Traumatic Brain Injury in Young Children: Post-Acute Effects on Cognitive and School Readiness Skills

Abstract

Previous studies have documented weaknesses in cognitive ability and early academic readiness in young children with traumatic brain injury (TBI). However, few of these studies have rigorously controlled for demographic characteristics, examined the effects of TBI severity on a wide range of skills, or explored moderating influences of environmental factors on outcomes. To meet these objectives, each of three groups of children with TBI (20 with severe, 64 with moderate, and 15 with mild) were compared with a group of 117 children with orthopedic injuries (OI group). The children were hospitalized for their injuries between 3 and 6 years of age and were assessed an average of 1½ months post injury. Analysis revealed generalized weaknesses in cognitive and school readiness skills in the severe TBI group and suggested less pervasive effects of moderate and mild TBI. Indices of TBI severity predicted outcomes within the TBI sample and environmental factors moderated the effects of TBI on some measures. The findings document adverse effects of TBI in early childhood on post-acute cognitive and school readiness skills and indicate that residual deficits are related to both injury severity and the family environment.

INTRODUCTION

Traumatic brain injury (TBI) is one of the most common causes of death and long-term disability in the pediatric age range (Gotschall, 1993; Kraus, 1995). According to a report on Emergency Department (ED) visit, hospitalizations, and deaths in the United States for the years 1995–2001 (Langlois et al., 2006), nearly half a million children 0–14 years of age had TBI each year during this period. Of this number, 91.6% were treated and released from an ED, 7.8 % were hospitalized, and .6% died. Hospitalization rates in this age range were higher for males than females by a ratio of nearly 2:1. Of those cases for which the external cause of injury was known, falls were a more common cause (39%) than motor vehicle-related events (11%), though these two causes were more equally distributed in children over 4 years of age. Defining severity based on the Glasgow Coma Scale (GCS, Jennett & Teasdale, 1974) score, Kraus (1995) estimated that most children sustain mild TBI (80–90%). Incidence rates are lower for moderate (7%–8%) and severe (5%–8%) TBI, but these more serious injuries are more often associated with long-term disability.

The consequences of TBI in children include physical conditions (e.g., neuromotor impairment, seizures, trauma-related orthopedic injuries), lowered cognitive and academic skills relative to age expectations or preinjury estimates, and problems in school performance, behavior, socialization, and adaptive functioning (Anderson et al., 2006; Ewing-Cobbs et al., 2004b; Stancin et al., 2002; Schwartz et al., 2003; Yeates & Taylor, 1997; Yeates, 2000). Studies of cognitive sequelae document global reductions in ability as measured by IQ testing, as well as impairments in the domains of language, memory, problem-solving, perceptual-motor skills, and attention and executive function (Anderson et al., 2006; Ewing-Cobbs et al., 2004b; Fay et al., 1994; Levin et al., 1995; Taylor et al., 1999; Yeates, 2000). Despite their pervasiveness, investigations of outcomes after TBI in school-age children indicate that effects may be especially pronounced on measures of perceptual and psychomotor speed, attention, planning and mental flexibility, discourse processing, and verbal recall (Anderson & Catroppa, 2005, Bawden et al., 1985; Chadwick et al., 1981; Donders, 2001; Levin & Eisenberg, 1979; Rutter, 1981; Taylor et al., 1999; Tremont et al., 1999; Yeates, 2000). Relative deficits are also observed on tasks with greater demands on self-organization, inferencing, or metacognition, as opposed to those providing external structure or tapping the child’s existing knowledge base (Dennis & Barnes, 2001; Dennis et al., 1996, 2001; Donders & Giroux, 2005; Ewing-Cobbs & Barnes, 2002; Hanten et al., 2002).

More negative outcomes are predicted by more severe TBI and less advantaged family environments (Anderson et al., 2006; Fletcher et al., 1995; Swartz et al., 2003; Taylor et al., 1999; Taylor et al., 2001). Cognitive deficits have been observed in both younger and older children with moderate to severe TBI (Anderson et al., 2004, 2005a, 2006; Taylor et al., 1999). Studies of children with mild TBI have yielded less consistent findings, with some studies demonstrating only transient cognitive deficits if any and others suggesting emerging consequences over time post injury (Anderson et al., 2001; Gronwald et al., 1997; Keenan et al., 2007; Ponsford et al., 1999). Children with more severe TBI show some recovery of cognitive and academic abilities but this occurs primarily during the first year post injury after which there is a slowing of any continued catch-up growth (Anderson et al., 2004; Chadwick et al., 1981; Ewing-Cobbs et al., 2004b; Jaffe et al., 1995; Taylor et al., 2004; Yeates et al., 2004).

Younger age at injury is another predictor of more negative outcomes. Specifically, children aged 2 to 7 years at the time of injury are more susceptible to deficits in expressive language, attention, and academic achievement and show less recovery of IQ compared with children injured at later ages (Anderson et al., 2005b; Barnes et al., 1999; Dennis et al., 1995; Ewing-Cobbs et al., 1997, 2004b; Kaufmann et al., 1993; Verger et al., 2000). Deficits that emerge over time are also reported in children injured at a young age (Anderson et al., 1999, 2000; Ewing-Cobbs et al., 2004a). Postinjury development and expressive language skills are particularly compromised in infants and toddlers (< 3 years) (Anderson et al., 2005b; Ewing-Cobbs et al., 1989). Researchers have speculated that the poorer outcomes in younger children may reflect a greater susceptibility to diffuse brain insult or abnormalities in neurogenesis, or a greater effect of injury on post-injury skill development (Anderson & Moore, 1995; Barnes et al., 1999; Ewing-Cobbs et al., 1997, 2004a, 2004b; Taylor & Alden, 1997; Wetherington & Hooper, 2006).

