Long-Term Behavior Problems Following Pediatric Traumatic Brain Injury: Prevalence, Predictors, and Correlates

Transcription

1 Long-Term Behavior Problems Following Pediatric Traumatic Brain Injury: Prevalence, Predictors, and Correlates Lisa Schwartz, 1 MA, H. Gerry Taylor, 1,2 PhD, Dennis Drotar, 1,2 PhD, Keith Owen Yeates, 3 PhD, Shari L. Wade, 4 PhD, and Terry Stancin, 1,5 PhD 1 Case Western Reserve University, 2 Rainbow Babies & Children s Hospital, 3 The Ohio State University and Columbus Children s Hospital, 4 University of Cincinnati and Children s Hospital Medical Center, 5 MetroHealth Medical Center Objective To study identified rates of long-term behavior problems in children with traumatic brain injury (TBI) compared to children with only orthopedic injuries and risk factors and correlates for new behavior problems following TBI. Methods Sample included children with severe TBI (n = 42), moderate TBI (n = 41), and orthopedic injuries only (ORTHO; n = 50). The baseline assessment measured child behavior, adaptation, and neuropsychological, academic, and family functioning. Follow-ups were conducted at 6 and 12 months and at an extended follow-up a mean of 4 years after injury. Results The prevalence of caseness, defined as elevated behavior problem ratings, was higher in one or both TBI groups than in the ORTHO group at each follow-up (e.g., 36% of severe TBI group, 22% of moderate TBI group, and 10% of ORTHO group at extended follow-up). Most instances of postinjury-onset caseness at the extended follow-up were evident within the first year after TBI. Predictors were severe TBI, socioeconomic disadvantage, and preinjury behavioral concerns. Concurrent correlates included weakness in working memory and adaptive behavior skills, poorer behavior and school competence, and adverse family outcomes. Conclusions Postinjury-onset caseness is persistent, risks are multifactorial, and correlates include child dysfunction and family sequelae. Key words functioning. traumatic brain injury; children; behavior problems; caseness; risk factors; family Pediatric traumatic brain injury (TBI) is a major public health problem in the United States, with an average of approximately 180 per 100,000 children hospitalized for head injury each year (Kraus, 1995). Studies of outcomes 1 to 3 years postinjury reveal that severe pediatric TBI leads to psychiatric and behavioral problems, difficulties in adaptive functioning, and deficits in academic and cognitive skills (Brown, Chadwick, Shaffer, Rutter, & Traub, 1981; Fay et al., 1994; Fletcher, Ewing-Cobbs, Miner, Levin, & Eisenberg, 1990; Jaffe, Polissar, Fay, & Liao, 1995; Max et al., 1997; Max, Castillo, Bokura, et al., 1998; Max, Castillo, Robin, et al., 1998; Max et al., 2000). Behavioral sequelae frequently fail to resolve over time and thus are of particular concern to pediatric psychologists (Taylor et al., in press). Several studies that have examined outcomes after TBI in relation to preinjury functioning have found high rates of emergent, or postinjury-onset, behavior problems. Based on interviewer and teacher ratings of pre- and postinjury behavioral status, Brown et al. (1981) identified these problems in 52% of a sample of 21 children with severe TBI examined at 1 year after injury. In contrast, only 14% of their control group with orthopedic injuries met criteria for new problems. Rates of postinjury-onset problems changed little over time. Max et al. (1997) also examined rates of new postinjury psychiatric problems, defined in terms of psychiatric diagnoses not present prior to injury. By the second year following the TBI, 15 out of 42 (36%) children met criteria for a new psychiatric disorder, 11 of whom (26%) had All correspondence should be sent to Lisa Schwartz, Department of Psychology, Case Western Reserve University, Euclid Avenue, Cleveland, Ohio lxs78@po.cwru.edu. Journal of Pediatric Psychology, Vol. 28 No. 4, 2003, pp Society of Pediatric Psychology DOI: /jpepsy/jsg013

2 252 Schwartz et al. a disorder that persisted beyond the 12-month assessment. Max et al. (2000) also examined rates of personality change disorder, a syndrome of behavioral change resulting from brain damage. Persistence of personality change disorder was found in about 40% of the TBI participants at an average of 2 years after injury. All of the children with persistent personality change disorder sustained a severe TBI. Bloom et al. (2001) investigated postinjury-onset problems in a manner similar to Max et al. (1997). In this study, psychiatric interviews conducted at least 1 year after injury revealed new problems in 27 of 46 children in the sample (58%). Twenty-two children (48%) had problems that arose after injury and persisted to the time of the assessment. Some children appear to be at greater risk than others to develop postinjury-onset behavior problems. Brown et al. (1981) found substantially higher rates of these problems after severe TBI than after milder TBI. Children with milder TBI had a rate of new problems comparable to the orthopedic injury control group. Additionally, the children with severe TBI who had new problems had greater family adversity and more subclinical preinjury behavior problems than the children from this group who did not have new problems. Likewise, Max and his associates (Max et al., 1997; Max, Castillo, Bokura, et al., 1998) reported that children with new postinjury psychiatric diagnoses had higher rates of preinjury psychiatric disturbance and greater family social disadvantage than children without new problems Children s postinjury cognitive functioning is a less consistent predictor of new behavior problems. Some studies have found a relationship between cognitive functioning and new psychiatric disorders (Max et al., 1999), while others have reported weak relationships between behavioral and cognitive outcomes of pediatric TBI (Anderson et al., 2001; Fletcher et al., 1990). Brown et al. (1981) found a relationship between new psychiatric disorders at 4 months postinjury and concurrent IQ scores, but not after 1.5 to 2 years. To our knowledge, no study has examined clinically significant postinjury-onset behavior problems beyond 3 years postinjury in relationship to comprehensive assessments of child and family outcomes. Assessment of these problems is important in estimating incidence rates, determining which children are at greatest risk, and advancing knowledge regarding possible etiologies. Investigation of concurrent correlates of these problems is required to understand associated child and family morbidity. Previous investigations have assessed behavioral changes after TBI in relation to relatively narrow sets of cognitive outcomes, such as IQ and memory measures (Brown et al., 1981; Max et al., 1999), rather