Outcomes From Pediatric Mild Traumatic Brain Injury. A dissertation presented to. the faculty of. the College of Arts and Sciences of Ohio University

Transcription

1 The Influence of Premorbid Attention and Behavior Problems on Neurobehavioral Outcomes From Pediatric Mild Traumatic Brain Injury A dissertation presented to the faculty of the College of Arts and Sciences of Ohio University In partial fulfillment of the requirements for the degree Doctor of Philosophy Erin M. Mark August Erin M. Mark. All Rights Reserved.

2 2 This dissertation titled The Influence of Premorbid Attention and Behavior Problems on Neurobehavioral Outcomes From Pediatric Mild Traumatic Brain Injury by ERIN M. MARK has been approved for the Department of Psychology and the College of Arts and Sciences by Julie A. Suhr Professor of Psychology Benjamin M. Ogles Dean, College of Arts and Sciences

3 3 ABSTRACT ERIN M. MARK, Ph.D., August 2011, Psychology The Influence of Premorbid Attention and Behavior Problems on Neurobehavioral Outcomes From Pediatric Mild Traumatic Brain Injury Director of Dissertation: Julie A. Suhr Hierarchical linear modeling was used to examine whether pre-injury attention and behavior problems (i.e. ADHD-spectrum problems) moderated postconcussive symptoms (PCS) and neuropsychological performance following a mild traumatic brain injury (TBI) in children 8-15 years old at the time of injury. Participants were 186 children with a mild TBI (32 children had a complicated mild TBI defined as abnormal MRI findings) and 99 children with an orthopedic-only injury (i.e. limb fracture), consecutively recruited from two pediatric hospitals in Ohio and followed for 12 months post-injury. Parent and child ratings of PCS were assessed within 3 weeks of injury, and again at 1, 3, and 12 months post-injury. Child performance on tests of list learning (CVLT-C Total Immediate Recall), spatial working memory (CANTAB Spatial Working Memory subtest), and planning and problem solving (CANTAB Stockings of Cambridge subtest) was evaluated within 3 weeks of injury, and again at 3 and 12 months postinjury. Retrospective parent ratings of a child's pre-injury ADHD-spectrum problems were assessed within 3 weeks of the injury using both the Child Behavior Checklist (CBCL)-Attention Problems Scale and the Disruptive Behavior Rating Scale (DBRS)- ADHD Index. Two indices of acute injury severity--loss of consciousness and MRI abnormality--were used uniquely as predictors of recovery trajectories in models also containing one of the two measures of pre-injury ADHD-spectrum problems (i.e. CBCL

4 4 or DBRS) and other covariates (e.g., race, SES, age-at-injury, etc.). Premorbid ADHDspectrum problems moderated parent-reported cognitive PCS, with children with high levels of pre-injury ADHD-spectrum problems and a complicated mild TBI showing the most cognitive PCS, across the first year post-injury. Postconcussive symptoms were more likely following mild TBI than orthopedic-only injury, with persistent PCS more likely following complicated mild TBI, than uncomplicated mild TBI. Pre-injury ADHD-spectrum problems independently predicted PCS and neuropsychological performance. Unexpectedly, children with mild TBI performed better than children with orthopedic-only injury on a spatial working memory and planning task. Clinicians should assess for common premorbid problems such as ADHD when working with children recovering from a mild TBI. Approved: Julie A. Suhr Professor of Psychology

5 5 ACKNOWLEDGMENTS There are many individuals that contributed to the completion of this document. I am very grateful for the professional support and guidance from my graduate advisor and dissertation director, Julie Suhr, Ph.D. Dr. Suhr has been an integral part of all aspects of my graduate training and professional development, as well as a responsive and dedicated partner in this writing process. I also wish to express my gratitude to my dissertation co-director, Keith O. Yeates, Ph.D., who gave me the opportunity to work with an excellent data set and provided expert support and supervision with data analysis and the write-up of the Results and Discussion sections. Thank you to the members of my dissertation committee, Christine Gidycz, Ph.D., Kenneth Holroyd, Ph.D., and Karen Montgomery-Reagan, M.D. for your critical review of my work during the proposal and defense stages--this manuscript is better as a result of your insightful suggestions. Finally, thank you to my husband, David, my mother, M. Catherine Dragisity and my mother- and father-in-law, Jean and Douglas Mark. The completion of this process would not have been possible without your support and sacrifice. Each of you made countless sacrifices so that I could complete this document and my graduate training. I am forever grateful. I love you all very much.

6 For David, Matthew, and Isabelle 6

7 7 TABLE OF CONTENTS Page Abstract... 3 Acknowledgments... 5 List of Tables... 9 List of Figures Introduction Methods Participants Procedure Measures Magnetic Resonance Imaging Pre-injury Attention and Behavior Problems Postconcussive Symptoms Neuropsychological Tests Data Analysis Hierarchical Linear Modeling (HLM) Results Preliminary Analyses Level 1 Unconditional Models Postconcussive Symptoms Neuropsychological Test Performance Level 2 Conditional Models Level 2 Models Predicting Child-rated Postconcussive Symptoms Level 2 Models Predicting Parent-reported Postconcussive symptoms Level 2 Models Predicting Neuropsychological Test Performance Discussion Evidence for Moderation of Parent-rated Cognitive Symptoms Main Effects of Pre-injury ADHD-spectrum Problems and Mild TBI Limitations... 54