However, methodological limitations of past studies make it difficult to interpret group differences in the cognitive effects of TBI in young children. Further research is needed to more clearly document injury sequelae and predictors of outcome in this age group. The primary limitation of most of the existing literature is failure to include groups of children without TBI but with other traumatic injuries as a means for estimating injury effects. Most studies of outcomes of early childhood TBI have examined differences between children of varying levels of TBI severity, rather than comparing these children to controls without TBI. The few studies that have included controls without TBI have recruited community samples of uninjured children (Anderson et al ). Evidence that children with TBI have more preinjury developmental problems or are from less advantaged family backgrounds than uninjured children raises questions as to whether group differences were present prior to TBI (Keenan et al., 2007; Goldstrohm & Arffa, 2005; Howard et al., 2005). Additional limitations include the age differences in test procedures and failure to control for background demographic characteristics (Taylor, 2004; Taylor & Alden, 1997). Studies of young children have examined associations of environmental factors with post-TBI outcomes (Anderson et al., 2006), but we are unaware of research exploring potential moderating effects of these factors. Evidence from studies of school-age children indicating that social disadvantage can exacerbate the effects of TBI justifies efforts to examine similar moderating influences in young children (Taylor et al., 1999; Yeates et al., 1997).

The primary objective of the present study was to investigate the effects of TBI during early childhood on post-acute neuropsychological and school readiness skills using methods that rigorously control for non-injury influences on outcomes. To provide an estimate of the effects of TBI that took into account pre-injury risk exposure as well as the experience of hospitalization for injury, children admitted to hospitals for orthopedic injuries but without TBI were recruited as a comparison group. Outcomes were assessed using comprehensive measures of cognitive and early academic skills that were applicable across all or at least a major portion of the 3- to 6-year-old age range. Findings suggesting that children’s self-regulatory or executive functions may be vulnerable to TBI and may play an important role in children’s ongoing development (Blair, 2002; Bronson, 2002; Anderson et al., 2005c; Ewing-Cobbs et al., 2004a) prompted inclusion of several experimental measures of this skill domain. Finally, group comparisons were made controlling for background factors.

We hypothesized deficits in young children hospitalized for TBI would have deficits in cognitive and school readiness skills relative to children with orthopedic injuries only. We anticipated that these deficits would be most pervasive in children with more severe TBI, but that impairments would also be found in children with moderate and mild TBI. Given the failure of many previous studies to investigate a wide range of outcomes in young children with TBI, we anticipated wide-ranging deficits and did not have expectations with respect to which skills would be more or less affected. Based on previous studies, we also explored the possibility that the effects of TBI severity would be exacerbated by less advantaged family environments and younger age at injury.

METHODS

Design

Data were collected as part of a longitudinal investigation of TBI in young children that employed a concurrent cohort/prospective research design. The study compared post-acute neuropsychological and early academic skills in children hospitalized for TBI at varying levels of severity with children hospitalized for orthopedic injuries not accompanied by TBI (OI group). The rationale for recruiting children with OI as a comparison group was to examine the consequences of TBI in relation to outcomes for children who also experienced hospitalization for their injuries and who, by virtue of being at risk for injury, were more likely that uninjured children to be have similar preinjury behavior and family characteristics (Anderson et al., 2004; Goldstrohm & Arffa, 2005). The initial assessment was conducted in the post-acute period and included tests of neuropsychological and early academic skills and data from parents regarding the family environment.

Sample

Children were recruited from consecutive admissions of children with mild to severe TBI or with OI at three tertiary care children's hospitals and a general hospital, all of which had Level 1 trauma centers. The study was approved by the ethics boards of all participating hospitals and informed consent was obtained prior to participation. Eligibility criteria included age at injury between 3 years, 0 months and 6 years, 11 months, no documentation in the medical chart or in parent interview of child abuse as a cause of the injury, and English as the primary spoken language in the home. Eligibility for the TBI group included a TBI due to blunt trauma requiring overnight admission to the hospital and either a GCS score <15 (suggesting altered neurological status) or evidence for TBI-related brain abnormalities from cortical tomography (CT) or magnetic resonance imaging (MRI).

Consistent with previous investigations (Anderson et al., 2006; Fletcher et al., 1990; Yeates et al., 2004), severe TBI was defined as one resulting in a GCS score of 8 or less. Moderate TBI was defined as a GCS score of 9–12 or a higher GCS score with abnormal neuroimaging. A final group of children with mild TBI comprised those participants with GCS scores of 13–14 without neuroimaging abnormalities. To insure that evidence for TBI was based on direct physical examination and not on history alone, children with GCS scores of 15 and normal neuroimaging were not recruited. Inclusion in the OI required a documented bone fracture to an area of the body other than the head that required an overnight hospital stay, and the absence of any evidence of loss of consciousness or other findings suggestive of brain injury (e.g., symptoms of concussion). Children with TBI or OI with previous diagnoses of autism or other developmental disabilities associated with generalized cognitive impairment were also excluded.

A total of 221 children (102 with TBI and 119 with OI) and their caregivers were enrolled in the study. Comparison of participants with children listed in the trauma registries of the participating hospitals who met age and injury severity criteria but were not enrolled indicated that our sample was representative of the eligible cohort in income and race distribution. At least a portion of the test battery was administered to 216 children (98%) at the baseline assessment. The final sample comprised 99 children with TBI (20 severe, 64 moderate, and 15 mild) and 117 with OI. Reasons for failure to test children included injuries that precluded testing (2 with severe TBI) and difficulties in arranging for travel for the assessment (1 with severe TBI and 2 with OI). Untested children did not differ from those assessed in parental marital status, median income, race, or sex.

As shown in Table 1, the groups did not differ in age at assessment, median neighborhood income based on the 2005 Census, distributions of sex, race, and maternal education levels, or parent resources and stressors as measured by the Life Stressors and Social Resources Inventory Adult Version (LISRES-A, Moos & Moss, 1994). Data collected in parent interview also failed to suggest group differences in preinjury developmental status, though did point to more pre-existing problems in boys than in girls. Parents reported special services prior to injury for 2% of girls compared with 14% of boys, χ² (df = 1, N = 213) = 8.61, p = .003. Concerns regarding children’s preinjury development, behavior, or learning were noted for 11% of girls compared with 29% of boys, χ² (df = 1, N = 214) = 9.93, p = .002.

Table 1.

Sample Demographic Characteristics.

