Traumatic brain injury (TBI) in children and adolescents is a major public health problem in the United States, involving the annual hospitalization for acute brain trauma of about 100,000 children under 15 years of age.1
Other than the present study, there has been only one prospective childhood psychiatric study of TBI in which standardized instruments were used.2 Subjects were studied over a 2¼-year follow-up. There was a marked increase in psychiatric disorders following severe TBI compared with control subjects. Among variables predicting new psychiatric disorder in children with severe TBI, the most important was increasing severity of injury, but also included were the children's preinjury behavior, their early postinjury intellectual level, and their psychosocial circumstances.
Another prospective study included psychiatric assessments but was conducted without standardized psychiatric instruments.3 The study population consisted of 105 children consecutively admitted for TBI, of whom most had mild TBI. Approximately 80% of the children showed no posttraumatic behavioral changes. The children who had an initial loss of consciousness had a higher frequency of behavioral difficulty at 1-year follow-up.3
We4 found that during the first 6 months of follow-up, severity of injury, family psychiatric history, and family functioning were significant in predicting the onset of a specific psychiatric disorder never before present in the individual ("novel"). In addition to these variables, during the first 3 months of follow-up only, lifetime preinjury psychiatric disorder and lower socioeconomic class/intellectual functioning were similarly predictive of a novel psychiatric disorder. Children with mild to moderate TBI who had a lifetime preinjury psychiatric disorder were more likely to have a novel psychiatric disorder during the first 3 months, but not the second, than those without such a preinjury history.
Our aim in this study was to assess severity of injury, lifetime psychiatric disorder, behavior/adaptive function, family psychiatric history, family function, socioeconomic class, and preinjury intellectual function as predictors of psychiatric outcome in the second 6 months following TBI in children and adolescents.
This prospective follow-up study included children and adolescents who suffered a TBI. Comprehensive psychiatric, family, and adaptive functioning assessments were conducted at "baseline" (as soon as possible following the injury) to assess preinjury functioning, and psychiatric assessments were repeated 12 months following the TBI.
Inclusion criteria were as follows: consecutively admitted TBI patients; age 6 years to 14 years at time of injury; patients admitted to a large tertiary care center and three regional hospitals; CT scan on admission as a threshold criterion of TBI severity; English as the primary language spoken at home. Exclusion criteria were as follows: patients who had injuries so serious that they had not emerged from posttraumatic amnesia (PTA) 3 months following the injury; penetrating TBI; documented history of child abuse; history of previous TBI involving at least one hospital admission longer than one night; history of mental retardation; other acquired or congenital central nervous system disorder; preexisting acute or chronic serious illness.
The lowest post-resuscitation score on the Glasgow Coma Scale (GCS)5 was recorded from clinical notes.
The initial CT scans were analyzed independently by W. S. (a pediatric radiologist) and by Y. S. (a pediatric neuroradiologist), and a consensus was reached by discussion in discrepant cases. Scans were classified as showing an intracranial traumatic lesion or not. Scans were additionally classified according to the Traumatic Coma Data Bank categorization,6 which incorporates the degree of brain edema and focal lesions into a single rating (1 through 6) of increasing severity. Interrater agreement for the presence or absence of a lesion was 100%; regarding the Traumatic Coma Data Bank categorization, agreement was 92%.
Our first predictive statistical model, "Severity of Injury," consists of three items: 1) lowest post-resuscitation GCS score, 2) Traumatic Coma Data Bank categorization, and 3) normal/abnormal initial CT scan.
For descriptive purposes, we have classified categories of severity of TBI as follows: severe injury was defined by a lowest post-resuscitation GCS score ≤8; moderate injury was defined by a lowest post-resuscitation GCS score of 9—12 or a score of 13—15 with an intracranial lesion or with a depressed skull fracture seen on the initial CT scan; and mild injury was defined by a lowest post-resuscitation GCS score of 13—15, irrespective of any associated linear skull fracture.
Another measure reflecting the severity of TBI included the duration of coma. This was defined as the time from injury to the attainment of a score of 6 (ability to follow commands) on the motor subscale of the GCS. PTA was estimated by using a combination of the Children's Orientation and Amnesia Test,7 nursing notes, and parents' report.
