Abstract

This study identified the subtypes of psychosocial functioning in children who had sustained traumatic brain injury using the Behavior Assessment System for Children, Second Edition. Participants (N = 91) were aged 6–20. Using hierarchical agglomerative clustering techniques, a reliable typology emerged that consisted of two subtypes, which were labeled as Normal and Pervasive Emotional Difficulties. Using further exploratory analyses, other less statistically reliable subtypes were also observed, which were thought to have clinical significance. These were labeled as Mild Externalizing/Depression, Mild Externalizing/Attention Problems, Mild Depression, and Mild Anxiety. The majority of participants were assigned to the Normal subtype. Relationships between subtypes and other variables, including gender, time elapsed since injury, age at injury, and age at testing were also analyzed, with time elapsed since injury being the only variable to significantly differentiate the subtypes.

Introduction

Humans are social beings, and social and emotional competence is one of the developmental tasks that must be mastered during childhood. These skills are important to a child's later mental health, as well as social and academic success. Given that a large proportion of the brain is involved in social interaction, it follows logically that any insult to the neurodevelopmental processes involved in acquiring these skills would result in disruption of both the structural and the functional development of the brain. In fact, it is generally posited that children who sustain traumatic brain injuries (TBIs) tend to manifest psychosocial and behavioral problems as a result of those injuries. Although the exact nature of the observed deficits may vary according to the specific combination of focal and diffuse damage, injury anywhere in the social brain network will disrupt the normal functioning of the system. This, in turn, will place the child at greater risk for deficits in social information processing (Yeates et al., 2007). In fact, even the milder forms of pediatric brain injury increase the risk for subsequent psychiatric problems (Luis & Mittenberg, 2002).

Studies have identified specific psychosocial sequelae of pediatric TBI, including personality change, ADHD, depression, and obsessive-compulsive symptoms (Bloom et al., 2001; Grados et al., 2008; Max et al., 2000). Children with TBI are also likely to be developmentally behind their same-age peers in their responses to problem solving, and they tend to struggle more with negotiation and collaboration (Hanten et al., 2008). Further, self-regulation has been found to be an area of deficit following childhood TBI (Ganesalingam, Sanson, Anderson, & Yeates, 2006, 2007). Research has also been conducted in an attempt to identify profiles of psychosocial functioning based on scores earned on general behavior rating scales (Butler, Rourke, Fuerst, & Fisk, 1997; Hayman-Abello, Rourke, & Fuerst, 2003). Hayman-Abello and colleagues used the Child Behavior Checklist (CBCL; Achenbach, 1991) and identified four subtypes: Normal, Attention, Delinquent, and Withdrawn-somatic. Although most children fell within the Normal subtype, parents of children with TBI observed more behavior problems than parents of non-referred children. The problems they observed did not, however, represent pathological psychosocial functioning.

Butler and colleagues (1997) used the Personality Inventory for Children-Revised (PIC-R; Wirt, Lachar, Klinedinst, & Seat, 1990) and also found most children to fall within the Normal subtype. Other subtypes described included cognitive deficit, somatic concern, mild anxiety, internalized psychopathology, Antisocial, and social isolation. Regarding relationships with severity of TBI injury, proportionately more severely injured children fell in the social isolation subtype, whereas more mildly and moderately injured children fell in the cognitive deficit subtype. Also, the Normal and Antisocial subtypes contained proportionately more moderately injured children, and the somatic concern subtype contained proportionately more moderately and severely injured children. Equal numbers of children comprised the mild anxiety and internalized psychopathology subtypes. The relationships between psychosocial functioning and both age at injury and time since injury were also examined. Generally, younger children were more likely to be included in the social isolation, cognitive deficit, and mild anxiety subtypes. More specifically, age at injury was greater for the Antisocial subtype than the mild anxiety, cognitive deficit, and social isolation subtypes; age at injury was greater for the Normal subtype than the cognitive deficit and social isolation subtypes; and age at injury was greater for the somatic concern subtype than the social isolation subtype. Thus, children injured at younger ages were significantly more likely to be assigned to the social isolation, cognitive deficit, and mild anxiety subtypes. For time since injury, children with a longer time since injury were more likely to be assigned to the social isolation subtype than the Antisocial and Normal subtypes.

To date, these are the only two studies that have attempted to identify profiles of psychosocial functioning of children and adolescents with TBI using widely available, general behavior rating scales. Unfortunately, both of these studies are somewhat dated in that they utilized instruments that have since been revised. The Behavior Assessment System for Children, Second Edition (BASC-2; Reynolds & Kamphaus, 2004) is a general behavior rating scale with more recently developed norms that was designed to be used as a clinical assessment tool. When compared with the CBCL, the BASC-2 possesses similar reliabilities; however, the BASC-2 is more closely linked to diagnostic definitions of behavioral disturbances. Further, although the empirical validity research of the BASC-2 is not nearly as extensive as that of the CBCL, the BASC-2 was designed to be used clinically instead of as a research tool.

Research informs us that increasing numbers of children are surviving TBIs (Thurman, Alverson, Dunn, Guerrero, & Sniezek, 1999), and professionals responsible for completing their evaluations, including school and clinical neuropsychologists, may be familiar with the BASC-2 in addition to the CBCL and PIC-2. Therefore, information regarding the characteristics of psychosocial problems in children suffering from TBI based on assessment using the BASC-2 would be helpful to these and other clinicians and, thus, represents research that is yet to be addressed.

Thus, the focus of this project was to provide parallel research to the studies already completed using the CBCL and PIC-R and to develop a BASC-2-based psychosocial profile of children with TBI. The following research questions were proposed: How do these derived BASC-2 profiles compare with known CBCL and PIC-R profiles? How do these derived BASC-2 profiles compare with known BASC-2 clinical profiles? What is the relationship between psychosocial outcome and variables such as gender, time elapsed since injury, age at injury, and age at testing? Formal hypotheses regarding answers to these research questions were not formulated, as the nature of this study was primarily exploratory. In addition, the purpose of this study was to replicate and extend the previous research studies just described. Thus, it was assumed that, as in the CBCL and PIC-R studies, a majority of participants would be classified within the normal range of psychosocial functioning.

Method

Research Participants

Participants were selected from a database maintained at a pediatric hospital in a large urban city providing treatment and rehabilitation for children and adolescents with TBI. Available demographic information included assessment number (e.g., initial assessment, second assessment), gender, ethnicity, age, handedness, diagnosis, time since injury, type of head injury (i.e., open or closed head injury), cause of injury (e.g., motor vehicle accident, fall, bicycle accident), whether or not the child was restrained in a motor vehicle accident, Glasgow Coma Scale (GCS) score, and neuropsychological test battery scores for children admitted to the hospital. Unfortunately, the severity of injury as measured by the GCS was not available for approximately half of the participants, and information regarding premorbid functioning was not available for any of the participants.

Participants were excluded from the current study if a BASC-2 Parent Rating Scale (PRS)—Child or BASC-2 PRS—adolescent had not been completed as a part of their comprehensive neuropsychological test battery, hence only those patients aged 6 years 0 months to 21 years 11 months were considered for inclusion. A note is also in order regarding items for which no answer or more than one answer were marked by the respondent, since an excess of these unscorable items can undermine the validity of the score generated for the scale to which it applies. Thus, for this study, if any participant's line of scores was missing a clinical scale score, the participant's scores were eliminated from further analysis.

Information was also gathered regarding gender, age at which participants sustained the TBI, chronological age at the time of the neuropsychological assessment, and time interval between the injury and the assessment. The final sample consisted of 91 children and adolescents (62 male and 29 female) between the ages of 6 years, 2 months and 20 years, 2 months at the time of neuropsychological assessment (M = 13.43, SD = 3.55). Ethnic composition, time elapsed since injury, age at testing, age at injury, and GCS score are summarized in Table 1.

Table 1.

Sample frequencies and descriptives

Variable n Percentage M SD Range 
Gender 91     
 Male 62 68.1    
 Female 29 31.9    
Ethnicity 88     
 Caucasian 67 73.6    
 African American 11 12.1    
 Hispanic 10 11.0    
Time since injury 90     
 <6 months 4.4    
 6–12 months 47 51.6    
 13–24 months 19 20.9    
 >24 months 20 22.0    
Age at testing 91  13.43 3.55 6.17–20.17 
Age at injury 90  11.76 4.00 2.25–18.50 
GCS score 47  6.30 2.97 3.00–13.00 
Variable n Percentage M SD Range 
Gender 91     
 Male 62 68.1    
 Female 29 31.9    
Ethnicity 88     
 Caucasian 67 73.6    
 African American 11 12.1    
 Hispanic 10 11.0    
Time since injury 90     
 <6 months 4.4    
 6–12 months 47 51.6    
 13–24 months 19 20.9    
 >24 months 20 22.0    
Age at testing 91  13.43 3.55 6.17–20.17 
Age at injury 90  11.76 4.00 2.25–18.50 
GCS score 47  6.30 2.97 3.00–13.00 

Note: GCS = Glasgow Coma Scale.

Measures

The BASC-2 (Reynolds & Kamphaus, 2004) is a widely used, multidimensional, multimethod system designed to facilitate differential diagnosis and educational classification of a variety of emotional and behavioral disorders in children. It is multidimensional in that it measures both positive, or adaptive, and negative, or clinical, aspects of behavior. It is multimethod in that it has five components, which can be used individually or in combination. These components include two rating scales, one for teachers and one for parents or caregivers; a self-report scale; a Structured Developmental History form; and a form for recording and classifying directly observed classroom behavior.