	Group
	Severe TBI (n = 20)	Moderate TBI (n = 64)	Mild TBI (n = 15)	OI (n = 117)
Age at assessment in years, M (SD)	4.86 (0.88)	5.19 (1.20)	4.68 (1.00)	5.21 (1.08)
Males n (%)	14 (70%)	37 (58%)	6 (40%)	67 (57%)
Non-white race, n (%)	7 (35%)	22 (44%)	6 (40%)	27 (33%)
Census median family income in dollars, M (SD)	52,767 (16,435)	57,096 (26,539)	51,556 (25,874)	63,888 (23,410)
Maternal education, n (%):
<2 years high school	0 (0%)	3 (5%)	0 (0%)	2 (2%)
2 years high school	5 (25%)	7 (11%)	1 (7%)	6 (5%)
High school degree/GED	10 (50%)	24 (38%)	7 (50%)	45 (38%)
2 years college	4 (20%)	11 (17%)	3 (21%)	23 (20%)
4 years college	1 (5%)	12 (19%)	3 (21%)	29 (25%)
Graduate degree	0 (0%)	6 (10%)	0 (0%)	12 (10%)

Note: TBI = traumatic brain injury; OI = orthopedic injury; SD = standard deviation; GED = General Education Diploma. All group differences non-significant.

Table 2 lists injury and medical characteristics for each of the groups. The time between injury and assessment was shorter for the OI group than for the TBI groups (significant for mild and moderate TBI groups, nonsignificant trend for severe TBI group). This difference was likely related to our willingness to extend recruitment somewhat beyond 3 months post injury (our initial recruitment window) in an effort to enroll as many of the children with TBI as possible. The groups also differed in their mean New Injury Severity Score (NISS, Osler et al., 1997), defined as the sum of the squares of the Abbreviated Injury Scale (AIS) scores for each child’s three most severely injured body regions. Post-hoc tests indicated higher NISS for the severe and moderate TBI groups compared with the mild TBI and OI groups. The groups also differed in mean “non-head-injury” NISS, computed as the NISS minus the AIS for the head region. According to post-hoc tests, the severity of injuries to regions other than the head was lower in the TBI groups than in the OI group. The distribution of causes of injury for the TBI group were consistent with national trends for young children, with a substantial proportion of both the TBI and OI group sustaining injuries due to falls (Langlois et al., 2006). A significant group difference in causes of injury reflects higher rates of transportation-related injuries in TBI groups compared with the OI group.

Table 2.

Injury and Medical Characteristics.

	Group
	Severe TBI (n = 20)	Moderate TBI (n = 64)	Mild TBI (n = 15)	OI (n = 117)
Age at injury in years, mean (SD)	4.74 (0.88)	5.06 (1.20)	4.55 (1.03)	5.11 (1.07)
External cause of injury, n (%):*a
Transportation	12 (60%)	22 (34%)	5 (33%)	10 (9%)
Bicycle crash	0 (0%)	3 (5%)	1 (7%)	7 (6%)
Fall	6 (30%)	33 (52%)	9 (60%)	84 (72%)
Other	2 (10%)	6 (10%)	0 (0%)	16 (14%)
Length of hospital Stay in days, mean (SD)*	6.70 (7.24)	2.89 (1.86)	1.60 (0.63)	1.63 (1.08)
NISS total, mean (SD)*	12.47 (8.57)	15.08 (7.81)	7.40 (5.87)	7.04 (2.66)
NISS non-head- related, mean (SD)*	1.24 (2.11)	2.44 (5.25)	1.60 (2.53)	7.04 (2.66)
Lowest GCS score, mean (SD)*	3.95 (1.79)	13.45 (2.00)	13.60 (0.51)
Neuroimaging abnormalities, n (%):*b
Absent	7/19 (37%)	12/63 (19%)	15/15 (100%)	117/117 (100%)
Mild	2/19 (11%)	14/63 (22%)	0/15 (0%)	0/117(0%)
Moderate	2/19 (11%)	13/63 (21%)	0/15 (0%)	0/117 (0%)
Severe	8/19 (42%)	24/63 (38%)	0/15 (0%)	0/117 (0%)
Coma duration, n (%):*
None	0 (0%)	64 (0%)	15 (100%)	117 (100%)
<24 hours	15 (75%)	0 (0%)	0 (0%)	0 (0%)
≥24 hours	5 (25%)	0 (0%)	0 (0%)	0 (0%)
Time since injury in months, mean (SD)*	1.51 (0.75)	1.51 (0.76)	1.62 (0.82)	1.16 (0.50)

Note: TBI = traumatic brain injury; OI = orthopedic injury; SD = standard deviation; NISS = New Injury Severity Score; GCS = Glasgow Coma Scale.

^*Significant difference between groups at p < .05.

^aInjuries due to “other” causes included those related to sports and recreation, rough-housing, and falling objects.

^bSee text for definition of severity of neuroimaging abnormality. Abnormality was absent in the mild TBI and OI groups by definition.

As anticipated based on the criteria for group assignment, the mean GCS score and duration of coma were higher for the severe TBI group than for the other TBI groups. Radiology reports were unavailable for 2 children with TBI. The group assignments of these children were based on information found elsewhere in the medical record. For the remainder of the TBI sample, these reports were used to classify neuroimaging findings into the categories of no lesion and three grades of lesion severity. Mild abnormalities were defined as a single subdural, subarachnoid or epidural hemorrhage, or a single intraparenchymal lesion, contusion or hemorrhage; moderate abnormalities as multifocal lesions without diffuse abnormality as identified by report of edema, mass effect, swelling, shift, volume loss, or diffuse axonal injury; and severe abnormalities as any diffuse abnormality, with or without focal lesions. These categories were consistent with research relating outcomes of TBI to the presence vs. absence and type of brain lesions (Bowen et al., 1997; Filley et al., 1987; Levin et al., 1992; Levin et al., 1997; Prasad et al., 2001; Williams et al. 1990). By definition, these abnormalities were absent in the mild TBI and OI groups. Selection criteria likely accounted for the high percentage of abnormalities in children with moderate TBI (81%).

Assessment Procedures

The child and family assessment procedures were administered as part of a more comprehensive evaluation of the child and family that also included assessment of family outcomes, ratings of child behavior, and video-taped parent-child interactions. Administered via parent interview, the LISRES-A has satisfactory internal consistency and was used to assess interpersonal supports and stressors experienced by the caregiver in a variety of social domains (e.g., with family members, friends, coworkers). Child tests were administered in a fixed order, with three separate but overlapping test batteries given to children in the age ranges: 3 years, 0 months to 3 years, 5 months; 3 years, 6 months to 5 years, 11 months; and 6 years, 0 months to 6 years, 11 months. Examiners were not aware of group assignment, and cross-site training was undertaken to insure consistency in test administration. Child assessment procedures are described below.