The Schedule for Affective Disorders and Schizophrenia for School-Age Children epidemiologic (K-SADS-E) version,8 supplemented by a posttraumatic stress disorder (PTSD) module, was used at baseline. Follow-up assessments employed the Schedule for Affective Disorders and Schizophrenia for School-Age Children present episode (K-SADS-P) version,9 supplemented by the K-SADS-E sections on attention-deficit/hyperactivity disorder (ADHD), oppositional defiant disorder (ODD), and alcohol and substance abuse, as well as the PTSD module. Parents typically were interviewed first, followed by the children and adolescents. DSM-III-R criteria were used. The interview was administered, in all cases, by the first author, J. M., who is a board-certified child psychiatrist. In addition, the Neuropsychiatric Rating Schedule,10 an interview with the same format as the K-SADS designed specifically to identify symptoms and subtypes of organic personality syndrome, was conducted at all evaluations by J. M.
Almost all interviews in the ongoing 2-year prospective study were videotaped, and every twentieth tape was rated by a co-author, C. C., for the purpose of interrater reliability measures. C. C., a child psychiatrist, was blind to all other data, including severity of injury. So far, 10 interviews have been assessed for interrater reliability. Agreement on diagnostic classification was achieved in 8 of 10 cases (80%).
The psychiatric assessment at baseline defined our second predictive model, "Lifetime Psychiatric Disorder." This assessment, together with the 12-month follow-up, which evaluated status of the 6- to 12-month interval, defined a "novel" psychiatric disorder. The designation of a novel psychiatric disorder was applied in one of two conditions. First, this could occur in a subject with no lifetime psychiatric disorders as of the baseline assessment who then manifests a psychiatric disorder. Second, this could occur in a subject with a lifetime psychiatric disorder who manifests a psychiatric disorder that was not present before the TBI. The term "novel" was used to avoid confusion when comparing our findings with those of an earlier study2 that focused on "new" psychiatric disorders corresponding only with the first condition in our definition. Because preinjury psychiatric disorder may be common in children who suffer TBI, we were interested in studying the influence of TBI on children with psychiatric disorders. Two modifications were made to the DSM-III-R diagnostic criteria. First, we abandoned the requirement of diagnosing an organic mental disorder if a subject met all criteria for a functional psychiatric disorder but an organic etiology was suspected. Second, we waived the upper age limit for the onset of ADHD.
Adaptive Functioning and Behavior Measures
Adaptive functioning assessment was accomplished with the Vineland Adaptive Behavior Scale interview.11 This involved a semistructured interview conducted by a trained research assistant with the primary caretaker.
The Pediatric Behavior Scale (PBS)12 is a behavioral rating scale designed specifically for use with neurological and other medical disorders. It has been demonstrated to have adequate reliability and validity. A parent and schoolteacher completed the relevant versions of this instrument. Raw scores were added to yield a total behavior score. The Vineland Adaptive Behavior Composite and the total score on the PBS constituted our third predictive model, "Behavior/Adaptive Function."
Family Psychiatric History
The Family History Research Diagnostic Criteria13 interview was conducted in most cases by J.M. and in other cases by a trained research assistant with 2 years of experience as a mental health caseworker. Criteria were modified to conform with DSM-III-R criteria. On the basis of videotaped interviews conducted by J. M., adequate interrater reliability was achieved (kappa values for the diagnoses screened ranged from 0.79 to 1.00). At least one parent acted as the informant. Family ratings were then summarized for first-degree relatives only and for a combined grouping of first- and second-degree relatives as follows: "0"=no family psychiatric disorder; "1"=at least one member of the family met criteria for a psychiatric disorder, but no treatment was received; "2"=a family member met criteria for a psychiatric disorder and has received outpatient treatment or been arrested for antisocial behavior; "3"=a family member met criteria for a psychiatric disorder and has had inpatient psychiatric treatment or has been incarcerated. The fourth predictive model, "Family Psychiatric History," consisted of two components: the rating for first-degree relatives only and the rating for first- and second-degree relatives combined.
