A Prospective Study of Short- and Long-Term Outcomes After Traumatic Brain Injury in Children: Behavior and Achievement

Neuropsychology
Copyright 2002 by the American Psychological Association, Inc.
2002, Vol. 16, No. 1, 15–27
0894-4105/02/$5.00
DOI: 10.1037//0894-4105.16.1.15
A Prospective Study of Short- and Long-Term Outcomes After
Traumatic Brain Injury in Children: Behavior and Achievement
H. Gerry Taylor
Keith Owen Yeates
Case Western Reserve University and
The Ohio State University and
Rainbow Babies & Children’s Hospital
Columbus Children’s Hospital
Shari L. Wade
Dennis Drotar
University of Cincinnati and
Case Western Reserve University and
Children’s Hospital Medical Center
Rainbow Babies & Children’s Hospital
Terry Stancin
Nori Minich
Case Western Reserve University and
Case Western Reserve University and
MetroHealth Medical Center
Rainbow Babies & Children’s Hospital
Longitudinal behavior and achievement outcomes of traumatic brain injury (TBI) were investi-
gated in 53 children with severe TBI, 56 children with moderate TBI, and 80 children with
orthopedic injuries not involving brain insult. Measures of preinjury child and family status and
of postinjury achievement skills were administered shortly after injury. Assessments were re-
peated 3 times across a mean follow-up interval of 4 years. Results from mixed model analysis
revealed persisting sequelae of TBI. Recovery of math skills was observed in the severe TBI
group but only for children from less stressed families. Social disadvantage in children with TBI
predicted more adverse behavioral sequelae and less favorable changes in some outcome mea-
sures. The ?ndings suggest that pediatric TBI has long-term effects on behavior and achievement
but that postinjury progress is in?uenced by the family environment.
Children who sustain severe traumatic brain injury (TBI)
ing, and educational performance (Ewing-Cobbs et al.,
are at high risk for problems in behavior, adaptive function-
1997; Fletcher, Ewing-Cobbs, Miner, Levin, & Eisenberg,
1990; Max, Koele, et al., 1998; Oddy, 1993; Perrott, Taylor,
& Montes, 1991; Rutter, Chadwick, & Shaffer, 1983).
Fletcher et al. (1996) documented postinjury behavior prob-
H. Gerry Taylor, Dennis Drotar, and Nori Minich, Department
lems in approximately 30% of a group of children with
of Pediatrics, Case Western Reserve University, and Department
severe TBI who were screened for preexisting neuropsychi-
of Pediatrics, Rainbow Babies & Children’s Hospital, Cleveland,
atric disorders. The children were assessed 3– 60 months
Ohio; Keith Owen Yeates, Department of Pediatrics, The Ohio
after injury. Other studies have observed even higher rates
State University, and Department of Pediatrics, Columbus Chil-
of postinjury behavior problems (Brown, Chadwick, Shaf-
dren’s Hospital, Columbus, Ohio; Shari L. Wade, College of
Medicine, University of Cincinnati, and Department of Physical
fer, Rutter, & Traub, 1981; Gerring et al., 1998; Kinsella,
Medicine and Rehabilitation, Children’s Hospital Medical Center,
Ong, Murtagh, Prior, & Sawyer, 1999; Max, Castillo, et al.,
Cincinnati, Ohio; Terry Stancin, Department of Pediatrics, Case
1998; Max, Robin, et al., 1997). Adverse effects on behav-
Western Reserve University, and Department of Pediatrics, Metro-
ior and achievement have been found in association with
Health Medical Center, Cleveland, Ohio.
both moderate and severe TBI, although sequelae are most
This work was supported by National Institute of Neurological
prominent in the latter group (Fay et al., 1994; Jaffe, Polis-
Diseases and Stroke Grant NS36335 from the National Institutes
sar, Fay, & Liao, 1995; Taylor, Yeates, Wade, Drotar,
of Health and by Grant MCJ-390611 from the Maternal and Child
Klein, & Stancin, 1999).
Health Bureau (Title V, Social Security Act, Health Resources and
Findings from longer term follow-ups of children with
Services Administration, Department of Health and Human Ser-
TBI suggest that postinjury behavior and scholastic prob-
vices). We acknowledge the contributions of Elizabeth Shaver,
Nichole Wood, Barbara Shapero, Madeline Polonia, and Matt
lems fail to resolve over time, despite at least partial recov-
Diamond in data collection and coding and of Mark Schluchter for
ery in cognitive functions (Fay et al., 1994; Fletcher et al.,
consultation regarding data analysis. We also acknowledge the
1990; Jaffe et al., 1995; Kinsella et al., 1999; Klonoff,
participation of the Children’s Hospital Medical Center of Akron,
Clark, & Klonoff, 1993; Perrott et al., 1991; Rutter et al.,
Akron, Ohio, and the collaboration of George Thompson, G. Dean
1983). We know little, however, about changes in sequelae
Timmons, and Dennis Weiner. Assistance with recruitment was
after injury. In one of the longest follow-up studies, Jaffe et
provided by the Rainbow Pediatric Trauma Center, Rainbow Ba-
al. (1995) documented sustained behavior and achievement
bies & Children’s Hospital, Columbus Children’s Hospital Trauma
sequelae over a 3-year postinjury interval. Injury sequelae
Program, and MetroHealth Center Trauma Registry.
were associated with TBI severity and were evident in
Correspondence concerning this article should be addressed to
matched comparisons of children with TBI with uninjured
H. Gerry Taylor, Department of Pediatrics, Rainbow Babies &
Children’s Hospital, 11100 Euclid Avenue, Cleveland, Ohio
controls. Tests of linear changes in behavioral outcomes
44106-6038. E-mail: hgt2@po.cwru.edu
from 1 to 3 years postinjury revealed continued effects of
15