Intelligence

The Differential Ability Scales (DAS, Elliott, 1990) was used to assess global intelligence. The DAS is a battery of cognitive tests for ages 2½ through 17 years. To obtain a measure of general cognitive ability, we administered the core subtests needed to compute the General Conceptual Ability score (GCA). Four subtests contributed to the GCA for children ages 3 years, 0 months to 3 years, 5 months; and 6 subtests for older children. This measure of intelligence was chosen for its excellent psychometric properties and because it has norms extending into the early childhood age range. Internal consistency indexes are .89 or higher across this age range and test-retest reliability is .90 over a 4-week interval. Standard scores for age were used in analysis of these and other measures with published norms.

Language

Tests of language skills included the Pragmatic Judgment subtest of the Comprehensive Assessment of Spoken Language (CASL, Carrow-Woolfold, 2000), and the Verbal Fluency subtest of the NEPSY: A Developmental Neuropsychological Assessment (NESPY, Korkman et al., 1998), both of which were assessed across the three age ranges. Additionally, the Verbal Comprehension and Naming Vocabulary subtests of the DAS were administered to the younger two age groups. Pragmatic Judgment measures social communication skills. Verbal Fluency, which requires the child to generate a list of different types of animals and foods/drinks as quickly as possible, is a measure of mental flexibility or the ability to shift from one conceptual set to another. Verbal Comprehension and Naming Vocabulary are tests of receptive and expressive language, respectively.

Memory

Tests of auditory and visual memory included DAS Recognition of Pictures and Recall of Digits, the NEPSY Sentence Memory subtest, and the Story Recall subtest of the Woodcock Johnson Tests of Achievement, Third Edition (WJ-III, Woodcock et al., 2001). Recall of Digits is a test of verbal working memory that requires the child to repeat back increasingly long strings of digits, first forward and then in reverse. In Sentence Memory, the child is asked to repeat orally presented sentences of increasing length. In Recognition of Pictures, the child is shown a picture of one or more objects for 5 or 10 seconds and is then asked to pick out these objects from a display that includes both target and distractor pictures. In Story Recall, the child is required to retell a series of brief stories.

Spatial Reasoning

Tests of visuospatial skills included DAS Copying Designs, Pattern Construction, and Picture Similarities. The former subtest was appropriate for only the middle age category and the latter two subtests for two of the three age categories. Pattern Construction and Copying Designs involve motor planning and spatial construction using blocks and paper and pencil, respectively. Picture Similarities is a motor-free test of nonverbal reasoning abilities involving identification of pictures and relationships between them.

Executive Function

Tests of executive function included Shape School (Espy, 1997), Delayed Alternation (DA, Espy et al., 1999), and the Delay of Gratification Task (Kochanska et al., 2000). Shape School is a Stroop-like measure of self regulatory abilities in young children. In this task, the child is first taught to name cartoon “pupils” by their shapes or colors. The child is then asked to name the color of each pupil who is “ready for lunch” (those with smiling faces), while inhibiting naming of each pupil who is “not ready” (those with frustrated faces). This test measures the ability to inhibit pre-potent responses and the mental flexibility to switch between color and shape names according to learned contingencies. An efficiency score was computed for each condition (Simple Naming, Inhibition, Switch, and Both inhibition and switching) using the formula number correct divided by completion time. In DA, the child is asked to retrieve a reward (e.g., an M&M or a Cheerio) hidden under one of two cups placed side by side. The contingency is then reversed with the reward hidden under the other cup. The child is not allowed to see where the reward is placed, but can learn to anticipate placement because the placement side is reversed after each correct response. Performance was defined in terms of number of consecutively correct alternations. Delay of Gratification requires the child to inhibit opening an attractive gift, with performance defined in terms of contact vs. no contact with the gift.

School readiness skills

To assess early achievement or academic “readiness” skills, we administered the 6 subtests comprising the School Readiness Composite (SRC) of the Bracken Basic Concept Scale - Revised (Bracken, 1998), as well as the WJ-III Letter/Word Identification, Spelling, and Applied Problems subtests. In addition, DAS Early Number Concepts was administered to the younger two age groups. The SRC assesses content found in most preschool and primary grade curricula (e.g., recognition of colors, letters, numbers, sizes, comparisons and shapes). The assessment is brief and requires only that the child point to the correct response. The WJ-III subtests administered measure letter/word recognition, pencil control and written spelling of letters and words, and knowledge of early math concepts. The latter subtests have test-retest reliabilities ranging from .85 to .96 for young children, with established validity in relation to other achievement tests. Early Number Concepts assesses the child’s ability to count, compare, and solve elementary number problems.

Data Analysis

Prior to analysis, raw scores on the outcome measures were converted to age-standardized scores using published norms. Because published age norms were unavailable for Shape School and DA, age-expected scores on these tests were generated based on regression analysis of data from the OI group. Age-adjusted z scores were then computed for each measure by dividing the differences between each observed and age-predicted score by the standard error of the estimate. Potential influences of extreme scores on results were limited by truncating, or windsorizing, standard scores to within 3 standard deviations of the mean score (Tabachnick & Fidell, 1989). Examination of the scores revealed normal distributions for all continuous measures with acceptable levels of skewness and kurtosis.

Group comparisons on the continuous measures of outcome were made using analysis of covariance (ANCOVA). Group effects were defined by preplanned contrasts of each TBI group with the OI group. Covariates included a measure of socioeconomic status (SES), sex, and race (white/non-white) as justified by evidence for associations of these factors with cognitive abilities in children (McDermott, 1995). Covariate-adjusted logistic analysis was used to examine group differences in odds of contact vs. no contact on the Delayed Gratification Task, with age at assessment entered as an additional covariate in analysis of this measure. To assess the dose-response relationship of TBI presence and severity with outcome, a further set of secondary analyses examined the linear trend between degree of TBI (none or 0 = OI; 1 = mild TBI, 2 = moderate TBI, and 3= severe TBI) and covariate-adjusted scores.

Preliminary analysis revealed that primary caregiver education level and median census tract income were positively correlated with each other and with most neuropsychological outcomes. SES was thus defined as the mean of the sample z-scores for these two variables. Additional sociodemographic variables (e.g., parent marital status and occupation) were examined but were excluded after initial analyses failed to reveal associations with outcomes independent of parent education and census income. Time since injury was not considered as analysis failed to reveal associations with the performance of the TBI sample.