The BASC-2 applies a triangulation method for gathering information by analyzing information from three different perspectives. A Self-Report of Personality is available for children and adolescents aged 6-0 through 21-11 to complete. A Teacher Rating Scale completed by a child's or adolescent's teacher, a Student Observation System, and a portable observation program are available to gather information from the school setting. Finally, a Parent Rating Scale, Structured Developmental History, and Parenting Relationship Questionnaire are available for parents to complete in order to provide information regarding the child's behavior in home and community settings, background information on the child, and information regarding parental relationships. Only the PRS was examined for the current study. Parents or caregivers can complete any one of three forms, depending on the child's age: preschool (ages 2–5), child (ages 6–11), or adolescent (ages 12–21). Forms can be completed in approximately 10–20 min using a four-choice response format (A for almost always, O for often, S for sometimes, and N for never).

Although the BASC-2 offers various adaptive and clinical scales to measure both adaptive and problem behaviors, analyses for the present study included consideration of only the clinical scales of the child and adolescent PRS forms. The adaptive scales were eliminated from the analysis for several reasons: (a) the adaptive scales measure positive or desirable traits rather than problem areas; (b) adaptive skills are impacted by secondary medical issues (e.g., physical disabilities), and independent living skills may not be validly assessed in a child who has spent much of the time preceding assessment in a hospital; and (c) this study parallels other studies (Butler et al., 1997; Hayman-Abello et al., 2003) using the CBCL and PIC, neither of which include scales measuring adaptive skills. The composite scales were eliminated from the analyses since they are aggregates of the clinical and adaptive scales. Therefore, any analyses performed with them would be duplicating data already in the analysis. Thus, only the clinical scales were included in the final analyses.

The PRS clinical scales on the child and adolescent forms measure maladaptive behaviors, with higher scores representing negative or undesirable characteristics that result in impaired functioning in home or community settings and/or peer relationships. Internalizing scales include Anxiety, Depression, and Somatization, whereas externalizing scales include Hyperactivity, Aggression, and Conduct Problems. The final three clinical scales include Attention Problems, Atypicality, and Withdrawal.

For the BASC-2 clinical scales, T-score elevations above the mean indicate an increased likelihood of significant psychopathology. A normalized T-score of <60 is considered to be within the normal range and suggests that the child does not display behaviors measured by that scale any more often than others his or her age. T-scores between 60 and 69 fall in the designated at-risk range, which indicates the presence of a significant problem that might require treatment but may not be severe enough to warrant a formal diagnosis. T-scores of ≥70 fall in the clinically significant classification range and represent a high level of maladaptive behavior.

Data Analysis

The primary purpose of this study was to group participants based on the similarity of their BASC-2 profiles; therefore, profiles for all participants using all nine of the BASC-2 clinical scales (i.e., Aggression, Anxiety, Attention Problems, Atypicality, Conduct Problems, Depression, Hyperactivity, Somatization, and Withdrawal) were subjected to agglomerative hierarchical cluster analyses. Whenever multivariate subtyping techniques are applied, even to random data, groups of subjects will always be produced. Therefore, it is important to determine the validity of the subtypes generated through cluster analysis by replicating subtypes either across different samples from the same population or across different subtyping techniques (Borgen & Barnett, 1987; Fuerst, Fisk, & Rourke, 1989). The latter method was employed in this study, as correspondence between the cluster analysis derived subtypes provides good evidence for the reliability of results.

A preliminary run using UPGMA (between-groups linkage) and Ward's (1963) methods was completed with no limits placed on the number of clusters that should be included in the solution. However, these two methods did not produce cluster solutions that demonstrated good agreement with one another; therefore, a broader, more exploratory, approach to the analysis was undertaken. To take full advantage of an exploratory approach, cluster analyses were run using each of the methods available in the SPSS statistical package, including between-groups linkage (also called UPGMA), within-groups linkage (also referred to as “average linkage within groups”), nearest neighbor or “single linkage” method, furthest neighbor or “complete linkage” method, centroid clustering (also called the “weighted pair-group method using centroid averages or WPGMC”), median clustering (also called the “unweighted pair-group method using centroid averages or UPGMC”), and Ward's method. The Euclidean distance was used as the interval measure; however, since squared Euclidean distance is preferred over simple Euclidean distance as the distance measure for the centroid, median, and Ward's methods, a second run using squared Euclidean distance was completed for these three methods. In addition, with these exploratory analyses, limits were placed on the minimum and maximum number of subtypes to be included in each solution. Given the seven-subtype solution derived in the Butler and colleagues (1997) study, which the current study is intended to parallel, a minimum of two and a maximum of seven clusters were specified. Thus, two-, three-, four-, five-, six-, and seven-cluster solutions were run and analyzed for each of the hierarchical cluster methods mentioned above. A decision regarding the most reliable and parsimonious solution was made based on the results of these analyses.

To determine statistical differences between clusters in the final solution, a repeated-measures multivariate analysis of variance (MANOVA) was calculated. Mean normalized total T-scores within each subtype were obtained to provide an indication of overall profile elevation for each psychosocial subtype. These mean normalized T-scores served as the repeated-measures dependent variables in the MANOVA, with the psychosocial subtypes serving as the between-subjects factors. Pairwise comparisons were calculated to determine which BASC-2 subscales were statistically higher than others within each factor.

To validate any cluster solution, it is important to compare the resulting subtypes on variables that were not included in the original clustering process. Various descriptive variables were examined for this purpose. Chi-squared analyses were conducted for gender, ethnicity, and months elapsed since injury because of their categorical nature. Months elapsed since injury were initially coded based on the following four categories: <6, 6–12, 13–24, and >24 months. Unfortunately, some cells contained a count <5; therefore, the four categories were collapsed into the following two categories: ≤12 and >12 months. Analysis of variance was calculated for age at injury, age at testing, and GCS given their continuous nature.

Visual inspection was used to compare the subtypes derived in this study with those derived in the Butler and colleagues (1997) and Hayman-Abello and colleagues (2003) studies. In addition, the subtypes derived in this study were statistically compared with the BASC-2 clinical sample profiles, as reported in the BASC-2 manual (Reynolds & Kamphaus, 2004). Each BASC-2 subscale mean was compared with the subscale mean derived in the current study using one-sample t-tests. It should be noted that, since clinical profiles in the BASC-2 manual are provided separately for children and adolescents, the sample in this study was divided into two groups (ages 6–11 and 12–21) prior to this statistical analysis.

Results

Descriptive Statistics

The current sample included 91 children and adolescents. Characteristics of the participants in terms of gender, ethnicity, time elapsed since injury, age at assessment, age at injury, and GCS score are reported in Table 1. Means and standard deviations for the BASC-2 PRS variables are reported in Table 2. All means reported in Table 2 fell within the average range.

Table 2.

Means and standard deviations across BASC-2 PRS variables

Variable M SD Range 
Hyperactivity 57.31 13.64 37–93 
Aggression 55.73 12.94 36–87 
Conduct Problems 53.37 13.37 34–99 
Anxiety 52.38 14.06 30–88 
Depression 58.70 15.90 39–109 
Somatization 51.05 13.14 35–100 
Atypicality 57.89 14.50 40–114 
Withdrawal 51.29 11.05 35–85 
Attention Problems 57.36 10.45 36–84 
Variable M SD Range 
Hyperactivity 57.31 13.64 37–93 
Aggression 55.73 12.94 36–87 
Conduct Problems 53.37 13.37 34–99 
Anxiety 52.38 14.06 30–88 
Depression 58.70 15.90 39–109 
Somatization 51.05 13.14 35–100 
Atypicality 57.89 14.50 40–114 
Withdrawal 51.29 11.05 35–85 
Attention Problems 57.36 10.45 36–84 

Psychosocial Subtype Derivation

As described previously, analyses using UPGMA and Ward's methods were attempted initially; however, there was not good agreement between the solutions derived using these methods. Therefore, a variety of agglomerative hierarchical techniques were explored in an effort to arrive at a more parsimonious solution. Two common methods used to establish the reliability of hierarchical techniques include replication of the subtypes in a different sample from the same population or replication using different subtyping techniques. Establishing internal reliability becomes especially important in exploratory studies using multivariate subtyping techniques since the structure of the data set is unknown or poorly understood. As this was the case in this study, a decision was made to establish internal reliability by replicating the derived cluster solution using additional clustering techniques. Inspection of the results of these exploratory analyses as well as cross-tabulations between each pair of analyses revealed best results with the Ward's and centroid methods using the squared Euclidean distance as the distance criterion for grouping cases. Based on Rand's statistic, agreement between the solutions derived from these two methods was good, and, thus, these solutions were retained.

Examination of the cluster solutions using these methods suggested the presence of two subtypes. Descriptive labels were assigned by the primary researcher to each subtype based on the major features of the profile: Normal and Pervasive Emotional Difficulties (PED).

Agreement between the two methods on the two-cluster solution was good, with only two cases being differentially placed when comparing the results generated from both methods. Rand's statistic was 0.98, demonstrating that these two subtypes were replicated with excellent accuracy by the two clustering techniques. (For Rand's statistic, a value of 0.0 indicates purely chance agreement, whereas 1.0 indicates complete agreement.) Although both methods produced highly similar solutions, Ward's method maximizes between-group differences and minimizes within-group distances and is preferred by some researchers for this reason (Borgen & Barnett, 1987). Therefore, the solution using Ward's method was selected for the purpose of reporting results. For each of the two subtypes, mean BASC-2 scores on all nine scales were calculated to obtain profiles (Fig. 1).

Fig. 1.

Mean BASC-2 profiles for the two-cluster solution. HYP = Hyperactivity; AGG = Aggression; CON = Conduct; ANX = Anxiety; DEP = Depression; SOM = Somatization; ATYP = Atypicality; WITH = Withdrawal; ATT = Attention Problems.