than broader neuropsychological batteries. Assessment of family outcomes has been similarly limited. The purpose of this study was to investigate rates, risk factors, and correlates of long-term postinjury-onset behavior problems in a sample of children with moderate to severe TBI. Behavior problems in children with moderate to severe TBI and in a control group of children with orthopedic injuries were assessed at a long-term, or extended, follow-up conducted a mean of 4 years after injury. One specific aim was to determine rates of significant postinjury behavior problems, or caseness, in children with TBI relative to children who had sustained orthopedic injuries only (ORTHO group). Although we focused on outcomes at the extended follow-up, we also examined the continuity of caseness across earlier follow-up assessments. A second aim was to identify risks for the development of long-term postinjury-onset caseness following TBI. To meet this objective, we compared children with TBI who satisfied criteria for a clinically significant postinjuryonset problem at the extended follow-up to children with TBI who did not satisfy these criteria. Risk factors considered included TBI severity, preinjury child behavior and family status, and neuropsychological performance at a baseline evaluation conducted soon after injury. Our third aim was to identify child and family outcomes that were concurrently associated with persisting postinjuryonset caseness. The first hypothesis was that rates of caseness at the extended follow-up would be higher in children with TBI than in the control group. We anticipated that group differences would be found even when the presence/absence of preinjury caseness and sociodemographic factors were taken into account. Based on findings suggesting that behavior problems emerge relatively soon after TBI and persist over time (Taylor et al., 2002), we anticipated that most postinjury-onset caseness at the extended followup would have been evident within the first year after injury. The second hypothesis was that predictors of postinjury-onset caseness in children with TBI would include greater TBI severity, higher levels of reported preinjury behavior problems, greater cognitive impairment soon after the injury, and lesser family social advantage (Brown et al., 1981). Our third hypothesis was that postinjury-onset caseness would be associated with cognitive, educational, and adaptive sequelae, as well as worse family outcomes (Anderson et al., 2001; Barry, Taylor, Klein, & Yeates, 1996; Max, Castillo, Robin, et al., 1998; Max et al., 2000).

3 Long-Term Behavior Problems 253 Method Sample Table I. Characteristics of Follow-up Sample The follow-up sample consisted of 134 of the 189 participants recruited for a longitudinal project (Taylor et al., 1999) who remained in the study at an extended follow-up a mean of 4.10 years after injury (SD = 0.91, range = ). Forty-two children had sustained severe TBI, 42 moderate TBI, and 50 orthopedic injuries only (ORTHO group). The participants were originally recruited from admissions to four hospitals in north-central Ohio between 1992 and The study was approved by the institutional review board at each institution, and informed consent was obtained prior to participation. Criteria for inclusion were (a) at least one night s hospitalization for a traumatic brain or orthopedic injury, (b) age at injury between 6 and 12 years, (c) the absence of evidence of child abuse or previous neurological problem, and (d) residence in an English-speaking household. Severe TBI was defined on the basis of a lowest postresuscitation Glasgow Coma Scale (GCS; Teasdale & Jennett, 1974) score < 9. Children with moderate TBI had a GCS score of 9 of 12, or a GCS score > 12 accompanied by a skull fracture, intracranial mass lesion or contusion, diffuse cerebral swelling, posttraumatic neurological abnormality, or loss of consciousness longer than 15 minutes. Children in the ORTHO group were free of symptoms suggestive of possible brain insult (e.g., concussion, severe facial trauma). Injury severity was further described using the Modified Injury Severity Score (Mayer, Matlak, Johnson, & Walker, 1980), defined as the sum of the squares of injury ratings for the three most affected body areas, including the head. To assess the severity of injuries not involving the head, we computed a partial Modified Injury Severity Score, defined as the sum of the squared ratings assigned to the three most affected body regions excluding the head. Families were characterized in terms of single- versus two-parent households, the Hollingshead Four-Factor Index (Hollingshead, 1975), and the Socioeconomic Composite Index (SCI; Yeates et al., 1997). The SCI, which served as the primary measure of family social status, was the average of the sample z scores for the Duncan Socioeconomic Index (Stevens & Featherman, 1981), annual family income as coded on the Life Stressors and Social Resources Inventory (Moos & Moos, 1988), and a 7-point maternal education scale. Higher SCI scores reflect greater socioeconomic disadvantage. Table I lists demographic and injury characteristics for the three groups of children in the follow-up sample. The groups were similar in terms of age at injury, gender, and sociodemographic characteristics. The ORTHO group had a higher proportion of minority participants than the other two groups, but this difference was not significant. Children with moderate TBI were hospitalized for fewer days than children in the severe TBI and ORTHO groups, but the latter two groups were hospitalized for similar periods. These two groups were also similar in injury severity to areas of the body other than the head. Mean GCS Severe TBI Moderate TBI ORTHO (n = 42) (n = 42) (n = 50) Age at injury in years, M (SD) 9.70 (2.07) 9.53 (1.83) 9.40 (1.88) Years between injury andextended 3.94 (0.94) 4.20 (0.86) 4.15 (0.92) follow-up Males (n, %) 33 (79%) 29 (69%) 30 (60%) Minorities (n, %) 10 (24%) 10 (24%) 22 (44%) Single-parent households (n, %) 16 (38%) 14 (33%) 27 (54%) Hollingshead Four Factor Index, M (SD) (12.57) (13.38) (13.08) Socioeconomic Composite Index, M (SD) 0.32 (0.77) 0.38 (0.83) 0.13 (0.83) Days hospitalized, M (SD)* (9.31) 6.69 (7.12) (14.06) Glasgow Coma Scale Score, M (SD)* 4.62 (1.86) (1.88) Days of impaired consciousness, M (SD) 5.41 (6.48) 0.19 (0.59) Modified Injury Severity Score, M (SD)* (12.21) (5.74) 7.06 (2.81) Partial modified injury severity 8.71 (10.78) 2.31 (3.52) 7.06 (2.81) score, M (SD) Hollingshead Four Factor Index see Hollingshead, 1975; Modified Injury Severity Score see Mayer et al., TBI = Traumatic brain injury; ORTHO = orthopedic injury. See text for definitions of modified injury severity scores and Socioeconomic Composite Index. *p <.01.

