Child Psychiatry Hum Dev (2013) 44:727–741

DOI 10.1007/s10578-013-0366-x

ORIGINAL ARTICLE

An Investigation of Complex Attachment- and Trauma-Related

Symptomatology Among Children in Foster and Kinship Care
Michael Tarren-Sweeney

Published online: 6 February 2013

Ó Springer Science+Business Media New York 2013

Abstract The paper reports an investigation into the nat- Introduction

ure, patterns and complexity of mental health symptom-
atology reported for a large (N = 347) population sample of More is known about the scale and prevalence of mental
children in foster and kinship care. Cluster analyses were health difficulties manifested by children and young people
performed on caregiver-reported Child Behavior Checklist in out-of- home care (referred to as looked after children in
(CBCL) and Assessment Checklist for Children (ACC) the UK and Ireland), than of the nature, patterns and com-
scores. The derived profile types are characterized more by plexity of their symptomatology. This is largely because
symptom complexity than specificity, and are delineated most available data were obtained as outcome measures in
more by elevation than shape. The analyses indicate that studies addressing non-clinical research questions. The
social and interpersonal relationship difficulties are hallmark present paper sets out to redress this imbalance by reporting
features of clinical presentations of children in care; that mental health symptom patterns and complexity for a large
anxiety is more often observed as a component of felt inse- sample of 4–11 year-old children in foster and kinship care.
curity than as generalized or trauma-specific anxiety; and
that attention-deficit hyperactivity is rarely manifested in What is Known About the Mental Health of Children
isolation from other difficulties. Whereas 35 % of children in Care?
had clinical difficulties that could plausibly be construed as
discrete mental disorders or comorbidity, another 20 % The scale of mental health difficulties among children in
displayed complex attachment- and trauma-related symp- care is exceptional for a non-clinical population. Over
tomatology that is not adequately conceptualized within twenty population studies of children in care in North
DSM or ICD classifications. America, Europe and Australia have measured their mental
health using standard caregiver-report rating scales—pri-
Keywords Foster care
Developmental marily the Child Behavior Checklist (CBCL) [1], the
psychopathology
Classification of mental disorders
Strengths and Difficulties Questionnaire (SDQ) [2], and the
Attachment
Trauma
Comorbidity
Rutter Scales [3]. Studies have consistently reported a scale
Developmental trauma disorder of mental health difficulties that more closely resembles
that of clinic referred children, than of children at large
[4–9]. Around half of children in care score in the clinical
range on one or more CBCL broadband or syndrome
M. Tarren-Sweeney (&)
scales, while around three quarters score above one or more
Health Sciences Centre, University of Canterbury,
Private Bag 4800, Christchurch 8140, New Zealand borderline range cut-points. Studies have also estimated
e-mail: Michael.Tarren-Sweeney@canterbury.ac.nz high prevalence of DSM-III-R and DSM-IV-TR Conduct
Disorder (17–45 %), Attention-deficit Hyperactivity Dis-
M. Tarren-Sweeney
order (10–30 %), Depression (4–36 %), Post Traumatic
Priority Research Centre for Translational Neuroscience
and Mental Health, University of Newcastle, Stress Disorder (40–50 %), and Generalized Anxiety (or
Newcastle, NSW, Australia DSM-III-R Overanxious) Disorder (4–26 %) among mixed

123
728 Child Psychiatry Hum Dev (2013) 44:727–741

samples of children and young people in foster and resi- constructing more complex symptomatology as a series of
dential care [10–15]. While older children in care have co-morbid diagnoses holds less conceptual validity.
poorer mental health, this age effect is accounted for by age One empirical method for investigating symptom pat-
at entry into care, whereby the older a child is in care, the terns and complexity is to construct symptom profiles, which
more likely they are to have entered care at an older age are graphical representations of the presence (e.g. in terms of
and with poorer pre-care mental health [16]. prevalence, frequency and/or severity) of any number of
symptom constructs—either as continuously distributed
Characteristic Difficulties Not Measured by Standard scores, or on multi-level dimensional scales. Derived using
Instruments cluster analysis, these profiles provide a means for exam-
ining the construct validity of existing taxonomies, and for
There are also considerable mental health difficulties that proposing alternative taxonomies. Findings from a small
are problematic for children and young people in care, number of studies suggest there is not one homogenous set of
which are not adequately measured by general mental symptom profile types, but rather, that various special pop-
health measures such as the CBCL and SDQ, including a ulations and clinical groupings may manifest either dis-
range of attachment- and trauma-related difficulties. The tinctly different or significantly varied profile types. For
Assessment Checklist for Children (ACC) is a caregiver- example, analyses of the behavioral difficulties of children
report psychiatric rating scale that was developed to mea- with Autistic Spectrum Disorders have identified several
sure this range of difficulties in the Children in Care Study mental health profile types that are largely delineated by
(CICS) in New South Wales, Australia [17]. ACC scores children’s cognitive and social functioning [22, 23]. A
obtained in the CICS suggest that around half of children in number of researchers have also proposed specific CBCL
care manifest one or more forms of clinically meaningful profile types for various clinical constructs. An example is
attachment-related interpersonal behavior difficulties, and the CBCL dysregulation profile (CBCL-DP), which is
sizeable proportions display clinically significant self-injury characterized by simultaneous marked clinical scores on the
(7 %) and abnormal responses to pain (5 %) [9]. A pattern anxious/depressed, attention problems, and aggressive
of excessive eating and food maintenance behavior without behavior scales [24]. However, these profiles are mostly
concurrent obesity was also identified, resembling the identified through deductive reasoning or heuristic means,
behavioral correlates of Hyperphagic Short Stature (Psy- rather than through cluster analysis or other empirically-
chosocial Dwarfism) [18]. Up to a third of children pre- validated methods. A hierarchical cluster analysis of CBCL
sented with problematic sexual behavior, which for some is syndrome scores for a population sample of 91 Dutch chil-
possibly mediated by attachment difficulties [19, 20]. The dren in foster care yielded a four-cluster solution, in which
present paper reports further findings from the CICS study. the derived profile types were delineated more by overall
profile elevation, than symptom type [25]. This is perhaps
Identifying Symptom Patterns and Complexity partly accounted for by the inclusion of scores for children
who did not have clinical-level difficulties, resulting in one
For children in care, there are big question marks around the cluster with a low, flat profile. However, another of the four
coherence and validity of clinical formulations that are based types consisted of clinical-level scores across all of the
on standard conceptualizations of psychopathology [21]. As syndromes except somatic complaints.
part of the development of the ACC instrument, 110 clinical In searching for a deeper understanding of the mental
assessment reports sourced from 50 psychological records of health symptomatology experienced by children in care,
children in care, and other child welfare clients, were two initial research questions are posed:
reviewed [17]. Most of these reports were written by psy-
1. To what extent do the patterns and complexity of
chologists and psychiatrists working in specialist public
mental health symptoms experienced by children and
health services, child welfare and alternate care agencies,
young people in care resemble that of children at large,
and private practice. The reports revealed considerable
and;
diagnostic disagreement and uncertainty, as well as a ten-
2. How adequate are existing and proposed diagnostic
dency to frame complex psychopathology as a series of
classification systems (ICD-10/11, DSM-IV-TR/V) for
discrete, comorbid disorders. At its simplest, patterns and
conceptualizing the mental health symptomatology of
complexity of children’s psychiatric symptoms can be con-
children and young people in care?
strued in terms of comorbidity—the co-occurrence of vari-
ous combinations of dichotomous, psychiatric disorders. The present paper addresses these research questions as
There are several common, well-observed co-morbid pre- they apply to children (not adolescents) by reporting an
sentations, notably ADHD with Oppositional-defiant disor- investigation of the patterns, and complexity of baseline
der, and co-occurring mood and anxiety disorders. However, mental health estimates obtained for a large (n = 347)

