoff JP, Meisels SJ. New York:
of Adelaide’s Faculty of Health Sciences. Cambridge University Press; 2000:651–682.
17. Luthar SS: Resilience in development: A synthesis of research across five
Author details decades. In Developmental Psychopathology: Risk, Disorder, and Adaptation.
1
Discipline of Paediatrics, School of Paediatrics and Reproductive Health, 2nd edition. Edited by Cicchetti D, Cohen DJ. New York: Wiley;
University of Adelaide, Adelaide, South Australia 5005, Australia. 2Research 2006:739–795. Volume 3.
and Evaluation Unit, Women’s and Children’s Health Network, 72 King 18. Masten AS, Cutuli JJ, Herbers JE, Reed M-GJ: Resilience in development. In
William Road, North Adelaide, South Australia 5006, Australia. 3Centre for Oxford Handbook of Positive Psychology. 2nd edition. Edited by Snyder CR,
Traumatic Stress Studies, School of Population Health, University of Adelaide, Lopez SJ. New York: Oxford University Press; 2009:117–131.
Miller-Lewis et al. Child and Adolescent Psychiatry and Mental Health 2013, 7:6 Page 21 of 23
http://www.capmh.com/content/7/1/6

19. Luthar SS, Zelazo LB: Research on resilience: An integrative review. In and adaptation in a high-risk sample. Dev Psychopathol 2010,
Resilience and vulnerability: Adaptation in the context of childhood adversities. 22(3):539–555.
Edited by Luthar SS. New York: Cambridge University Press; 2003:510–549. 42. Zaslow MJ, Dion MR, Morrison DR, Weinfield N, Ogawa J, Tabors P:
20. Cicchetti D: Resilience under conditions of extreme stress: a multilevel Protective factors in the development of preschool-age children of
perspective. World Psychiatry 2010, 9(3):145–154. young mothers receiving welfare. In Coping with Divorce, Single Parenting,
21. Yates TM, Egeland B, Sroufe LA: Rethinking resilience: A developmental and Remarriage: A risk and resiliency perspective. Edited by Hetherington E.
process perspective. In Resilience and vulnerability: Adaptation in the context New Jersey: Lawrence Erlbaum; 1999:193–223.
of childhood adversities. Edited by Luthar SS. New York: Cambridge 43. Lengua LJ: The Contribution of emotionality and self-regulation to the
University Press; 2003:243–266. understanding of children’s response to multiple risk. Child Dev 2002,
22. Bronfenbrenner U, Morris PA: The ecology of developmental processes. In 73(1):144–161.
Handbook of child psychology: Theoretical models of human development. 5th 44. Maughan A, Cicchetti D, Toth SL, Rogosch FA: Early-occurring maternal
edition. Edited by Damon W, Lerner RM. New York, NY: Wiley; depression and maternal negativity in predicting young children’s
1998:993–1028. Volume 1. emotion regulation and socioemotional difficulties. J Abnorm Child
23. Masten AS, Powell JL: A resilience framework for research policy and Psychol 2007, 35(5):685–703.
practice. In Resilience and vulnerability: Adaptation in the context of childhood 45. Maughan A, Cicchetti D: Impact of child maltreatment and interadult
adversities. Edited by Luthar SS. New York: Cambridge; 2003:1–25. violence on children’s emotion regulation abilities and socioemotional
24. Kolar K: Resilience: revisiting the concept and its utility for social adjustment. Child Dev 2002, 73(5):1525–1542.
research. Int J Ment Health Addict 2011, 9(4):421–433. 46. Buckner JC, Mezzacappa E, Beardslee WR: Characteristics of resilient youths
25. Werner EE, Smith RS: Vulnerable but invincible: A longitudinal study of resilient living in poverty: The role of self-regulatory processes. Dev Psychopathol
children and youth. New York: Adams, Bannister, and Cox; 1989. 2003, 15(1):139–162.
26. Werner EE, Smith RS: Overcoming the odds: High risk children from birth to 47. Martinez-Torteya C, Bogat AG, von Eye A, Levendosky AA: Resilience
adulthood. Ithaca: Cornell University; 1992. among children exposed to domestic violence: The role of risk and
27. Elder GH, Conger RD: Children of the land: Adversity and success in rural protective factors. Child Dev 2009, 80(2):562–577.