4 254 Schwartz et al. scores and duration of impaired consciousness for the TBI groups are presented for descriptive purposes. Compared to the follow-up sample, participants who dropped out of the study prior to the extended follow-up had lower socioeconomic status as measured by the SCI. Attrition was also higher in the ORTHO group than in the TBI groups. The dropouts did not differ from the follow-up sample in terms of sex distribution, age at injury, race, or preinjury behavior problems as rated by parents soon after injury. Measures and Procedure Overview. The extended follow-up consisted of a half-day session with the child and parent. While the child was tested, the parent completed interviews and ratings of the child and family. Similar assessments had been completed on three previous occasions, including a baseline at an average of 3 weeks postinjury and follow-ups at 6 and 12 months after the baseline (Taylor et al., 1999, 2001; Wade, Taylor, Drotar, Stancin, & Yeates, 1998; Yeates et al., 1997). Table II lists the child and family measures considered for this study. Child testing consisted of measures of IQ, language skills, perceptual motor abilities, memory, attention and executive function, and academic achievement. Other child characteristics were assessed by administering the Child Behavior Checklist (CBCL; Achenbach, 1991a) and Vineland Adaptive Behavior Scales (Sparrow, Balla, & Cichetti, 1984) to parents and the Teacher s Report Form (TRF; Achenbach, 1991b) to teachers. CBCL and TRF Behavior Problem scales provide ratings of a broad spectrum of child behavior problems; CBCL competence is a composite of ratings of participation in activities, social relations, and school performance; and TRF academic performance is a rating of performance in academic subject areas. The Vineland is a semistructured parent interview that assesses adaptive functioning in the domains of communication, daily living skills, and socialization. To assess global school outcomes, we also monitored grade repetitions and special education placements. Family measures included parent self-report ratings of family functioning, parent psychological distress, and injury-related family burden (see Wade et al., 1998). Family functioning was assessed by the McMaster Family Assessment Device (Miller, Bishop, Epstein, & Keitner, 1985), a self-report measure with demonstrated reliability and validity (Stevenson-Hinde & Akister, 1995). The 12-item Family Assessment Device General Functioning Scale was used as a summary measure of family functioning. The Global Severity Index of the Brief Symptom Inventory (Derogatis & Melisaratos, 1983) served as a summary measure of parent psychological adjustment. Family burden was measured by the Family Burden of Injury Interview (Burgess et al., 1999) and the Impact on Family Scale, Version G (Stein & Jessop, 1985). The Family Burden of Injury Interview was developed to assess the unique burdens and challenges of pediatric TBI for families. A summary of overall injury-related stress was defined as the mean of subscales pertaining to (a) concerns with the child s recovery and adjustment, (b) the reactions of extended family and friends, and (c) spouse s reactions. The Impact on Family Scale, Version G provided a previously validated measure of the global impact of pediatric disability (in this case, injury) on the family. The Impact on Family Scale, Version G Total Negative Impact score served as a summary of family burden. Most of the battery of measures were given at baseline and the extended follow-up. However, testing at the extended follow-up excluded several of the tests given previously (Test of Nonverbal Intelligence 2; Token Test: Part V; Clinical Evaluation of Language Fundamentals Revised, Sentence Structure; Grooved Pegboard Test; and Woodcock Johnson Test of Achievement Revised, Passage Comprehension and Dictation) due to lack of discrimination between the groups at earlier points postinjury. In place of these tests, we administered the Oral Expression and Figurative Language subtests of the Test of Language Competence-Expanded Version (Wiig & Secord, 1985) to further assess language skills, Consonant Trigrams (Paniak, Millar, Murphy, & Keizer, 1997) to assess memory, and Rey-Osterrieth Complex Figure (Bernstein & Waber, 1996) to assess executive function. The child examiner was not directly informed of the child s group assignment, and tests were given in counterbalanced order across children to minimize test order effects. Teacher ratings were mailed to schools after each assessment. Respondents completed the baseline CBCL, TRF, Vineland, and Family Assessment Device on the basis of preinjury child or family status. Age-adjusted standard scores were used in analysis of measures with published norms. Raw scores for other measures were transformed to age- and gender-corrected z scores based on the performance of the ORTHO group. Copy and recall scores for the Rey-Osterrieth Complex Figure were scored as described by Taylor, Klein, Minich, and Hack (2000). Child Behavior Problems. Clinically significant child behavior problems, or caseness, were defined as a T score > 63 on the CBCL Behavior Problem scale, corresponding to ratings obtained by < 10% of the normative sample. Thus, children with a T score > 63 on the CBCL Behavior Problem scale prior to injury had preinjury caseness. Achen-

5 Long-Term Behavior Problems 255 Table II. Child and Family Measures Domain/Measure Reference Assessment Child Measures IQ WISC-III for ages <17 years; Wechsler (1991) B, F WAIS-III for ages > 17 years: Prorated IQ based on Vocabulary, Similarities, Block Design, and Object Assembly TONI-2 Brown, Sherbenou, & Johnsen (1990) B Language Boston Naming Test Yeates (1994) B, F Controlled Oral Word Association Test Spreen & Strauss (1991) B, F Token Test, Part V DiSimoni (1978) B CELF-R: Sentence Structure, Recalling Sentences Semel, Wiig, Secord, & Sabers (1987) B, F TLC-E: Screening Composite Wiig & Secord (1985) F Perceptual-Motor Skills Grooved Pegboard Test: time per peg averaged across hands Knights & Norwood (1980) B Developmental Test of Visual-Motor Integration Beery (1989) B, F Memory CVLT-C: trials 1-5 total Delis, Kramer, Kaplan, & Ober (1986) B, F Rey-Osterrieth Complex Figure, recall Bernstein & Waber (1996) F Attention and Executive function Continuous Performance Test -3: omissions & commissions Lindgren & Lyons (1984) B, F Underlining Test: total for subtests 2, 4, & 9 Rourke & Orr (1977) B, F