8 8 References Appendix A: Measures and Psychometric Information Ratings of Pre-injury Child Behavior Child Behavior Checklist (CBCL) Disruptive Behavior Rating Scale (DBRS) Ratings of Post-injury Child Behavior Post-Concussive Symptom Interview (PCS-I) Health and Behavior Inventory (HBI) Neuropsychological Measures of Post-injury Functioning California Verbal Learning Test-Children s Version (CVLT-C) Cambridge Neuropsychological Test Automated Battery (CANTAB) Appendix B: Statistical Analyses Growth Curve Analysis and HLM Hypotheses Hypothesis Hypothesis

9 9 LIST OF TABLES Page Table 1: Participant Characteristics by Injury Group...65 Table 2: Schedule of Administration of Measures...67 Table 3: Correlations Among Measures of Pre-injury Functioning...68 Table 4: Level 1 Unconditional Models for Postconcussive Outcomes...69 Table 5: Level 1 Unconditional Models for Neuropsychological Tests...72 Table 6: Level 2 Models of Child-reported HBI-Cognitive (LOC/CBCL)...74 Table 7: Level 2 Models of Child-reported HBI-Cognitive (LOC/DBRS)...75 Table 8: Level 2 Models of Child-reported HBI-Cognitive (MRI/CBCL)...76 Table 9: Level 2 Models of Child-reported HBI-Cognitive (MRI/DBRS)...77 Table 10: Level 2 Models of Parent-reported HBI-Cognitive (LOC/CBCL)...78 Table 11: Level 2 Models of Parent-reported HBI-Cognitive (LOC/DBRS)...80 Table 12: Level 2 Models of Parent-reported HBI-Cognitive (MRI/CBCL)...81 Table 13: Level 2 Models of Parent-reported HBI-Cognitive (MRI/DBRS)...83 Table 14: Level 2 Models of Child-reported PCS-I (LOC/CBCL)...85 Table 15: Level 2 Models of Child-reported PCS-I (LOC/DBRS)...86 Table 16: Level 2 Models of Child-reported PCS-I (MRI/CBCL)...87 Table 17: Level 2 Models of Child-reported PCS-I (MRI/DBRS)...88 Table 18: Level 2 Models of Parent-reported PCS-I (LOC/CBCL)...89 Table 19: Level 2 Models of Parent-reported PCS-I (LOC/DBRS)...90 Table 20: Level 2 Models of Parent-reported PCS-I (MRI/CBCL)...91 Table 21: Level 2 Models of Parent-reported PCS-I (MRI/DBRS)...92 Table 22: Level 2 Models of Child-reported HBI-Somatic (LOC/CBCL)...93 Table 23: Level 2 Models of Child-reported HBI-Somatic (LOC/DBRS)...94 Table 24: Level 2 Models of Child-reported HBI-Somatic (MRI/CBCL)...95 Table 25: Level 2 Models of Child-reported HBI-Somatic (MRI/DBRS)...96 Table 26: Level 2 Models of Parent-reported HBI-Somatic (LOC/CBCL)...97 Table 27: Level 2 Models of Parent-reported HBI-Somatic (LOC/DBRS)...98 Table 28: Level 2 Models of Parent-reported HBI-Somatic (MRI/CBCL)...99 Table 29: Level 2 Models of Parent-reported HBI-Somatic (MRI/DBRS) Table 30: Level 2 Models of CVLT-C Total Immediate Recall (LOC/CBCL) Table 31: Level 2 Models of CVLT-C Total Immediate Recall (LOC/DBRS) Table 32: Level 2 Models of CVLT-C Total Immediate Recall (MRI/CBCL) Table 33: Level 2 Models of CVLT-C Total Immediate Recall (MRI/DBRS) Table 34: Level 2 Models of CANTAB SWM (LOC/CBCL) Table 35: Level 2 Models of CANTAB SWM (LOC/DBRS) Table 36: Level 2 Models of CANTAB SWM (MRI/CBCL) Table 37: Level 2 Models of CANTAB SWM (MRI/DBRS) Table 38: Level 2 Models of CANTAB SOC (LOC/CBCL) Table 39: Level 2 Models of CANTAB SOC (LOC/DBRS) Table 40: Level 2 Models of CANTAB SOC (MRI/CBCL) Table 41: Level 2 Models of CANTAB SOC (MRI/DBRS)...118

10 Table 42: Internal Consistencies for Child PCS Measures Table 43: Internal Consistencies for Parent PCS Measures

11 11 LIST OF FIGURES Page Figure 1: Moderation of Parent-rated Cognitive symptoms by CBCL with LOC Figure 2: Moderation of Parent-rated Cognitive symptoms by CBCL with MRI Figure 3: Moderation of Parent-rated Cognitive symptoms by DBRS with MRI...122

12 12 INTRODUCTION Traumatic brain injury (TBI) is a common cause of mortality and morbidity in children and adults, accounting for approximately 10% of emergency room visits annually. Approximately 85% of TBIs can be classified as mild (Centers for Disease Control and Prevention [CDC] 2003; Yeates et al., 1999, Yeates, 2000). According to the American Congress of Rehabilitation Medicine (ACRM; 1993), a mild TBI is associated with a Glasgow coma scale (GCS) score of 13-15, and can include 30 or less minutes of loss of consciousness (LOC), less than 24 hours of posttraumatic amnesia (PTA), altered mental status (e.g. dizziness, confusion) and focal neurological signs (e.g. sensitivity to light and noise; Teasdale & Jennett, 1974). As the term mild implies, the duration and severity of signs and symptoms are less than those seen in either moderate or severe TBI. Neurobehavioral deficits are uncommon after a mild TBI, with 85-90% of mild TBI patients experiencing no persistent symptoms or cognitive deficits after 3 months post-injury (Alexander, 1995; Bazarian et al., 2005; Satz et al, 1997; Yeates and Taylor, 2005). Although most mild TBI cases recover within 3 months (Satz et al., 1997; Yeates & Taylor, 2005), not all cases of mild TBI are of equivalent severity or produce equivalent sequelae. Some injuries present clinically as a mild TBI (e.g., an injury that is accompanied by a GCS of 13-15) but are accompanied by structural or functional imaging abnormalities. Such mild TBIs are referred to as complicated, as first proposed by Levin and Williams (1987). Population estimates of the incidence of complicated mild TBI vary based on the method used to define complicated status (e.g., MRI versus CT scan), but range from approximately 5% - 15% (Coombs and Davis, 2000; Haydel et