123
Child Psychiatry Hum Dev (2013) 44:727–741 729

population sample of children in long-term foster and CBCL syndrome analyses were having at least one score in
kinship care. the clinical range on any CBCL syndrome or broadband
scale, or having a CBCL total problem score above 29.
This yielded scores for 239 children (132 boys, 107 girls).
Method The inclusion criterion for ACC–DSM-oriented analyses
was having at least one score in the clinical range on any
The present analyses were carried out on caregiver-reported CBCL syndrome, broadband or DSM-oriented scale, or on
mental health data obtained in the CICS baseline survey. any ACC scale. This yielded scores for 236 children (128
The CICS is a prospective, epidemiological study of the boys, 108 girls). All of these 236 children are among the
mental health and development of children and young peo- former group of 239 children.
ple in court-ordered foster and kinship care, in New South
Wales (NSW), Australia. The baseline survey obtained Mental Health Measures
caregiver-reported mental health estimates; concurrent and
retrospective measures of likely risk and protective factors; The mental health and socialization of subject children were
reports of health, developmental, language and educational estimated using two caregiver-report checklists, the CBCL
problems; and use of mental health and related services. [1, 27], and the ACC [17]. The reliability of foster parent
Data were collected from a mail-out caregiver question- reports of children’s problems remains somewhat uncertain,
naire, and from the state child welfare and alternate care although there is evidence that in respect of children in
database (Client Information System, CIS). Children did not long-term care, foster parents are at least as reliable as
actively participate in the survey. The University of New- parents [28]. The CBCL measures child problem behavior
castle Human Research Ethics Committee provided ethics across eight empirically-derived clinical sub-scales, with
approval for this baseline survey. two higher-order scales approximating spectrums of
depressive/anxious symptoms (internalizing) and disruptive
Study Sample behavioral symptoms (externalizing). The CBCL also
includes DSM-oriented scales that were derived through
Questionnaires were sent to the caregivers of 621 eligible expert ratings of items’ conformity to DSM-IV-TR diag-
children (the study sample), with 347 returns (56 % nostic criteria [1]. The present study used the 1991 version
response rate). The study sample and non-contactable of the CBCL (for ages 4–16) [27], allowing for comparison
children had similar age and gender distributions. Partici- with clinical profiles previously identified among clinic
pant and non-participant children did not differ in terms of referred children at large [26]. Scores for three DSM-ori-
the distribution of gender, age, regional location, ethnicity, ented scale scores and three 2001 syndrome scales could not
care type (foster vs. kinship care), or being previously be wholly calculated from the available item scores, but
restored to their birth parents. However, non-participant instead were estimated using an acceptable pro-rating pro-
children entered care at a younger age, had less exposure to cedure provided by the instrument’s developer.
maltreatment, and were more likely to have spent the larger The ACC is a 120-item caregiver-report psychiatric
part of their life with their current caregivers. This bias is rating scale, measuring behaviors, emotional states, traits,
likely to have resulted in a slight over-estimation of mental and manners of relating to others, as manifested by children
health problems in the study population. in care [17]. The instrument has 10 clinical scales that were
The 347 participant children are henceforth referred to empirically-derived though factor analysis of CICS baseline
as the aggregate sample. With the exception of the cluster scores, named I. Sexual behavior, II. Pseudomature inter-
analyses, all other statistical analyses reported in this paper personal behavior, III. Non-reciprocal interpersonal
were carried out on data for the aggregate sample. The behavior, IV. Indiscriminate interpersonal behavior, V.
cluster analyses were restricted to sub-samples containing Insecure interpersonal behavior, VI. Anxious—Distrustful,
children with clinical-level difficulties. The purpose of VII. Abnormal pain response, VIII. Food maintenance, IX.
conducting the cluster analyses was to identify and delin- Self-injury and X. Suicide discourse. Scales II to VI mea-
eate between characteristic clinical profiles. Scores of sure difficulties that are relatively common among children
children who were not reported as having clinically in care (high prevalence scales), while scale I and scales VII
meaningful difficulties were therefore excluded from the to X measure less common difficulties (low prevalence
analyses. Achenbach [26] included clinic referred children scales). For each scale, two cut-points were selected to
with CBCL total problem scores above 29. However, this identify children who have ‘possible’ and ‘probable’ clini-
could have the effect of excluding children who have very cally-meaningful mental health problems. Scores above the
discrete symptomatology that is nonetheless clinically higher cut-points constitute clinical ranges that are highly
significant. The inclusion criteria selected for the present predictive of psychiatric impairment (highly specific),