Contingency Naming Test: efficiency score Anderson, Anderson, Northam, & Taylor (2000) B, F Consonant Trigrams: total recall Paniak, Millar, Murphy, & Keizer (1997) F Academic Achievement WJ-R Woodcock & Mather (1989) Passage Comprehension, Dictation B Letter-Word Identification B, F Writing Samples B, F Calculation B, F Behavior Problems CBCL Behavior Problem Scale Achenbach (1991a) B, a F TRF Behavior Problem Scale Achenbach (1991b) B, a F School Performance TRF School Performance Scale Achenbach (1991b) B, a F Competence and Adaptive Behavior CBCL Competence Achenbach (1991a) B, a F Vineland Adaptive Behavior scales, Adaptive Behavior Sparrow, Balla, & Cicchetti (1984) B, a F Composite Family Measures Family Functioning FAD-GF Miller, Bishop, Epstein, & Keitner (1985) B, a F Parent Distress BSI: Global Severity Index Derogatis & Melisarator (1983) B, F Family Burden FBII: total score Burgess et al. (1999) B, F Impact on Family Scale: total negative impact Stein & Jessop (1985) B, F B = baseline assessment; F = extended follow-up. WISC-III = Wechsler Intelligence Scale for Children, 3rd ed.; WAIS-III = Wechsler Adult Intelligence Scale, 3rd ed.; TONI-2 = Test of Nonverbal Intelligence-2; CELF-R = Clinical Evaluation of Language Fundamentals-Revised; TLC-E = Test of Language Competence-Expanded Version; CVLT-C= California Verbal Learning Test-Children s Version; WJ-R = Woodcock-Johnson Test of Achievement-Revised; CBCL = Child Behavior Checklist; TRF = Child Behavior Checklist-Teacher s Report Form; FAD-GF = Family Assessment Device, General Functioning Scale; BSI = Brief Symptom Inventory; FBII = Family Burden of Injury Interview. a Reflects preinjury child or family status.

6 256 Schwartz et al. bach (1991a) recommends this cutoff as a means for optimizing classification into clinical versus nonclinical groups. The CBCL, a widely used behavior rating scale, has good reliability and validity in assessing children s behavior and is normed for school-age children and adolescents. Prior research supports the validity of the CBCL in identifying behavior problems (Biederman et al., 1993; Jensen et al., 1996). We excluded the TRF in defining caseness due to previous demonstration of the insensitivity of teacher ratings to behavior problems following TBI (Taylor et al., 1995) and because different teachers rated the children at baseline and at the extended follow-up, thereby reducing the sensitivity of teacher ratings to change. Table III. Prevalence of Caseness by Group at Each Assessment (Total Sample) Analysis Group differences in rates of caseness were examined via chi-square analysis and logistic regression. Differences between each TBI group and the ORTHO group were represented in logistic analysis by dummy contrasts. The presence/absence of preinjury problem, gender, race, and the SCI were included as covariates in logistic regression to control for the influences of these factors. Preliminary analysis failed to reveal differences in caseness in relation to age at injury; thus, this factor was not considered. Only 41 children from the moderate TBI group were considered in analysis due to failure of one parent to complete the CBCL at the extended follow-up. Missing data resulted in further reductions in the sample size for some measures. The largest reductions were due to missing teacher ratings (5% missing TRFs at the extended follow-up). To identify factors associated with postinjury-onset caseness, we considered only children with TBI who did not meet criteria for preinjury caseness. We then divided these children into groups with and without postinjury-onset caseness at the extended follow-up. Predictors were identified by comparing the two groups on injury measures, preinjury behavior and academic performance, family measures obtained at baseline, and baseline neuropsychological and achievement skills. Injury measures considered were TBI group (severe vs. moderate), duration of impaired consciousness defined by days until the child could follow verbal commands, lowest GSC score, nonhead Modified Injury Severity Score, and computed tomography (CT) scan findings as classified according to the presence/ absence of both focal frontal lesions and other lesions. Concurrent correlates were examined by comparing the groups on child and family outcomes at the extended follow-up. To identify predictors and correlates unrelated to TBI group, we included this factor as a covariate in analysis. Additional covariates were gender, race, and the SCI. To reduce potential Type I errors, p level was set at.05 for each of the domains listed in Table II, with Bonferroni adjustments made in evaluating multiple measures within domains. Results Total Sample Prevalence of Caseness. Table III presents rates of caseness prior to injury (baseline ratings) and at each of the follow-ups. The table also summarizes findings from logistic regressions that compared each TBI group with the ORTHO group, controlling for gender, race, and the SCI. Despite the absence of group differences in preinjury caseness, rates were higher in the severe TBI group relative to the ORTHO group at each follow-up, and rates were higher in the moderate TBI group relative to ORTHO group at the 6-month and extended follow-up. The pattern of rates over time suggests that there was an increase in caseness in the severe group relative to the other two groups. Higher SCI scores, reflecting greater socioeconomic disadvantage, were associated with higher rates of caseness at each follow-up (p <.01). Race failed to predict caseness independently of the SCI at any of the assessments. A gender difference was found only at the 6-month follow-up, with a higher rate of caseness in boys than girls. Odds Ratio (95% CI) Severe TBI Moderate TBI Severe TBI Moderate TBI ORTHO χ 2 (2) vs. ORTHO vs. ORTHO Baseline (preinjury) 4/42 (10%) 10/41 (24%) 10/50 (20%) (0.12, 1.62) 1.78 (0.59, 5.39) 6-month follow-up 9/39 (23%) 9/41 (22%) 5/46 (11%) (1.64, 95.92)*10.48 (1.40, 77.64)* 12-month follow-up 12/39 (31%) 9/40 (23%) 6/46 (13%) (3.03, )** 4.88 (0.77, 30.63) Extended follow-up 15/42 (36%) 9/41 (22%) 5/50 (10%) 8.85* (3.00, 49.03)** 3.96 (1.01,.55)* TBI = traumatic brain injury; ORTHO = orthopedic injury. Odds ratios (confidence intervals) were obtained from logistic regressions that controlled for the Socioeconomic Composite Index (SCI), gender, and race. The presence of preinjury caseness was also included as a covariate in logistic analyses of the follow-up data. *p <.05. **p <.01.