---

16
TAYLOR ET AL.
TBI throughout the follow-up period, with no evidence for
from this group also permitted control for either practice
either recovery or worsening of these sequelae over time.
effects or changes due to repeated exposure to the outcome
This same pattern of persistent behavioral and academic
assessments. Division of follow-up into short-term and
sequelae has been observed in other follow-up investiga-
long-term phases was justi?ed by evidence suggesting that
tions. Ewing-Cobbs, Fletcher, Levin, Iovino, and Miner
children with TBI may show initial recovery of function
(1998) compared achievement skills in children who had
over the ?rst 1–2 years after injury, followed by stable or
severe versus mild–moderate TBI at follow-ups con-
even worsening outcomes (Brown et al., 1981; Jaffe et al.,
ducted 6, 12, and 24 months after a baseline evaluation.
1995; Kinsella et al., 1999).
Although the children with more severe TBI had lower
On the basis of previous literature, we anticipated that
achievement scores across the follow-ups, repeated mea-
recovery would be more evident on achievement tests than
sures analysis failed to reveal changes in academic sequelae
on measures of behavior or classroom performance (Jaffe et
over time. Similar results were obtained in a study of
al., 1995; Perrott et al., 1991; Rutter et al., 1983). However,
outcomes over the 1st postinjury year in children with
we had no basis for predicting other measure-speci?c vari-
moderate to severe TBI (Taylor et al., 1999). Compared
with the children who had orthopedic injuries only, the
ations either in sequelae or in susceptibility to environmen-
children who had TBI had poorer behavioral outcomes and
tal in?uences.
writing skills. However, multivariate analysis of variance
Speci?c hypotheses are listed below:
(MANOVA) failed to reveal changes in these sequelae
1. Moderate to severe TBI leads to more behavior prob-
across 6- and 12-month postbaseline follow-ups.
lems and lower adaptive functioning and scholastic
Because outcomes vary dramatically within samples of
competence than do orthopedic injuries not involving
children with TBI, patterns of recovery may be dif?cult
either to determine on the basis of group comparisons alone
insult to the central nervous system.
or to determine without using analytic methods sensitive to
2. Children with TBI show some recovery of function
longitudinal change (Fletcher, Ewing-Cobbs, Francis, &
(i.e., improvement over time relative to children with
Levin, 1995; Taylor & Alden, 1997). Initial improvements
orthopedic injuries only), but little recovery is evident
in behavior and academic skills may be most prominent in
after the 1st postinjury year.
children with severe TBI (Barry, Taylor, Klein, & Yeates,
3. More severe TBI and less advantaged family environ-
1996; Jaffe et al., 1995; Max, Robin, et al., 1997; Thompson
ments are associated with more pronounced behav-
et al., 1994). Persisting or even worsening functional de?-
ioral and academic sequelae and with less favorable
cits later in follow-up may also be characteristic of these
changes over time.
children (Ewing-Cobbs et al., 1998; Fay et al., 1994; Rutter
et al., 1983). The family environment is another factor to
Method
consider in predicting recovery from TBI, especially given
the more marked behavioral and adaptive sequelae found in
Overview of Design and Procedures
children from more dysfunctional or disadvantaged back-
grounds (Brown et al., 1981; Gerring et al., 1998; Kinsella
In keeping with a concurrent cohort-prospective design, groups
et al., 1999; Max, Castillo, et al., 1997; Max, Robin, et al.,
of children with moderate to severe TBI or with orthopedic injuries
1997; Rivara et al., 1993, 1994; Taylor et al., 1999; Yeates
not involving brain insult (ORTHO) were recruited at the time of
hospitalization for their injuries. Children and their families were
et al., 1997).
invited to participate only after the children were medically stable
The major objectives of this study were to investigate
and the parents gave informed consent. Parent measures of the
short- and long-term changes in behavioral and academic
child’s preinjury behavior and adaptive functioning were obtained
sequelae for children with moderate to severe TBI and to
shortly after recruitment and at a baseline assessment conducted an
determine the relationship of TBI severity and the family
average of 3 weeks after injury. The baseline assessment also
environment to patterns of change. Previous reports have
included collection of family data, child testing, and ratings of
described child outcomes for this sample across the 1st
child behavior and school performance from teachers. Teachers
postinjury year (Taylor et al. 1999; Yeates et al., 1997). In
were asked to complete these forms on the basis of the child’s
the present study, we followed the sample to a mean of 4
preinjury status. A neuropsychological battery was also adminis-
years after injury and examined changes both during the
tered, results from which are presented in a separate report (Yeates
initial postinjury year and across the later, extended fol-
et al., 2001). To avoid testing children with posttraumatic amnesia,
low-up interval. As recommended in assessing the sequelae
we undertook baseline testing only after children with TBI scored
of TBI (Fletcher et al., 1995; Oddy, 1993; Taylor & Alden,
within two standard deviations of the mean for age on the Chil-
1997), we used an orthopedic-injury-only comparison
dren’s Orientation and Amnesia Test (Ewing-Cobbs, Levin,
group, took the children’s preinjury status into account in
Fletcher, Miner, & Eisenberg, 1990) for at least 2 consecutive
evaluating injury consequences when feasible, considered
days. The assessments were then repeated at short-term follow-ups
conducted 6 and 12 months later and again at an extended fol-
the outcomes in relation to TBI severity, and followed the
low-up that took place an average of 4.10 years postinjury
children at multiple time points over a long postinjury
(range
2.37–5.84 years, SD
0.91). Because baseline data from
interval. The orthopedic-injury-only group was included to
parents and teachers pertained to the child’s preinjury status,
control for the effects of a traumatic hospitalization and for
measures of postinjury child behavior, adaptive functioning, and
the characteristics associated with the children’s predispo-
classroom performance were available only at the three
sition to accidental injury. Collection of longitudinal data
follow-ups.