123
730 Child Psychiatry Hum Dev (2013) 44:727–741

while scores above the lower cut-point ranging up to and interpersonal behavior, Insecure, and Anxious-distrustful).
including the high cut-point constitute elevated ranges that While the CBCL syndrome scales have stronger empirical
retain high sensitivity for detecting psychiatric impairment. validity than the DSM-oriented scales, the latter were selected
The clinical and elevated (i.e. borderline clinical) ranges for so that the identified clusters might be better referenced to
the ACC clinical scales were each separately defined using standard classifications of child mental disorders. Impor-
a combination of criterion-referenced procedures and clin- tantly, there is evidence that the DSM-oriented scales effec-
ical reasoning [29]. The ACC’s content validity was tively delineate between young people with identified discrete
reviewed by clinicians and foster parents, while construct DSM-IV-TR disorders [31]. The DSM-oriented Somatic
validity is strongly supported by factor analysis findings and Problems scale was excluded because few children in the
concurrent validity obtained in the CICS [17]. The ACC present study were reported with clinical-level somatic
also demonstrated criterion-related validity in the CICS, in problems. The remaining ACC scales were excluded because
that: 1. it differentiates between referred and non-referred of the need to contain the number of variables included in
children in care; 2. scores are highly sensitive to children’s the analyses (the excluded ACC scales measure less common
risk exposure; and 3. scores are strongly differentiated by difficulties).
children’s age at entry into care [17]. As yet, no further The first consideration in selecting a method to identify
validation data have been published from other studies. psychopathology clusters is the extent to which variation in
children’s symptomatology can be subsumed within a
Reference CBCL Symptom Profiles hierarchical structure. Evidence of a hierarchical structure
is found in common comorbidity patterns, and the existence
Using cluster analysis, Achenbach [26] identified charac- of higher-order CBCL internalizing and externalizing fac-
teristic symptom profile types from carer-report (CBCL) and tors. However, these examples translate as just two levels
teacher-report (Teacher Report Form) syndrome scale scores within a nested structure. At increasingly higher levels, a
for clinic referred 6–11 year-old children at large, with a hierarchical analysis is likely to combine clusters that are
view to proposing an empirically-based taxonomy of com- not clinically coherent or meaningful. It was thus decided
mon child mental disorders. Six profile types were identified that a non-hierarchical method was preferable for final
from cluster analyses of boys’ and girls’ CBCL scores, identification of child psychopathology clusters in both sets
consisting of: four cross-informant, cross-gender types of analyses. The main drawback for a non-hierarchical
(labeled Withdrawn, Somatic, Social, and Delinquent- method is that it provides no ready indication of the optimal
Aggressive); one CBCL-specific, cross-gender type (Delin- number of clusters or likely relationships between clusters.
quent); and one CBCL-specific type that was only found For this reason, various authors have recommended a two-
among boys (Withdrawn-Anxious/Depressed-Aggressive). stage enquiry, in which exploratory cluster analysis is
Four of the six profile types point to relatively non-complex, performed using a hierarchical method, after which con-
discrete symptomatology. However, the Social and With- firmatory cluster analysis is performed using a non-hierar-
drawn-Anxious/Depressed-Aggressive profile types have chical method [32, 33]. Exploratory hierarchical cluster
elevated scores on most syndrome scales, indicating highly analyses were performed using both the weighted-average
complex symptomatology that is inadequately conceptual- linkage (WPGMA) method employed by Edelbrock and
ized within standard classifications of mental disorders. Achenbach [34] with Euclidean distance measure, and the
Ward’s linkage method with Euclidean squared distance
Cluster Analysis Procedure measure. It was decided that any clustering solution should
exclude clusters containing less than five children. The
Two sets of gender-specific cluster analyses were carried WPGMA method was found to be unworkable, yielding
out, using STATA 12 statistical software [30]. The first set many clusters that contained only one or two children,
identified symptom clusters across the eight 1991-version whereas the Ward’s method yielded more equal-sized
CBCL syndrome scales. ACC scores were withheld from clusters.
these cluster analyses so that the syndrome profiles could Non-hierarchical cluster analysis was performed using
be directly compared with Achenbach’s reference profile the K-Means agglomerative method with Euclidean dis-
types [26]. tance measure, and randomly generated starting centers.
The second set identified symptom clusters across five While Euclidean distances provide the best clustering
CBCL DSM-oriented scales (Affective problems, Anxiety solutions for hierarchical analyses, they are sensitive to
problems, Attention-deficit/hyperactivity problems, Opposi- variations in score range across the scales [33]. Hierarchical
tional defiant problems, and Conduct problems) and six ACC analyses were therefore performed on scores that were
scales (Sexual behavior, Pseudomature interpersonal behav- rescaled using a range-bound transformation that eliminates
ior, Nonreciprocal interpersonal behavior, Indiscriminate this distortion, namely Z = X/[Max (X)-Min (X)]. This

123
Child Psychiatry Hum Dev (2013) 44:727–741 731

method is preferable to other range-based methods and to (r = .66 to .70 for boys, and .67 to .75 for girls). Attention
standard z-score transformations, when carrying out hier- problems correlated highly with each of the externalizing
archical analyses [35, 36]. Conversely, non-transformed sub-scales and with the social problems and thought prob-
raw scores yield more accurate agglomerative clustering lems sub-scales (r = .60 to .70 for boys, and .66 to .75 for
solutions than either standardized (i.e. z or T) scores or girls), suggesting they serve as markers for a range of
range-defined transformations [36]. Hence, raw scores were problems observed among children in care. Correlations of
used for the non-hierarchical analyses. the withdrawn-depressed and anxious-depressed internal-
izing scales were more moderate, while somatic problems
scores correlated relatively poorly with most other scales.
Results Correlations among ACC clinical scale scores for boys
and girls are set out in Table 1. Children’s scores on the
The study participants (N = 347) were 176 boys and 171 pseudomature, non-reciprocal, indiscriminate and insecure
girls aged 4–11 years, with a mean (median) age of 7.8 (7.7) interpersonal behavior scales were closely correlated
years. The number of children in foster and kinship care (r = .56 to .65). These attachment-related problem scales
were 297 (86 %) and 50 (14 %) respectively. The mean also correlated highly with scores on age-inappropriate
(median) age at entry into care was 3.5 (3.2) years, and the sexual behavior. As expected, scores on the anxious-
mean (median) time that children had been in care was 4.3 depressed and insecure scales were also closely related
(4.1) years. The mean (median) number of placements since (r = .62), since both scales measure anxiety symptoms.
entering care was 3.1 (2), and the mean (median) length of However, there was low to moderate correlation (r = .16 to
their present placements was 3.3 (3.0) years. Subject chil- .45) between the various low prevalence clinical scales.
dren mostly endured substantial and complex maltreatment Correlations of ACC clinical scale scores with CBCL
prior to entering care. Their mean (median) number of syndrome, broadband and DSM-oriented raw scale scores
notifications of maltreatment confirmed by child welfare are set out in Table 2. Excluding the CBCL somatic scales
services was 3.5 (3). The mean (median) time between a (syndrome and DSM-oriented), the ACC high prevalence
child’s first confirmed maltreatment event, and their entry scale scores (numbers II to VI) had moderate to high corre-
into court-ordered care was 1.7 (1.1) years. lation (ranging from .37 to .76) with the various CBCL
The aggregate sample (n = 347) mean CBCL and ACC scores, while the ACC total clinical and CBCL total problem
scores (with standard deviations) have been reported pre- scores were strongly correlated (r = .88). There were low to
viously [9]. Mean CBCL broadband and sub-scale raw moderate correlations between CBCL scales and the sexual
scores for boys and girls ranged from .3 to 1.9 standard behavior, abnormal pain response, food maintenance, self-
deviations above Australian community means (all differ- injury and suicide discourse scales. These correlations are
ences significant at P \ 0.001) [37]. Fifty-three percent of partially deflated because these ACC scales measure rela-
girls and 57 % of boys had at least one CBCL scale score tively low prevalence problems. The non-reciprocal and
in the clinical range. Equivalent proportions of children indiscriminate scales correlated highly with the externaliz-
with any score in the borderline plus clinical ranges were ing and social, attention and thought problems sub-scales, as
65 % (girls) and 74 % (boys). The scale of their mental well as with the equivalent DSM-oriented scales (i.e. atten-
health difficulties was at the upper end of the range of tion-deficit, oppositional-defiant, and conduct problems).
previously reported estimates for children in foster care.
One explanation for their relatively high level of difficul- Internal Consistency of Aggregate Sample Scores
ties is that they included children who, in other jurisdic-
tions, would have resided in residential care.1 The internal consistency (Cronbach’s alpha) of CBCL
syndrome scale scores ranged from .71 to .92, and that of
Inter-Scale Correlations for the Aggregate Sample DSM-oriented scale scores ranged from .71 to .89, which
are comparable to alpha ranges reported for US normative
There were high correlations of 2001-version CBCL samples [38]. The CBCL alphas are listed in Table 2. The
aggressive and rule-breaking raw scale scores (r = .82 for internal consistency of the ACC clinical scales ranged from
boys and .78 for girls) and among the social problems, .70 to .86, and have been reported previously [17].
attention problems and thought problems sub-scales
Higher Order Factor Analysis of CBCL Syndrome
1
Scores
The NSW government closed all residential care facilities several
years prior to the start of the CICS. The present sample thus included
a sizeable number of children who, in previous times and in other The higher-order factor structure of CBCL scores for the
jurisdictions, would not be placed with families. present sample was investigated through principal