7 Long-Term Behavior Problems 257 Continuity of Caseness. Few children with TBI had postinjury-onset caseness at the extended follow-up that was not present earlier in follow-up. Considering only participants who had data for all follow-ups, we found that 8 of 12 (75%) children with severe TBI with postinjuryonset caseness at the extended follow-up also satisfied the criterion for caseness at one or both of the earlier followups, compared to 2 of 3 (67%) children with moderate TBI and 0 of 2 (0%) children in the ORTHO group, χ 2 (2, N = 17) = 3.24, p =.14. Children With TBI Who Did Not Have Preinjury Caseness Incidence of Postinjury-Onset Caseness. Fifteen (22%) of the 69 children with TBI without caseness met criteria for caseness at the extended follow-up, including 12 of 38 (32%) with severe TBI and 3 of 31 (11%) with moderate TBI, χ 2 (1, N = 69) = 3.61, p =.06. Of the 15 children with postinjury-onset caseness, 13 met criteria for externalizing problems, 8 met criteria for internalizing problems, and 7 met criteria for both. Only 4 of 14 children in the postinjury-onset group met the criterion of a Behavior Problem T score > 63 on the TRF (1 TRF not completed), suggesting a lack of agreement of the teacher and parent ratings. Significance of Postinjury-Onset Caseness. To determine if postinjury-onset caseness reflected meaningful changes in behavior, we assessed the extent to which postinjury behavior ratings differed from ratings of preinjury behavior, as well as associations of postinjury-onset caseness with postinjury child or family counseling. If we considered only children in the sample without preinjury caseness, the mean change in the CBCL Behavior Problem T score at the extended follow-up, relative to the preinjury score, was points (SD = 4.76), or more than 1.5 SDs, for the children with postinjury-onset caseness. This compares to a mean change of only 0.22 points (SD = 9.53) for the children with TBI who failed to meet the criterion for caseness. The minimum change in the children with postinjury-onset caseness was 10 points, or 1 SD. Rates of child or family counseling received after injury, but not prior to it, were also higher in the children with postinjuryonset caseness. Specifically, a history of postinjury-only counseling was identified in 11/15 (73%) of the children with caseness, compared to 24/54 (44%) of those without caseness, χ 2 (1, N = 69) = 5.52, p <.05. Predictors of Postinjury-Onset Caseness. Logistic regressions indicated that the rate of caseness was higher in the severe TBI group than in the moderate TBI group (odds ratio = 4.85, 95% confidence interval = , p <.05). A higher rate of caseness was also associated with greater socioeconomic disadvantage and higher preinjury CBCL Behavior Problem ratings, but not with race or gender. No injury measure besides TBI severity (severe vs. moderate) discriminated between the children with and without caseness, whether analysis was carried out with the total sample of children with TBI or with each TBI group separately. Analyses also failed to reveal differences between children with and without caseness in preinjury teacher ratings of behavior problems or school performance, adaptive behavior, neuropsychological functioning, academic achievement, or any of the preinjury or early postinjury family measures. Concurrent Correlates of Postinjury-Onset Caseness. Table IV presents the results of comparisons of children with and without caseness on concurrent child and family outcomes. According to these results, the children with caseness had lower scores on the Vineland Adaptive Behavior Composite and Consonant Trigrams. The children with caseness also had poorer family outcomes, including higher parent distress and greater injury-related family burden, as well as poorer family functioning. Although children with and without caseness did not differ in rates of postinjury grade repetitions, 10 of 15 (67%) of the children with caseness were in special education at the extended follow-up, compared to 10 of 54 (19%) of the children without caseness, χ 2 (1, N = 69) = 10.99, p <.01. Controlling for covariates, we found that logistic regression revealed only a marginally significant group difference in rates of special education (odds ratio = 4.19, 95% confidence interval = 0.98, 17.90, p =.05). This difference was significant when analysis was restricted to the severe TBI group (odds ratio = 8.97, 95% confidence interval = 1.40, 57.41, p <.05). Discussion This study extends the knowledge of long-term behavioral sequelae of pediatric TBI by describing the prevalence of such problems and identifying risk factors and correlates of postinjury-onset caseness an average of 4 years after injury. Using an orthopedic injury control group allowed us to substantiate increased rates of caseness in children with TBI compared to children who sustained other injuries. Another strength of this study was our examination of a wide range of predictors and correlates of caseness, including injury characteristics, preinjury child status, family measures, and neuropsychological outcomes. Consistent with study hypotheses, children with moderate to severe TBI had higher rates of caseness at the extended follow-up than the ORTHO group, even after controlling for presence/absence of preinjury caseness and