Fig. 1.

Mean BASC-2 profiles for the two-cluster solution. HYP = Hyperactivity; AGG = Aggression; CON = Conduct; ANX = Anxiety; DEP = Depression; SOM = Somatization; ATYP = Atypicality; WITH = Withdrawal; ATT = Attention Problems.

The profile for the Normal subtype did not contain any clinically significant or at-risk level score elevations. In other words, no score fell within the range of scores ≥60 T. Parents of children in this subtype did not express undue concern about their child's psychosocial functioning. This subtype was the largest, accounting for 86% of the sample.

The PED profile was characterized by clinically significant elevations on all subscales except Anxiety, Withdrawal, and Attention Problems. Scores on these three subscales fell within the at-risk range. Children falling within this subtype appear to have significant difficulties with externalizing behavioral control as well as symptoms related to depression. In addition, these children behave in ways that are often viewed by others as being strange or odd. As a result, these children may not interact or integrate well with other children in the school setting and may require specialized assistance regarding management of acting out behaviors and depressive symptoms. This subtype accounted for 14% of the sample.

Although the two-cluster solution appeared most parsimonious, the analyses using Ward's method retained their integrity throughout the three-, four-, and five-cluster solutions. The drawback to retaining a three-, four-, or five-cluster solution using Ward's method in place of the two-cluster solution is that none of the other clustering techniques provided support in the form of replicability. For example, Rand's statistic fell to 0.57, 0.55, and 0.34 for the three-, four-, and five-cluster solutions, respectively, when Ward's method and centroid clustering technique results were compared, suggesting that there is not good agreement between cluster solutions. However, because valuable information of clinical significance could be lost otherwise, the subtypes composing the three-, four-, and five-cluster solutions are described below. For each of the three-, four-, and five-cluster solutions, mean BASC-2 scores on all nine scales were calculated to obtain profiles (Figs. 2–4). It should be noted that the Normal and PED subtypes were retained virtually unchanged throughout each of the cluster solutions.

Fig. 2.

Mean BASC-2 profiles for the three-cluster solution. HYP = Hyperactivity; AGG = Aggression; CON = Conduct; ANX = Anxiety; DEP = Depression; SOM = Somatization; ATYP = Atypicality; WITH = Withdrawal; ATT = Attention Problems. Subtype accounted for 32% of the sample; within the five-cluster solution, it accounted for 16% of the sample.

Fig. 2.

Mean BASC-2 profiles for the three-cluster solution. HYP = Hyperactivity; AGG = Aggression; CON = Conduct; ANX = Anxiety; DEP = Depression; SOM = Somatization; ATYP = Atypicality; WITH = Withdrawal; ATT = Attention Problems. Subtype accounted for 32% of the sample; within the five-cluster solution, it accounted for 16% of the sample.

Fig. 3.

Mean BASC-2 profiles for the four-cluster solution. HYP = Hyperactivity; AGG = Aggression; CON = Conduct; ANX = Anxiety; DEP = Depression; SOM = Somatization; ATYP = Atypicality; WITH = Withdrawal; ATT = Attention Problems.

Fig. 3.

Mean BASC-2 profiles for the four-cluster solution. HYP = Hyperactivity; AGG = Aggression; CON = Conduct; ANX = Anxiety; DEP = Depression; SOM = Somatization; ATYP = Atypicality; WITH = Withdrawal; ATT = Attention Problems.

Fig. 4.

Mean BASC-2 profiles for the five-cluster solution. HYP = Hyperactivity; AGG = Aggression; CON = Conduct; ANX = Anxiety; DEP = Depression; SOM = Somatization; ATYP = Atypicality; WITH = Withdrawal; ATT = Attention Problems.

Fig. 4.

Mean BASC-2 profiles for the five-cluster solution. HYP = Hyperactivity; AGG = Aggression; CON = Conduct; ANX = Anxiety; DEP = Depression; SOM = Somatization; ATYP = Atypicality; WITH = Withdrawal; ATT = Attention Problems.

The “Mild Externalizing/Depression subtype” profile included at-risk level elevations on the Hyperactivity, Aggression, Depression, Atypicality, and Attention Problems subscales with the mean score for the Depression subscale reaching the clinically significant level in the five-cluster solution. In contrast, the “Mild Externalizing/Attention Problems subtype” profile included at-risk level elevations on the Hyperactivity, Aggression, Conduct, and Attention Problems subscales. The behavioral characteristics of children in these two subtypes include difficulties with externalizing behaviors and inattention, with the differentiation between the two being the observation of depressive symptoms in the former subtype.

The third subtype, the “Mild Depression subtype,” emerged only in the four-cluster solution and was characterized by a relatively normal BASC-2 profile, with the exception of at-risk level elevations on the Hyperactivity and Atypicality subscales and a single clinically significant elevation on the Depression subscale. The “Mild Anxiety subtype” was also characterized by a relatively normal BASC-2 profile, although the Anxiety and Somatization subscale scores were statistically higher. This subtype emerged exclusively in the five-cluster solution.

Subtype Differences across Psychosocial Variables

A multivariate general linear model (GLM) was computed with the BASC-2 subscales as repeated dependent measures (within-subjects factors) and the two psychosocial subtypes derived from the Ward's method cluster analysis serving as the between-subjects factors. Homogeneity of variance/covariance matrices were analyzed to determine if the data met the criteria for univariate or multivariate approaches to the analysis. In this case, a multivariate approach to the data could not be completed due to the violation of the equality of homogeneity of the covariance matrices of the dependent variable as determined by Box's M test, F(45, 1,496.84) = 1.81, p < .01. Therefore, a univariate GLM was undertaken. The assumption of sphericity as tested by Mauchly's test of sphericity was not met (p < .001). Given this violation of sphericity, the Greenhouse–Geisser epsilon adjustment was applied to the data. Greenhouse–Geisser was chosen because this is considered a conservative adjustment, especially when the sample size is low. Using these adjusted values, tests of within-subjects effects showed an overall significant main effect for the BASC-2 PRS subscales across the psychosocial subtypes, F(5.18, 460.80) = 10.76, p < .01, η2 = 0.11. This was a moderate effect. The interaction of BASC-2 subscales and the psychosocial subtypes was also significant, F(5.18, 460.80) = 4.71, p < .01, η2 = 0.05, although this constituted a weak effect with only 5% of the variance in the BASC-2 subscales accounted for by the psychosocial subtypes. Because the interaction was significant, simple effects of each cluster were also examined. Simple effects for both the Normal and PED clusters were also significant, F(5.61, 432.18) = 9.24, p < .01, η2 = 0.11 for the Normal subtype and F(3.21, 38.57) = 3.81, p < .05, η2 = 0.24 for the PED subtype.

Univariate between-subjects tests showed that the psychosocial subtypes were significantly related to the BASC-2 subscales, F(1, 89) = 126.38, p < .01, η2 = 0.59. The effect size was large, with 59% of the variance accounted for in the BASC-2 subscales by the psychosocial subtypes. Table 3 demonstrates the significant differences between the subtypes across the psychosocial variables based on pairwise comparisons (p < .05).

Table 3.

Means and standard deviations for BASC-2 variables for psychosocial subtypes using Ward's two-cluster solution

 Normal (n = 78; M [SD]) PED (n = 13; M [SD]) 
Hyp 53.54 (10.06) 79.92 (10.03)** 
Agg 52.68 (10.79) 74.00 (9.32)** 
Con 50.59 (9.18) 73.54 (17.66)** 
Anx 50.10 (12.24) 66.08 (16.87)** 
Dep 54.37 (11.35) 84.69 (14.69)** 
Som 47.45 (8.23) 72.69 (16.40)** 
Atyp 53.76 (10.04) 82.69 (12.33)** 
Wdw 48.68 (8.35) 66.92 (12.63)** 
Att 55.31 (9.53) 69.69 (6.55)** 
 Normal (n = 78; M [SD]) PED (n = 13; M [SD]) 
Hyp 53.54 (10.06) 79.92 (10.03)** 
Agg 52.68 (10.79) 74.00 (9.32)** 
Con 50.59 (9.18) 73.54 (17.66)** 
Anx 50.10 (12.24) 66.08 (16.87)** 
Dep 54.37 (11.35) 84.69 (14.69)** 
Som 47.45 (8.23) 72.69 (16.40)** 
Atyp 53.76 (10.04) 82.69 (12.33)** 
Wdw 48.68 (8.35) 66.92 (12.63)** 
Att 55.31 (9.53) 69.69 (6.55)** 

Notes: Hyp = Hyperactivity; Agg = Aggression; Con = Conduct Problems; Anx = Anxiety; Dep = Depression; Som = Somatization; Atyp = Atypicality; Wdw = Withdrawal; Att = Attention Problems; PED = Pervasive Emotional Difficulties. Column differences >10.23 for PED and 2.63 for Normal indicate significant subscale differences by group.

**p < .01 indicates significant group differences by subscale.

Because clinical significance can be drawn from the less statistically reliable three-, four-, and five-cluster solutions derived using Ward's method, a more detailed analysis of their statistical foundations appears warranted. Multivariate GLM statistics were again computed. In all three cluster solutions, Box's test of the equality of covariance matrices was significant, indicating that univariate, as opposed to multivariate, approaches were more appropriate. In addition, the assumption of sphericity was violated in all cases; therefore, again, the Greenhouse–Geisser epsilon adjustment was applied. For all three solutions, tests of within-subjects effects showed overall significant main effects for the BASC-2 PRS subscales across the psychosocial subtypes, F(5.17, 454.89) = 14.37, p < .01, η2 = 0.14 for the three-cluster solution, F(5.43, 472.22) = 17.61, p < .01, η2 = 0.17 for the four-cluster solution, and F(5.62, 483.24) = 14.95, p < .01, η2 = 0.15 for the five-cluster solution. Effect size was moderate in all cases, with 14%, 17%, and 15% of the variance in the BASC-2 subscales accounted for by the psychosocial subtypes in the three-, four-, and five-cluster solutions, respectively.