13 13 al., 2000; McCrea, 2008; Levin et al., 1987; Williams et al., 1990). Studies reporting on clinic-referred participants describe higher rates, some as high as 50% (Levin et al., 1987; Williams et al., 1990). A somewhat surprising finding is that common indices of TBI severity --such as ratings of GCS, LOC, and PTA acutely-- are not necessarily predictive of which patient will have brain damage detectable by MRI (Levin et al., 1987; Williams et al., 1990). In the current study, approximately 18% of the participants had injuryrelated abnormalities on MRI. Most of the mild TBI outcome literature has focused on neuropsychological outcomes. Although some child studies have demonstrated group differences on some cognitive measures (e.g., Anderson, Catroppa, Morse, Haritou, & Rosenfeld, 2001; Belanger et al., 2005), reviews of children and adults have not found consistent group differences on neuropsychological tests of attention, memory, or executive functioning (Babikian & Asarnow, 2009; Bazarian et al., 2005; Satz et al, 1997). However, a small but significant group of individuals with a mild TBI (both children and adults) continue to present with postconcussive symptoms (PCS) or subjective somatic, cognitive, and emotional problems (e.g., Jakola et al., 2007; Mittenberg, Wittner, & Miller, 1997; Yeates et al., 1999). Additional research is needed to help identify child and injury characteristics that are associated with an increased risk of poor outcomes following a mild TBI. A possible explanation for some of the inconsistency in mild TBI outcome studies is the failure of prior investigations to separately analyze outcomes for different levels of severity of mild TBI. Specifically, there is some evidence that complicated mild TBI, as defined above, may be associated with worse outcomes or slower rates of recovery than

14 14 an uncomplicated mild TBI. For example, in a study comparing the performance of youth following a mild TBI, Levin et al. (2008) found that performance on cognitive tests measuring cognitive inhibition, calculation, working memory, verbal learning, episodic memory, and psychomotor processing speed was worse among children with complicated mild TBI relative to those with uncomplicated TBI. More recently, Yeates et al. (2009) reported that children with a greater number of acute clinical features that are typically predictive of intracranial injury (including children with complicated mild TBI), in other words a more severe mild TBI, were more likely to report acute and persistent PCS symptoms than children with less severe mild TBI and control group children with an orthopedic injury. Specifically, children with evidence of intracranial injury on MRI, as well as children with a greater number of acute clinical features of brain injury (such as persistent post-traumatic amnesia, LOC, disorientation, and other mental status changes) had greater odds of having PCS symptoms both acutely and chronically. Understanding of the sequelae of pediatric mild TBI is also limited by the lack of examination of pre-injury child risk factors and their potential impact on outcomes. Most previous studies have not evaluated premorbid child characteristics, such as attention or behavior problems. One issue complicating the evaluation of the potential influence of premorbid child characteristics, such as attention and behavior problems, and recovery from TBI is the considerable overlap between reported symptoms seen in individuals with ADHD and those with mild or moderate TBI (i.e. worse performance on cognitive measures and subjective behavioral and cognitive complaints, relative to normal controls; McCrea, 2008; Pelham, Fabiano, & Massetti, 2005; Taylor, 2004). Consideration of

15 15 premorbid attention or behavior problems is particularly relevant to any discussion of pediatric TBI because such problems not only increase a child s risk for sustaining a traumatic injury in the first place (orthopedic or brain injury), but such behaviors may also interact with a mild TBI, thus exacerbating the negative impact of TBI on the developing brain (Dennis, 2000; Taylor, 2004). It may be that attention and behavior problems, such as those commonly associated with ADHD, may be indicative of increased vulnerability to the effects of TBI, reflecting diminished reserve capacity and a reduced ability to efficiently compensate for the effects of even a mild TBI (Dennis, 2000; Satz, 1993; Stern, 2003). Alternatively, a mild TBI in the context of premorbid behavior problems may exacerbate or worsen such premorbid inattentive and impulsive behavior problems (Yeates et al., 2005). This relationship, however, has not been adequately investigated in TBI studies due to design limitations that the present study attempted to address. Indeed, in their meta-analytic review of pediatric mild TBI outcomes, Satz and colleagues (1997) concluded that the design of most studies obfuscated interpretations about whether TBI risk factors such as ADHD symptom behaviors interact with mild TBI to lower the minimal threshold of TBI necessary to adversely affect outcomes (p. 125). Most studies of mild TBI have methodological limitations that preclude the examination of premorbid child factors. First, most studies are retrospective--gathering data months or longer after the injury, which prevents the reliable collection of parentreport data on premorbid child functioning. Second, most studies lack an adequate control group of participants, such as children who experience a traumatic injury other than that involving the brain, who may share the same premorbid characteristics. Third,