8 258 Schwartz et al. Table IV. Child and Family Measures Associated With Postinjury-Onset Caseness at the Extended Follow-Up: Concurrent Correlates With Caseness W/out Caseness (n = 15) (n = 54) Group Effect Domain/Measure M (SD) M (SD) from ANCOVA Child Measures IQ WISC-III/WAIS-III (16.95) (20.19) F(1, 62) = 0.76 Language Boston Naming Test 0.30 (0.99) 0.02 (0.95) F(1, 62) = 0.21 Controlled Oral Word Association 0.67 (0.81) 0.25 (1.05) F(1, 62) = 0.23 CELF-R Recalling Sentences 7.86 (3.18) 9.37 (2.75) F(1, 59) = 0.67 TLC-E Screening Composite (14.44) (15.69) F(1, 62) = 0.06 Perceptual Motor Skills Developmental Test of Visual-Motor Integration (14.08) (16.80) F(1, 61) = 1.88 Memory CVLT-C Trials 1-5 total 0.48 (1.45) 0.04 (1.10) F(1, 62) = 0.26 Rey-Osterreith Complex Figure, recall 0.08 (1.05) 0.04 (0.86) F(1,60) = 0.00 Attention and Executive Function Continuous Performance Test, omissions and commisions 0.01 (0.62) 0.16 (0.83) F(1, 51) = 0.27 Underlining Test, total for subtests 2, 4, & (1.15) 0.10 (1.06) F(1,62) = 1.15 Contingency Naming Test efficiency score 0.44 (0.94) 0.10 (1.16) F(1, 62) = 0.06 Consonant Trigrams 0.97 (1.44) 0.08 (1.13) F(1, 62) = 7.17** Academic Achievement WJ-R Letter-Word Identification (14.98) (16.26) F(1, 62) = 0.01 WJ-R Writing Samples (13.49) (20.90) F(1, 62) = 0.86 WJ-R Calculations (11.64) (15.90) F(1, 60) = 0.40 School Performance TRF School Performance Scale (6.02) (8.76) F(1, 56) = 2.61 Competence and Adaptive Behavior CBCL Competence (5.79) (10.80) F(1, 63) = 3.88 Vineland Adaptive* Behavior Composite (17.33) (14.41) F(1,63) = 4.68* Family Measures Family Functioning Family Assessment Device-General Functioning 2.17 (0.40) 1.74 (0.44) F(1, 63) = 7.07* Parent Distress BSI Global Severity Index (10.38) (8.83) F(1,62) = 8.54** Family Burden FBII, z score for total score 1.19 (0.83) 0.32 (0.56) F(1,63) = 11.00** Impact on Family Scale, total negative impact (5.42) (5.13) F(1, 63) = 9.25** ANCOVA = analyses of covariance; WISC-III = Wechsler Intelligence Scale for Children, 3rd ed.; WAIS-III = Wechsler Adult Intelligence Scale, 3rd ed.; CELF-R = Clinical Evaluation of Language Fundamentals-Revised; TLC-E = Test of Language Competence-Expanded Version; WJ-R = Woodcock-Johnson Tests of Achievement-Revised; TRF = Child Behavior Checklist-Teacher s Report Form; CBCL = Child Behavior Checklist; FAD-GF = Family Assessment Device, General Functioning Scale; BSI = Brief Symptom Inventory; FBII = Family Burden of Injury Interview. Only children with TBI who did not have preinjury caseness were considered in the analyses. Covariates included in ANCOVA were TBI group (severe vs. moderate), the Socioeconomic Composite Index (SCI), race, and gender. Only significant differences are listed. Higher scores on all family measures reflect poorer outcomes. *p <.05, Bonferroni corrected. **p <.01, Bonferroni corrected. for sociodemographic factors. As in past studies, children with severe TBI were most often affected, and most children with TBI with postinjury-onset caseness had a history of caseness dating back to the first year after injury (Max et al., 1997; Max et al., 2000). Relative to children without caseness either before or after TBI, those with postinjuryonset caseness had more severe TBI, were less socially advantaged, and had higher ratings of preinjury behavior concerns. These differences were demonstrated with TBI severity taken into account, confirming the multifactorial nature of influences on postinjury behavior (Brown et al., 1981; Gerring et al., 1998; Max et al., 1997; Max, Castillo, Bokura, et al., 1998). Postinjury-onset caseness was concurrently associated with weaknesses on a test of working

9 Long-Term Behavior Problems 259 memory, less optimal school and adaptive outcomes, and greater family distress and burden. The results accord with previous findings showing that most postinjury-onset behavior disorders emerge soon after injury and that, once present, these disorders become discouragingly persistent (Bloom et al., 2001; Brown et al., 1981; Max et al., 1997, 2000). The findings also implicate brain insult as the primary basis for the elevated rate of postinjury-onset caseness in the severe TBI group. The behavior problems present at follow-up represented marked change relative to ratings of preinjury behavior. Moreover, differences were found in comparison to another injury control group that was closely similar to the severe TBI group in terms of length of hospitalization and severity of injury to nonhead body regions. Associations of social disadvantage with behavior problems after TBI have been demonstrated previously (Brown et al., 1981; Gerring et al., 1998; Kinsella, Ong, & Murtagh, 1999; Max et al., 1997; Max, Castillo, Bokura, et al., 1998; Max et al., 2000). These associations indicate that the behavioral consequences of TBI are not due to organic impairment alone but that environmental factors also play a role (Brooks, 1990; Rutter, Chadwick, & Shaffer, 1983). Although the mechanisms responsible for these associations are unclear, the family and professional resources needed to effectively manage cognitive and behavioral consequences of TBI may be less available in disadvantaged settings. Children with TBI also may be more vulnerable than other children to the stresses in these environments (Taylor et al., 2002). Past studies also document relationships between preinjury child functioning and postinjury-onset problems (Brown et al., 1981; Gerring et al., 1998; Max et al., 1997; Max, Castillo, Bokura, et al., 1998). One explanation for these relationships is that children with more parentreported pre-existing behavioral concerns are more vulnerable to neurological insults. This vulnerability may be related to greater difficulties coping with or compensating for the adverse effects of injury on their abilities or motivation. An alternative explanation for our findings is that smaller changes in behavior were required to detect postinjury-onset caseness in children with higher levels of pre-existing problems. However, most children with postinjury-onset caseness showed substantial pre- to postinjury changes in behavior. With regard to concurrent correlates, several past investigations have reported associations of post-tbi behavior problems with behavior competence and adaptive functioning (Fletcher et al., 1990, 1996), as well as with substandard adaptive and school functioning (Anderson et al., 2001; Kinsella et al., 1995, 1997; Max, Castillo, Bokura, et al., 1998). These associations may reflect either the broader influences of factors responsible for the behavior problems, or the effects of the behavior problems themselves on other aspects of functioning. Associations between postinjury-onset behavior problems and adverse long-term family outcomes after TBI have also been reported (Barry et al., 1996; Max, Castillo, Robin, et al., 1998; Rivara et al., 1994). In this