Regression analysis was used to examine moderating effects of SES, LISRES-A stressors and resources scores, and age at injury on the group differences. To test for moderation, each of these factors was entered separately into a regression along with the TBI-OI group contrast terms and the interaction of each contrast with that factor. Models that included the covariates were used to subsequently examine the moderating effects of LISRES-A resources and stressors and age at injury. Moderating effects of race and sex were not examined due to small cell sizes for the severe and mild TBI groups. Logistic regression was conducted to investigate moderating effects on the Delayed Gratification Task, with age at assessment again included as an additional covariate.

Classification of children into mild, moderate, and severe TBI groups was based on traditional distinctions but did not permit empirical examination of the combined effects of different dimensions of TBI severity. To further investigate the relation of injury severity to outcomes for the children with TBI, we conducted regression analyses of data from the three TBI groups combined. Severity measures included the GCS score, coma duration (none, <24 hours, ≥24 hours), NISS, and the presence and severity of neuroimaging abnormality as defined above. The combined effects of these indices on outcomes, the GCS score, NISS, and neuroimaging abnormality were examined by entering these factors into a hierarchical regression following entry of SES, race, and sex. Because the GCS score was closely related to coma duration (r = .85, p < .001), the latter variable was not included in these regressions. The NISS and neuroimaging abnormality were also correlated (r = .60, p < .001) but analysis did not indicate high levels of collinearity.

Sample size for each measure varied due to a restricted age range for some of the tests and the inability of some children to complete some of the tests. An alpha of .05 for all comparisons was justified by our interest in examining the effects of TBI on specific measures rather than testing a study-wide hypothesis (Brandt, 2007). Effect sizes provided an indication of the magnitude of group differences.

RESULTS

Group Differences

Table 3 summarizes group performance on the tests by domain. Significant group contrasts from the ANCOVAs are also reported. As shown in the table, the severe TBI group had lower scores than the OI group on the GCA, contrast estimate (CE) (standard error [se]) = −10.58 (3.25), p = .001, Sentence Memory, CE = −1.62 (0.68), p = .018, Story Recall, CE = −8.75 (3.60), p = .016; Recognition of Pictures, CE = −6.22 (2.62), p = .018, Recall of Digits, CE = −6.40 (2.51), p = .012, Pattern Construction, CE = −9.46 (2.35), p < .001, Copying Designs, CE = −5.52 (2.22), p = .014, Early Number Concepts, CE = −6.63 (2.31), p = .005, and SRC, CE = −11.40 (3.36), p = .001. The moderate TBI group had lower scores than the OI group on Pragmatic Judgment, CE = −4.62 (2.03), p = .024, Sentence Memory, CE = −1.19 (0.44), p = .007, Recall of Digits, CE = −4.55 (1.66), p = .007, Shape School Both condition, CE = −0.37 (0.18), p =.044, and SRC, CE = −4.56 (2.14), p = .034. The only test on which the mild TBI group had lower scores than the OI group was Pragmatic Judgment, CE = −4.62 (2.03), p =.037. It is unclear why adverse effects on Pragmatic Judgment were found for children with moderate and mild TBI but not for those with severe TBI. However, the ordering or group means for most measures were in the expected direction.

Table 3.

Neuropsychological Variables

	Group
Raw mean (SD), effect size for Contrast with OI groupa	Severe TBI (n = 20)		Moderate TBI (n = 64)		Mild TBI (n = 15)		OI (n = 117)
General ability: GCAb*	85.00 (15.63)	0.68	97.37(16.49)	0.10	95.36(11.78)	0.19	101.78 (15.11)
Language:
Pragmatic Judgmentb†‡	101.00 (13.59)	0.01	99.16(14.75)	0.34	97.29(14.49)	0.56	105.10 (12.95)
Verbal Fluencyd	7.89 (2.02)	0.29	8.42 (2.93)	0.21	7.67 (2.74)	0.53	9.21 (3.00)
Verbal Comprehensionc	38.53 (10.03)	0.34	43.05 (9.88)	0.11	42.92 (7.61)	0.09	45.87 (10.86)
Naming Vocabularyc	47.35 (12.17)	0.22	50.26(11.07)	0.06	50.77 (8.60)	0.04	52.45 (10.16)
Memory:
Recognition of Picturesc*	42.05 (10.73)	0.56	49.26(12.22)	0.04	45.93 (9.75)	0.34	51.09 (10.70)
Recall of Digitsc*†	44.45 (11.88)	0.61	47.75(10.23)	0.43	47.46 (6.91)	0.50	53.07 (10.69)
Story Recallb*	101.93 (14.60)	0.65	108.72(13.62)	0.26	108.00(13.06)	0.38	113.18 (13.29)
Sentence Memoryd*†	7.95 (3.08)	0.59	8.74 (2.01)	0.43	9.50 (1.51)	0.17	10.20 (3.12)
Spatial reasoning:
Pattern Constructionc*	41.78 (10.37)	0.97	53.76 (9.81)	0.07	52.50 (7.57)	0.11	54.47 (9.77)
Copying Designsc*	39.31 (5.45)	0.63	48.27(10.16)	0.15	50.60 (8.76)	0.37	48.08 (8.59)
Picture Similaritiesc	42.41 (11.81)	0.47	49.19(11.81)	0.01	50.08(11.26)	0.10	50.61 (11.47)
Executive Function:
Gift Delay, no contact, n (%)	8 (44%)	----	23 (38%)	----	7 (50%)	----	32 (29%)
DA, consecutive alternationse	−0.06 (0.90)	0.01	0.11 (1.02)	0.14	0.08 (1.16)	0.13	0.00 (1.00)
Shape School efficiencye:
Simple Naming condition	−0.68 (0.84)	0.51	−0.10 (0.85)	0.02	−0.55 (1.12)	0.53	(0.99)
Inhibit condition	−0.46 (0.91)	0.19	−0.25 (0.94)	0.12	−0.30 (0.86)	0.23	−0.01 (0.97)
Switch condition*	−0.76 (0.68)	0.61	−0.44 (0.88)	0.35	−0.61 (0.65)	0.62	−0.02 (0.97)
Both condition*†	−0.82 (0.71)	0.76	−0.47 (0.83)	0.40	−0.62 (0.80)	0.57	−0.00 (0.99)
School Readiness Skills:
Letter/Word Identificationb	101.21 (13.55)	0.06	103.30(16.18)	0.07	108.43(17.98)	0.15	105.64 (16.74)
Spellingb	93.58 (10.90)	0.29	98.10(13.21)	0.06	101.00(11.28)	0.07	99.46 (13.47)
Applied Problemsb	94.10 (13.25)	0.46	102.47(13.55)	0.01	106.70(13.57)	0.32	103.81 (13.71)
Early Number Conceptsc*	38.59 (7.30)	0.70	46.66(10.29)	0.02	44.82 (6.76)	0.27	48.38 (9.92)
SRCb*†	89.20 (15.20)	0.73	99.34(17.47)	0.29	103.93(14.88)	0.03	106.50 (14.80)