16 16 most studies rely on clinic-referred participants instead of recruiting prospectively, thereby biasing their sample toward individuals with persistent or more severe complaints. Finally, most studies report on only two time points and do not follow children longitudinally. Examination of the interaction of premorbid child factors with injury factors with a longitudinal, prospective and controlled design will enhance understanding of the factors that best predict recovery following mild TBI in children, and may ultimately lead to improved outcomes for large numbers of children. The primary purpose of the present study was to investigate the relationship between pre-injury ADHD-spectrum behaviors and mild TBI severity, as these factors may influence neurobehavioral outcomes following mild TBI in children. There were two main study hypotheses. The first hypothesis was that pre-injury ADHD-spectrum behaviors would moderate child and parent report of postconcussive symptoms, as well as performance on cognitive tests (i.e. list learning, spatial working memory, and planning) following a mild TBI, such that the difference between children with a mild TBI and children with only orthopedic injuries will be greater for children with more preinjury ADHD-spectrum behaviors than children with fewer ADHD-spectrum problems. The second main study hypothesis was that the influence of pre-injury ADHDspectrum behaviors on the moderation of outcomes following mild TBI would be more pronounced in children with complicated versus uncomplicated mild TBI. The present study examined two different indices of mild TBI severity in analyses: 1) presence or absence of loss of consciousness (LOC) and 2) presence or absence of MRI abnormalities for classification of complicated versus uncomplicated mild TBI.

17 17 METHODS Participants The present study analyzed data collected as part of the Ohio Mild Head Injury study (OMHI), a five-year project funded by the National Institute of Child Health and Development examining neurobehavioral outcomes in children following mild TBI (Yeates & Taylor, 2005). 186 children with mild TBI and 99 children with minor orthopedic injuries (OI), aged 8-15 years old at the time of their injuries, were prospectively recruited over a 36-month period from consecutive Emergency Department admissions from two Ohio children s hospitals: Columbus Children s Hospital (now Nationwide Children s Hospital) and Rainbow Babies and Children s Hospital in Cleveland. For a more comprehensive description of study design and procedures please see Yeates and Taylor (2005). Children were eligible for inclusion in the mild TBI group if they sustained blunt head trauma that resulted in either an observed loss of consciousness for 30 minutes or less, a Glasgow Coma Scale (GCS; Teasdale & Jennett, 1974) score of 13-15, or at least two symptoms of concussion as noted by Emergency Department medical staff (i.e., persistent post-traumatic amnesia, transient neurological deficits, vomiting, nausea, headache, diplopia, dizziness). Hospitalization and evidence of skull fracture on imaging did not exclude a child from participating. A child was excluded from the study in any of the following circumstances: If the child s neurological status deteriorated (e.g., if the GCS fell below 13) or the child required surgical intervention; if the child sustained any extra-cranial injury acquired at the same time as the TBI with an Abbreviated Injury Scale (AIS; American Association for the Advancement of Automotive Medicine, 1990)

18 18 score > 3; if the child sustained an associated injury that could interfere with neuropsychological testing (e.g., broken arm); if the child suffered hypoxia, hypotension, or shock during or following the injury; if drugs or alcohol were involved with the injury; if there was a medical history of previous TBI, premorbid neurological disorder or mental retardation (not including ADHD or Learning disability), or psychiatric hospitalization; or if the TBI was the result of child abuse or assault. Children were eligible for the orthopedic injury control group if they sustained upper or lower extremity fractures with AIS scores < 3. Exclusion criteria for the orthopedic control group were similar to those for the mild TBI group. Children in the orthopedic injury control group were also excluded if they complained of or displayed any symptoms of a head injury or concussion. Forty-eight percent (48%) of the children meeting criteria for mild TBI participated in the study, while 35% of orthopedic control children participated. Four children in the mild TBI group did not complete the MRI and were, therefore, excluded from the current study. Participants did not differ from non-participants in either group with respect to age, gender, ethnic/racial minority status, or socioeconomic status (i.e. a composite of maternal education, median family income, and occupational prestige). The mild TBI group (186 children) and orthopedic injury group (99 children) did not differ with respect to age, gender, ethnic/racial minority status, or socioeconomic status. In both groups, sports and recreation-related injuries were the most common types of injury (57% of the mild TBI group and 62% of the orthopedic injury group), followed by falls (20% of the mild TBI group and 21% of the orthopedic injury group). Participants in the mild TBI group were 71% male and 73% White. Participants in the orthopedic control group

19 19 were 65% male and 65% White. The mild TBI and orthopedic injury control groups had similar retrospective parent-ratings of pre-injury ADHD-spectrum problems. See Table 1 for characteristics of the participants in the three injury groups. Of the children in the mild TBI group, 18% (32) evidenced injury-related intracranial abnormalities on MRI and were classified as having a complicated mild TBI in the current study, while 39% (73) reported a loss of consciousness, and 13% had a GCS score of 13/15 (i.e. the lowest score available in the mild range). The three resulting groups--complicated mild TBI, uncomplicated mild TBI, and orthopedic injury control group--did not differ in age-at-injury, gender, or socioeconomic status. Although the 2 mild TBI groups had similar GCS scores, the complicated mild TBI group had significantly more children who experienced a loss of consciousness. Unexpectedly, the complicated mild TBI group had significantly fewer children of minority race/ethnicity than the uncomplicated mild TBI or orthopedic injury groups. The orthopedic control group had the greatest overall injury to the body-- not including injury to the head--as measured by the Modified Injury Severity Scale (MISS) which is calculated from the AIS score (i.e. the sum of the squares of the three most severely injured areas of the body; Mayer, Matlak, Johnson, and Walker, 1980). When head injury was included in the MISS score, the uncomplicated mild TBI group had greatest overall injury severity. Procedure The medical records of all children who presented with TBI or orthopedic injury were reviewed daily by a research assistant to determine if the children met inclusion criteria. An initial baseline assessment appointment was completed within 3 weeks of the injury, with 80% of the baseline assessments completed between 1 and 2 weeks post-