The interactions of BASC-2 subscales and the psychosocial subtypes were also significant in each cluster solution, F(10.34, 454.89) = 5.20, p < .01, η2 = 0.11 for the three-cluster solution, F(16.28, 472.22) = 6.71, p < .01, η2 = 0.19 for the four-cluster solution, and F(22.48, 483.24) = 6.25, p < .01, η2 = 0.23 for the five-cluster solution. Effect size increased with each successive cluster solution, reaching almost 25% with the five-cluster solution. Because these interactions were significant, simple effects of each subtype in each solution were also examined. Within the three-cluster solution, the simple effects for each subtype were also significant, F(5.25, 252.17) = 3.30, p < .01, η2 = 0.06 for the Normal subtype, F(3.21, 38.57) = 3.81, p < .05, η2 = 0.24 for the PED subtype, and F(4.84, 135.59) = 9.31, p < .01, η2 = 0.25 for the Mild Externalizing/Depression subtype. Within the four-cluster solution, all simple effects were again significant, F(5.25, 252.17) = 3.30, p < .01, η2 = 0.06 for the Normal subtype, F(3.21, 38.57) = 3.81, p < .05, η2 = 0.24 for the PED subtype, F(3.50, 45.48) = 11.51, p < .01, η2 = 0.47 for the Mild Externalizing/Attention Problems subtype, and F(3.94, 55.13) = 8.75, p < .01, η2 = 0.39 for the Mild Depression subtype. Within the five-cluster solution, all simple effects were again significant, F(5.16, 149.77) = 8.51, p < .01, η2 = 0.23 for the Normal subtype, F(3.21, 38.57) = 3.81, p < .05, η2 = 0.24 for the PED subtype, F(3.50, 45.48) = 11.51, p < .01, η2 = 0.47 for the Mild Externalizing/Attention Problems subtype, F(3.94, 55.13) = 8.75, p < .01, η2 = 0.39 for the Mild Externalizing/Depression subtype, and F(5.14, 92.55) = 3.00, p < .05, η2 = 0.14 for the Mild Anxiety subtype. It is interesting to note that the size of effect of the PED subtype did not change throughout all four cluster solutions, suggesting that this is a statistically stable subtype and that it was the original Normal subtype that was further partialled in order to derive the additional subtypes in the three-, four-, and five-cluster solutions.

Finally, univariate between-subjects tests showed that the psychosocial subtypes in each cluster solution were significantly related to the BASC-2 subscales, F(2, 88) = 168.37, p < .01, η2 = 0.79 for the three-cluster solution, F(3, 87) = 129.08, p < .01, η2 = 0.82 for the four-cluster solution, and F(4, 86) = 131.14, p < .01, η2 = 0.86 for the five-cluster solution. The effect sizes in each case were large (Tables 4–6).

Table 4.

Means and standard deviations for BASC-2 variables for psychosocial subtypes using Ward's three-cluster solution

 Normal (n = 49; M [SD]) PED (n = 13; M [SD]) ME/D (n = 29; M [SD]) 
Hyp 47.80 (6.53)a 79.92 (10.03)b 63.24 (7.06)c 
Agg 46.82 (5.96)a 74.00 (9.32)b 62.59 (9.86)c 
Con 46.49 (6.14)a 73.54 (17.66)b 57.52 (9.37)c 
Anx 47.31 (9.51)a 66.08 (16.87)b 54.83 (14.85)a 
Dep 49.61 (7.23)a 84.69 (14.69)b 62.41 (12.58)c 
Som 47.00 (8.15)a 72.69 (16.40)b 48.21 (8.45)a 
Atyp 49.18 (6.43)a 82.69 (12.33)b 61.48 (10.39)c 
Wdw 46.14 (7.63)a 66.92 (12.63)b 52.97 (7.86)c 
Att 51.02 (8.09)a 69.69 (6.55)b 62.55 (7.15)b 
 Normal (n = 49; M [SD]) PED (n = 13; M [SD]) ME/D (n = 29; M [SD]) 
Hyp 47.80 (6.53)a 79.92 (10.03)b 63.24 (7.06)c 
Agg 46.82 (5.96)a 74.00 (9.32)b 62.59 (9.86)c 
Con 46.49 (6.14)a 73.54 (17.66)b 57.52 (9.37)c 
Anx 47.31 (9.51)a 66.08 (16.87)b 54.83 (14.85)a 
Dep 49.61 (7.23)a 84.69 (14.69)b 62.41 (12.58)c 
Som 47.00 (8.15)a 72.69 (16.40)b 48.21 (8.45)a 
Atyp 49.18 (6.43)a 82.69 (12.33)b 61.48 (10.39)c 
Wdw 46.14 (7.63)a 66.92 (12.63)b 52.97 (7.86)c 
Att 51.02 (8.09)a 69.69 (6.55)b 62.55 (7.15)b 

Notes: Hyp = Hyperactivity; Agg = Aggression; Con = Conduct Problems; Anx = Anxiety; Dep = Depression; Som = Somatization; Atyp = Atypicality; Wdw = Withdrawal; Att = Attention Problems; PED = Pervasive Emotional Difficulties; ME/D = Mild Externalizing/Depression. Column differences ≥10.23 for PED, 2.63 for ME/D, and 2.61 for Normal indicate significant subscale differences by group. Row means with different superscripts indicate significant differences between groups, p < .01.

Table 5.

Means and standard deviations for BASC-2 variables for psychosocial subtypes using Ward's four-cluster solution

 Normal (n = 49; M [SD]) PED (n = 13; M [SD]) ME/AP (n = 29; M [SD]) MD (n = 14; M [SD]) 
Hyp 47.80 (6.53)a 79.92 (10.03)b 61.14 (5.80)c 65.20 (7.75)c 
Agg 46.82 (5.96)a 74.00 (9.32)b 61.86 (8.59)c 63.27 (11.18)c 
Con 46.49 (6.14)a 73.54 (17.66)b 61.43 (7.51)a 53.87 (9.68)a 
Anx 47.31 (9.51)a 66.08 (16.87)b 44.00 (9.45)a 64.93 (12.16)b 
Dep 49.61 (7.23)a 84.69 (14.69)b 54.00 (4.80)a 70.27 (12.57)c 
Som 47.00 (8.15)a 72.69 (16.40)b 48.71 (9.27)a 47.73 (7.90)a 
Atyp 49.18 (6.43)a 82.69 (12.33)b 57.14 (11.15)a 65.53 (8.02)a 
Wdw 46.14 (7.63)a 66.92 (12.63)b 48.50 (5.88)a 57.13 (7.28)c 
Att 51.02 (8.09)a 69.69 (6.55)b 62.71 (7.29)b 62.40 (7.28)b 
 Normal (n = 49; M [SD]) PED (n = 13; M [SD]) ME/AP (n = 29; M [SD]) MD (n = 14; M [SD]) 
Hyp 47.80 (6.53)a 79.92 (10.03)b 61.14 (5.80)c 65.20 (7.75)c 
Agg 46.82 (5.96)a 74.00 (9.32)b 61.86 (8.59)c 63.27 (11.18)c 
Con 46.49 (6.14)a 73.54 (17.66)b 61.43 (7.51)a 53.87 (9.68)a 
Anx 47.31 (9.51)a 66.08 (16.87)b 44.00 (9.45)a 64.93 (12.16)b 
Dep 49.61 (7.23)a 84.69 (14.69)b 54.00 (4.80)a 70.27 (12.57)c 
Som 47.00 (8.15)a 72.69 (16.40)b 48.71 (9.27)a 47.73 (7.90)a 
Atyp 49.18 (6.43)a 82.69 (12.33)b 57.14 (11.15)a 65.53 (8.02)a 
Wdw 46.14 (7.63)a 66.92 (12.63)b 48.50 (5.88)a 57.13 (7.28)c 
Att 51.02 (8.09)a 69.69 (6.55)b 62.71 (7.29)b 62.40 (7.28)b 

Notes: Hyp = Hyperactivity; Agg = Aggression; Con = Conduct Problems; Anx = Anxiety; Dep = Depression; Som = Somatization; Atyp = Atypicality; Wdw = Withdrawal; Att = Attention Problems; PED = Pervasive Emotional Difficulties; ME/AP = Mild Externalizing/Attention Problems; MD = Mild Depression. Column differences >10.23 for PED, 5.29 for ME/AP, 8.07 for MD, and 2.75 for Normal indicate significant subscale differences by group. Row means with different superscripts indicate significant differences between groups, p < .05.

Table 6.