123
732 Child Psychiatry Hum Dev (2013) 44:727–741

components factor analyses (with varimax rotations) of

discourse
syndrome raw scale scores, for each gender. The most

Suicide
meaningful higher-order model for boys contains four fac-

.27***
.23**

.23**
.20**

.22**
.43
.37

.31

.39

.48
tors, accounting for 86 % of the score variance. These are
two internalizing factors (anxious-depressed/somatic; and

injury
anxious-depressed/withdrawn-depressed), a social-atten-
.25***
.29***

Self-
tion-thought problems factor (SAT), and an externalizing
.59

.33
.43
.38
.43
.47

.47

.39
factor. The most meaningful higher-order model for girls

maintenance
contains only two factors, accounting for 75 % of the score
variance. These are an internalizing factor and a combined
externalizing—SAT factor. For girls, three- and four-factor

Food
.59
.31
.35
.45
.40
.43
.46
.44

.44
.41
models progressively dismantle the internalizing factor,
Table 1 Correlation matrices of ACC clinical scale scores for girls (above diagonal, n = 171), and boys (below diagonal, n = 176) with P values

while retaining a clearly delineated externalizing—SAT

Abnormal pain factor.
response

Combinations of Clinically Significant CBCL

Syndrome Scores
***

ns
.17*
.62
.35
.40
.48
.43
.51
.37

.28
.05

The proportions of children with various combinations of

All P values are less than 0.0001, except where indicated by: *** P \ 0.001, ** P \ 0.01, * P \ 0.05, ns (not significant)

clinically significant CBCL syndrome scores were calcu-

distrustful
Anxious-

lated, with a view to identifying variations in the type and

***

complexity of children’s symptomatology. Fifty-seven

.75
.47
.52
.55
.49
.65

.27

.34
.31
.32

percent of boys (101/176) and 51 % of girls (88/171) had at

least one clinically significant CBCL score across the eight
Insecure
***

syndrome scales. Table 3 lists: the proportions of those 189

.83
.45
.60
.63
.65

.58
.40

.30
.34
.29

clinical cases having clinical-level scores on one through to

a maximum of eight syndrome scales; the proportions
Indiscriminate

having clinical-level scores that are confined within each of

three broadband symptom domains (internalizing, exter-
**

nalizing, and social-attention-thought problems); and the

.83
.56
.63
.64

.57
.39
.35

.41
.35
.19

proportions having clinical-level scores that fall across two,

reciprocal

as well as all three domains. Forty two percent of cases had

clinical-level scores confined to a single domain (i.e. either
Non-
**

pure internalizing, pure SAT or pure externalizing), 32 %

.83
.62
.62

.56
.65
.51
.46

.39
.34
.23

had clinical-level scores within two domains, and 25 % had

Pseudomature

clinical-level scores within all three domains.

Symptom Profile Types: CBCL Syndromes

**
.77
.53

.62
.48
.56
.52
.34

.32
.22
.32

For the Ward’s analyses of CBCL syndrome scores, stop-

ping rules provided inconclusive guidance on the optimal
behavior

number of clusters. Dendograms of the hierarchical cluster

Sexual
ns
.19**
**

structures indicated that a seven-cluster solution for both

.70

.40
.54
.45
.40
.45
.34

.21
.13

boys and girls yielded the greatest number of distinct

clinical

symptom clusters containing five or more children.

K-Means agglomerative cluster analyses were then carried
Total
.61
.72
.85
.76
.83
.69
.56

.55
.54
.39

out, specifying seven clusters. The derived cluster sizes

ranged from 8 to 28 girls and 9 to 26 boys. The symptom
Anxious-distrustful

Food maintenance

Suicide discourse

profile types, depicting variation in profile shape and overall

Sexual behavior

Abnormal pain
Non-reciprocal
Indiscriminate
Pseudomature
Total clinical

elevation of scores, are presented for boys and girls in

Self-Injury
response

Figs. 1 and 2 respectively. The symptom profiles place

Insecure

cluster mean scores for each of the CBCL syndrome scales

within a four-range measurement scale, defined by CBCL

123
Child Psychiatry Hum Dev (2013) 44:727–741 733

Table 2 Correlation matrix of ACC and CBCL scale scores (age 6–11, n = 296)

ACC SCALES
I. Sexual behavior .30 .37 .27 .40 .47 .33 .54 .41 .35 .30 .21 .30 .35 .43 .38 .48 .50
II. Pseudomature .46 .36 .30 .50 .50 .37 .54 .54 .42 .40 .24 .47 .53 .44 .47 .57 .60
III. Non-reciprocal .48 .62 .24 .67 .68 .64 .76 .71 .53 .48 .18 .57 .62 .72 .55 .76 .78
IV. Indiscriminate .44 .39 .28 .70 .61 .64 .59 .62 .48 .47 .22 .63 .56 .56 .46 .64 .71
V. Insecure .69 .68 .41 .66 .65 .54 .47 .50 .63 .62 .35 .46 .42 .42 .74 .51 .70
VI. Anxious-distrustful .66 .51 .51 .50 .60 .38 .46 .50 .65 .70 .39 .41 .43 .42 .70 .51 .65
VII. Abnormal pain response .30 .41 .21 .35 .49 .43 .40 .40 .32 .30 .16 .40 .35 .39 .37 .42 .47
VIII. Food maintenance .28 .31 .30 .45 .50 .36 .53 .44 .45 .29 .22 .36 .40 .51 .35 .50 .53
IX. Self-injury .41 .40 .43 .43 .54 .40 .41 .44 .57 .36 .41 .37 .38 .44 .49 .45 .55
*** **
X. Suicide discourse .40 .27 .24 .31 .36 .21 .34 .36 .47 .33 .15 .23 .31 .39 .38 .37 .40