20 20 injury (mean days = 11.35, SD=3.42). Informed written consent was obtained from parents prior to additional interviewing or child testing. Children were asked to provide written assent. A mild TBI was classified as either complicated or uncomplicated based on evidence of trauma-related intracranial injury on MRI. Children were assessed at four time points: baseline (i.e. within 3 weeks of the injury), 1 month, 3 months, and 12 months after the injury. Parent s retrospective ratings of children's pre-injury ADHD-spectrum problems and post-concussive symptoms were obtained at the baseline assessment. Parent and child ratings of children's current postconcussive symptoms were also assessed at baseline, and at 1, 3, and 12 months postinjury. Children were administered standardized cognitive measures (described below) at 1, 3, and 12 months post-injury. See Table 2 for a schedule of measure administration. Of the 285 children who completed the initial assessment, 98% (n=280, 98% with mild TBI and 98% with orthopedic injury) completed the one month assessment, 94% (n=268, 96% with mild TBI and 91% with orthopedic injury) completed the three-month assessment, and 91% (n=25, 91% with mild TBI and 85% with orthopedic injury) completed the 12-month assessment. Children completing each assessment session did not differ significantly in demographic characteristics from children who did not return for a session, except at the 12-month assessment, at which time children who completed the 12-month assessment tended to be of higher SES and were more likely to be of nonminority race/ethnicity than children who did not. Measures More comprehensive psychometric information for measures used in the present study can be found in Appendix A.

21 21 Magnetic Resonance Imaging The MRI pulse sequences included the following images: Axial T2-weighted and proton density fast spin echo, sagittal T1-weighted spin echo, coronal 2-dimensional gradient echo, axial diffusion-weighted echo planar, and coronal FLAIR (i.e. fluid attenuated inversion recovery). Board-certified radiologists with specialized training in pediatric neuroradiology reviewed and coded the MRI images in terms of presence or absence of intracranial abnormalities. The radiologists were blind to group membership and assessment results. The MRI procedure lasted approximately 30 minutes. Children were not sedated prior to imaging, but instead, were taught behavioral relaxation techniques. Pre-injury Attention and Behavior Problems Parents were administered the Child Behavior Checklist (CBCL) and the Disruptive Behavior Rating Scale (DBRS; Barkley, 1997) at the baseline appointment to assess retrospective ratings of children's premorbid functioning. In the current study, the CBCL-Attention Problems scale was used as a dimensional measure of child attention and behavior problems (T score; M=50, SD=10), while the ADHD index from the Disruptive Behavior Rating Scale was used as a measure of DSM-IV TR symptom criteria for ADHD. In the current study, parent ratings on these two child behavior measures were very highly correlated, r =0.80, p<.01. The psychometric properties of the CBCL are excellent and are supported by decades of research. Scales were empirically constructed and confirmed with factor analysis. All items on the CBCL significantly discriminate clinic-referred from nonreferred children at p <.01 (Achenbach and Rescorla, 2001, p. 109). The Attention

22 22 Problems scale correlates highly with Diagnostic and Statistical Manual of Mental Disorders (DSM-IV TR; American Psychiatric Association, 2000) criteria for ADHD (r =.80, p <.001) and highly with the ADHD Index of the Connors Parent Rating Scale (r =.77, p <.001). Eight-day test-retest reliability for the Attention Problems scale is good (r =.92) and internal item-scale consistency is adequate (Cronbach s alpha =.86). The ADHD index of the Disruptive Behavior Rating Scale assesses parent ratings of child behavior occurring in the last 6 months with respect to DSM-IV TR symptom criteria for ADHD using behaviorally-based descriptions (Barkley, 1997). Participating parents were instructed by the research assistant to rate the child s behavior prior to the injury. Parents rated the frequency of child behavior on 18 ADHD symptoms (i.e. nine inattentive, nine hyperactive-impulsive) on a four point scale (0 = never or rarely, 1 = sometimes, 2 = often, 3 = very often). Ratings of two or three indicate a problem for that item (Barkley, 1997). The Disruptive Behavior Rating Scale was standardized on a nationally representative sample. Internal consistency for the ADHD scale when completed by mothers was alpha =.89. The test retest reliability over a 4-week was.89 (DuPaul, Anastopoulos et al., 1998; DuPaul, Power, et al., 1997). Postconcussive Symptoms Retrospective parent ratings of pre-injury postconcussive symptoms and parent and child ratings of current postconcussive symptoms were assessed with the Postconcussive Symptom Interview (PCS-I; Mittenberg, Miller, and Luis, 1997) and the Health and Behavior Inventory (HBI; Yeates et al., 1999). Both measures inquire about a variety of somatic, cognitive, emotional, and behavioral symptoms that children may

23 23 experience after a brain injury. The child versions were changed slightly to reflect ageappropriate language. The parent versions were changed to reflect the third person. The Postconcussive Symptom Interview is a structured, 15-item (yes/no) interview assessing parent and child ratings of a variety of cognitive, somatic, and emotional problems that may be the result of concussion or TBI (Mittenberg et al., 1997). The variable used in the current analyses is the total number of problems endorsed, with higher scores indicating more postconcussive symptoms. In previous studies, children with mild TBI reported a greater number of postconcussive symptoms and had higher composite scores than healthy and orthopedic-controls (Mittenberg et al., 1997; Yeates et al., 1999). In previous research using an adult version of the Postconcussive Symptom Interview (i.e. same items with language modifications), scores on the interview demonstrated satisfactory internal consistency and reliability (Fergusson & Mittenberg, 1995). In the present study, the internal consistencies for parent-rated current symptoms on the Postconcussive Symptom Interview ranged from alpha = , depending on the assessment occasion (see Appendix A). The internal consistencies for child-rated current symptoms on the Postconcussive Symptom Interview ranged from alpha = , depending on the assessment occasion. The internal consistency for pre-injury parent rated Postconcussive Symptom Interview was alpha =.71. The Health and Behavior Inventory is a research measure that inquires about a variety of somatic, cognitive, emotional, and behavioral symptoms that children may experience after a brain injury. Parents and children rate the frequency of each symptom on a 4-point scale, (0 = never, 1 = rarely, 2 = sometimes, 3 = often) during the last week.