Means and standard deviations for BASC-2 variables for psychosocial subtypes using Ward's five-cluster solution

 Normal (n = 49; M [SD]) PED (n = 13; M [SD]) ME/AP (n = 14; M [SD]) ME/D (n = 15; M [SD]) MA (n = 19; M [SD]) 
Hyp 46.70 (6.15)a 79.92 (10.03)b 61.14 (5.80)c 65.20 (7.75)c 49.53 (6.91)a 
Agg 45.53 (5.35)a 74.00 (9.32)b 61.86 (8.59)c 63.27 (11.18)c 48.84 (6.44)a 
Con 45.10 (5.30)a 73.54 (17.66)b 61.43 (7.51)c 53.87 (9.68)c 48.68 (6.86)a 
Anx 42.00 (6.38)a 66.08 (16.87)b 44.00 (8.45)a 64.93 (12.16)b 55.68 (7.38)a 
Dep 47.87 (7.91)a 84.69 (14.69)b 54.00 (4.80)a 70.27 (12.57)b 52.37 (5.06)a 
Som 42.23 (4.38)a 72.69 (16.40)b 48.71 (9.27)a 47.73 (7.90)a 54.53 (6.98)a 
Atyp 46.37 (5.03)a 82.69 (12.33)b 57.14 (11.15)a 65.53 (8.02)c 53.63 (5.95)a 
Wdw 43.23 (5.48)a 66.92 (12.63)b 48.50 (5.88)a 57.13 (7.28)c 50.74 (8.39)a 
Att 51.50 (9.09)a 69.69 (6.55)b 62.71 (7.29)b 62.40 (7.28)b 50.26 (6.37)a 
 Normal (n = 49; M [SD]) PED (n = 13; M [SD]) ME/AP (n = 14; M [SD]) ME/D (n = 15; M [SD]) MA (n = 19; M [SD]) 
Hyp 46.70 (6.15)a 79.92 (10.03)b 61.14 (5.80)c 65.20 (7.75)c 49.53 (6.91)a 
Agg 45.53 (5.35)a 74.00 (9.32)b 61.86 (8.59)c 63.27 (11.18)c 48.84 (6.44)a 
Con 45.10 (5.30)a 73.54 (17.66)b 61.43 (7.51)c 53.87 (9.68)c 48.68 (6.86)a 
Anx 42.00 (6.38)a 66.08 (16.87)b 44.00 (8.45)a 64.93 (12.16)b 55.68 (7.38)a 
Dep 47.87 (7.91)a 84.69 (14.69)b 54.00 (4.80)a 70.27 (12.57)b 52.37 (5.06)a 
Som 42.23 (4.38)a 72.69 (16.40)b 48.71 (9.27)a 47.73 (7.90)a 54.53 (6.98)a 
Atyp 46.37 (5.03)a 82.69 (12.33)b 57.14 (11.15)a 65.53 (8.02)c 53.63 (5.95)a 
Wdw 43.23 (5.48)a 66.92 (12.63)b 48.50 (5.88)a 57.13 (7.28)c 50.74 (8.39)a 
Att 51.50 (9.09)a 69.69 (6.55)b 62.71 (7.29)b 62.40 (7.28)b 50.26 (6.37)a 

Notes: Hyp = Hyperactivity; Agg = Aggression; Con = Conduct Problems; Anx = Anxiety; Dep = Depression; Som = Somatization; Atyp = Atypicality; Wdw = Withdrawal; Att = Attention Problems; PED = Pervasive Emotional Difficulties; ME/AP = Mild Externalizing/Attention Problems; ME/D = Mild Externalizing/Depression; MA = Mild Anxiety. Column differences >10.23 for PED, 5.29 for ME/AP, 5.68 for MA, 8.07 for ME/D, and 2.87 for Normal indicate significant subscale differences by group. Row means with different superscripts indicate significant differences between groups, p < .05.

Relationship to Known Subtypes

Subtypes derived in the current study were statistically compared with the BASC-2 clinical profiles reported in the manual (Reynolds & Kamphaus, 2004). One-sample t-tests were used to accomplish this. In addition, the current sample had to be divided into two groups based on age, since, in the manual, means are provided separately for scores from the child and adolescent forms. Further, the BASC-2 manual divides the clinical sample into several subgroups, including Attention-Deficit/Hyperactivity Disorder, Autism/Asperger's/Pervasive Developmental Disorder, Bipolar Disorder, Depression Disorders, Emotional/Behavioral Disturbance, Hearing Impairment, Learning Disability, Mental Retardation/Developmental Delay, Motor Impairment, and Speech/Language Disorder. The clinical subgroups that most directly relate to this study include Bipolar Disorder, Depression Disorders, and Emotional/Behavioral Disturbance; therefore, statistical comparisons were completed using means reported for these BASC-2 clinical subgroups. Means and results of t-test analyses for the child portion of the sample are reported in Table 7, whereas results for the adolescent portion of the sample are reported in Table 8.

Table 7.

Subtype comparisons to BASC-2 PRS clinical subgroups, child sample

 Current sample
 
 M [SDM [SDt 
Hyp 
 BD 73.30 (13.40) 57.16 (12.85) 7.11** 
 DD 67.50 (9.20) 57.16 (12.85) 4.55** 
 EBD 65.50 (13.10) 57.16 (12.85) 3.67** 
Agg 
 BD 68.00 (9.90) 56.97 (15.27) 4.09** 
 DD 74.00 (1.40) 56.97 (15.27) 6.31** 
 EBD 64.80 (11.70) 56.97 (15.27) 2.90** 
Con 
 BD 71.10 (14.50) 54.78 (12.91) 7.15** 
 DD 78.50 (7.80) 54.78 (12.91) 10.39** 
 EBD 64.60 (13.90) 54.78 (12.91) 4.30** 
Anx 
 BD 57.00 (10.90) 51.78 (13.90) 2.12* 
 DD 57.50 (20.50) 51.78 (13.90) 2.33* 
 EBD 54.40 (13.30) 51.78 (13.90) 1.07 
Dep 
 BD 67.80 (12.00) 57.19 (14.48) 4.15** 
 DD 79.50 (21.90) 57.19 (14.48) 8.72** 
 EBD 65.40 (16.10) 57.19 (14.48) 3.21** 
Som 
 BD 59.60 (15.10) 50.00 (12.55) 4.33** 
 DD 57.00 (5.70) 50.00 (12.55) 3.15** 
 EBD 55.50 (14.60) 50.00 (12.55) 2.48* 
Atyp 
 BD 72.10 (9.70) 61.56 (16.68) 3.57** 
 DD 71.50 (4.90) 61.56 (16.68) 3.37** 
 EBD 66.60 (15.30) 61.56 (16.68) 1.71 
Wdw 
 BD 57.40 (9.50) 51.94 (11.59) 2.67* 
 DD 56.50 (20.50) 51.94 (11.59) 2.23* 
 EBD 60.90 (13.70) 51.94 (11.59) 4.38** 
Att 
 BD 67.00 (6.70) 59.91 (10.46) 3.84** 
 DD 68.00 (1.40) 59.91 (10.46) 4.38** 
 EBD 63.90 (7.60) 59.91 (10.46) 2.16* 
 Current sample
 
 M [SDM [SDt 
Hyp 
 BD 73.30 (13.40) 57.16 (12.85) 7.11** 
 DD 67.50 (9.20) 57.16 (12.85) 4.55** 
 EBD 65.50 (13.10) 57.16 (12.85) 3.67** 
Agg 
 BD 68.00 (9.90) 56.97 (15.27) 4.09** 
 DD 74.00 (1.40) 56.97 (15.27) 6.31** 
 EBD 64.80 (11.70) 56.97 (15.27) 2.90** 
Con 
 BD 71.10 (14.50) 54.78 (12.91) 7.15** 
 DD 78.50 (7.80) 54.78 (12.91) 10.39** 
 EBD 64.60 (13.90) 54.78 (12.91) 4.30** 
Anx 
 BD 57.00 (10.90) 51.78 (13.90) 2.12* 
 DD 57.50 (20.50) 51.78 (13.90) 2.33* 
 EBD 54.40 (13.30) 51.78 (13.90) 1.07 
Dep 
 BD 67.80 (12.00) 57.19 (14.48) 4.15** 
 DD 79.50 (21.90) 57.19 (14.48) 8.72** 
 EBD 65.40 (16.10) 57.19 (14.48) 3.21** 
Som 
 BD 59.60 (15.10) 50.00 (12.55) 4.33** 
 DD 57.00 (5.70) 50.00 (12.55) 3.15** 
 EBD 55.50 (14.60) 50.00 (12.55) 2.48* 
Atyp 
 BD 72.10 (9.70) 61.56 (16.68) 3.57** 
 DD 71.50 (4.90) 61.56 (16.68) 3.37** 
 EBD 66.60 (15.30) 61.56 (16.68) 1.71 
Wdw 
 BD 57.40 (9.50) 51.94 (11.59) 2.67* 
 DD 56.50 (20.50) 51.94 (11.59) 2.23* 
 EBD 60.90 (13.70) 51.94 (11.59) 4.38** 
Att 
 BD 67.00 (6.70) 59.91 (10.46) 3.84** 
 DD 68.00 (1.40) 59.91 (10.46) 4.38** 
 EBD 63.90 (7.60) 59.91 (10.46) 2.16* 

Notes: BD = Bipolar Disorder; DD = Depression Disorders; EBD = Emotional/Behavioral Disturbance; Hyp = Hyperactivity; Agg = Aggression; Con = Conduct Problems; Anx = Anxiety; Dep = Depression; Som = Somatization; Atyp = Atypicality; Wdw = Withdrawal; Att = Attention Problems.

*p < .05.

**p < .01.

Table 8.