---

LONGITUDINAL CHANGES
17
Sample Characteristics
mean times between baseline and follow-up assessments. Attrition
was due to family moves, unwillingness to continue with the study,
The sample consisted of 189 children recruited from four hos-
and multiple missed appointments. A comparison of children who
pitals in North-Central Ohio, including 53 with severe TBI, 56
had participated in the extended follow-up with those who had
with moderate TBI, and 80 with orthopedic injuries only. Recruit-
dropped out previously failed to reveal differences in age at injury,
ment criteria included (a) hospitalization involving at least one
gender or ethnic distribution, measures of preinjury child function-
overnight stay either for moderate to severe TBI (with or without
ing, or postinjury test results obtained at baseline. However, attri-
orthopedic injury) or for orthopedic injury only, (b) age at time of
tion was highest in the ORTHO group, and those who dropped out
injury from 6 to 12 years, (c) the absence of indications of previous
had lower family socioeconomic status than did the children who
neurological disorder or child abuse, and (d) residence in an
remained in follow-up. We found similar differences when com-
English-speaking household. Children either with symptoms of
paring those who dropped out at the 6- or 12-month follow-ups
mild head injury or with brain insults due to causes other than
with the children who remained in the study at those times.
blunt trauma were excluded. Following other investigators
(Fletcher et al., 1990), we de?ned severe TBI by a lowest post-
Child Measures
resuscitation Glasgow Coma Scale (Teasdale & Jennett, 1974)
score of 8 or less. Moderate TBI was de?ned by either a score of
Measures of child functioning included the Child Behavior
9 –12 or a higher score accompanied by a skull fracture, intracra-
Checklist (CBCL; Achenbach, 1991a), Child Behavior Check-
nial mass lesion or contusion, diffuse cerebral swelling, posttrau-
list—Teacher’s Report Form (TRF; Achenbach, 1991b), Vineland
matic neurological abnormality, or loss of consciousness in excess
Adaptive Behavior Scales (Sparrow, Balla, & Cicchetti, 1984), and
of 15 min. Because the determination of intracranial pathology was
Letter–Word Identi?cation, Calculation, and Writing Samples
based primarily on computed tomography (CT) scan ?ndings, and
subtests of the Woodcock-Johnson Tests of Achievement—Re-
given the limitations of CT scans in this regard (Koelfen et al.,
vised (Woodcock & Mather, 1989). These measures have good
1997), we did not categorize children according to lesion
reliability and validity in assessing behavior and achievement and
localization.
are normed for school-age children and adolescents. These mea-
Table 1 summarizes group characteristics at the baseline assess-
sures also provide comprehensive assessment of behavioral and
ment. The groups were similar in terms of age at injury and
scholastic functioning and have been used widely in studies of
measures of the family environment. The ORTHO group had
outcomes of early brain insults (Jaffe et al., 1995; Max, Koele,
proportionally fewer male participants and fewer White partici-
et al., 1998; Taylor, Klein, Minich, & Hack, 2000; Taylor,
pants than the TBI groups, but only the difference in ethnicity was
Schatschneider, & Minich, 2000).
signi?cant. As anticipated, the two TBI groups differed markedly
The CBCL and TRF Total Behavior scales are parent- and
in injury severity. The ORTHO group had longer hospitalizations
teacher-based ratings of behavior problems. The CBCL Compe-
and greater severity of injury to nonhead areas than did the
tence scale is a composite of parent ratings of school performance,
moderate TBI group. Length of hospitalization and severity of
activities, and social relationships. The TRF Academic Perfor-
nonhead injuries were similar in the severe TBI and ORTHO
mance scale is a rating of the child’s performance in academic
groups. Motor-vehicle-related accidents were the most common
subject areas. The Vineland is a semistructured parent interview
cause of injury in the severe TBI group, whereas sports and
that assesses multiple domains of children’s adaptive functioning,
recreational accidents were more frequent causes of injury in the
including communication, daily living skills, and socialization.
moderate TBI group (Taylor et al., 1999).
The Woodcock-Johnson Letter–Word Identi?cation, Calculation,
Table 2 presents information on group composition and attrition
and Writing Samples subtests measure word recognition, written
at each of the follow-ups. No group differences were found in
math computation, and writing skills, respectively.
Table 1
Sample Characteristics at the Baseline Assessment (Total Sample Recruited)
Group
Severe TBI
Moderate TBI
ORTHO
(n
53)
(n
56)
(n
80)
Variable
M
SD
M
SD
M
SD
Age at injury (years)
9.4
2.1
10.0
1.9
9.3
1.9
No. of boys (%)
39
74
41
73
47
59
No. of Whites (%)
40
76
43
77
45
56
Four Factor Index score
(Hollingshead, 1975)
33.2
15.4
33.9
14.9
33.2
15.2
Days hospitalized
13.4**
10.0
7.0**
7.5
13.8**
13.7
Modi?ed injury severity score
(Mayer et al., 1980)
20.2**
11.9
12.7**
5.9
7.4**
3.2
Partial Modi?ed Injury Severity score
8.8**
10.4
2.4**
3.8
7.4**
3.2
Glasgow Coma Scale score
4.8*
1.9
14.0*
1.8
Days of unconsciousness
5.6*
6.7
0.2*
0.6
Note.
Modi?ed Injury Severity scores were de?ned as the sum of the squared scores for the three
most affected body regions, including the head. Partial Modi?ed Injury Severity scores were de?ned
similarly but excluded the head region. Duration of unconsciousness was de?ned as the number of
days from injury until the child was able to follow a simple verbal command. TBI
traumatic brain
injury; ORTHO
orthopedic injury.
* p
.05.
** p
.01.

---

18
TAYLOR ET AL.
Table 2
Sample Composition, Children’s Ages, and Baseline to Follow-up Intervals
at Each Follow-Up
Group
Total
sample
Severe TBI
Moderate TBI
ORTHO
Assessment
(n
53)
(n
56)
(n
80)
(n
189)
6-month follow-up
n
46
53
63
162
Attrition (%)
13
5
21
14
Child’s age in years
M
10.0
10.5
10.0
10.2
SD
2.1
1.9
1.9
2.0
Follow-up interval in years
M
0.5
0.5
0.5
0.5
SD
.04
.05
.05
.05
12-month follow-up
n
44
53
65
162
Attrition (%)
17
5
19
14
Child’s age in years
M
10.6
11.0
10.4
10.6
SD
2.1
1.9
1.9
2.0
Follow-up interval in years
M
1.0
1.0
1.0
1.0
SD
.1
.1
.1
.1
Extended follow-up
n
42
42
50
134
Attrition (%)
18*
25*
38*
28*
Child’s age in years
M
13.7
13.7
13.5
13.6
SD
2.2
1.7
1.8
1.9
Follow-up interval in years
M
3.9
4.1
4.1
4.0
SD
0.9
0.9
0.9
0.9
Note.
We excluded two cases, one with persistent coma and one with a second traumatic brain
injury (TBI) prior to the extended follow-up. The follow-up interval is measured from baseline for
the 6- and 12-month follow-ups and from injury for the extended follow-up. ORTHO
orthopedic
injury.
* Group differences signi?cant, p
.05.
Scores entered into analysis were the raw sums for the CBCL
based only on the relevant scales. Correlations between the family
and TRF scales, the raw domain scores for the Vineland, and the
factors for the total sample at baseline were as follows: SCI with
interval-scaled raw score equivalents (W scores) for the Wood-
Family Stressors, r(188)
.11, p
.10; SCI with Family
cock-Johnson. The use of raw or W scores was justi?ed by their
Resources, r(188)
.35, p
.01; Family Stressors with Family
sensitivity to patterns of change over time; standard scores obscure
Resources, r(188)
.27, p
.01.
these changes by adjusting for age-related variation (Francis,
Fletcher, Stuebing, Davidson, & Thompson, 1991).
Analysis
Family Measures
General linear mixed model analysis, also referred to as hierar-
chical linear or growth modeling, was used to examine child
Family characteristics were evaluated across assessments using
outcomes longitudinally (Burchinal, Bailey, & Snyder, 1994; Jen-
both distal and proximal measures of the family environment. The
nrich & Schluchter, 1986). In the mixed model approach, models
distal environment was measured by the Socioeconomic Compos-
for means and variance– covariance parameters are ?t to repeated
ite Index (SCI; Yeates et al., 1997). We computed this measure by
measures, or nested data. The essential aim of analysis is to make
averaging sample z scores for (a) the Duncan Socioeconomic
inferences about how factors that vary across individuals (i.e.,
Index (Stevens & Featherman, 1981), (b) annual family income as
within- and between-subjects ?xed effects) are related to variation
coded on the Life Stressors and Social Resources Inventory—
in the dependent measure.
Adult Form (Moos & Moos, 1994), and (c) years of maternal
The mixed model approach has several advantages over more
education. The proximal environment was measured by scales
traditional repeated measures analysis of variance or MANOVA
from the latter inventory that assessed family social stressors and
(Francis et al., 1991; Spikman, Timmerman, van Zomeren, &
resources. The Family Stressors score was de?ned as the mean of
Deelman, 1999; Thompson et al., 1994). This approach makes use
the T scores for ?ve stressors scales (Health, Work, Spouse,
of the actual ages of the participants at each assessment and does
Extended Family, and Friends); the Family Resources score was
not require the constant spacing of assessments. Mixed model
de?ned as the mean of the T scores for four resources scales
analysis also incorporates estimates of intrasubject correlations
(Work, Spouse, Extended Family, and Friends). Scores for single-
across repeated assessments and is thus a sensitive method for
parent families or for families in which parents did not work were
assessing change. Additional advantages are that all data can be