Note: TBI = traumatic brain injury; OI = orthopedic injury; SD = standard deviation; GCA: Differential Ability Scales General Conceptual Composite; SRC = Bracken Basic Concept Scale School Readiness Composite.

^aEffect sizes defined by Cohen’s d: difference between estimated (covariate-adjusted) means/estimate of pooled within-group SD.

^bstandard score

^cT score

^dscaled score

^eage-standardized z score

^*Significant difference, severe TBI vs. OI (p<.05).

^†Significant difference, moderate TBI vs. OI (p<.05).

^‡Significant difference, mild TBI vs. OI (p<.05).

Many of the effect sizes for the severe TBI-OI group contrasts were of medium magnitude (Cohen’s d = .5-.8), whereas most effects for the moderate TBI-OI group contrasts were small (Cohen’s d = .2-.5). Effect sizes for several of the mild TBI-OI group contrasts were in the medium range and in the hypothesized direction despite a lack of statistical significance, with nonsignificant trends (p < .1) for Verbal Fluency, Recall of Digits, and the Shape School Simple Naming and Switch conditions.

Results from the multiple regressions also revealed that all three covariates accounted from unique variance in at least some outcomes. Higher SES was associated with better performance on all tests except for the Gift Delay, DA, and the Shape School Both condition. Whites scored higher than nonwhites on Verbal Comprehension, Naming Vocabulary, and the Shape School Both condition. Females outperformed males on all the GCA, Pragmatic Judgment, Verbal Fluency, Verbal Comprehension, all four memory tests, Pattern Construction, Shape School Simple Naming and Inhibition conditions, Letter/Word Identification, and Spelling.

In regressions examining the effect of group along a continuum from OI to severe TBI, significant linear trend effects were found for GCA, unstandardized beta (B) (se) = −2.08 (0.84), p = .014, Sentence Memory, B = −0.57 (0.18), p = .001, Story Recall, B = −2.25 (0.87), p = .011, Recall of Digits, B = −2.21 (0.66), p = .011, Pattern Construction, B = −1.38 (0.63), p = .029, Shape School Switch condition, B = −0.17 (0.07), p = .015, Shape School Both condition, B = −0.21 (0.07), p = .004, Early Number Concepts, B = −1.36 (0.66), p = .039, and SRC, B = −2.99 (0.86), p = .011. For each of these outcomes, scores decreased with increasing TBI severity.

Indices of TBI severity as Predictors of Outcomes within the TBI Sample

Controlling for the covariates and considering only children with TBI, a lower GCS score predicted worse performance on the GCA, B = 0.99 (0.32), p = .002, Applied Problems, B = 0.79 (0.30), p = .011, Pattern Construction, B = 0.92 (0.23), p = < .001, Copying, B = 0.80 (0.22), p = .001, Picture Similarities, B = 0.72 (0.28), p = .012, Early Number Concepts, B = 0.67 (0.21), p = .003, and SRC, B = 0.69 (0.34), p = .046. Increased coma duration predicted lower scores on the GCA, B = −6.75 (2.53), p = .009, Pattern Construction, B = −7.38 (1.99), p < .001, Copying, B = −5.46 (1.68), p = .002, Picture Similarities, B = −4.49 (2.17), p = .042, Applied Problems, B = −5.47 (2.49), p = .031, and Early Number Concepts, B = −4.52 (1.64), p = .008. Higher NISS predicted lower scores on Letter/Word Identification, B = −0.48 (0.22), p = .031, DA, B = −0.03 (0.01), p = .019, and SRC, B = −0.42 (0.20), p = .034. Greater severity of neuroimaging abnormalities also predicted lower scores on DA, B = −0.20 (0.09), p = .025. The GCS score, NISS, and neuroimaging abnormality when entered after the covariates accounted additional variance in the GCA, R² change = .10, F (3, 85) = 4.57, p = .005, Recall of Pictures, R² change = .09, F (3, 84) = 10.45, p < .025, Pattern Construction, R² change = .22, F (3, 69) = 8.78, p < .001, Copying, R² change = .16, F (3, 50) = 4.97, p = .004, Applied Problems, R² change = .10, F (3, 81) = 4.14, p = .009, Early Number Concepts, R² change = .15, F (3, 51) = 4.91, p = .004, and SRC, R² change = .08, F (3, 87) = 3.92, p = .011.

Moderators of TBI Effects

Moderating effects included SES x severe TBI-OI group contrast interactions for Spelling and Naming Vocabulary. Follow-up analyses using mixed model analysis revealed that lower SES was associated with less adverse effects of severe TBI on Spelling, B = −10.35 (4.71), p = .029, but with more adverse effects on Naming Vocabulary, B = 9.46 (3.39), p = .006. Further follow-up of these interactions confirmed a adverse effects of severe TBI on Naming Vocabulary for children with low SES (defined as 1 SD below mean SES for the total sample) (p = .016) and on Spelling for children with high SES (defined as 1 SD above the sample mean) (p = .014).

Group differences in some outcomes were also moderated by LISRES-A stressors. These effects included a stressors x severe TBI-OI group contrast for Pattern Construction, B = 0.66 (0.28), p = .021, and stressors x moderate TBI-OI group contrast interactions for Spelling, B = −0.78 (0.29), p = .008, and SRC, B = −0.72 (0.30), p = .018. According to results from follow-up analyses, the source of the moderating effect on Pattern Construction was an association of increasing stressors with lesser effects of severe TBI. The negative consequences of severe TBI on Pattern Construction were nevertheless evident at both lower and higher levels of stressors. In contrast, increasing stressors were associated with more negative consequences of moderate TBI on Spelling and SRC. Further follow-up analyses these effects indicated adverse effect of moderate TBI for children at high stressors (1 SD below the sample mean) for both Spelling (p = .022) and SRC (p = .002).