Total clinical score .65 .65 .45 .77 .81 .68 .75 .73 .70 .64 .36 .66 .64 .68 .72 .77 .88
Withdrawn / depressed

Oppositional-defiant
Anxious / depressed

Attention problems

Affective problems
Somatic complains

Conduct problems
Thought problems

Somatic problems
Anxiety problems

CBCL SCALES
Social problems

ADH problems

Total problems
Rule-breaking

Externalizing
Internalizing
Aggressive

Internal consistency (alpha) .81 .71 .77 .78 .79 .84 .79 .92 .71 .71 .78 .80 .82 .89
Syndrome scales DSM-oriented scales Broadband
All P values are less than 0.0001, except where indicated by: *** P \ 0.001 or ** P \ 0.01

T-score ranges. The four ranges are labeled: normative (T- shape and overall elevation of scores, are presented for
scores\63, representing scores that are clearly normative); boys and girls in Figs. 3 and 4 respectively. For the DSM-
elevated (T-scores = 63–69, representing sub-clinical oriented scales, the four ranges were defined using the
scores that are less clearly normative, with the upper end same normative T-score cut-points as described earlier for
encompassing the CBCL borderline clinical range); clini- the CBCL syndrome profiles. For the ACC scales, exist-
cally indicated (T-scores = 70–73, the moderate end of the ing cut-points delineating between the normative and
clinical range); and marked clinical (T-scores [73, the elevated ranges and between the elevated and clinical
severe end of the clinical range). ranges were utilized. Given a relative lack of research on
the clinical meaning of high ACC scale scores, it was
Symptom Profile Types: ACC and DSM-Oriented reasoned that assignment of scores to a marked clinical
Scales range should be conservative. Consequently, in compari-
son with the CBCL scales, greater proportions of the
For the Ward’s analyses of the ACC and DSM-oriented ACC score ranges were defined within the clinically
scores, stopping rules provided inconclusive guidance on indicated range, and lesser proportions within the marked
the optimal number of clusters, but pointed to 13- and clinical range.
12-cluster solutions for boys and girls respectively. Dend-
ograms of the hierarchical cluster structures indicated 8-
and 9-cluster solutions yielded the best number of distinct Discussion
symptom profiles containing five or more children, for girls
and boys respectively. K-Means agglomerative cluster The present findings provide new insight into the nature,
analyses were then carried out for girls and boys, speci- patterns and complexity of the mental health symptom-
fying 8 and 9 clusters respectively. The cluster sizes ranged atology of school-aged children in foster and kinship care.
from 5 to 22 girls and 7 to 24 boys. A high prevalence of social and interpersonal relationship
Symptom profile types for this combination of ACC difficulties was found among the present population sam-
and DSM-oriented scales, depicting variation in profile ple, both in absolute terms, and relative to other mental

123
734 Child Psychiatry Hum Dev (2013) 44:727–741

Table 3 Distributions of
Proportion (n)
number and type of clinical
range CBCL syndrome scores Total (n = 189) Girls (n = 88) Boys (n = 101)
among clinical cases (n = 189)
# syndrome scores in clinical range
1 32.3 % (61) 33.0 % (29) 31.7 % (32)
2 17.5 % (33) 20.5 % (18) 14.9 % (15)
3 12.2 % (23) 10.2 % (9) 13.9 % (14)
Clinical cases defined as 4 12.2 % (23) 9.1 % (8) 14.9 % (15)
children having at least one
5 10.1 % (19) 10.2 % (9) 9.9 % (10)
CBCL syndrome score in the
clinical range (189/347) 6 9.0 % (17) 10.2 % (9) 7.9 % (8)
a
Location of clinical range 7 4.8 % (9) 4.5 % (4) 5.0 % (5)
syndrome scores within 8 2.1 % (4) 2.3 % (2) 2.0 % (2)
broadband symptom domains: Location of clinical-range syndrome scores within symptom domainsa
Internalizing encompasses
withdrawn-depressed, anxious- Internalizing only 9.0 % (17) 11.4 % (10) 6.9 % (7)
depressed and somatic SAT only 10.6 % (20) 10.2 % (9) 10.9 % (11)
complaints syndromes; SAT Externalizing only 22.8 % (43) 20.5 % (18) 24.8 % (25)
encompasses social problems,
Internalizing ? SAT 5.8 % (11) 6.8 % (6) 5.0 % (5)
thought problems and attention
problems syndromes; and Internalizing ? externalizing 1.6 % (3) 2.3 % (2) 1.0 % (1)
Externalizing encompasses SAT ? externalizing 24.9 % (47) 22.7 % (20) 26.7 % (27)
rule-breaking and aggressive Internalizing ? SAT ? externalizing 25.4 % (48) 26.1 % (23) 24.8 % (25)
behavior syndrome

health concerns. The findings suggest such difficulties are a insecure attachment style, but equally may represent a
hallmark feature of the clinical presentations among a large child’s adaptation to previous losses (including foster
proportion of children in care. placement breakdowns). Furthermore, there are likely to be
many and varied systemic pressures affecting the felt
Attachment-Related Difficulties security of children in care.
Inhibited and disinhibited attachment disorder behaviors
Inhibited and disinhibited forms of attachment disturbance are presently conceptualized as distinct sub-types of DSM-
were hypothesized during the development of the ACC, IV-TR Reactive Attachment Disorder (RAD). This distinc-
and scales that broadly refer to these phenomena (‘non- tion is maintained within the proposed DSM-V delineation
reciprocal’ and ‘indiscriminate’ interpersonal behavior) between RAD (representing the inhibited form) and Disin-
were derived through factor analysis. It is likely that high hibited Social Engagement Disorder (the disinhibited form).
scores on these scales represent, in most instances, socially It has been suggested these sub-types may not be symp-
determined attachment-related disturbances, since sample tomatically distinct [41, 42], and an alternative classification
scores on these scales were strongly predicted by children’s of infant mental health disorders (the Diagnostic Classifi-
exposure to social adversity, and their age at entry into care cation: 0-3R) includes a mixed pattern of inhibition and
[16]. Nonetheless, one should be mindful that social disinhibition [43]. The present study found moderate to high
relatedness problems among children in care may in some correlations of scores on the pseudomature, non-reciprocal,
instances be determined more by brain injury, intellectual indiscriminate and insecure scales, as well as symptom
disability, autistic spectrum disorders, or temperament profiles that include simultaneous high scores on combina-
[39]. The author has also observed cases in which non- tions of two to four of those scales. Furthermore, few chil-
reciprocal interpersonal behavior is situation-specific. For dren were reported with discrete (or pure) forms of social or
example, these difficulties may be apparent in a child’s interpersonal relationship difficulties. Together, this sug-
relationship with their foster parents, but not with their gests that many children in care present with a complex array
birth parents. The emergence of the ‘pseudomature’ factor of attachment-related difficulties that are more correctly
had not been anticipated. This pattern of interpersonal conceptualized as individual profiles of attachment-related
behavior has been described previously [40], but is not difficulties, rather than discrete types of attachment disorder.
presently thought of as an attachment-related clinical
phenomenon. Finally, the meaning of high scores on the Emotional Difficulties of Children in Care
insecure scale remains somewhat uncertain, because the
scale does not delineate between state and trait insecurity. The present study provides several insights into the charac-
Insecure and overly conforming behaviors may indicate an teristics and patterns of emotional difficulties experienced by