24 24 Evidence for validity comes from studies showing that children with moderate, severe, and mild TBI had significantly higher total scores than orthopedic injury control children (Yeates et al., 1999). In the present study, only the parent- and child-reported cognitive and the somatic scales from the Health and Behavior Inventory were used in analyses, as they were derived from factor analyses of the HBI, shown to be reliable across raters and time (Ayr, Yeates, Taylor, & Browne, 2009), and are most likely to be tapping problems associated with ADHD-spectrum symptoms and TBI (e.g., trouble concentrating). In the current study, both the Cognitive and Somatic scales of the HBI demonstrated good internal consistency across all assessment points, with Cronbach's alpha ranging from for current parent ratings, and for current child ratings. Internal consistencies for pre-injury parent ratings on the HBI-Cognitive Scale was alpha =.94, and pre-injury parent ratings on the HBI-Somatic scale was alpha =.80. Neuropsychological Tests The California Verbal Learning Test-Children s Version (CVLT-C) Delis, Kramer, Kaplan, & Ober, 1994) was used to assess verbal learning. The CVLT-C is a well-validated measure assessing verbal list learning and memory in children ages Children are presented with a spoken list of 15 words from three semantic categories. The list is presented five times, with the number of immediately recalled items summed across five trials (Total Recall). In the present study, Total Recall score was used in analyses, as it is associated with moderate to large effects in mild TBI and ADHD research studies (e.g., Seidman, 2006; Yeates et al., 1999). Internal consistency is good (Sherman, Sprauss, & Spreen, 2006) and 1-month test-retest reliability is adequate (n = 106, r = ) and reflects expected practice effects. The CVLT-C has been shown to

25 25 be sensitive to impairments related to TBI (e.g., Thompson et al, 1994; Yeates, Blumenstein, Patterson, & Delis, 1995) and ADHD (Delis et al., 1994). Two subtests from the Cambridge Neuropsychological Test Automated Battery (CANTAB; Sahakian & Owen, 1992) were used to assess spatial working memory and planning: Spatial Working Memory (SWM) and Stockings of Cambridge (SOC). Both the Spatial Working Memory and Stockings of Cambridge subtests have been shown to be sensitive to frontal dysfunction and to correctly identify lesion patients versus otherclinical controls (Robbins, 1996) and take approximately 5 10 minutes to administer. The Spatial Working Memory subtest is a self-ordered search task that assesses spatial working memory. The total error raw score on the Spatial Working Memory test was used in analyses. The Stockings of Cambridge, which is very similar to the Tower of London task, assesses spatial planning, spatial problem solving, and response inhibition. For the Stockings of Cambridge, the total number of problems solved in the minimum number of moves (raw score) was used in analyses. CANTAB subtests used in the current study appear to be a valid instruments for children age 5-12 years old (Luciana and Nelson, 2002; Robbins, 1999) in assessing frontal function and activating the prefrontal cortex (Owen, Doyon, Petrides, & Evans, 1996; Robbins, 1999). Due to the design of stimulus items and response format (i.e. touch screen), the CANTAB also appears to be a good test for children with poor verbal skills and articulation problems (Luciana & Nelson, 2002). Data Analysis The present study examined the following hypotheses: 1) Pre-injury ADHDspectrum behavior problems will moderate ratings of postconcussive symptoms and

26 26 performance on neuropsychological tasks of memory, working memory, and executive functioning following mild TBI, such that differences between the mild TBI and orthopedic injury control groups will be greater for children with more ADHD-spectrum behavior problems than children with fewer ADHD-spectrum problems; 2) the moderating effects of pre-injury ADHD-spectrum problems will be most apparent in children with complicated mild TBI. Hierarchical Linear Modeling (HLM) Hierarchical linear modeling was used to examine children s performance on cognitive measures at baseline, 3 months, and 12 months post-injury, as well as child and parent ratings of postconcussive symptoms at baseline, 1 month, 3 months, and 12 months post-injury. The HLM-6.8 software program was used for analyses (Raudenbush, Bryk, Cheong, Congdon, & du Toit, 2004). Analyses were conducted for nine dependent variables: 1) Health and Behavior Inventory-Cognitive scale child-report, 2) Health and Behavior Inventory-Cognitive scale parent-report,3) Health and Behavior Inventory- Somatic scale child-report, 4) Health and Behavior Inventory-Somatic scale parentreport, 5) Postconcussive Symptom Interview child- report, 6) Postconcussive Symptom Interview parent-report, 7) California Verbal Learning Test-Children's Version Total Immediate Recall, 8) CANTAB Stockings of Cambridge problems solved in minimum number of moves, and 9) CANTAB Spatial Working Memory total errors. Each of the nine outcome variables was run in four different series of models, each series containing one of the two injury severity variables and one of the two pre-injury ADHD-spectrum behavior measures: LOC and CBCL, LOC and DBRS, MRI and CBCL, and MRI and DBRS. In all, 36 Level 2 models were estimated.