Subtype comparisons to BASC-2 PRS clinical subgroups, adolescent sample

 Current sample
 
 M [SDM [SDt 
Hyp 
 BD 73.40 (9.80) 57.39 (14.16) 8.68** 
 DD 62.00 (17.90) 57.39 (14.16) 2.50* 
 EBD 69.00 (14.70) 57.39 (14.16) 6.30** 
Agg 
 BD 68.90 (7.80) 55.05 (11.57) 9.20** 
 DD 61.80 (16.70) 55.05 (11.57) 4.48** 
 EBD 65.30 (13.90) 55.05 (11.57) 6.81** 
Con 
 BD 68.10 (11.40) 53.37 (13.70) 8.26** 
 DD 68.50 (21.70) 53.37 (13.70) 8.48** 
 EBD 67.70 (15.00) 53.37 (13.70) 8.03** 
Anx 
 BD 59.40 (10.10) 52.71 (14.26) 3.60** 
 DD 54.20 (16.00) 52.71 (14.26) 0.80 
 EBD 56.40 (12.60) 52.71 (14.26) 1.99 
Dep 
 BD 68.80 (7.70) 59.53 (16.68) 4.27** 
 DD 76.10 (26.60) 59.53 (16.68) 7.63** 
 EBD 66.40 (15.40) 59.53 (16.68) 3.17** 
Som 
 BD 58.30 (10.00) 51.63 (13.52) 3.79** 
 DD 55.80 (11.30) 51.63 (13.52) 2.37* 
 EBD 56.70 (14.4) 51.63 (13.52) 2.88* 
Atyp 
 BD 55.90 (10.00) 55.90 (12.89) 0.00 
 DD 61.40 (18.00) 55.90 (12.89) 3.28** 
 EBD 62.90 (15.40) 55.90 (12.89) 4.17** 
Wdw 
 BD 62.50 (4.70) 50.93 (10.84) 8.20** 
 DD 59.30 (11.70) 50.93 (10.84) 5.93** 
 EBD 60.00 (12.00) 50.93 (10.84) 6.43** 
Att 
 BD 62.90 (5.10) 55.98 (10.26) 5.18** 
 DD 53.30 (12.00) 55.98 (10.26) 2.01 
 EBD 63.70 (9.10) 55.98 (10.26) 5.78** 
 Current sample
 
 M [SDM [SDt 
Hyp 
 BD 73.40 (9.80) 57.39 (14.16) 8.68** 
 DD 62.00 (17.90) 57.39 (14.16) 2.50* 
 EBD 69.00 (14.70) 57.39 (14.16) 6.30** 
Agg 
 BD 68.90 (7.80) 55.05 (11.57) 9.20** 
 DD 61.80 (16.70) 55.05 (11.57) 4.48** 
 EBD 65.30 (13.90) 55.05 (11.57) 6.81** 
Con 
 BD 68.10 (11.40) 53.37 (13.70) 8.26** 
 DD 68.50 (21.70) 53.37 (13.70) 8.48** 
 EBD 67.70 (15.00) 53.37 (13.70) 8.03** 
Anx 
 BD 59.40 (10.10) 52.71 (14.26) 3.60** 
 DD 54.20 (16.00) 52.71 (14.26) 0.80 
 EBD 56.40 (12.60) 52.71 (14.26) 1.99 
Dep 
 BD 68.80 (7.70) 59.53 (16.68) 4.27** 
 DD 76.10 (26.60) 59.53 (16.68) 7.63** 
 EBD 66.40 (15.40) 59.53 (16.68) 3.17** 
Som 
 BD 58.30 (10.00) 51.63 (13.52) 3.79** 
 DD 55.80 (11.30) 51.63 (13.52) 2.37* 
 EBD 56.70 (14.4) 51.63 (13.52) 2.88* 
Atyp 
 BD 55.90 (10.00) 55.90 (12.89) 0.00 
 DD 61.40 (18.00) 55.90 (12.89) 3.28** 
 EBD 62.90 (15.40) 55.90 (12.89) 4.17** 
Wdw 
 BD 62.50 (4.70) 50.93 (10.84) 8.20** 
 DD 59.30 (11.70) 50.93 (10.84) 5.93** 
 EBD 60.00 (12.00) 50.93 (10.84) 6.43** 
Att 
 BD 62.90 (5.10) 55.98 (10.26) 5.18** 
 DD 53.30 (12.00) 55.98 (10.26) 2.01 
 EBD 63.70 (9.10) 55.98 (10.26) 5.78** 

Notes: BD = Bipolar Disorder; DD = Depression Disorders; EBD = Emotional/Behavioral Disturbance; Hyp = Hyperactivity; Agg = Aggression; Con = Conduct Problems; Anx = Anxiety; Dep = Depression; Som = Somatization; Atyp = Atypicality; Wdw = Withdrawal; Att = Attention Problems.

*p < .05.

**p < .01.

Tables 7 and 8 demonstrate that the BASC-2 clinical subgroup profile means are largely significantly different from the profile means derived from the current TBI sample. Further, it is interesting to note that, not only are the means significantly different from each other, but, in some cases, the variability in scores is quite different as well. This seems to occur most often when comparing the current sample to the BASC-2 child sample's Depressive Disorders group. For example, for the Aggression subscale, the current sample standard deviation is 15.27, whereas the BASC-2 sample standard deviation is 1.40. Therefore, there is quite a bit more variability in scores with the current sample than the BASC-2 clinical sample. One exception is for the Withdrawal scale, where the reverse is true. For the adolescent portion of the sample, there is less difference in variability. The only striking differences are for the Depression scale, where variability is quite different between the current sample and both the Bipolar Disorders and Depressive Disorders groups. Interestingly, the current sample has greater variability when compared with the Bipolar Disorders group but less variability when compared with the Depressive Disorders group.

Exceptions to these differences between the current sample and the BASC-2 clinical samples include similarities in the child portion of the sample with the Emotional/Behavioral Disturbance group in terms of level of anxiety and atypicality. The adolescent portion of the current sample was similar to the BASC-2 Depression Disorders group in terms of level of anxiety and attention and the Emotional/Behavioral Disturbance group in terms of level of anxiety. In addition, the means for the adolescent portion of the current sample and the BASC-2 Bipolar Disorders sample were exactly the same in terms of atypicality.

Relationship of Other Variables with Psychosocial Functioning

Cross-tabulations and one-way ANOVAs were calculated to determine the relationship of various variables with psychosocial functioning. Specifically, variables including gender, ethnicity, age at assessment, age at injury, GCS score, and time elapsed since injury were investigated. It should be noted that the relationship of these variables was compared with subtypes generated within the two-cluster solution.

Cross-tabulations of gender and ethnic classification with the psychosocial subtype were constructed. Results of the Pearson chi-squared analysis demonstrated no significant differences between either gender or ethnicity and the psychosocial subtype to which subjects were assigned—for gender, χ2(1, N = 91) = 0.01, p = .93; for ethnicity, χ2(2, N = 88) = 0.31, p = .86.

One-way ANOVAs with GCS score, age at assessment, and age at injury as independent variables and the psychosocial subtype as the dependent variable were calculated. Once again, this analysis revealed no differences between these variables and the psychosocial subtype to which subjects were assigned—for GCS, F(1, 45) = 0.17, p = .69; for age at assessment, F(1, 89) = 1.24, p = .27; for age at injury, F(1, 88) = 0.28, p = .60.

The only significant difference in psychosocial subtype classification was observed with time elapsed since injury—χ2(1, N = 90) = 4.83, p < .05. Specifically, the Pearson chi-squared revealed a significant difference only for the PED subtype. An examination of the relative frequencies suggested that a disproportionate number of children injured within the last 12 months were assigned to this subtype. More specifically, within the first year following injury, a greater percentage of children were classified within the PED subtype than within the Normal subtype, whereas after the first year, a greater percentage was classified as normal than as having emotional difficulties.

Summary

The results presented in this section are summarized below in terms of the research questions outlined previously.

  • How do these derived BASC-2 profiles compared with known CBCL and PIC-R profiles? Mixed results were found when profiles derived from the CBCL (Hayman-Abello et al., 2003) and PIC-R (Butler et al., 1997) were compared with the current BASC-2 profiles. The most striking similarity is the presence of a Normal subtype containing a majority of the participants, whereas the most striking difference seems to be the number of subtypes derived. Although the current study found two reliable profiles, the PIC-R study found seven, and the CBCL study found four. When these PIC-R and CBCL subtypes were compared with the current study's subtypes generated in three-, four-, and five-cluster solutions, additional similarities and differences between subtypes were noticed. These will be described in more detail in the following section.

  • How do these derived BASC-2 profiles compared with known BASC-2 clinical profiles? The profiles derived with a TBI sample were largely significantly different from the BASC-2 clinical subgroup profiles.

  • What is the relationship between psychosocial outcome and variables such as gender, time elapsed since injury, age at injury, age at testing, and GCS? No significant differences were found between subtypes based on gender, ethnicity, age at assessment, age at injury, or GCS. The only significant difference found was for time elapsed since injury for those participants classified within the PED subtype.

Discussion

The primary goal of this investigation was to develop a profile of psychosocial functioning in children who had sustained a TBI using cluster analysis of the BASC-2 PRS clinical scales. The BASC-2 was selected as the measure of choice because it is a widely used instrument in clinical and school settings; however, very little research exists in the current literature regarding its use with TBI children or adolescents. In addition, two other studies investigating the psychosocial functioning of children with TBI have been completed (Butler et al., 1997; Hayman-Abello et al., 2003) and parallel research seemed necessary to assess the correspondence of their research using the PIC-R and CBCL with research using the BASC-2.

Characteristics of the Typology

Two robust clusters or subtypes, labeled as Normal and PED, were derived. This allowed for descriptions of these children in terms of heterogeneous psychosocial functioning. The pattern of scale elevations indicated that children assigned to the Normal group had significantly lower overall profile elevations. No clinically significant elevations were observed in the profile for the Normal subtype, resulting in a flat profile across all BASC-2 clinical scales and suggesting that these children were relatively free of pathology. Thus, parents were not expressing any concerns about their child's psychosocial functioning. This subtype was largest, accounting for 86% of the sample.

In contrast, the PED group's mean profile contained elevations in the at-risk or clinically significant range on all BASC-2 clinical scales. These children are perceived by parents as having significant externalizing behaviors, symptoms of depression, and behaviors viewed as atypical or odd. In addition, they appear to be showing either subclinical or emerging symptoms of anxiety, social withdrawal, and attention problems. As a result, these children may not interact or integrate well with other children in the school setting and may require specialized assistance regarding management of acting out behaviors and depressive symptoms.