study, children with postinjury-onset caseness did not differ from children without caseness in measures of preinjury or early postinjury family status. The relationship between postinjury-onset caseness and negative family outcomes at the extended follow-up thus appears to have emerged after TBI. The direction of potential child-family influences is difficult to establish. Although child behavior problems may have contributed to postinjury parent distress and family burden, adverse family reactions to injury may have exacerbated child behavior problems. Previous analysis of child and family data from this study supports the possibility of such bidirectional relationships (Taylor et al., 2001). Neuropsychological testing generally failed to discriminate children with postinjury-onset caseness from those without caseness, whether given soon after TBI or several years later. The only neuropsychological difference observed was on Consonant Trigrams at the extended follow-up. The latter finding suggests that children with postinjury-onset caseness may have weaknesses in verbal working memory. Armstrong, Mangeot, Colvin, Yeates, and Taylor (2002) found associations between working memory using Consonant Trigrams and parent ratings of executive functioning, a construct likely related to behavioral self-regulation. Nevertheless, the general absence of associations between cognitive and behavior outcomes is consistent with observations from past studies (Brown et al., 1981; Fletcher et al., 1990). One explanation for the absence of associations is that many neuropsychological tests are insensitive to the types of cognitive problems that accompany post-tbi behavior problems, such as subtle deficits in metacognition or pragmatic judgment (Dennis, Guger, Roncadin, Barnes, & Schachar, 2001). Another possibility is that brain systems that mediate behavior and self-regulation are distinct from those that mediate cognitive functions of the sort measured by most neuropsychological tests (Eslinger, Biddle, & Grattan, 1997), and TBI differentially affects these systems. Limitations One major limitation of this study is that attrition was highest in the ORTHO group and in children from lower SES. Higher dropout among children in the ORTHO group

10 260 Schwartz et al. may have diminished statistical power to detect group differences. A higher dropout rate among children with lower socieoeconomic status may have also reduced or obscured differences between the groups of children with and without postinjury-onset caseness. Although sociodemographic factors were included as covariates in the analyses, caution is advised in generalizing the findings to the broader population of children with TBI. A further study limitation is that we relied on parent ratings to identify caseness. Our use of the CBCL was justified by its validity in distinguishing clinical from nonclinical samples, as well as by the fact that children with postinjury-onset caseness had higher rates of child and family counseling than children without caseness. Nevertheless, some investigators have reported that the CBCL has a low sensitivity to behavioral sequelae of pediatric TBI (Fletcher & Ewing-Cobbs, 1991; Kinsella et al., 1995; Knights et al., 1991). Two recent studies found that structured interviews identified more post-tbi behavior problems than did parent ratings and that parent ratings were least effective in identifying internalizing problems (Bloom et al., 2001; Green, Foster, Morris, Muir, & Morris, 1998). Our procedure of excluding children who had preinjury caseness may also have resulted in underestimating rates of postinjury-onset disorders, as previous studies found new forms of behavior problems among children who met criteria for pre-existing diagnoses (Bloom et al., 2001; Max et al., 1997). Other concerns are that the parent ratings were used to assess both child and family outcomes and that ratings of preinjury child behavior were retrospective. Reliance on parent ratings in assessing both child behavior and family outcomes also raises the possibility that shared method variance may have contributed to associations between these measures. With respect to the preinjury ratings, parents perceptions of their child s preinjury behavior may have been affected by the injury itself. Although the parent s ratings were collected soon after the injury and the efficacy of this method has been demonstrated in previous studies (Rivara et al., 1994; Rutter et al., 1983), biases due to retrospective recall cannot be ruled out. Our small sample size is an additional concern. Despite the relatively large sample of children with moderate to severe TBI initially recruited, some were lost to followup, a number of those followed had preinjury caseness, and postinjury-onset caseness was identified in only a fraction of the remaining children. Although this is the first study to comprehensively examine factors associated with long-term behavioral status, larger samples are needed to identify the predictors and correlates of postinjury-onset disorders with greater precision. Significance and Future Directions Findings that the majority of children with moderate to severe TBI did not have long-term postinjury-onset caseness may provide some reassurance to many families. Assessment of the injured child s preinjury behavior and the family resources may assist in the identification of the children who most need monitoring. Our data indicate that children with severe TBI, previously identified behavior concerns, and limited family resources are most at risk and may deserve careful follow-up and more intensive school and support services, even if neuropsychological testing fails to reveal deficits. Psychological interventions with the child and family within the first year after injury may help to reduce the incidence or extent of long-term problems. Although there is a scarcity of research demonstrating the effectiveness of interventions for children with TBI, family-based approaches deserve consideration in treating the behavioral sequelae of TBI (Wade, Drotar, Taylor, & Stancin, 1995). Continued monitoring