---

LONGITUDINAL CHANGES
19
used in analysis, even from participants not seen on all occasions,
manner. Main effects for SCI were included in all models, but
and that analysis allows for both continuous and categorical pre-
main effects for Family Stressors and Family Resources were
dictors. Inclusion of time-varying covariates is a further virtue in
retained only if they predicted outcomes independently of the SCI.
investigations in which predictors of outcome, such as family
To adjust for multiple comparisons, we grouped child outcomes
circumstances, change over time.
into ?ve domains: behavior problems (CBCL and TRF Behavior
Data from this study were analyzed using SAS Proc Mixed
Problem scales), competence (CBCL Competence scale), adaptive
(SAS Institute, Inc., 1990). This application used unstructured
functioning (Vineland Communication, Daily Living Skills, and
covariance matrices, allowing variances and covariances to vary
Socialization scales), school performance (TRF Academic Perfor-
across time points rather than to conform to a priori constraints.
mance scale), and achievement skills (Woodcock-Johnson Letter–
Group (severe TBI, moderate TBI, and ORTHO), gender, and race
Word Identi?cation, Calculation, and Writing Samples subtests).
were included in the models as discrete predictors. Age, age
Domainwise alpha level was set at p
.05, with Bonferroni
squared (to account for nonlinear age effects), and family factors
corrections applied in analysis of measures from multiple-measure
(SCI, Family Stressors and Family Resources) served as time-
domains. Alpha for examining simple effects was set at p
.05.
varying covariates. In analysis of the CBCL, TRF, and Vineland,
the preinjury score (obtained retrospectively at baseline) was in-
cluded as a non-time-varying covariate. By controlling for prein-
Results
jury status, we could more readily interpret group differences in
Results con?rmed group main effects for CBCL Compe-
outcome as injury related.
Postinjury changes in outcomes were modeled as linear splines,
tence, F(2, 160)
11.31, p
.01, Vineland Communica-
or connecting slopes, between successive follow-ups. Short-term
tion, F(2, 166)
7.48, p
.01, Vineland Daily Living
changes were de?ned by splines between the 6- and 12-month
Skills, F(2, 166)
4.96, p
.05, and Woodcock-Johnson
follow-ups for the CBCL, TRF, and Vineland and by splines
Writing Samples, F(2, 177)
8.05, p
.01. (Signi?cance
between the baseline and the 12-month follow-up for the Wood-
levels listed are Bonferroni corrected.) Table 3 summarizes
cock-Johnson subtests. Long-term changes were de?ned by splines
results from contrasts between each pairing of the groups on
between the 12-month and the extended follow-ups. Our decision
these measures. Simple effects tests revealed that outcomes
to use a single spline from the baseline to the 12-month follow-up
were poorer for the severe TBI group than for the ORTHO
for the Woodcock-Johnson subtests was based on lack of evidence
group on all four measures. A poorer outcome for the
for differential change across the two halves of the 1st postinjury
moderate TBI group relative to the ORTHO group was
year. We took into account intersubject variability in the time
between the 12-month and the extended follow-ups by including
found only on Woodcock-Johnson Writing Samples.
this interval as a model parameter.
Analysis also revealed a Group
SCI interaction for the
Group differences in short-term (Spline 1) and long-term
CBCL Behavior Problem scale, F(2, 166)
5.70, p
.01.
(Spline 2) changes were examined by tests of Group
Spline
Follow-up tests showed that differences between the TBI
interactions. To investigate moderating in?uences of family factors
groups and the ORTHO group increased with decreasing
on group differences in change, we also included triple interactions
levels of socioeconomic advantage. At low levels of the
of group, each spline, and a given family factor in the models. We
SCI, both TBI groups had more problems than the ORTHO
identi?ed signi?cant effects after models were trimmed by elimi-
group.
nating nonsigni?cant higher level and then lower level interaction
Analysis generally failed to reveal interactions of group
terms. To preserve power, the SCI was the only family factor
with short-term or long-term changes, suggesting a pattern
included in the initial or base models. Once the models involving
the SCI had been trimmed, Family Stressors and interactions of
of stable sequelae across follow-up. One exception to this
this factor with group and the splines were added to examine these
pattern was a marginally signi?cant interaction of group
effects. Models involving Family Stressors, in turn, were trimmed,
with long-term change for Woodcock-Johnson Writing
and the in?uences of Family Resources were examined in a similar
Samples, F(2, 177)
3.11, p
.05, before Bonferroni
Table 3
Sources of Signi?cant Group Differences
Group contrast
Child outcome and
Severe TBI
Severe TBI
Moderate TBI
domain/measure
versus ORTHO
versus moderate TBI
versus ORTHO
Child competence
CBCL Competence
2.07 (0.45)**
1.60 (0.46)**
0.47 (0.44)
Adaptive functioning
Vineland Communication
2.78 (0.72)**
1.73 (0.74)*
1.05 (0.69)
Vineland Daily Living
2.98 (0.95)**
1.64 (0.97)
1.33 (0.90)
Achievement skills
Woodcock-Johnson
Writing Samples
7.70 (1.93)**
3.51 (2.00)
4.19 (1.88)*
Note.
Data presented are model estimates of the differences in mean raw scores (standard errors)
pooled across the postinjury assessments, controlling for all other effects in the model. TBI
traumatic brain injury; ORTHO
orthopedic injury; CBCL
Child Behavior Checklist; Vine-
land
Vineland Adaptive Behavior Scales; Woodcock-Johnson
Woodcock-Johnson Tests of
Achievement—Revised.
* p
.05.
** p
.01.