Although these two subtypes accurately describe the data in a statistical manner, they are quite broad and provide little in the way of increased understanding of the psychosocial functioning of children who have sustained a TBI. It is simply not enough for a neuropsychologist to describe a child with a brain injury as either emotionally disturbed or not. Further, treatment planning and behavioral management of these school-age children must be individualized to a greater degree in order to provide the most effective learning environment possible. Therefore, although the lack of statistical significance suggests that their value should be examined cautiously, the clinical significance of the subtypes generated using further exploratory cluster analytic methods appears important.

Four further differentiated subtypes emerged in the exploratory portion of the cluster analysis process. Children categorized within the Mild Externalizing/Depression subtype were perceived by parents as exhibiting a range of mild externalizing behaviors and behaviors viewed as odd or strange. In addition, however, these children are experiencing significant feelings of depression, which they are likely expressing via their externalizing behaviors. Thus, these children may be viewed by others as having mild behavior problems when, in reality, these behaviors are masking significant symptoms of depression, which deserve special attention and treatment planning. Not surprisingly, this cluster of symptoms also appears to affect their ability to focus and concentrate, resulting in mild attention problems as well.

Children grouped within the Mild Externalizing/Attention Problems subtype were characterized by mild elevations on a range of externalizing behaviors and attention problems. These children may, on the surface, appear quite similar to the children in the Mild Externalizing/Depression subtype. However, in contrast, their externalizing behaviors are not masking underlying internalizing symptoms as described above. Thus, their behavior problems must be approached in a different manner with stronger behavior management plans.

The Mild Depression subtype is, again, similar to the Mild Externalizing/Depression subtype, given its focus on mild hyperactivity and atypicality with underlying, clinically significant, symptoms of depression. Children falling within this subtype differ, in that their externalizing behaviors are not as aggressive and they do not possess the attention problems observed in other subtypes.

Finally, the Mild Anxiety subtype is characterized by scores contained within the average range, although it is important to note that these average-range scores were consistently higher than those observed in the Normal subtype. Higher (although still within the average range) scores were observed on scales measuring anxiety and somatization. Even though the elevations of scores in this profile were <70 T, the shape is still important because the profile is representative of the “average” of all children in the subtype (Butler et al., 1997). Thus, the delineation of this separate subtype is important because it likely alerts neuropsychologists to a group of children who may appear to be functioning normally. However, disregard for their worries, fears, and health concerns may contribute to greater psychosocial problems in the future.

Validity of the Typology

The current study supports previous research (e.g., Bloom et al., 2001; Max et al., 2004), in that there was no relationship between a child's level of injury severity and that child's classification within a particular subtype, although this result needs to be interpreted cautiously since GCS scores were not available for almost half of the sample. Another significant finding generated by this study was that, within the first year following injury, a greater percentage of children were classified within the PED subtype than within the Normal subtype, whereas after the first year, a greater percentage was classified as normal than as having emotional difficulties. Thus, it appears that either there was some improvement in psychosocial functioning with recovery or, alternatively, parents became more accustomed to their children's behavior and were more equipped and better able to manage more difficult behaviors.

Descriptions of the range of profiles developed in this study beyond the two most reliable subtypes revealed that, when children did experience psychosocial difficulties, they were most often externalizing problems. When internalizing problems did occur, they were either quite mild in nature or were comorbid with other externalizing behaviors. Other researchers have found similar results. For example, Hawley (2003) found that parents were not overly concerned with symptoms of anxiety and depression but were observing more behavioral and temper problems in their children. Ganesalingam and colleagues (2006) reported more externalizing behaviors, including defiance, destructiveness, impulsivity, and poor emotional awareness, while generalized emotional difficulties emerged as the most problematic area in a study conducted by Doherty and McCusker (2005). An explanation for results such as these might be that externalizing behaviors are more problematic for parents, more observable, and more likely to generate parental concern than internalizing behaviors. Symptoms related to internalizing problems are often less tangible and, as a result, less recognizable and easier to ignore or miss (Bidaut-Russell et al., 1995).

Comparison to Previous Research

Relationship to known subtypes

Visual comparisons between the current BASC-2 subtypes and the PIC-R and CBCL subtypes previously described reveal a major similarity in that most participants were categorized within a Normal subtype, indicating no psychosocial concerns from a parental point of view. Another similarity was observed between the PIC-R and BASC-2 Mild Anxiety subtypes in that a higher score was observed on anxiety subscales while other scores remained within the average range. One difference, however, between the PIC-R and BASC-2 Mild Anxiety subtypes was that the PIC-R subtype was characterized by a significantly elevated score on the Anxiety subscale, whereas the BASC-2 subtype's anxiety score remained within the average range but was elevated beyond the other scores in the profile. One possible explanation for this difference between scores on the two scales is the manner in which each scale measures anxiety. The BASC-2 anxiety subscale measures characteristics such as excessive worry, fears and phobias, self-deprecation, and nervousness (Reynolds & Kamphaus, 2004). The PIC-R anxiety subscale, on the other hand, measures moodiness, fearfulness, worry, common fears, poor self-concept, insecurity, and over-sensitivity (Wirt et al., 1990). Thus, the two scales seem to be measuring slightly different aspects of anxiety.

Regarding internalized psychopathology, the PIC-R identified a subtype characterized by these symptoms. On the BASC-2, internalized symptoms were most often observed in conjunction with externalized symptoms (i.e., Mild Externalizing/Depression subtype). In the four-cluster solution, a subtype characterized by mild depression was delineated; however, this subtype was not retained on any other cluster solution. Interestingly, the CBCL solutions also did not identify a subtype dedicated to internalized psychopathology. The most similar is likely the Withdrawn-Somatic subtype; however, this subtype may be more similar to the PIC-R Social Isolation subtype than the Internalized Psychopathology subtype.

Regarding externalized behaviors, both the PIC-R and CBCL studies identified subtypes with elevated scores on subscales measuring these behaviors (i.e., Antisocial subtype on the PIC-R and Delinquent subtype on the CBCL). On the BASC-2, externalizing behaviors were often observed alongside either internalized behaviors (i.e., Mild Externalizing/Depression subtype) or attention problems (i.e., Mild Externalizing/Attention Problems subtype) or both (i.e., PED subtype). Although the CBCL study identified a specific subtype characterized by primary attention problems, the PIC-R did not. As already inferred, attention problems on the BASC-2 were most often observed alongside externalizing problems. One significant difference between the BASC-2 profiles and the PIC-R and CBCL profiles is the lack of a BASC-2 subtype containing somatic concerns. Both the PIC-R and CBCL studies identified specific subtypes with elevations on their respective somatic subtypes, whereas, on the BASC-2, scores on the Somatization subscale were never elevated beyond the average range except on the PED subtype. A hypothesis regarding this observation, which would require further investigation in order to confirm or disconfirm, is that the PIC-R and CBCL may be more sensitive than the BASC-2 to the influence of physical problems associated with TBI.

Relationship with a clinical sample

Finally, the current typology was compared with the profiles of various clinical samples presented in the BASC-2 manual. Specifically, the current typology was compared with the BASC-2 profiles for the Bipolar Disorder, Depressive Disorder, and Emotional/Behavioral Disturbance groups. The child portion of the sample was significantly different from all BASC-2 clinical groups except for the Emotional/Behavioral Disorder group on the Anxiety and Atypicality scales. The adolescent sample was similar to the Bipolar Disorder group in terms of level of atypicality, the Depressive Disorders group in terms of level of anxiety and attention problems, and the Emotional/Behavioral Disorder group in terms of level of anxiety.

Thus, it is apparent that, for the most part, the children with TBI were dissimilar to groups of children diagnosed with psychiatric and behavioral disorders. This is likely the case because, when taken together, all means for the TBI profile fell within the average range. In other words, when the TBI sample is viewed as a whole, it looks more normal than abnormal. Thus, it is not surprising that, when the profile of average-range scores from the TBI sample was compared with clinical profiles containing many clinically significant scores, significant differences were observed.

Limitations and suggestions for future research

The sample in the current study consisted of 91 participants. This sample size is just large enough to make cluster analysis a viable statistical option but may have been too small to conclude that certain relationships between variables actually do exist. In other words, type II errors could have resulted, thus obscuring true relationships between variables. Further, because there were so few children classified with PED, cell sizes in some of the cross-tabulations were small. In addition, GCS scores were available for only 47 participants, thus limiting the conclusions that could be drawn regarding the effects of injury severity on psychosocial functioning. Given the variability in the effects of injury severity on psychosocial functioning reported in this and other studies of children with TBI, this relationship deserves and requires clarification. Future studies could also remedy problems with sample bias that may have been present in this study. The source of participants for this investigation was an urban children's hospital. Therefore, only children treated in this one hospital in this one city were included. Using a study design where neuropsychological test results from children in multiple hospitals or centers are used is needed for generalization of these results.

Information regarding premorbid functioning was not available in the data set used for these analyses; therefore, children in the current study were not excluded on the basis of premorbid psychosocial problems as was the case in the Butler and colleagues (1997) study. As a result, conclusions that the psychosocial sequelae of pediatric TBI are causally related to the brain injuries sustained during head trauma cannot be made in this study. Given the research reviewed previously, the importance of examining variables associated with pre-injury child and family functioning is important. Thus, future subtyping research should accommodate for this type of inquiry in order that the sequelae of TBI beyond the influence of premorbid psychosocial status can be investigated.