America. Chicago: University of Chicago Press; 2000. 48. Mathiesen KS, Prior M: The impact of temperament factors and family
28. Cowen EL, Work WC, Wyman PA, Parker GR, Wannon M, Gribble PA: Test functioning on resilience processes from infancy to school age. Eur J Dev
comparisons among stress-affected, stress-resilient and nonclassified Psychol 2006, 3(4):357–387.
fourth through sixth grade urban children. J Community Psychol 1992, 49. Obradović J: Effortful control and adaptive functioning of homeless
20:200–214. children: Variable-focused and person-focused analyses. J Appl Dev
29. Hoyt-Meyers LA, Cowen EL, Work WC, Wyman PA, Magnus KB, Fagen DB, Psychol 2010, 31:109–117.
Lotyczewski BS: Test correlates of resilient outcomes among highly 50. Kim J, Cicchetti D: Longitudinal pathways linking child maltreatment,
stressed 2nd and 3rd grade urban children. J Community Psychol 1995, emotion regulation, peer relations, and psychopathology. J Child Psychol
23:326–338. Psychiatry 2010, 51(6):706–716.
30. Smith J, Prior M: Temperament and stress resilience in school-age 51. Armstrong S, Buckley S, Lonsdale M, Milgate G, Bennetts Kneebone L,
children: A within-families study. J Am Acad Child Adolesc Psychiatry 1995, Cook L, Skelton F: Starting school: A strengths-based approach towards
34(2):168–179. Aboriginal and Torres Strait Islander children. A report prepared by the
31. Cicchetti D, Rogosch FA: The role of self-organization in the promotion of Australian Council for Educational Research, for the Department of
resilience in maltreated children. Dev Psychopathol 1997, 9:797–815. Families, Housing, Community Services and Indigenous Affairs. Australia:
32. Gaylord-Harden NK, Ragsdale BL, Mandara J, Richards MH, Petersen AC: Australian Council for Educational Research; 2012.
Perceived support and internalizing symptoms in African American 52. Lengua LJ, Bush NR, Long AC, Kovacs EA, Trancik AK: Effortful control
adolescents: Self-esteem and ethnic identity as mediators. J Youth as a moderator of the relation between contextual risk factors
Adolesc 2007, 36:77–88. and growth in adjustment problems. Dev Psychopathol 2008,
33. Kim J, Cicchetti D: A longitudinal study of child maltreatment, 20:509–528.
mother-child relationship quality and maladjustment: The role of 53. Erickson MF, Sroufe LA, Egeland B: The relationship between quality of
self-esteem and social competence. J Abnorm Child Psychol 2004, attachment and behavior problems in preschool in a high-risk sample.
32(4):341–354. Monogr Soc Res Child Dev 1985, 50(1/2):147–166.
34. Losel F, Bliesener T: Some high-risk adolescents do not develop conduct 54. Egeland B, Kalkoske M, Gottesman N, Erickson MF: Preschool
problems: A study of protective factors. Int J Behav Dev 1994, 17:753–777. behavior problems: stability and factors accounting for change.
35. Sandler IN, Wolchik S, Davis C, Haine R, Ayers T: Correlational and J Child Psychol Psychiatry 1990, 31(6):891–909.
experimental study of resilience in children of divorce and parentally 55. Egeland B, Pianta R, Obrien MA: Maternal intrusiveness in infancy and
bereaved children. In Resilience and vulnerability: Adaptation in the context child maladapation in early school years. Dev Psychopathol 1993,
of childhood adversities. Edited by Luthar SS. New York: Cambridge 5(3):359–370.
University Press; 2003:213–240. 56. Vanderbilt-Adriance E, Shaw DS: Protective factors and the development
36. Pargas RCM, Brennan PA, Hammen C, Le Brocque R: Resilience to maternal of resilience in the context of neighborhood disadvantage. J Abnorm
depression in young adulthood. Dev Psychol 2010, 46(4):805–814. Child Psychol 2008, 36(6):887–901.