27 27 Dummy variables were created to represent the three injury groups (i.e. complicated mild TBI, uncomplicated mild TBI, and orthopedic injury controls) as defined by MRI abnormality or loss of consciousness; one contrasted the two mild TBI groups with the orthopedic injury control group and the other contrasted the complicated mild TBI and uncomplicated TBI groups. Level 1 models examining intra-individual growth curves yielded parameters for intercept, slope, and curvature. Time-since-injury was scaled or centered at 90 days post-injury (i.e. measurement occasion 90 days/365), given evidence that most individuals experiencing a mild TBI experience full symptom resolution by 90 days post-injury (Satz et al., 1997; Yeates & Taylor, 2005). The analysis of each dependent variable consisted of two steps--estimation of the Level 1 and Level 2 models. First, the unconditional Level 1 model was estimated for each dependent variable to examine the mean and variance for the growth curve parameters: intercept, slope, and curvature. The Level 1 intercept parameter represented the expected level of the dependent variable (i.e. postconcussive symptoms or cognitive performance) at 3 months post-injury. The Level 1 slope parameter represented the linear rate of change in the dependent variable at 3 months post-injury. Some Level 1 models also included a quadratic parameter that represented the amount of curvature (i.e. change in rate of change) in the dependent variable over time. In the Level 1 unconditional model, parameters for which the mean and variance differed significantly from zero were retained and used in Level 2 models. In the second step, Level 2 models were tested to estimate the unique contributions of each predictor variable (i.e. injury group as measured by loss of consciousness or MRI abnormality, Child Behavior Checklist, Disruptive Behavior Rating Scale, and the interaction of injury

28 28 group with Child Behavior Checklist or Disruptive Behavior Rating Scale) to the level of the dependent variable at 3 months post-injury (i.e. intercept), the linear rate of change at 3 months post-injury (i.e. slope), and the degree of curvature in the recovery trajectory (i.e. quadratic) across the first 12 months post-injury. Race, SES, and age-at-injury were included as covariates in all models. In models with parent-rated postconcussive symptoms as the dependent variable (i.e. Postconcussive Symptom Interview-Total symptoms, Health and Behavior Inventory- Cognitive Scale, Health and Behavior Inventory-Somatic Scale), parent-rated pre-injury postconcussive symptoms was also included as a covariate. Parent-ratings of pre-injury postconcussive symptoms were not included as a covariate in models examining childrated postconcussive symptoms. All continuous predictor variables (including the covariates named above) were centered in analyses to aid interpretation of the results by reducing problems with multicolinearity (i.e. predictor grand sample mean). To answer the main hypotheses, eight interaction terms were created to test for moderation of postconcussive symptoms (i.e. parent and child ratings on the Postconcussive Symptom Interview and the Health and Behavior Inventory) and cognitive performance (i.e. child's performance on the CANTAB and California Verbal Learning Test-Children's version) over time by two measures of pre-injury ADHDspectrum problems (i.e. CBCL and DBRS) and two indices of injury severity defined by presence of loss of consciousness and presence of abnormal MRI findings. General linear hypothesis tests were conducted to determine whether the inclusion of interaction terms explained significant variance. If the results of the general linear hypothesis tests were significant, then interaction terms were retained in the full model. If the interaction

29 29 terms were not found to be significant in initial analyses, then the models were reanalyzed without the interaction terms. When significant effects were detected for predictors involving injury group, pre-injury ADHD spectrum problems, or their interaction, then the percentage of parameter variance accounted for by each significant predictor was estimated using the Raudenbush and Bryk (2002) method, whereby the percent of variance- accounted-for equals the variance accounted for by the full model (i.e. the model containing all predictors) minus the variance accounted for by the restricted model (i.e. the model without the predictors in question).

30 30 RESULTS Preliminary Analyses Table 3 presents the simple correlations between the parent rated pre-injury child behavior and symptom measures used in the current study (i.e. DBRS, CBCL, Health and Behavior Inventory-retrospective, and Postconcussive Interview-retrospective), with correlations above the diagonal representing the mild TBI group and below the diagonal representing the control group. As expected, parent-ratings of pre-injury behavior on the CBCL-Attention Problems scale were highly correlated with parent-ratings of ADHD symptoms on the DBRS for both groups. The two ADHD-spectrum behavior ratings were also very highly correlated with the HBI Cognitive scale for both groups, likely reflecting the considerable overlap between the three measures. Accordingly, the two ADHD-spectrum behavior ratings were less correlated with the HBI Somatic scale, which likely reflects the lack of item-overlap between these three measures. Level 1 Unconditional Models The results of the Level 1 unconditional models for postconcussive symptoms are summarized in Table 4. The results of the Level 1 unconditional models for performance on neuropsychological measures are summarized in Table 5. Postconcussive Symptoms As expected, the estimated mean intercepts were significant for all measures of postconcussive symptoms. The results of the Level 1 unconditional models indicate that children exhibited significant variation in the number of child- and parent-reported postconcussive symptoms reported at 3months post-injury (intercept parameter) and in the rate of linear change (slope parameter) of child- and parent-reported postconcussive