123
Child Psychiatry Hum Dev (2013) 44:727–741 735

Fig. 1 Boys’ symptom profile types: mean CBCL syndrome scale scores (1991 profile) for 7 groups identified through K-means cluster analysis
(n = 132)

children in care. First, although the ACC insecure and anx- measurement error, is unclear. It may be for example, that
ious-distrustful scales appear to respectively measure felt some caregivers have difficulty recognizing depressive
security and trauma-related anxiety, they correlated highly symptoms experienced by children who manifest complex
with the CBCL anxious-depressed and withdrawn-depres- symptomatology. Finally, a characteristic feature of the
sed sub-scales, and with the DSM-oriented affective prob- reported symptomatology was the relative absence of
lems and anxiety problems scales. Together, these findings somatic complaints. This phenomenon perhaps partly
describe an array of emotional difficulties experienced by accounts for the absence of a differentiated ‘internalizing’
many children in care. Second, it is important to note that higher-order factor among the sample CBCL scores.
clinically meaningful emotional difficulties were rarely
identified in isolation from other forms of psychopathology What do the Symptom Profile Types Represent?
(few children had pure internalizing disorders). Third, the
ratio of internalizing to externalizing difficulties reported for CBCL Syndrome Profiles
children in care differs to that reported for normative and
clinic referred populations at large [1]. The extent to which The CBCL syndrome profile types are characterized more
this represents a real difference in symptomatology, versus by symptom complexity than specificity, and are delineated

123
736 Child Psychiatry Hum Dev (2013) 44:727–741

Fig. 2 Girls’ symptom profile types: mean CBCL syndrome scale scores (1991 profile) for 7 groups identified through K-means cluster analysis
(n = 107)

more by elevation than shape. It is notable that the four aggressive behavior; and marked, clinical-level social and
symptom clusters identified within the small Dutch sample attention problems. For both genders: profile #6 describes
were also markedly delineated by elevation, rather than children with clinical-level difficulties across the SAT
shape [25]. These symptom profiles lack the degree of (except girls’ thought problems) and externalizing
symptom specificity that has historically defined psycho- domains; and profile #7 describes children with clinical-
pathology typologies. For both genders, CBCL profile level difficulties across every syndrome (except boys’
types #1 to #3 are characterized by normative-level diffi- somatic scores).
culties across most syndromes, coupled with various How then do these profile types compare with those
combinations of elevated, sub-clinical inattention and identified for clinic referred children at large? First, it is
externalizing behavior. Boy profiles #4 and #5 respectively notable that the four Achenbach profiles that describe rel-
depict: clinical-level thought problems with elevated, sub- atively discrete, non-complex symptomatology were not
clinical social and attention problems; and clinical-level located in the present study, with one exception (girl profile
inattention and externalizing behavior. Girl profiles #4 and #4 is comparable to Achenbach’s delinquent profile).
#5 respectively depict: clinical-level delinquent behavior, Conversely, variations of the two Achenbach profiles
with elevated, sub-clinical thought problems and describing relatively complex and severe symptomatology

123
Child Psychiatry Hum Dev (2013) 44:727–741 737

Fig. 3 Boys’ symptom profile types: mean DSM-oriented and ACC scale scores for 9 groups identified through K-means cluster analysis
(n = 128)

were identified among children in care. To varying degrees, point to discrete, non-complex disorders. Profile #1
boy profile #6 and girl profiles #5 and #6 resemble describes normative behavior with elevated, sub-clinical
Achenbach’s social type, which has no diagnostic coun- conduct problems. Profile #2 represents an anxious—
terpart in either the present or previous versions of DSM insecure type for girls, and an anxious—insecure—sexual
[26], while for both genders, profile #7 resembles the problems type for boys, both of which are fairly specific to
‘withdrawn—anxious/depressed—aggressive’ type. these symptoms. Boy profile #3 is a relatively simple
conduct/oppositional-defiant type. However, the other
ACC–DSM-Oriented Profiles profile types reveal varying degrees of symptom com-
plexity. Girl profile #3 and boy profile #4 consist of one or
The ACC–DSM-oriented profile types are also delineated two clinical-level scores combined with elevated sub-
more by elevation than shape. Only a few of the profiles clinical difficulties across most of the remaining scales.

123
738 Child Psychiatry Hum Dev (2013) 44:727–741

Fig. 4 Girls’ symptom profile types: mean DSM-oriented and ACC scale scores for 8 groups identified through K-means cluster analysis
(n = 108)

Boy profile #5 and girl profile #5 are more complex entering care. The present findings point to the possibility
externalizing types, characterized by co-occurring inter- that anxiety amongst children in care tends to be more
personal difficulties (as well as sexual behavior problems strongly experienced as a component of felt insecurity (i.e.
for girls). While the #5 profiles are possibly ‘accommo- as measured by the ACC insecure scale).
dated’ within present DSM and ICD classifications, the #6,
#7 and #8 profiles and boy profile #9 all describe complex Complexity Revealed but not Yet Adequately
symptom patterns that are inadequately conceptualized Understood
within present nosology.
With the exception of profile #2, all of the ACC–DSM- Among the present aggregate sample of children in care,
oriented symptom profile types have relatively low DSM- roughly 30 % of children were reported with normative
oriented affective problems and anxiety problems, and difficulties; another 15 % had elevated, sub-clinical check-
ACC anxious-distrustful behaviors. This was somewhat list scores [9]; around 35 % had relatively non-complex,
unexpected given the extent of social adversity and clinically significant psychopathology that can be concep-
developmental trauma that children experience before tualized as discrete mental disorders or comorbidity within