---

20
TAYLOR ET AL.
correction. Examination of the source of this interaction,
justi?ed by our interest in identifying recovery patterns,
revealed catch up in writing skills in the moderate TBI
group relative to the ORTHO group between the 12-month
and the extended follow-ups. Rates of change in writing
skills during this period did not distinguish between the
severe TBI and the ORTHO groups.
Other exceptions included four interactions involving
group, short- or long-term change, and family factors. Spe-
ci?cally, the SCI moderated group differences in short-term
change in Vineland Socialization, F(2, 166)
4.38, p
.05, and in long-term change in TRF Academic Perfor-
mance, F(2, 129)
3.58, p
.05. Family Stressors mod-
erated group differences in short-term change in TRF Aca-
demic Performance, F(2, 129)
4.46, p
.05, and in
Woodcock-Johnson Calculation, F(2, 178)
4.22, p
.05.
According to follow-up analyses, group differences in
short-term change in Vineland Socialization were more
pronounced at lower levels of the SCI. To assist in deter-
mining the source of this interaction, we estimated group
means (see Figure 1) at the 6- and 12-month follow-ups at
high and low levels of SCI (i.e., for children from advan-
taged vs. disadvantaged families), de?ned in terms of values
of one standard deviation above and one standard deviation
below the sample mean, respectively (Figures 1a and 1b).
Growth in socialization skills was similar across groups
when the SCI was high, whereas growth was slower in the
two TBI groups relative to the ORTHO group when the SCI
was low. Socioeconomic disadvantage, thus, appeared to
hinder development of socialization over this interval more
Figure 1.
Model estimates of group means in Socialization
for the TBI groups than for the ORTHO group. Simple
subtest raw scores, Vineland Adaptive Behavior Scales (Sparrow
effects tests revealed that both the TBI groups had lower
et al., 1984), at the 6- and 12-month follow-ups for children with
(a) high and (b) low socioeconomic status (i.e., those from advan-
scores than the ORTHO group at the 12-month follow-up
taged vs. disadvantaged families) as de?ned by values one stan-
when the SCI was low.
dard deviation above and one standard deviation below the sample
Follow-up analyses indicated that group differences in
mean on the Socioeconomic Composite Index (Yeates et al.,
long-term change in TRF Academic Performance also were
1997). The plots show poorer development of socialization skills
more marked at lower levels of the SCI. Figure 2 plots
in the traumatic brain injury (TBI) groups than in the orthopedic
estimated group means at the 12-month and extended fol-
injury (ORTHO) group, but only when socioeconomic advantage
low-ups at high and low levels of SCI (i.e., for children from
was low.
advantaged vs. disadvantaged families) as de?ned above
(Figures 2a and 2b, respectively). Academic performance
ratings remained stable across groups when the SCI was
group showed a more rapid growth rate than the other
high but declined more for the severe TBI group than for the
groups when family stress was low. Low stress, thus, ap-
other groups when the SCI was low. Declining academic
peared to facilitate recovery in the severe TBI group. Simple
performance, thus, was not characteristic of all children
effects tests revealed lower scores for the severe TBI group
with severe TBI but only of those from disadvantaged
than for the ORTHO group. This difference was found at the
families. Simple effects tests showed that the severe TBI
baseline when stress was low and at the 12-month follow-up
group had lower ratings than the ORTHO group at all
when stress was high.
assessments when the SCI was high but that these groups
Follow-up analysis of short-term change in TRF Aca-
differed only at the extended follow-up when the SCI was
demic Performance indicated more pronounced group dif-
low.
ferences in change at lower levels of Family Stressors.
In follow-up analysis of Woodcock-Johnson Calculation
Inspection of means indicated that this effect was due to an
scores, group differences in short-term change increased
anomalous pattern of ?ndings for children with moderate
with decreases in Family Stressors. The latter interaction is
TBI. Whereas low family stress was associated with posi-
illustrated in Figure 3, which graphs estimated group means
tive gains over time in the severe TBI and ORTHO groups,
at the baseline and the 12-month follow-up at low and high
high family stress was associated with these gains in the
levels of Family Stressors (i.e., for children from unstressed
moderate TBI group. In simple effects tests, the moderate
vs. stressed families), de?ned by values of one standard
TBI group had lower ratings than the ORTHO group at the
deviation below and one standard deviation above the sam-
12-month follow-up when stress was low, and the severe
ple mean, respectively (Figures 3a and 3b). The severe TBI
TBI group had lower ratings than the ORTHO group at all