The only moderating effects of age at injury were on the mild TBI-OI group contrasts for Pragmatic Judgment, B = 7.83 (3.56), p = .029, and Letter/Word Identification, B = −10.63 (4.36), p = .016. Follow-up analyses of these interactions revealed that younger age at injury predicted lower scores for the mild TBI group on Pragmatic Judgment, while older age at injury predicted lower scores on Letter/Word Identification. Similar but less dramatic age-at-injury effects were evident for the other TBI groups. In contrast, age at injury was unrelated to scores on these tests for the OI group. Follow-up tests revealed an adverse effect of mild TBI on Pragmatic Judgment only in younger children (defined as 1 SD below the sample mean age) (p = .005), and a trend for an adverse effect on Letter/Word Identification only in older children (defined as 1 SD above the sample mean) (p = .09).

DISCUSSION

Effects of Severe, Moderate, and Mild TBI in Young Children

Consistent with findings from past research on young children, the participant’s in this study with either severe or moderate TBI performed more poorly than the OI group on a wide range of neuropsychological and achievement tests (Anderson et al., 1997, 2006; Anderson & Catroppa, 2005; Ewing-Cobbs et al., 1989, 1997, 2004a, 2004b; Morse et al., 1999). Compared with the OI group, children with severe TBI had poorer general cognitive ability as measured by the GCA and lower scores on tests of memory, spatial reasoning, executive function, and school readiness skills. Although group comparisons failed to indicate deficits on some verbal tests, a negative effect of severe TBI on Naming Vocabulary was observed in children with low SES. The moderate TBI group performed more poorly than the OI group on tests of pragmatic language, memory, executive function, and school readiness concepts, but not in general ability. The effect sizes for the latter differences were generally smaller than those for the severe TBI vs. OI group contrasts. Similar to previous findings, moderate TBI in young children thus had more selective and less pronounced consequences than severe TBI (Anderson et al., 1997, 2005; Ewing-Cobbs et al., 1997; Goldstrohm & Arffa, 2005).

The only measure on which the mild TBI group performed significantly less well than the OI group was Pragmatic Judgment. However, this difference yielded a medium effect size, as did differences between the mild TBI and OI groups on Verbal Fluency, Recall of Digits, and the Shape School Simple Naming condition. In view of the small size of the mild TBI group, these results are interpreted as at least tentative support for post-acute neurocognitive effects of these injuries. Previous studies have yielded inconsistent effects of mild TBI, with some studies demonstrating adverse consequences for cognition or achievement (Anderson et al., 2001; Dennis & Barnes, 2001; Gronwall et al., 1997) and others not (Anderson et al., 2005a; 2006). Results may vary depending on when after injury children are assessed and criteria used to define mild TBI (Satz et al., 1997). We recruited children who were hospitalized for at least 1 day and had some impairment in consciousness as defined by a GCS score of 13 or 14. For this reason, our mild TBI group may have had more significant trauma than children discharged from out-patient emergency departments or with GCS scores of 15.

Additional Evidence for Effects of Injury Severity

As further evidence for a relationship between TBI severity and outcomes, lower test scores on many of the same tests that discriminated the severe TBI and OI groups was linearly related to the extent of TBI, from none in the OI group to mild, moderate, and severe TBI. This approach to analysis demonstrated that TBI severity was associated with the Switch and Both conditions of Shape School, offering further evidence for negative consequences of TBI on executive function in young children. In analyses of data from the aggregate TBI sample, decreasing GCS score, increasing coma duration, increasing injury severity as measured by the NISS, and increasing degree of neuroimaging abnormality were all associated with lower scores on many of the same tests that discriminated the TBI and OI groups. Both increasing injury severity as measured by the NISS and increasing severe neuroimaging abnormality were associated with lower scores on DA, adding again to evidence for effects on TBI on measures of early developing executive functions. The results of these analyses are consistent with past findings indicating that multiple measures of TBI severity are useful in predicting outcomes of TBI (Bowen et al., 1997; Foreman et al., 2007; Levin, 1995; Levin & Eisenberg, 1979; Prasad et al., 2002).

Non-injury Factors Related to Outcomes of TBI

For the total sample, the covariates SES, race, and sex was each independently associated with some test scores. Consistent with previous studies of young children, higher SES was positively correlated with performance on most of the tests (Anderson et al., 2006). With SES in the regression models, white race predicted higher scores on only three of the measures. The finding of higher scores in girls than boys for the majority of the measures was unexpected given the general absence of sex differences in on our previous research on a school-age TBI cohort (Yeates et al., 2002). The greater prevalence of developmental disorders among males (Thompson et al., 2003) may help to explain this finding, but only if one assumes that associated cognitive dysfunctions are more pronounced in younger children. Alternatively, a male predilection for preinjury developmental problems may have been exacerbated in our sample if these disorders contributed more to risk of injury in boys than in girls. While we are not aware of any relevant data from previous studies, higher rates of pre-existing developmental problems in males is consistent with either interpretation.

The findings also documented moderating influences of non-injury factors on some of the group differences. In keeping with results of our previous study of older children with TBI (Taylor et al., 1995, 2002; Yeates et al., 1997), weaknesses in Naming Vocabulary in the severe TBI group were evident only at lower levels of SES, and weaknesses in Spelling and the SRC in the moderate TBI group were found only at higher levels of parent stressors. At least for measures of verbal and early academic skills, the sequelae of TBI appeared to be exacerbated by environment disadvantage. Possible explanations for these findings are that these skills were not as well established in children from less advantaged backgrounds and thus more easily disrupted by TBI, or that fewer resources were available after injury to support postinjury recovery (Taylor et al., 2002). The fact that all three measures involve knowledge-based skills is consistent with these interpretations. Contrary to expectations, the adverse effects of severe TBI on Pattern Construction and Spelling were more evident in children with higher SES. Similar moderating effects have been found in studies of cognitive outcomes of very low birth weight (Taylor et al., 2006) and can be explained in two ways: either children with severe TBI are less able to benefit from environmental supports than children without TBI, or the effects of severe TBI are dampened or obscured by social disadvantage.