37. Curtis WJ, Cicchetti D: Emotion and resilience: A multilevel 57. Egeland B, Carlson E, Sroufe LA: Resilience as process. Dev Psychopathol
investigation of hemispheric electroencephalogram asymmetry and 1993, 5:517–528.
emotion regulation in maltreated and nonmaltreated children. 58. Zaslow MJ, Weinfield NS, Gallagher M, Hair EC, Ogawa JR, Egeland B,
Dev Psychopathol 2007, 19(3):811–840. Tabors PO, De Temple JM: Longitudinal prediction of child
38. LeBuffe PA, Naglieri JA: Devereux Early Childhood Assessment Program (DECA). outcomes from differing measures of parenting in a low-income
Lewisville, NC: Kaplan Press; 1999. sample. Dev Psychol 2006, 42(1):27–37.
39. Miller AL, Seifer R, Stroud L, Sheinkopf SJ, Dickstein S: Biobehavioral 59. Chang H, Olson SL, Sameroff AJ, Sexton HR: Child effortful control as
indices of emotion regulation relate to school attitudes, a mediator of parenting practices on externalizing behavior:
motivation, and behavior problems in a low-income preschool evidence for a Sex-differentiated pathway across the transition
sample. In Resilience in Children. Edited by Lester BM, Masten AS, from preschool to school. J Abnorm Child Psychol 2011, 39(1):71–81.
McEwen B. Oxford: Blackwell Publishing; 2006:325–329. Volume 1094. 60. Dunn J, Cheng H, O’Connor TG, Bridges L: Children’s perspectives on their
40. Myers SS, Morris AS: Examining associations between effortful relationships with their nonresident fathers: Influences, outcomes and
control and teacher-child relationships in relation to Head Start implications. J Child Psychol Psychiatry 2004, 45(3):553–566.
children’s socioemotional adjustment. Early Educ Dev 2009, 61. Wolchik SA, Tein JY, Sandler IN, Ayers TS: Stressors, quality of the child-
20(5):756–774. caregiver relationship, and children’s mental health problems after
41. Yates TM, Obradovic J, Egeland B: Transactional relations across parental death: The mediating role of self-system beliefs. J Abnorm Child
contextual strain, parenting quality, and early childhood regulation Psychol 2006, 34(2):221–238.
Miller-Lewis et al. Child and Adolescent Psychiatry and Mental Health 2013, 7:6 Page 22 of 23
http://www.capmh.com/content/7/1/6

62. Ge X, Natsuaki MN, Neiderhiser JM, Reiss D: The longitudinal effects of 87. Kim-Cohen J, Moffitt TE, Caspi A, Taylor A: Genetic and environmental
stressful life events on adolescent depression are buffered by parent– processes in young children’s resilience and vulnerability to
child closeness. Dev Psychopathol 2009, 21:621–635. socioeconomic deprivation. Child Dev 2004, 75(3):651–668.
63. Crittenden PM: Maltreated infants - vulnerability and resilience. J Child 88. Bowes L, Maughan B, Caspi A, Moffitt TE, Arseneault L: Families promote
Psychol Psychiatry Allied Discip 1985, 26(1):85–96. emotional and behavioural resilience to bullying: evidence of an
64. O’Grady D, Metz JR: Resilience in children at high risk for psychological environmental effect. J Child Psychol Psychiatry 2010, 51(7):809–817.
disorder. J Pediatr Psychol 1987, 12(1):3–23. 89. Sharkey JD, You S, Schnoebelen K: Relations among school assets,
65. Malmberg L-E, Flouri E: The comparison and interdependence of individual resilience, and student engagement for youth grouped
maternal and paternal influences on young Children’s behavior by level of family functioning. Psychol Sch 2008, 45(5):402–418.
and resilience. J Clin Child Adolesc Psychol 2011, 40(3):434–444. 90. Schoon I: Risk and resilience: Adaptations in changing times. Cambridge:
66. Calkins SD, Blandon AY, Williford AP, Keane SP: Biological, Cambridge University Press; 2006.
behavioural, and relational levels of resilience in the context of 91. Luthar SS, D’Avanzo K, Hites S: Maternal drug abuse versus other
risk for early childhood behavior problems. Dev Psychopathol 2007, psychological disturbances: Risks and resilience among children. In
19:675–700. Resilience and vulnerability: Adaptation in the context of childhood adversities.