---

LONGITUDINAL CHANGES
21
base models. Interactions of this factor with group and
short-and long-term changes were also included. Results
generally failed to indicate moderating effects of age at
injury on group differences. The only exception was an
interaction involving age at injury, group, and short-term
change in TRF Academic Performance. The source of the
interaction was the absence of a group difference in this
rating at the 6-month follow-up for the youngest children in
the sample. A group difference was found for older children
at the 6-month follow-up; differences for younger children
were present at the 12-month and extended follow-ups. The
absence of group differences among younger children at the
6-month follow-up may have re?ected either a lack of
academic expectations for children in the early grades or
their teachers’ unwillingness to make negative appraisals of
scholastic competence. The results, therefore, do not indi-
cate age differences in TBI sequelae per se.
For descriptive purposes, we examined means and stan-
dard deviations for age-standardized scores by group across
Figure 2.
Model estimates of group means in academic perfor-
mance raw ratings, Teacher’s Report Form (TRF, Achenbach,
1991b), at the 12-month and extended follow-ups for children with
(a) high and (b) low socioeconomic status (i.e., those from advan-
taged vs. disadvantaged families) as de?ned by values one stan-
dard deviation above and one standard deviation below the sample
mean on the Socioeconomic Composite Index (Yeates et al.,
1997). Extended follow-ups were conducted at a mean of 4 years
postinjury. The plots show decline in the severe traumatic brain
injury (TBI) group relative to the other two groups but only when
socioeconomic advantage was low. ORTHO
orthopedic injury.
assessments when stress was high. Although these mean
differences are consistent with expectations, the pattern of
change over time is dif?cult to interpret.
In addition to revealing group differences, results con-
?rmed the importance of taking multiple predictors into
account in modeling child outcomes. Increases in age or age
squared predicted higher scores on CBCL Competence, on
all three Vineland domains, and on the three Woodcock-
Johnson subtests. Main effects were also found for gender,
Figure 3.
Model estimates of group mean W scores on the
race, and each of the three family factors. The family factor
Calculation subtest of the Woodcock-Johnson Tests of Achieve-
most consistently associated with child outcome was the
ment—Revised (Woodcock & Mather, 1989) at the baseline and
SCI, with higher SCI scores predicting better outcomes.
the 12-month follow-up for children with (a) low and (b) high
levels of family stressors (i.e., those from unstressed vs. stressed
Poorer preinjury functioning on the CBCL, TRF, and Vine-
families) as de?ned by scores one standard deviation below and
land was uniformly related to poorer postinjury functioning
one standard deviation above the sample mean. The plots show
on the corresponding measure.
catch-up in math skills in the severe traumatic brain injury (TBI)
To explore potential effects of age at injury (6 –12 years)
group relative to the orthopedic injury (ORTHO) group but only
on group differences, we added this factor to the trimmed
when family stressors were low.

---

22
TAYLOR ET AL.
the assessments (table available from H. Gerry Taylor).
interference with recovery related to early postinjury cog-
Although mean scores indicated poorer outcomes for the
nitive de?cits, or to disruptions in school attendance (Taylor
severe TBI group than for the ORTHO group, all but one
et al., 2001).
group mean fell within one standard deviation of the nor-
A second exception to long-term continuity of impair-
mative mean. The sole exception was a mean T score on
ment was the later decline in teacher ratings of academic
CBCL Competence of 39.93 for the severe TBI group at the
performance in children with severe TBI from less advan-
extended follow-up. These data argue against pervasive
taged backgrounds. Trends over time are likely to have been
de?ciencies in behavior or achievement, even in the latter
obscured by completion of ratings by different teachers at
group.
each assessment; hence, demonstration of signi?cant long-
term declines in this outcome suggests potentially robust
increases in academic dif?culties. We are not aware of other
Discussion
group data showing delayed emergence of educational prob-
lems, but this phenomenon is described by Ylvisaker and
Support for Study Hypotheses
Feeney (1998), and it parallels ?ndings of decreasing qual-
ity of life over time following adult TBI (Klonoff et al.,
In support of Hypothesis 1, outcomes were poorer for
1986).
children with moderate to severe TBI than for children with
Hypothesis 3 was supported by evidence for moderating
orthopedic injuries only. Compared with the ORTHO
in?uences of TBI severity and environmental factors on
group, at least some subset of the severe TBI group had
outcomes. Overall, sequelae were more evident in the se-
more behavior problems and lower competence by parent
vere TBI group than in the moderate TBI group. The effects
report, lower teacher ratings of academic performance,
of TBI on developmental change were also generally more
poorer adaptive functioning, and weaker math and writing
marked in children with severe TBI. Family factors moder-
skills. These sequelae were evident across multiple post-
ated group differences in several outcomes. As noted, short-
injury assessments. Consistent with previous research, more
term catch-up growth in math was evident only when family
limited sequelae were observed in the moderate TBI group
stress was low, and long-term decline in academic perfor-
(Fay et al., 1994; Jaffe et al. 1995; Taylor et al., 1999;
mance was found only in children from more disadvantaged
Yeates et al., 1997). Neither of the TBI groups differed from
backgrounds. Socioeconomic disadvantage was also associ-
the ORTHO group in either word recognition skills or
ated with more behavioral sequelae in children with TBI
teacher ratings of behavior problems. However, these mea-
across follow-up and with less rapid short-term progress in
sures have also been insensitive to TBI in other studies,
their socialization skills. The latter ?ndings suggest that the
perhaps because word recognition skills are relatively
adverse effects of TBI on behavior and the development of
spared after TBI in school-age children (Catroppa & Ander-
social skills were exacerbated by unfavorable family cir-
son, 1999; Yeates et al., 1997) and because teachers are
cumstances. These results con?rm and extend evidence for
reluctant to endorse postinjury behavior problems (Kinsella
moderating in?uences reported in our earlier follow-up of
et al., 1995).
the sample (Taylor et al., 1999; Yeates et al., 1997).
Consistent with Hypothesis 2, children with severe TBI
from lower stress environments showed short-term recovery
Possible Mechanisms Underlying Longitudinal
of function, although only in math skills. Furthermore, there
Outcomes of TBI
was little evidence of recovery after the 1st postinjury year.
As in past investigations, sequelae were observed over a
We can only speculate as to mechanisms responsible for
follow-up interval that extended well beyond the 1st postin-
the facilitative effects of social advantage on TBI sequelae.
jury year (Ewing-Cobbs et al., 1998; Kinsella et al., 1999;
Recovery of math skills may be attributed either to envi-
Max, Koele, et al., 1998; Max, Robin, et al., 1997; Perrott et
ronmentally mediated neural reorganization or to children’s
al., 1991; Rutter et al., 1983). The results suggest chronic
use of residual capacities. Enriched environments can facil-
functional problems similar to those found in adolescents
itate behavioral and neural reorganization in laboratory an-
and adults (Klonoff, Snow, & Costa, 1986; Levin, Gross-
imals following experimental brain lesions (Greenough,
man, Rose, & Teasdale, 1979; Oddy, Coughlan, Tyerman,
Black, Klintsova, Bates, & Weiler, 1999; Kolb, 1989;
& Jenkins, 1985; Putnam & Adams, 1992; Thomsen, 1989).
Rosenzweig, 1999). Applied to human recovery of function,
One exception to this pattern of stable long-term sequelae
a similar process could explain the catch-up growth in math
was the catch-up growth shown by the moderate TBI group
skills we observed in children with severe TBI from fami-
between the 12-month and the extended follow-ups on a
lies with few stressors. Neural reorganization would also
writing test. Although in need of replication, this result
help to account for moderating effects of family factors on
implies that some functions may improve after the 1st
the earlier cognitive outcomes of TBI in this sample (Taylor
postinjury year, at least in children with less severe TBI.
et al., 1999). Alternatively, the bene?t of living in an ad-
Demonstrations of long-term post-TBI recovery of cogni-
vantaged environment may have stemmed from opportuni-
tive abilities are consistent with this possibility (Chadwick,
ties to enhance behavioral adjustment and learn compensa-
Rutter, Shaffer, & Shrout, 1981; Klonoff, Low, & Clark,
tory skills. The environmental conditions that facilitated
1977). The absence of earlier catch-up growth in the mod-
recovery are unclear but could have included greater stim-
erate TBI group is uncertain but may re?ect limited oppor-
ulation or more appropriate support from parents and others
tunity for improvement during this 6-month interval, or
working with the child.