123
Child Psychiatry Hum Dev (2013) 44:727–741 739

standard diagnostic classifications; and the remaining 20 % analyses yielded statistically delineated symptom profile
manifested complex attachment- and trauma-related types, their clinical distinctness is questionable. Close
psychopathology that is not adequately conceptualized inspection of children’s individual score profiles shows a
within standard classification. The presence of hallmark lack of clear delineation between the derived profile types,
social and interpersonal relationship difficulties among a and considerable variability between children’s symptom
large proportion of children in care, which overlays and profiles. This implies that any taxonomy of complex
possibly mediates children’s experience of other types of attachment- and trauma-related disorders is likely to have
symptoms, adds to symptom complexity. This complexity poor specificity. It may be that much of the symptom-
accounts for some surprising results. Despite a high preva- atology identified in the present study is too complex, and
lence of clinical-level DSM-oriented attention-deficit shows too much variability across children’s individual
hyperactivity scores among the present sample [9], an profiles, to allow for traditional classification.
ADHD profile type was not located for either gender.
Instead, the findings suggest that clinically significant inat- Limitations and Suggestions for Further Research
tention/over-activity is largely manifested as a component of
complex symptomatology that includes social impairments. The present study was hampered by a number of method-
They are also consistent with recent findings that children ological and conceptual limitations. Firstly, children’s
who present with comorbid ADHD have greater social mental health difficulties were solely measured using carer-
impairment than do children who experience ADHD without report instruments. A definitive study of children’s symp-
co-morbid symptoms [44], and speculation that some social tomatology should utilize multiple informants, and obtain
impairment among children with comorbid ADHD is at least information directly from the subject children. Secondly,
partly determined by comorbid symptomatology [45]. Sim- carer-report rating scales measure the number and fre-
ilarly, a lack of symptom specificity precluded the presence quency of children’s mental health symptoms, but not
of the CBCL dysregulation profile among the present sam- symptom intensity or extremity, or functional impairment.
ple. Each child reported with clinical-level scores on the High checklist scores thus only provide proxy estimates of
three CBCL-DP indicator scales (n = 16, 4.6 %) had clini- severity. Thirdly, the use of Pearson’s r correlation statistic
cal-level scores on one or more other CBCL syndrome scales as a measure of association is not ideal for the present
as well! The psychopathology described by the most com- analyses, as correlations appear to be sensitive to variations
plex and elevated symptom profiles mirrors somewhat the in the prevalence of different types of difficulties. Fourthly,
diagnostic criteria and clinical presentations of Bessel van the identification of clusters of children with high scores on
der Kolk and colleagues’ proposed developmental trauma almost all symptom scales raises the possibility that some
disorder (DTD) [46, 47]. Notably, the present study sample scores have over-estimated children’s true difficulties
had almost universal exposure to severe and chronic social because of respondent bias.
adversity and trauma prior to their entry into care. The
present findings partially support the rationale for introduc-
ing DTD or a similar construct into DSM and ICD classifi- Summary
cation systems—in that these findings demonstrate that DTD
provides a more accurate conceptualization of complex The present paper describes an investigation into the nat-
attachment- and trauma-related psychopathology than that ure, patterns and complexity of mental health symptom-
provided by existing taxonomies. The present findings also atology reported for a large (N = 347) population sample
caution against thinking of developmental trauma disorder of children in long-term foster and kinship care. Separate
(or complex symptomatology in general) as a discrete, cluster analyses were performed on CBCL syndrome scale
symptom-specific, clinical phenomenon—suggesting scores, and on a combination of ACC clinical scales and
instead the need for a symptom profile approach to clinical CBCL DSM-oriented scales for sub-samples of children
formulation. with clinically significant difficulties. The derived profile
There are some distinct differences between the derived types are characterized more by symptom complexity than
symptom profiles that merit further research, and which specificity, and are delineated more by elevation than
suggest new ways of conceptualizing complex attachment- shape. Of six CBCL syndrome profile types found among
and trauma-related disorders. Beyond this however, the US clinic-referred children at large, only two complex
profiles do not provide a basis for a taxonomy of complex types were located within the present sample. The study
attachment- and trauma-related disorders. As stated previ- findings confirm that social and interpersonal relationship
ously, the clusters are differentiated more by profile ele- difficulties are a hallmark feature of the clinical presenta-
vation than profile shape or pattern, suggesting an absence tions of children in care. They also suggest that anxiety is
of typology. Furthermore, whilst the present cluster more often observed as a component of felt insecurity than

123
740 Child Psychiatry Hum Dev (2013) 44:727–741

as generalized or trauma-specific anxiety; and that atten- 7. Burns BJ, Phillips S, Wagner H, Barth RP, Kolko D, Campbell Y,
tion-deficit hyperactivity is rarely manifested in isolation Landsverk J (2004) Mental health need and access to mental
health services by youths involved with child welfare: a national
from other difficulties. survey. J Am Acad Child Adolesc Psychiatry 43(8):960–970
Whereas 35 % of children had relatively non-complex 8. Cappelletty G, Brown M, Shumate S (2005) Correlates of the
clinical difficulties, pointing towards discrete mental dis- Randolph Attachment Disorder Questionnaire (RADQ) in a
orders or plausible co-morbidity, another 20 % displayed sample of children in foster placement. Child Adol Soc Work J
22(1):71–84
complex attachment- and trauma-related symptomatology 9. Tarren-Sweeney M, Hazell P (2006) The mental health of chil-
that is not adequately conceptualized within DSM-IV-TR dren in foster and kinship care in New South Wales, Australia.
or ICD-10, or the proposed DSM-V and ICD-11 diagnostic J Paediatr Child Health 42:91–99
classifications. The 20 % of children who presented with 10. Stein E, Rae-Grant N, Ackland S, Avison W (1994) Psychiatric
disorders of children ‘‘in care’’: methodology and demographic
complex attachment- and trauma-related symptomatology correlates. Can J Psychiatry 39(6):341–347
experienced more profound developmental adversity, 11. McCann J, Wilson S, Dunn G (1996) Prevalence of psychiatric
endured greater placement instability, and had greater disorders in young people in the care system. Br Med J
evident need for mental health services, than children with 313:1529–1530
12. Blower A, Addo A, Hodgson J, Lamington L, Towlson K (2004)
less complex disorders. This presents a particular quandary Mental health of ‘looked after’ children: a needs assessment. Clin
for clinical practice with children in care—namely, that Child Psychol Psychiatry 9:117–129
standard clinical reasoning is likely to produce inaccurate 13. McMillen JC, Zima BT, Scott L, Auslander W, Munson M,
formulation of the mental health difficulties of those chil- Ollie M, Spitznagel E (2005) Prevalence of psychiatric disorders
among older youths in the foster care system. J Am Acad Child
dren who have greatest need for our professional opinions Adolesc Psychiatry 44(1):88–95
and interventions. Accurate formulation of clinical phe- 14. Famularo R, Augustyn M (1996) Persistence of pediatric post
nomena is an essential component of evidence-based psy- traumatic stress disorder after 2 years. Child Abuse Negl
chological practice. Until further progress is achieved, it is 20(12):1245–1248
15. Dubner AE, Motta RW (1999) Sexually and physically abused
important that clinicians understand the particular limita- foster care children and posttraumatic stress disorder. J Consult
tions of standard classifications for children in care who Clin Psychol 67(3):367–373
have complex symptomatology, and modify their clinical 16. Tarren-Sweeney M (2008) Retrospective and concurrent predic-
reasoning and formulations to account for this. When tors of the mental health of children in care. Child Youth Serv
Rev 30:1–25
complex attachment- and trauma-related difficulties are not 17. Tarren-Sweeney M (2007) The assessment checklist for chil-
adequately explained using standard diagnostic classifica- dren—ACC: a behavioral rating scale for children in foster,
tion, clinicians should employ a symptom profile approach kinship and residential care. Child Youth Serv Rev 29:672–691
to clinical formulation, taking account of the forms, 18. Tarren-Sweeney M (2006) Patterns of aberrant eating among pre-
adolescent children in foster care. J Abnorm Child Psychol
severity, and complexity of a child’s total symptomatology. 34:623–634
19. Tarren-Sweeney M (2008) Predictors of problematic sexual
Acknowledgments This study was funded and supported by the behavior among children with complex maltreatment histories.
NSW Department of Family Community Services. Child Maltreat 13(2):182–198
20. Friedrich WN, Baker A, Parker R, Schneiderman M, Gries L,
Archer M (2005) Youth with problematic sexualised behaviors in
the child welfare system: a one-year longitudinal study. Sex
Abuse 17(4):391–406
21. DeJong M (2010) Some reflections on the use of psychiatric
References diagnosis in the looked after or ‘in care’ child population. Clin
Child Psychol Psychiatry 15(4):589–599
1. Achenbach T, Rescorla L (2001) Manual for ASEBA school-age 22. Prior MR, Eisenmajer R, Leekman S, Wing L, Gould J, Ong B,
forms and profiles. University of Vermont, Research Center for Dowe D (1998) Are there subgroups within the austistic spec-
Children, Youth, & Families, Burlington trum? A cluster analysis of a group of children with Autistic
2. Goodman R (2001) Psychometric properties of the strengths and Spectrum Disorders. J Child Psychol Psychiatry 39(6):893–902
difficulties Questionnaire. J Am Acad Child Adolesc Psychiatry 23. Bitsika V, Sharpley C, Orapeleng S (2008) An exploratory
40:1337–1345 analysis of the use of cognitive, adaptive and behavioural indices
3. Elander J, Rutter M (1996) Use and development of the Rutter for cluster analysis of ASD subgroups. J Intellect Disabil Res
parents’ and teachers’ scales. Int J Methods Psychiatr Res 52(11):973–985
6(2):63–78 24. Holtmann M, Buchmann A, Esser G, Schmidt M, Banaschewski
4. Armsden G, Pecora PJ, Payne VH, Szatkiewicz JP (2000) Chil- T, Laucht M (2011) The child behavior checklist—dysregulation
dren placed in long-term foster care: an intake profile using the profile predicts substance use, suicidality, and functional
child behavior checklist/4-18. J Emot Behav Dis 8(1):49–64 impairment: a longitudinal analysis. J Child Psychol Psychiatry
5. Heflinger CA, Simpkins CG, Combs-Orme T (2000) Using the 52(2):139–147
CBCL to determine the clinical status of children in state custody. 25. Strijker J, Zandberg T, Van der Meulen BF (2002) Taxonomy of
Child Youth Serv Rev 22(1):55–73 foster children in foster homes. Int J Child Fam Welf 1–2:18–27
6. Pilowsky D (1995) Psychopathology among children placed in 26. Achenbach T (1993) Empirically based taxonomy: how to use
family foster care. Psychiatr Serv 46(9):906–910 syndromes and profile types derived from the CBCL/4-18, TRF,