67. Peisner-Feinberg ES, Burchinal MR, Clifford RM, Culkin ML, Howes C, Kagan Edited by Luthar SS. New York: Cambridge University Press; 2003:104–129.
SL, Yazejian N: The relation of preschool child-care quality to children’s 92. Owens EB, Shaw DS: Poverty and early childhood adjustment. In Resilience
cognitive and social developmental trajectories through second grade. and vulnerability: Adaptation in the context of childhood adversities. Edited by
Child Dev 2001, 72(5):1534–1553. Luthar SS. New York: Cambridge University Press; 2003:267–292.
68. Howes C, Ritchie S: Attachment organizations in children with difficult life 93. Condly SJ: Resilience in children: A review of literature with implications
circumstances. Dev Psychopathol 1999, 11(2):251–268. for education. Urban Educ 2006, 41(3):211–236.
69. Johnson B: Teacher-student relationships which promote resilience at 94. Leontopoulou S: Resilience of greek youth at an educational transition
school: a micro-level analysis of students’ views. British J Guid Couns 2008, point: The role of locus of control and coping strategies as resources.
36(4):385–398. Soc Indic Res 2006, 76:95–126.
70. Theron LC, Engelbrecht P: Caring Teachers: Teacher–Youth Transactions 95. Masten AS, Burt KB, Roisman GI, Obradović J, Long JD, Tellegen A:
to Promote Resilience. In The Social Ecology of Resilience. Edited by Ungar Resources and resilience in the transition to adulthood: Continuity and
M. New York: Springer; 2012:265–280. change. Dev Psychopathol 2004, 16:1071–1094.
71. Sameroff AJ, Gutman LM, Peck SC: Adaptation among youth facing 96. Li H, Martin AJ, Armstrong D, Walker R: Risk, protection, and resilience in
multiple risks: prospective research findings. In Resilience and Chinese adolescents: A psycho-social study. Asian J Soc Psychol 2011,
vulnerability: Adaptation in the context of childhood adversities. Edited 14(4):269–282.
by Luthar SS. New York: Cambridge University Press; 2003:364–391. 97. Australian Bureau of Statistics: Catalogue 4442.0: Family Characteristics
72. Howard S, Dryden J, Johnson B: Childhood resilience: Review and critique Australia, 2003. Canberra: Commonwealth of Australia; 2004.
of literature. Oxf Rev Educ 1999, 25(3):307–323. 98. Goodman R: The strengths and difficulties questionnaire: a research
73. Liebenberg L, Ungar M: Resilience in Action: Working With Youth Across note. J Child Psychol Psychiatry 1997, 38:581–586.
Cultures and Contexts. Toronto: University of Toronto Press; 2008. 99. Goodman R: Psychometric properties of the strengths and difficulties
74. Ungar ME: The Social Ecology of Resilience. New York: Springer; 2012. questionnaire. J Am Acad Child Adolesc Psychiatry 2001, 40:1337–1345.
75. Emerson E, Einfeld S: Emotional and behavioural difficulties in 100. Stone LL, Otten R, Engels RCME, Vermulst AA, Janssens JMAM:
young children with and without developmental delay: a bi- Psychometric properties of the parent and teacher versions of the
national perspective. J Child Psychol Psychiatry 2010, 51(5):583–593. strengths and difficulties questionnaire for 4- to 12-year-olds: a
76. Mutimer A, Reece JE, Matthews J: Child Resilience: Relationships review. Clin Child Fam Psychol Rev 2010, 13:254–274.
between stress, adaptation and family functioning. E J Appl Psychol 101. Warnick EM, Bracken MB, Kasl S: Screening efficiency of the child behavior
2007, 3(1):16–25. checklist and strengths and difficulties questionnaire: a systematic
77. Smart D, Vassallo S, Sanson A, Richardson N, Dussuyer I, McKendry B, review. Child Adolesc Ment Health 2008, 13(3):140–147.
Toumbourou J, Prior M, Oberklaid F: Patterns and precursors of adolescent 102. Fall M, McLeod E: Identifying and assisting children with low self-efficacy.
antisocial behaviour: Types, resiliency and environmental influences. Prof Sch Couns 2001, 4:334–342.