---

LONGITUDINAL CHANGES
23
The disruptive in?uences of social disadvantage on be-
(Taylor et al., 2001) suggests that this same process may
havior and on growth in socialization skills and academic
constrain recovery of function after TBI. In a similar vein,
performance may be explained in several ways. Parents
Wachs (2000) proposed a multitude of environmental and
from disadvantaged circumstances may face many dif?cul-
biological in?uences on development, with links between
ties in addition to those related to children’s injury-related
these in?uences contributing to stable individual character-
behavior and academic problems. Because of these dif?cul-
istics. Consistency over time in the family environment, as
ties, parents may have limited resources to invest in efforts
illustrated by the high correspondence of SCI scores at
to enhance the child’s recovery, or they may be burdened by
baseline and extended follow-up, r(134)
.93, p
.01,
other stressors that distract them from remedial efforts.
may also limit opportunities for positive change. Continuity
Socioeconomic circumstances may also have limited their
of sequelae is not unique to children with TBI but is char-
children’s access to special interventions. Because family
acteristic of diverse neurodevelopmental disorders (Keogh,
dysfunction, ineffective child management, negative par-
Bernheimer, & Guthrie, 1997; Mof?tt, 1993; Sigman &
ent– child interactions, and deviant models of behavior are
Kim, 1999; Streissguth, Barr, Sampson, & Bookstein, 1994;
more common in disadvantaged environments, another pos-
Taylor, Klein, et al., 2000; Taylor, Schatschneider, &
sibility is that these characteristics mediated the effects of
Minich, 2000).
social disadvantage on child outcomes (DeGarmo, For-
gatch, & Martinez, 1999; Gutman & Eccles, 1999; McLoyd,
Limitations
1990; Rutter, 1985; Spieker, Larson, Lewis, Keller, & Gil-
christ, 1999; Wachs, 2000).
The major limitation of this study is that a substantial
If children with brain insults are more dependent on a
number of children dropped out over follow-up, with dis-
positive and supportive family environment than are neuro-
proportional dropout of children of lower socioeconomic
logically normal children (Greenberg & Crnic, 1988;
status. Attrition bias is a common concern in longitudinal
Landry, Smith, Miller-Loncar, & Swank, 1997, 1998; Rut-
studies and must be taken into account in interpreting results
ter, 1981), then those with severe TBI may be especially
(Cicchetti & Nelson, 1994; Francis, Copeland, & Moore,
vulnerable to family adversity. The reasons for this suscep-
1994; Taylor, Klein, et al., 2000). The effects of family
tibility may relate to neuropsychological sequelae of early
factors on outcomes may have been obscured by differential
brain insults. For example, the weaknesses shown by chil-
dropout of the more disadvantaged participants. It seems
dren with severe TBI in executive functions and problem
unlikely, however, that attrition bias would have produced
solving may make it dif?cult for them to meet classroom
spurious ?ndings with regard to recovery trends or the
performance expectations without individual educational
moderating effects of family factors. This bias is also un-
assistance and accommodations (Ewing-Cobbs & Bloom,
likely to have altered the ?nding of continuity in longer term
1999; Ylvisaker & Feeney, 1998). The decline in academic
sequelae for many children with severe TBI. Caution is
performance observed among disadvantaged children with
nevertheless advised in generalizing the ?ndings to the
severe TBI may also have re?ected weaker neuropsycho-
broader population of children with TBI and in using these
logical skills among this speci?c subset of participants. We
results to estimate the magnitude of social in?uences on
are currently analyzing neuropsychological sequelae in a
outcomes.
manner similar to behavioral and achievement outcomes.
Additional limitations include the use of generic rating
The results of these analyses, to be presented in a separate
scales to assess child behavior problems and the fact that
report (Yeates et al., 2001), shed light on cognitive changes
most of the outcome measures, with the exception of the
after injury and the in?uences of social factors on these
Woodcock-Johnson subtest W scores, were not interval
sequelae.
scaled. Researchers have reported low sensitivity of the
A further possibility is that the selective decline in aca-
CBCL and TRF to behavioral sequelae of TBI in children
demic skills was an artifact of differential teacher biases.
and have suggested that interview procedures would be
Teachers of disadvantaged children may have been either
preferable (Fletcher & Ewing-Cobbs, 1991; Kinsella et al.,
less aware of the child’s history of TBI or less willing to
1995). Ratings of neurobehavioral symptoms are also wor-
alter expectations of the child in light of this information
thy of consideration in assessing the functional conse-
than teachers of children from more advantaged back-
quences of TBI (Barry et al., 1996; Rivara et al., 1994). The
grounds. If so, teachers of the disadvantaged children may
virtue of interval-scaled instruments is that they are equally
have been less lenient in their ratings. The latter explana-
sensitive to change throughout the measurement range, per-
tion, of course, would require that differences in teacher
mitting accurate comparisons of developmental change in
bias become more exaggerated with advancing grade levels.
children of different ages and ability levels (Thompson et
The persistence of most sequelae across follow-up also
al., 1994). Because of the relatively small differences in
requires explanation. Although residual brain insults likely
group means and evidence for catch-up of the severe TBI
contribute to long-term de?cits, a multitude of factors may
group in the interval-scaled measure of math skills, the early
be responsible for the relatively poor recovery observed in
recovery trends evident in the severe TBI group are not
this and past studies. Rutter (1994) hypothesized an “indi-
likely to have been spurious. Nonetheless, the use of non-
rect chain of events” (p. 930), including entrenched person–
interval-scaled measures in this study may have obscured
environment interactions, that serve to perpetuate the
group differences in change.
child’s disorder. Evidence from our project for bidirectional
A further weakness of the study is that outcomes were not
relationships between child and family outcomes over time
tracked at points between the 12-month and the extended