Findings revealed little support for our expectation that outcomes of TBI would be worse for children injured at younger ages. An adverse effect of mild TBI on Pragmatic Judgment was found only in the children injured at younger ages. However, this finding is difficult to explain in isolation and may be an artifact of the lower mean age at injury of the mild TBI group. Such an age disparity could arise because younger children are generally more difficult to engage in testing (Anderson et al., ) or because the expressive language skills tapped by Pragmatic Judgment may only be emerging these children. Analysis also revealed that age at injury moderated differences between the mild TBI and OI groups on Letter/Word Identification. For this measure, the adverse consequences of mild TBI were more evident with increasing age at injury. Because the initial items on this test require only that the child point to a named symbol whereas the later items require name retrieval, this finding may be another artifact of the group differences in age at injury. The lack of evidence for age at injury effects contrasts with results from previous studies that have examined TBI across wider age spans (Anderson et al., 2005; Barnes et al., 1999; Dennis et al., 1995; Ewing-Cobbs & Barnes, 2002; Levin et al., 1995; Verger et al., 2000), suggesting that age at injury may be less consequential in children injured before 7 years of age (Ewing-Cobbs et al., 1997). Alternatively, the impact of younger age at injury may become more pronounced with increasing time since injury (Anderson et al., 2000; Ewing-Cobbs et al., 2004).

Methodological Strengths and Limitations

To our knowledge, this study is among the first to demonstrate effects of TBI in a hospitalized sample of young children using a comparison group of children hospitalized for other traumatic injuries. Goldstrohm and Arffa (2005) conducted a similar study involving young children with mild to moderate TBI. In their study, the TBI group was compared with a group of children hospitalized for injuries to other body regions, as well as with a group of non-injured controls. Similar to our findings, the TBI group had lower scores than the other-injury group on the SRC and on tests of verbal memory. Although the authors found many additional deficits in the TBI group relative to non-injured controls, the latter group had more preinjury behavior and developmental problems. Although Goldstrohm and Arffa’s other-injury group was not limited to traumatic orthopedic injuries and they did not include children with severe TBI, their study underscores the importance of an appropriate comparison group. Other previous studies have compared young children with TBI with uninjured community controls or have contrasted moderate to severe TBI with mild TBI. The benefit of recruiting children with other injuries is that this permits comparison of children with vs. without TBI while helping to insure similarity in preinjury child characteristics. Other strengths include adjustment for demographic characteristics in analysis of group differences and TB severity, administration of an extensive battery of tests most of which were appropriate across the 3- to 6-year-old age range, and a sample of children with TBI who were representative of admissions to the participating hospitals.

One major study limitation is that the only clinical neuroimaging, in most instances CT scans, were available to determine the presence and nature of the brain insults in the TBI sample. Imaging methods with greater sensitivity to white matter damage and focal lesions would likely have provided information useful in enhancing prediction of variations in outcomes of TBI (Scheibel & Levin, 1997; Wilde et al., 2006). A further limitation is that the mild and severe TBI groups were relatively small despite recruitment from multiple hospitals. Larger group sizes would have increased statistical power for detection of moderating effects of environmental variables and age at injury. Analysis of individual tests rather than ability constructs is another potential weakness. Although this approach was justified by the lack of established methods for assessing ability constructs in this age range, the use of multiple-indicator latent constructs would have allowed us to more clearly distinguish the skills most and least affected by TBI and reduced the probability of Type I error.

Conclusion

Study findings suggest several conclusions with respect to the post-acute effects of TBI on young children. While these conclusions are tentative given the exploratory nature of the study, they provide a basis for formulating more specific hypotheses about how TBI affects young children:

1. Severe TBI sustained during early childhood can result in generalized cognitive impairment and deficits in school readiness skills. Furthermore, memory, spatial reasoning, and executive function may be more affected than some language skills.

2. Children with moderate TBI have more specific impairments than those with severe TBI, and similar impairments may be present in children with mild TBI.

3. Specific cognitive measures are sensitive to injury consequences. Based on effect sizes, the DAS GCA, DAS Pattern Construction, Copying Designs, Recall of Pictures, and Recall of Digits subtests, NEPSY Sentence Memory subtest, WJ-III Story Recall subtest, and Shape School Switch and Both conditions are useful in detecting cognitive impairment in children with severe TBI. The memory measures and Shape School are also sensitive to less severe TBI. In light of the lack of group differences on the DA and Delay of Gratification Task, these tasks are less likely to be useful in examining injury effects.

4. Indices of TBI severity, including the GCS score, NISS, coma duration, and neuroimagining abnormality, are related to cognitive and school readiness skills.

5. Age at injury is not a critical factor in predicting outcomes of TBI sustained during the early childhood years.

6. Although measures of environmental disadvantage predict lower scores on most tests in children with and without TBI, these factors may amplify the effects of TBI on some tests but dampen or obscure effects on other measures.

The major clinical implication of these findings is that young children with TBI are at risk for post-acute deficits in cognition and school readiness skills and that measures used in this study, or similar ones, are sensitive to these deficits. The results also encourage examination of both injury-related and environmental factors as determinants of injury consequences. We can only speculate about the reasons for environmental moderation of injury effects, but they underscore the complexity of influences on recovery (Taylor, 2004). Additional research is needed to better understand the nature of the effects mild to severe TBI in young children on brain development and how brain pathology maps onto neurobehavioral outcomes (Taylor, 2004). Larger scale studies of outcome will also be required, especially in examining the consequences of mild TBI. Further research on this subset of children might explore factors contributing to the decision to hospitalize these children and track acute changes in cognitive functioning. Based on the presumption that mild TBI results in residual deficits in only a minority of children with mild TBI (Bigler, 2007; Kirkwood et al., in press), it may be more helpful to find ways to distinguish affected from unaffected individuals than to demonstrate group differences. The development of tests that clarify the effects of TBI at all levels of severity and the study of the factors that contribute to skill development following injury will be essential in improving our capacity to identify and treat the sequelae of TBI in young children. We are currently following the present cohort to determine the nature of changes in cognitive and achievement outcomes over time post injury, evaluate behavioral sequelae, and investigate the effects of family factors on outcomes.