123
Child Psychiatry Hum Dev (2013) 44:727–741 741

and YSR. University of Vermont Department of Psychiatry, component of the national survey of mental health and well-
Burlington being. Aust N Z J Psychiatry 35(6):806–814
27. Achenbach T (1991) Manual for the child behavior checklist/ 38. Achenbach T, Dumenci L, Rescorla L (2003) DSM-oriented and
4–18, and 1991 profile. University of Vermont, Burlington empirically based approaches to constructing scales from the
28. Tarren-Sweeney M, Hazell P, Carr V (2004) Are foster parents same item pools. J Clin Child Adolesc Psychol 32(3):328–340
reliable informants of children’s behaviour problems? Child Care 39. Richters MM, Volkmar FR (1994) Reactive attachment disorder
Health Dev 30(2):167–175 of infancy or early childhood. J Am Acad Child Adolesc Psy-
29. Tarren-Sweeney M (in press) Clinician’s guide to the assessment chiatry 33(3):328–332
checklist series: specialized mental health measures for children 40. Lieberman AF, Zeanah CH (1995) Disorders of attachment in
in care. Routledge, London infancy. Child Adolesc Psychiatr Clin N Am 4(3):571–587
30. StataCorp (2011) Stata. 12.0 edn. StataCorp LP, College Station, 41. Minnis H, Keck G (2003) A clinical/research dialogue on reactive
TX attachment disorder. Attachment Hum Dev 5(3):297–301
31. Nakamura B, Ebesutani C, Bernstein A, Chorpita B (2009) 42. Smyke A, Dumitrescu A, Zeanah CH (2002) Attachment distur-
A psychometric analysis of the child behavior checklist DSM- bances in young children. I: the continuum of caretaking. J Am
oriented scales. J Psychopathol Behav Assess 31:178–189 Acad Child Adolesc Psychiatry 41(8):972–982
32. Steele R, Aylward B (2007) The use of cluster analytic tech- 43. Zero to Three (2005) Diagnostic classification of mental health and
niques in developmental and behavioral research. J Dev Behav developmental disorders of infancy and early childhood: revised
Pediatr 28(4):327–329 edition (DC: 0–3R). Zero To Three Press, Washington, DC
33. Henry D, Tolan P, Gorman-Smith D (2005) Cluster analysis in 44. Becker S, Langberg J, Vaughn A, Epstein J (2012) Clinical utility
family psychology research. J Fam Psychol 19(1):121–132 of the Vanderbilt ADHD diagnostic parent rating scale comor-
34. Edelbrock C, Achenbach T (1980) A typology of child behavior bidity screening scales. J Dev Behav Pediatr 33:221–228
profile patterns: distribution and correlates for disturbed children 45. Becker S, Luebbe A, Langberg J (2012) Co-occurring mental
aged 6–16. J Abnorm Child Psychol 8(4):441–470 health problems and peer functioning among youth with atten-
35. Milligan G, Cooper M (1988) A study of standardization of tion-deficit/hyperactivity disorder: a review and recommenda-
variables in cluster analysis. J Classif 5:181–204 tions for future research. Clin Child Fam Psychol Rev 15:
36. Schaffer C, Green P (1996) An empirical comparison of variable 279–302
standardization methods in cluster analysis. Multivar Behav Res 46. van der Kolk B (2005) Developmental trauma disorder. Psychiatr
31(2):149–167 Ann 35(5):401–408
37. Sawyer MG, Arney FM, Baghurst PA, Clark JJ, Graetz BW, 47. D’Andrea W, Ford J, Stolbach B, Spinazzola J, van der Kolk B
Kosky RJ, Nurcombe B, Patton GC, Prior MR, Raphael B, Rey (2012) Understanding interpersonal trauma in children: why we
JM, Whaites LC, Zubrick SR (2001) The mental health of young need a developmentally appropriate trauma diagnosis. Am J
people in Australia: key findings from the child and adolescent Orthopsychiatry 82(2):187–200

123