---

24
TAYLOR ET AL.
follow-ups. Multiple follow-ups after the 1st postinjury year
sciousness and the extent and localization of neuropathol-
would have allowed a more precise depiction of patterns of
ogy, are likewise warranted (Taylor et al., 2001). Age at
long-term change. For example, catch-up growth in the
injury, although not related to postinjury changes in behav-
severe TBI group during the 1st postinjury year may have
ior or achievement in the present sample, deserves consid-
continued for some time after the 12-month follow-up but
eration as well. TBI during the preschool years, as com-
then slowed, or outcomes even worsened, nearer to the
pared with later in childhood, has been associated with less
extended follow-up. Multiple follow-ups during this inter-
growth in cognitive abilities and with more dif?culties
val would have permitted tests of these nonlinear (e.g.,
acquiring reading skills after injury (Anderson, Catroppa,
quadratic or cubic) changes.
Morse, Haritou, & Rosenfeld, 2000; Barnes, Dennis, &
Wilkinson, 1999; Ewing-Cobbs et al., 1997).
Signi?cance and Future Directions
A better understanding of factors that moderate outcomes
of TBI would be useful in identifying high-risk children and
Despite these shortcomings, we followed a relatively
in enhancing awareness of ways to facilitate recovery. Dis-
large sample of children with moderate to severe TBI pro-
covery of relationships of family factors to postinjury
spectively and for a substantial time postinjury. As advised
changes in child outcomes provides impetus for including
by other researchers, we assessed recovery by comparing
family treatment and other environmental modi?cations as
children with TBI with an other-injury group, we took
components of rehabilitation (Oddy, 1993; Taylor et al.,
preinjury status into account when feasible, we examined
1999; Wade, Taylor, Drotar, Stancin, & Yeates, 1998). If
both short- and long-term phases of the recovery, we used
processes critical for postinjury progress can be clari?ed
growth modeling to assess recovery trends, and we inves-
and incorporated into treatment, children may show greater
tigated measures of the family environment as moderators
potential for positive change than is currently recognized
of postinjury changes in functioning (Fletcher et al., 1995;
(Bach-y-Rita, 1990; Greenough et al., 1999; Prigatano,
Goldstein & Levin, 1985; Klonoff et al., 1977; Oddy, 1993;
1987; Stein, 1988).
Rutter, Chadwick, Shaffer, & Brown, 1980).
A primary contribution of this study is the evidence it
References
provides for substantial and long-term residual disability in
Achenbach, T. M. (1991a). Manual for the Child Behavior Check-
children with severe TBI, with a more favorable outlook for
list/4-18 and 1991 Pro?le. Burlington, VT: University of Ver-
children with moderate TBI. The ?ndings also show that the
mont.
development of behavioral and academic competence after
Achenbach, T. M. (1991b). Manual for the Teacher’s Report Form
TBI is better for children from more advantaged environ-
and 1991 Pro?le. Burlington, VT: University of Vermont.
ments and that both distal and proximal family factors are
Anderson, V., Catroppa, C., Morse, S., Haritou, F., & Rosenfeld,
associated with postinjury changes in sequelae. Observa-
J. (2000). Recovery of intellectual ability following traumatic
tions of more positive changes after TBI in children from
brain injury in childhood: Impact of injury severity and age at
more advantaged backgrounds parallels past reports of
injury. Pediatric Neurosurgery, 32, 282–290.
Bach-y-Rita, P. (1990). Brain plasticity as a basis for recovery of
lower rates of new psychiatric disorders in children with
function in humans. Neuropsychologia, 28, 547–554.
TBI from more favorable family circumstances (Brown et
Barnes, M. A., Dennis, M., & Wilkinson, M. (1999). Reading
al., 1981; Gerring et al., 1998; Max, Koele, et al., 1998). As
after closed head injury in childhood: Effects on accuracy,
far as we are aware, however, this is the ?rst study to show
?uency, and comprehension. Developmental Neuropsychol-
that environmental advantage is associated with more pos-
ogy, 15, 1–24.
itive age-related changes in functioning following injury
Barry, C. T., Taylor, H. G., Klein, S., & Yeates, K. O. (1996). The
and not merely with lesser impairment at a given time
validity of neurobehavioral symptoms in children with traumatic
postinjury.
brain injury. Child Neuropsychology, 2, 213–226.
Additional follow-up studies are needed to replicate and
Brown, G., Chadwick, O., Shaffer, P., Rutter, M., & Traub, M.
extend the present ?ndings. A more complete picture of the
(1981). A prospective study of children with head injuries: III.
Psychiatric sequelae. Psychological Medicine, 11, 63–78.
natural history of outcomes would be provided by following
Burchinal, M. R., Bailey, D. B., & Snyder, P. (1994). Using
children over longer periods of time, examining cognitive
growth curve analysis to evaluate child changes in longitudinal
outcomes, and conducting more in-depth assessments of
investigations. Journal of Early Intervention, 18, 422– 442.
environmental in?uences on postinjury development. It is
Catroppa, C., & Anderson, V. (1999). Recovery of educational
also important to clarify the nature of residual sequelae and
skills following paediatric traumatic brain injury. Pediatric Re-
the types of outcomes most subject to environmental mod-
habilitation, 3, 167–175.
i?cation. The present results suggest that recovery is more
Chadwick, O., Rutter, M., Shaffer, D., & Shrout, P. E. (1981). A
likely to be found in math skills than in behavior or school
prospective study of children with head injuries: IV. Speci?c
performance, but further studies are required to examine this
cognitive de?cits. Journal of Clinical Neuropsychology, 3, 101–
issue. Another need is to consider other moderators of
120.
Cicchetti, D. V., & Nelson, L. D. (1994). Re-examining threats to
change, including children’s developmental and behavioral
the reliability and validity of putative brain-behavior relation-
status prior to injury and the extent of initial neuropsycho-
ships: New guidelines for assessing the effects of patients lost to
logical impairments (Brown et al., 1981; Gerring et al.,
follow-up. Journal of Clinical and Experimental Neuropsychol-
1998; Kinsella et al., 1999; Max, Robin, et al., 1997; Yeates
ogy, 16, 339 –343.
et al., 1997). Investigations of injury parameters as predic-
DeGarmo, D. S., Forgatch, M. S., & Martinez, C. R., Jr. (1999).
tors of development after TBI, such as duration of uncon-
Parenting of divorced mothers as a link between social status

---