31 31 symptoms reported for all the measures of postconcussive symptoms. Quadratic parameters also were significant for parent-rated HBI Somatic symptom score, childrated HBI Cognitive symptom score, and child-rated PCS-I score, indicating significant variation in the rate of acceleration/deceleration for these measures across the first year following the injury. As there was not significant variation in the quadratic change parameter for the Level 1 model involving child-rated HBI Cognitive symptoms, it was removed from the model. Likelihood-ratio tests comparing deviance statistics for restricted versus more complex Level 1 unconditional models (i.e. multiple parameter test of the variancecovariance components) were also conducted for each rating of postconcussive symptoms. Based on the results of these likelihood-ratio tests, the quadratic change parameter was treated as non-randomly varying in Level 2 model estimation (i.e. the residual variance of the quadratic parameter was fixed to zero) for models of childreported HBI Somatic symptoms. Neuropsychological Test Performance As expected, the estimated mean intercepts were significantly different from zero for each of the three cognitive test variables. The results of the Level 1 unconditional models indicate that children exhibited significant variation in cognitive test performance at 3 months post-injury (intercept parameter) for each of the cognitive test scores analyzed. Unexpectedly, however, children did not appear to exhibit significant variation in the rate of linear change (slope parameter) at 3 months, indicating that their scores were changing at similar rates at 3 months post injury. Because the quadratic change parameter did not display significant variance in the models for any of the cognitive test

32 32 scores, the quadratic parameter was removed from each Level 1 model. Additionally, based on likelihood-ratio tests comparing deviance statistics for restricted versus more complex Level 1 unconditional models of cognitive performance, the linear change parameter was treated as non-randomly varying for all the Level 2 models of cognitive performance. Level 2 Conditional Models The details of the 36 Level 2 models for postconcussive symptoms and cognitive test performance are presented in Tables Four of the 36 Level 2 models showed significant interactions between pre-injury ADHD-spectrum behaviors and injury group at the p <.05 significance level. Three of the 4 models with significant interactions were models predicting parent-reported Cognitive symptoms on the Health and Behavior Inventory. A significant interaction was also detected for a model predicting Total Immediate Recall on the California Verbal Learning Test-Children's Version. In general, as expected, orthopedic injury controls had significantly fewer postconcussive symptoms, as measured by child and parent-rated Postconcussive Symptom Interview and the Health and Behavior Inventory, than children with a mild TBI across the 4 time points, with the exception of child responses on the Health and Behavior Inventory-Cognitive scale. (i.e. 20 of 24 postconcussive models). None of the 4 models using the Health and Behavior Inventory child-rated cognitive symptoms showed group differences in recovery trajectories. Unexpectedly, for 8 of the 12 neuropsychological models, there was either no group difference on performance or there were group differences in the opposite direction than was expected. Specifically, children in the control group performed worse that the children in mild TBI group on 2 of the 4 models with CANTAB Stockings of

33 33 Cambridge as the dependent variable and all 4 models with CANTAB Spatial Working Memory as the dependent variable. Another general pattern to the findings was that pre-injury ADHD-spectrum behavior problems was a significant independent predictor for both postconcussive symptoms and neuropsychological test performance (i.e. 31 of 36 models), with the exception that the DBRS was not a significant predictor of child-responses on either the Postconcussive Symptom Interview or the Health and Behavior Inventory-Somatic scale and the CBCL was not a significant predictor of performance on the CANTAB Stockings of Cambridge. The results were fairly consistent across the two injury severity variables (i.e. LOC and MRI abnormality), as well as the two different premorbid ADHD-spectrum behavior measures (i.e. CBCL-Attention Problems Scale and DBRS-ADHD index). In other words, results were not dependent on how injury severity was measured or whether premorbid ADHD-spectrum behavior was measured empirically (i.e. with the CBCL- Attention problems scale) versus categorically (DBRS-ADHD index). Below, the results are presented for Level 2 models predicting the 3 child-rated postconcussive symptom variables, the 3 parent-rated postconcussive symptom variables, and 3 cognitive test performance variables, respectively. In each of the 3 sections (e.g., Child-rated Postconcussive Symptoms), a summary of the findings is presented first, followed by detailed results for each of the 4 models estimated for each of the 3 outcome variables (e.g., Postconcussive Symptom Interview, Health and Behavior-Cognitive scale, Health and Behavior-Somatic scale). The child-rated and parent-rated outcome variables are presented in the following order: Health and Behavior Inventory-Cognitive Scale, Health and Behavior Inventory-

34 34 Somatic Scale, and the Postconcussive Symptom Interview. The cognitive test outcome variables are presented in the following order: CVLT-C, CANTAB Spatial Working Memory, and CANTAB Stockings of Cambridge. For each outcome, the Level 2 models are presented in the following order: 1) Model with LOC and CBCL - complicated mild TBI is defined by loss of consciousness and pre-injury ADHD-spectrum problems are measured by the CBCL Attention problems scale; 2) Model with LOC and DBRS - complicated mild TBI is defined by loss of consciousness and pre-injury ADHDspectrum problems are measured by the DBRS-ADHD index; 3) Model with MRI and CBCL - complicated mild TBI is defined by presence or absence of MRI abnormality and pre-injury ADHD-spectrum problems are measured by the CBCL Attention problems scale; and 4) Model with MRI and DBRS - complicated mild TBI is defined by presence or absence of MRI abnormality and pre-injury ADHD-spectrum problems are measured by the DBRS-ADHD index. Level 2 Models Predicting Child-rated Postconcussive Symptoms No significant interactions were detected for child-reported postconcussive symptoms on any of the 3 measures. The vast majority of models predicting child-rated postconcussive symptoms showed main effects for injury severity and pre-injury ADHDspectrum behavior problems. The CBCL-Attention problem scale was significant for all child-report symptom models, whereas the DBRS-ADHD index was only predictive for child-reported Cognitive symptoms on the Health and Behavior Inventory. Both injury severity indices (i.e. loss of consciousness and MRI abnormality) were significant predictors of recovery parameters for child-reported Postconcussive Symptom Interview and the Health and Behavior Inventory-Somatic scale. Injury severity was not a