of behavior beyond the first year is also advised to identify children with later-emerging problems. Future research is needed to examine risk factors for postinjury-onset behavior disorders in greater detail, including predisposing child characteristics, type of brain insult, and environmental influences. Additionally, we need to better understand the mechanisms responsible for both the emergence of behavior problems after injury and behavioral resiliency (Masten, 2001). Comprehensive assessment of behavioral outcomes, their evolution over time postinjury, and their cognitive correlates will also be useful (Dennis et al., 2001). The development and testing of clinical interventions for those at risk for or those who have developed long-term behavioral problems following TBI are also recommended. Advances will require large, multisite samples and study designs that help to isolate the effects of TBI. Acknowledgments This work was supported by NINDS grant NS36335, National Institutes of Health, grant MCJ , Maternal and Child Health Bureau (Title V, Social Security Act, Health Resources and Services Administration, Department of Health and Human Services), and by National Institute of Mental Health grant 18330, research training in pediatric psychology. We acknowledge the contributions of Elizabeth Shaver, Anne Birnbaum, Nichole Wood, Sharon Armstrong, Barbara Shapero, Madeline Polonia, and Matt Diamond in data collection; Nori Minich in data management and analysis; and Anne Birnbaum for her

11 Long-Term Behavior Problems 261 assistance in manuscript preparation. We also acknowledge the participation of the Children s Hospital Medical Center of Akron and the collaboration of George Thompson, MD, G. Dean Timmons, MD, and Dennis Weiner, MD. Assistance with recruitment was provided by the Rainbow Pediatric Trauma Center, Rainbow Babies & Children s Hospital, Columbus Children s Hospital Trauma Program, and MetroHealth Center Trauma Registry. Received February 11, 2002; revisions received May 2, 2002; accepted May 29, 2002 References Achenbach, T. M. (1991a). Manual for the Child Behavior Checklist/4 18 and 1991 Profile. Burlington: University of Vermont, Department of Psychiatry. Achenbach, T. M. (1991b). Manual for the Teacher s Report Form and 1991 Profile. Burlington: University of Vermont, Department of Psychiatry. Anderson, P. Anderson, V., Northam, E., & Taylor, H. G. (2000). Standardization of the Contingency Naming Test (CNT) for school-aged children: A measure of reactive flexibility. Clinical Neuropsychological Assessment, 1, Anderson, V. A., Catroppa, C., Haritou, F., Morse, S., Pentland, L., Rosenfeld, J., et al. (2001). Predictors of acute child and family outcome following traumatic brain injury in children. Pediatric Neurosurgery, 34, Armstrong, K., Mangeot, S., Colvin, A., Yeates, K. O., & Taylor, H. G. (2002, February). Long-term executive deficits in children with traumatic brain injuries. Paper presented at the Thirteenth Annual Meeting of the International Neuropsychological Society, Toronto, Ontario, Canada. Barry, C. T., Taylor, H. G., Klein, S., & Yeates, K. O. (1996). The validity of neurobehavioral symptoms in children with traumatic brain injury. Child Neuropsychology, 2, Beery, K. E. (1989). Revised administration, scoring, and teaching manual for the Developmental Test of Visual- Motor Integration. Cleveland, OH: Modern Curriculum Press. Bernstein, J. H., & Waber, D. P. (1996). Developmental scoring system for the Rey-Osterrieth Complex Figure: Professional manual. Odessa, FL: Psychological Assessment Resources. Biederman, J., Faraone, S. V., Doyle, A., Lehman, B. K., Kraus, I., Perrin, J., et al. (1993). Convergence of the Child Behavior Checklist with structured interview-based psychiatric diagnoses of ADHD children with and without comorbidity. Journal Child Psychology and Psychiatry, 34, Bloom, D. R., Levin, H. S., Ewing-Cobbs, L., Saunders, A. E., Song, J., Fletcher, J. M., et al. (2001). Lifetime and novel psychiatric problems after pediatric traumatic brain injury. Journal of the American Academy of Child and Adolescent Psychiatry, 40, Brooks, D. N. (1990). Behavioral and social consequences of severe head injury. In B. Deelman, R. Soan, & A. van Zomeren (Eds.), Traumatic brain injury: Clinical, social and rehabilitational aspects (pp ). Amsterdam: Swets & Zeitlinger. Brown, G., Chadwick, O., Shaffer, P., Rutter, M., & Traub, M. (1981). A prospective study of children with head injuries: III. Psychiatric sequelae. Psychological Medicine, 11, Brown, L., Sherbenou, R., & Johnsen, S. (1990). Test of Nonverbal Intelligence: 2nd ed. Austin, TX: PRO ED. Burgess, E. S., Drotar, D., Taylor, H. G., Wade, S., Stancin, T., & Yeates, K. O. (1999). The Family Burden of Injury Interview: Reliability and validity studies. Journal of Head Trauma Rehabilitation, 14, Delis, D., Kramer, J., Kaplan, E., & Ober, B. (1986). The California Verbal Learning Test: Children s Version. San Antonio, TX: Psychological Corporation. Dennis, M., Guger, S., Roncadin, C., Barnes, M., & Schachar, R. (2001). Attentional-inhibitory control and social-behavioral regulation after childhood closed head injury: Do biological, developmental, and recovery variables predict outcome? Journal of the International Neuropsychological Society, 7, Derogatis, L., & Melisaratos, N. (1983). The Brief Symptom Inventory: An introductory report. Psychological Medicine, 13, DiSimoni, F. G. (1978). The Token Test for Children. Boston: Teaching Resources. Eslinger, P. J., Biddle, K. R., & Grattan, L. M. (1997). Cognitive and social development in children with prefrontal cortex lesions. In N. A. Krasnegor, G. R. Lyon, & P. S. Goldman-Rakic (Eds.), Development of the prefrontal cortex: Evolution, neurobiology, and behavior (pp ). Baltimore, MD: Paul H. Brookes. Fay, G. C., Jaffe, K. M., Polissar, M. L., Liao, S., Rivara, J. B., & Martin, K. M. (1994). Outcome of pediatric traumatic brain injury at three years: A cohort study. Archives of Physical Medicine and Rehabilitation, 75, Fletcher, J. M., & Ewing-Cobbs, L. (1991). Head injury in children. Brain Injury, 5,