Melbourne: Crime Prevention Victoria; 2003. 103. Hughes HM, Pugh R: The behaviour rating form-revised: A parent-report
78. Smart D, Sanson A, Baxter J, Edwards B, Hayes A: Home-to-school transitions measure of children’s self-esteem. J Clin Psychol 1984, 40:1001–1005.
for financially disadvantaged children. Sydney: Australian Institute of Family 104. Plake BS, Impara JC, Spies RA: The 15th mental measurement yearbook.
Studies & The Smith Family; 2008. Lincoln: Buros Institute of Mental Measurements; 2003.
79. Treyvaud K, Inder TE, Lee KJ, Northam EA, Doyle LW, Anderson PJ: Can the 105. Lebuffe PA, Shapiro VB: Lending “strength” to the assessment of
home environment promote resilience for children born very preterm in preschool social-emotional health. Calif Sch Psychol 2004, 9:51–61.
the context of social and medical risk? J Exp Child Psychol 2012, 106. NICHD Early Child Care Research Network: NICHD study of Early Child Care
112(3):326–337. and Youth Development: Phase 2 Instrument Document. ; http://secc.rti.org/
80. Luthar SS, Cicchetti D, Becker B: The construct of resilience: A Phase2InstrumentDoc.pdf.
critical evaluation and guidelines for future work. Child Dev 2000, 107. Pianta RC: Child–Parent Relationship Scale. Charlottesville, Virginia: Curry
71(3):543–562. School of Education, University of Virginia; 1995.
81. Davydov DM, Stewart R, Ritchie K, Chaudieu I: Resilience and mental 108. Pianta RC: Student-Teacher Relationship Scale. Florida: Psychological
health. Clin Psychol Rev 2010, 30(5):479–495. Assessment Resources Inc.; 2001.
82. Rutter M: Psychosocial resilience and protective mechanisms. Am J 109. NICHD Early Child Care Research Network: Trajectories of physical
Orthopsychiatry 1987, 57(3):316–331. aggression from toddlerhood to middle childhood. Monogr Soc Res Child
83. Garmezy N, Masten AS, Tellegen A: The study of stress and competence in Dev 2004, 69(4):vii–129.
children: A building block for developmental psychopathology. Child Dev 110. NICHD Early Child Care Research Network: Does class size in first
1984, 55(1):97–111. grade relate to children’s academic and social performance or
84. Luthar SS, Cushing G: Measurement issues in the empirical study of observed classroom processes? Dev Psychopathol 2004, 40:651–664.
resilience: an overview. In Resilience and development: Positive life 111. Goldberg DP, Williams P: A user’s guide to the General Health Questionnaire.
adaptations. Edited by Glantz MD, Johnson JL. New York: Kluwer Windsor: NFER-Nelson; 1991.
Academic/Plenum Publishers; 1999:129–160. 112. Goldberg DP, Gater R, Sartorius N, Ustun TB, Piccinelli M, Gureje O, Rutter C:
85. Baldwin AL, Baldwin CP, Kasser T, Zax M, Sameroff A, Seifer R: Contextual The validity of two versions of the GHQ in the WHO study of mental
risk and resiliency during late adolescence. Dev Psychopathol 1993, illness in general health care. Psychol Med 1997, 27:191–197.
5:741–761. 113. Donath S: The validity of the 12-item general health questionnaire in
86. Borman GD, Overman LT: Academic resilience in mathematics among Australia: a comparison between three scoring methods. Aust N Z J
poor and minority students. Elem Sch J 2004, 104:177–195. Psychiatry 2001, 35:231–235.
Miller-Lewis et al. Child and Adolescent Psychiatry and Mental Health 2013, 7:6 Page 23 of 23
http://www.capmh.com/content/7/1/6

114. Brugha T, Cragg D: The list of threatening experiences. Acta Psychiatr

Scand 1990, 82:77–81.
115. Brugha T, Bebbington P, Tennant C, Hurry J: The list of threatening
experiences: a subset of 12 life event categories with considerable long-
term contextual threat. Psychol Med 1985, 15(01):189–194.
116. Australian Institute of Family Studies: Growing up in Australia, the
Longitudinal Study of Australian Children: Wave 1 content sources and
supporting rationale. Melbourne: Australian Institute of Family Studies,
Australian Government; 2008.