Family Assessment Measures
Family assessment measures were collected by both interview and questionnaire methods. The McMaster Structured Interview of Family Functioning is a research interview of the family.14 The interviewer uses the Clinical Rating Scale14 to rate each of seven domains, including global family functioning, from 1 to 7 on a Likert scale where lower scores indicate poorer function. The interviews were initially conducted by J. M. and subsequently by two trained research assistants. J. M. achieved an interrater reliability Pearson correlation coefficient of 0.83 for the global functioning rating with the team of the developers of the interview. Interrater reliability Pearson correlation coefficients for the global assessment domain of 0.87 and 0.96, respectively, were achieved between J. M. and the two trained research assistants. The Family Assessment Device (FAD)14 questionnaire was completed by family members at least 12 years of age. It measures the same domains of family functioning, including a global rating. We calculated a mean global functioning dimension score for each family. Higher scores on this instrument indicate more dysfunction. The fifth predictive model, "Family Function," consisted of two ratings, the global family functioning score from the interview and from the questionnaire.
Socioeconomic Class and Intelligence Measures
Socioeconomic class assessment was accomplished through the Four Factor Index.15 Other measures were from the baseline teacher's version of the PBS, which has items rating the child's preinjury intellectual ability and academic achievement on 5-point scales ranging from "far below average" to "far above average." The preinjury national percentile rank for vocabulary on the Iowa Tests of Basic Skills (ITBS) was included here because it is highly correlated with verbal IQ.16 The sixth predictive model, "Socioeconomic Class and Preinjury Intellectual Function," consisted of four items: Four Factor Index rating, PBS teacher's rating of preinjury intelligence, PBS teacher's rating of preinjury school achievement, and the ITBS national percentile rank for vocabulary. Socioeconomic class and intellectual function were linked in one model because our sample size limited the number of models we could construct and because the measures are often related.
To assess the representativeness of the sample with respect to those eligible patients who did not participate, we conducted t-tests for continuous variables and chi-square analyses for categorical variables. Logistic regression analyses were conducted on each of the six predictive models described above, to account for the development of novel psychiatric disorders. If the model was significant, we examined the significance of each of the variables within that model. Significant variables (P<0.05) and those tending to significance (P<0.1) from each significant model were selected for post hoc analysis using a forward stepwise logistic regression. In addition, we tested for the presence of an additive effect between models. This involved the standardization of all the variables and construction of 6 corresponding composite models. These were all entered into a logistic regression. The most important variable was then eliminated and the regression was repeated. The difference in the —2 log likelihood chi-square was the value tested for significance of the additive effect between models. SPSS17 was used for statistical analyses.
Eighty-seven TBI patients met criteria for entry into the study. The parents of 50 patients, after a complete description of the study, provided written informed consent to participate in the study and completed at least the psychiatric assessment component of the baseline assessment. Assent was obtained from the children when they were competent to give it.
Coma duration was as follows: no coma in 22, <15 minutes in 10, 1—24 hours in 6, 1—2 days in 2, 2—7 days in 6, and >7 days in 4. Subjects' PTA durations were as follows: no PTA in 6, <15 minutes in 1, 15—59 minutes in 8, 1—24 hours in 18, 1—2 days in 3, 2—7 days in 5, >7 days in 9. One subject remained in a vegetative state beyond the 3 months and has been dropped from subsequent analyses. By the 12-month follow-up, all other subjects were ambulatory, and gross neuromotor dysfunction was never worse than mild. Categorical classification of severity was as follows: 21/43 (48.8%) mild TBI, 9/43 (20.9%) moderate TBI, and 13/43 (30.2%) severe TBI. Causes of injury were as follows: motor vehicle accidents (MVA), 10 (20%); bicycle-car accidents, 6 (12%); falls from bicycles, 9 (18%); other falls, 10 (20%); sports and recreation, 7 (14%); pedestrian-MVA, 2 (4%); motorcycle/all-terrain vehicle accidents, 2 (4%); other, 4 (8%). At the 12-month follow-up, 44/49 subjects (90%) returned for reassessment. Two of the 44 children were siblings, and therefore 1 of the siblings was dropped from most analyses because certain variables could not be considered independent for the sibling pair.