It must also be emphasized that the BASC-2 PRS is a parent-response questionnaire. Thus, the PRS provides only the parent's or primary caregiver's perception of the child's psychosocial functioning, and perception alone cannot be equated to actual behavior or psychiatric disturbance. Parental perceptions also often differ drastically from a child's or adolescent's self-reported behaviors or a teacher's perceptions of behaviors (e.g., DiBartolo & Grills, 2006).

In addition, the existence of post-injury psychosocial problems may be underestimated for several reasons. First, in the initial months following injury, parents may, first and foremost, be grateful that their child is alive and thus overlook the severity of any behavioral problems that may exist. Second, due to their neurological and physical disabilities, these children may not be mobile or cognizant enough to engage in some of the more serious psychosocial problems. Thus, although the BASC-2 is a well-normed, reliable instrument that allows researchers to formulate hypotheses regarding patterns of children's behavior, the use of semi-structured clinical interviews can also be incorporated to provide a more comprehensive assessment of post-injury psychosocial functioning. It is conceivable that behavior rating scales such as the BASC-2, PIC-R, and CBCL are relatively insensitive in identifying psychosocial problems in children with TBI given their rather general nature. Alternatively, the sample used in this study may have been too mildly affected with regard to psychosocial sequelae, and thus cluster analytic methods were not effective in describing post-injury psychosocial problems.

Future research is also needed to answer the question regarding whether or not there is a set of characteristics that differentiates children with TBI who develop adaptive psychosocial functioning from those who develop maladaptive psychosocial functioning. Another question that requires investigation is whether there exists a set of characteristics that differentiates children with TBI who develop particular patterns of psychosocial functioning. In other words, are there relationships between psychosocial disturbance and various cognitive abilities, neuropsychological skills, or locations of lesions? Such multidimensional classifications of a child's psychopathology may better reflect overall post-injury clinical presentation and allow for the identification of characteristics that may be potential buffers or exacerbaters of dysfunction (Saunders, Hall, Casey, & Strang, 2000). Further, a strong link has been found between school performance and behavioral problems (Hawley, 2003). With a multidimensional classification of psychopathology, prognosis can be more accurately determined, and school psychologists will be better able to formulate effective treatment protocols. However, this also implies that early recognition and understanding via assessment is necessary. According to parents, few children suffering TBIs are assessed and receive clinical input from a psychologist following the injury (Hawley). Thus, if no assessment is conducted, it is more likely that teachers will not be aware of the TBI, particularly if the injury occurred more than a year ago and, thus, will not associate behavioral problems and poor school performance with the TBI (Hawley).

Changes in psychosocial functioning over time can only be discerned through further investigation. One major contribution of this study is confirmation of the conclusion that the psychosocial sequelae of pediatric TBI cover a variety of patterns. Although most children appear to fall well within the normal range of functioning, potential psychosocial problems range from rather severe, pervasive disturbances characterized by both internalizing and externalizing behaviors, to mild disturbances in relatively circumscribed areas. In addition, the present discussion of psychosocial profiles following TBI and the comparison to established developmental profiles for other psychiatric groups potentially enable the practitioner to differentiate psychosocial disturbance relative to TBI from psychiatric disturbance that is developmentally acquired. Such a categorization as is presented here creates a template against which individuals can be compared and leads to a better understanding of prognosis and treatment associated with that particular pattern of functioning. Continued replication of these results as well as more complex and detailed investigations will be imperative.

Conflict of Interest

None declared.

References

Achenbach
T. M.
Manual for the Child Behavior Checklist/4-18 and 1991 profile
 , 
1991
Burlington, VT
University of Vermont, Department of Psychiatry
Bidaut-Russell
M.
Reich
W.
Cottler
L. B.
Robins
L. N.
Compton
W. M.
Mattison
R. E.
The diagnostic interview schedule for children (PC-DISC v.3.0): Parents and adolescents suggest reasons for expecting discrepant answers
Journal of Abnormal Child Psychology
 , 
1995
, vol. 
23
 (pg. 
641
-
659
)
Bloom
D. R.
Levin
H. S.
Ewing-Cobbs
L.
Saunders
A. E.
Song
J.
Fletcher
J. M.
, et al.  . 
Lifetime and novel psychiatric disorders after pediatric traumatic brain injury
Journal of the American Academy of Child and Adolescent Psychiatry
 , 
2001
, vol. 
40
 (pg. 
572
-
579
)
Borgen
F. H.
Barnett
D. C.
Applying cluster analysis in counseling psychology research
Journal of Counseling Psychology
 , 
1987
, vol. 
34
 (pg. 
456
-
468
)
Butler
K.
Rourke
B. P.
Fuerst
D. R.
Fisk
J. L.
A typology of psychosocial functioning in pediatric closed-head injury
Child Neuropsychology
 , 
1997
, vol. 
3
 (pg. 
98
-
133
)
DiBartolo
P. M.
Grills
A. E.
Who is best at predicting children's anxiety in response to a social evaluative task? A comparison of child, parent, and teacher reports
Anxiety Disorders
 , 
2006
, vol. 
20
 (pg. 
630
-
645
)
Doherty
N. N.
McCusker
C. G.
Assessment of functional and psychosocial outcomes in children four years post head injury
Educational and Child Psychology
 , 
2005
, vol. 
22
 (pg. 
29
-
38
)
Fuerst
D. R.
Fisk
J. L.
Rourke
B. P.
Psychosocial functioning of learning-disabled children: Replicability of statistically derived subtypes
Journal of Consulting and Clinical Psychology
 , 
1989
, vol. 
57
 (pg. 
275
-
280
)
Ganesalingam
K.
Sanson
A.
Anderson
V.
Yeates
K. O.
Self-regulation and social and behavioral functioning following childhood traumatic brain injury
Journal of the International Neuropsychological Society
 , 
2006
, vol. 
12
 (pg. 
609
-
621
)
Ganesalingam
K.
Sanson
A.
Anderson
V.
Yeates
K. O.
Self-regulation as a mediator of the effects of childhood traumatic brain injury on social and behavioral functioning
Journal of the International Neuropsychological Society
 , 
2007
, vol. 
13
 (pg. 
298
-
311
)
Grados
M. A.
Vasa
R. A.
Riddle
M. A.
Slomine
B. S.
Salorio
C.
Christensen
J.
, et al.  . 
New onset obsessive-compulsive symptoms in children and adolescents with severe traumatic brain injury
Depression and Anxiety
 , 
2008
, vol. 
25
 (pg. 
398
-
407
)
Hanten
G.
Wilde
E. A.
Menefee
D. S.
Li
X.
Lane
S.
Vasquez
C.
, et al.  . 
Correlates of social problem solving during the first year after traumatic brain injury in children
Neuropsychology
 , 
2008
, vol. 
22
 (pg. 
357
-
370
)
Hawley
C. A.
Reported problems and their resolution following mild, moderate, and severe traumatic brain injury amongst children and adolescents in the UK
Brain Injury
 , 
2003
, vol. 
17
 (pg. 
105
-
129
)
Hayman-Abello
S. E.
Rourke
B. P.
Fuerst
D. R.
Psychosocial status after pediatric traumatic brain injury: A subtype analysis using the Child Behavior Checklist
Journal of the International Neuropsychological Society
 , 
2003
, vol. 
9
 (pg. 
887
-
898
)
Luis
C. A.
Mittenberg
W.
Mood and anxiety disorders following pediatric traumatic brain injury: A prospective study
Journal of Clinical and Experimental Neuropsychology
 , 
2002
, vol. 
24
 (pg. 
270
-
279
)
Max
J. E.
Koele
S. L.
Castillo
C. C.
Lindgren
S. D.
Arndt
S.
Bokura
H.
, et al.  . 
Personality change disorder in children and adolescents following traumatic brain injury
Journal of the International Neuropsychological Society
 , 
2000
, vol. 
6
 (pg. 
279
-
289
)
Max
J. E.
Lansing
A. E.
Koele
S. L.
Castillo
C. S.
Bokura
H.
Schachar
R.
, et al.  . 
Attention deficit hyperactivity disorder in children and adolescents following traumatic brain injury
Developmental Neuropsychology
 , 
2004
, vol. 
25
 (pg. 
159
-
177
)
Reynolds
C. R.
Kamphaus
R. W.
Behavior Assessment System for Children
 , 
2004
2nd ed.
Circle Pines, MN
American Guidance Service
Saunders
C. D.
Hall
E. J.
Casey
J. E.
Strang
J. D.
Subtypes of psychopathology in children referred for neuropsychological assessment
Child Neuropsychology
 , 
2000
, vol. 
6
 (pg. 
129
-
143
)
Thurman
D. J.
Alverson
C.
Dunn
K. A.
Guerrero
J.
Sniezek
J. E.
Traumatic brain injury in the United States: A public health perspective
Journal of Head Trauma Rehabilitation
 , 
1999
, vol. 
14
 (pg. 
602
-
615
)
Ward
J. H.
Hierarchical grouping to optimize an objective function
Journal of the American Statistical Association
 , 
1963
, vol. 
58
 (pg. 
236
-
244
)
Wirt
R. D.
Lachar
D.
Klinedinst
J. K.
Seat
P. D.
Multidimensional description of child personality: A manual for the Personality Inventory for Children
 , 
1990
Los Angeles
Western Psychological Service
Yeates
K. O.
Bigler
E. D.
Dennis
M.
Gerhardt
C. A.
Rubin
K. H.
Stancin
T.
, et al.  . 
Social outcomes in childhood brain disorder: A heuristic integration of social neuroscience and developmental psychology
Psychological Bulletin
 , 
2007
, vol. 
133
 (pg. 
535
-
556
)

Author notes

The research presented in this manuscript is based on the first author's doctoral dissertation. The first author wishes to acknowledge Dr. René Paulson for her assistance with the statistical analyses involved in this study.