117. McCubbin HI, Patterson JM: Stress: The family inventory of life events and
changes. In Marriage and Family Assessment: A Source Book for Family
Therapy. Edited by Filsinger EE. Beverly Hills: Sage Publications;
1983:275–297.
118. McCubbin HI, Patterson JM, Bauman E, Harris LH: Adolescent-family
inventory of life events and changes. In Family Inventories. Edited by Olson
DH, McCubbin HI, Barnes HL, Larsen A, Muxen M, Wilson M. St. Paul:
University of Minnesota Press; 1985:109–119.
119. Masten AS, Morison P, Pellegrini D, Tellengen A: Competence under stress:
Risk and protective factors. In Risk and protective factors in the development
of psychopathology. Edited by Rolf JE, Masten AS, Cicchetti D, Nuecherlein
KH, Weintraub S. New York: Cambridge Unviersity Press; 1990:236–256.
120. Obradović J, Bush NR, Stamperdahl J, Adler NA, Boyce WT: Biological
sensitivity to context: The interactive effects of stress reactivity and
family adversity on socio-emotional behavior and school readiness.
Child Dev 2010, 81:270–289.
121. Cohen J: Statistical power analysis for the behavioral sciences. 2nd edition.
New Jersey: Lawrence Erlbaum; 1988.
122. Fergusson CJ: An effect size primer: a guide for clinicians and
researchers. Prof Psychol Res Pr 2009, 40(5):532–538.
123. Sink CA, Stroh HR: Practical significance: the Use of effect sizes in school
counseling research. Prof Sch Couns 2006, 9(5):401–411.
124. De Los Reyes A, Kazdin AE: Informant discrepancies in the assessment of
childhood psychopathology: A critical review, theoretical framework,
and recommendations for further study. Psychol Bul 2005, 131(4):483–509.
125. Aiken LS, West SG: Multiple regression: Testing and interpreting interactions.
Newbury Park, CA: Sage; 1991.
126. Luthar SS, Doernberger CH, Zigler E: Resilience is not a unidimensional
construct: Insights from a prospective study of inner-city adolescents.
Dev Psychopathol 1993, 5:703–717.
127. Davis E, Sawyer MG, Lo SK, Priest N, Wake M: Socioeconomic risk factors
for mental health problems in 4–5-year-Old children: Australian
population study. Acad Pediatr 2010, 10(1):41–47.
128. Garson GD: “Regression Analysis”, from Statnotes: Topics in Multivariate
Analysis. ; http://faculty.chass.ncsu.edu/garson/pa765/statnote.htm.
129. Judge TA, Erez A, Bono JE, Thoresen CJ: Are measures of self-esteem,
neuroticism, locus of control, and generalised self-efficacy indicators of a
common core construct? J Pers Soc Psychol 2002, 83(3):693–710.
130. Masten AS, Garmezy N, Tellegen A, Pellegrini DS, Larkin K, Larsen A:
Competence and stress in school children: The moderating effects of
individual and family qualities. J Child Psychol Psychiatry 1988,
29(6):745–764.
131. Kraemer HC, Kazdin AE, Offord DR, Kessler RC, Jensen PS, Kupfer DJ: Coming
to terms with the terms of risk. Arch Gen Psychiatry 1997, 54(4):337–343.
132. Luthar SS, Brown PJ: Maximizing resilience through diverse levels of
inquiry: Prevailing paradigms, possibilities, and priorities for the future.
Dev Psychopathol 2007, 19(3):931–955.
133. Steering Committee for the Review of Government Service Provision: Report
on Government Services 2008: Children's Services. Canberra: Productivity
Commission; 2008. 3.1-3.80. Submit your next manuscript to BioMed Central
and take full advantage of:
doi:10.1186/1753-2000-7-6
Cite this article as: Miller-Lewis et al.: Resource factors for mental health • Convenient online submission
resilience in early childhood: An analysis with multiple methodologies.
• Thorough peer review
Child and Adolescent Psychiatry and Mental Health 2013 7:6.
• No space constraints or color figure charges
• Immediate publication on acceptance
• Inclusion in PubMed, CAS, Scopus and Google Scholar
• Research which is freely available for redistribution

Submit your manuscript at

www.biomedcentral.com/submit