The participants had the following demographic characteristics: age at injury (mean±SD), 10.12±2.49 years; 65.1% male; 98% white; social class 2.81±1.03. Current and lifetime preinjury psychiatric disorders were evident in 16/43 (37.2%) and 22/43 (51.2%) of participants, respectively. The frequency of family psychiatric history classifications in first-degree relatives where there was no family psychiatric history was 8/42 (19.0%); where a relative met diagnostic criteria for a disorder but had no treatment, 8/42 (19.0%); where a relative had outpatient treatment or was arrested, 14/42 (33.3%); and where a relative had inpatient treatment or was incarcerated, 12/42 (28.6%). The mean Family Assessment Device (questionnaire) score was 1.87±0.30, and the mean Clinical Rating Scale (family assessment interview) score was 4.02±1.52. Fifty percent of families were in the clinical range defined by either measure, compared with 54% with high "psychosocial adversity" in a previous study.2 The mean Vineland adaptive behavior composite was 94.98±13.55. The severe TBI and mild/moderate TBI groups were not significantly different with respect to age, gender, socioeconomic class, preinjury psychiatric status, or preinjury family, intellectual, and adaptive functioning.
The participating children (n=50) were not significantly different from those who declined enrollment (n=37) with respect to age (10.30±2.43 vs. 10.19±2.41), gender (64% male vs. 59.5% male), race (98% white vs. 100% white), social class (2.75±1.04 [n=48] vs. 2.53±1.14), history of behavior/psychiatric disorder or treatment elicited on screening instrument (14/50 [28%] vs. 7/33 [21.2%]), but were significantly different regarding the distribution of mild, moderate, and severe TBI (χ2=8.808, df=2, P<0.02). This was due to greater participation among severely injured subjects.
The presence of a novel psychiatric disorder in the second 6 months following TBI occurred in 16 of 43 subjects. Of these 16 subjects, novel disorders in 7 had persisted from the previous assessment 6 months following TBI. In another 2 subjects, the earlier novel disorder or disorders had been replaced by a different novel disorder, and the other 7 subjects developed their disorder during this 6-month interval. Novel disorders were varied and consisted of ODD in 7 children, ADHD in 5 children, organic personality syndrome in 4 children (1 resolved), simple phobia in 2 children, separation anxiety disorder in 2 children, hypomania in 1 child, obsessive-compulsive disorder (resolved) in 1 child, major depression in 1 child (resolved), depressive disorder not otherwise specified in 1 child (resolved), panic disorder in 1 child (resolved), and adjustment disorder with mixed disturbance of emotions and conduct in 1 child (resolved). Two of the 16 subjects, in addition to their novel disorder, also suffered a recurrence of and recovery from an episodic disorder that had been present before TBI (1 with major depression and 1 with adjustment disorder with depressed mood). The determination of the classification of "novel" psychiatric disorder did not depend on the diagnosis of Organic Personality Syndrome exclusively in any case.
The six models we wished to test to account for the development of a novel psychiatric disorder in the second 6 months following injury were entered into a logistic regression analysis. Four of the models were significant: "Family Function," "Family Psychiatric History," "Socioeconomic Class and Preinjury Intellectual Function," and "Behavior/Adaptive Function." The results are presented in T1. A post hoc forward stepwise logistic regression analysis to determine which of the variables within the model significantly contributed to the development of a novel psychiatric disorder was conducted. This involved entering the following four variables, in which significance levels were P<0.1: family psychiatric history in first-degree relatives (χ2=6.154, df=1, P=0.013); global family functioning score derived from the family assessment interview (χ2=4.970, df=1, P=0.026); teacher's PBS score of preinjury achievement (χ2=4.321, df=1, P=0.038); and the teacher's PBS score of preinjury intellectual ability (χ2=2.851, df=1, P=0.091). The results of post hoc testing revealed that the only significant variable was the global family function rating from the family assessment interview (χ2=6.835; df=1; P=0.009), which correctly predicted 71.4% of novel disorders. We were unable to demonstrate a significant additive effect among the six models.
This study was designed to prospectively assess risk factors related to the presence of novel psychiatric disorders in the second 6 months following TBI in children and adolescents. The major finding of this study is that the following four models, out of the six models we tested, significantly predicted novel psychiatric disorder: "Family Psychiatric History," "Family Function," "Socioeconomic Class and Preinjury Intellectual Function," and "Behavior/Adaptive Function." Post hoc forward stepwise logistic regression analyses revealed that a significant predictive variable was the global family functioning score derived from the family assessment interview.
This study has several limitations. The representativeness of the sample of children hospitalized for TBI is an issue. We were able to follow up at 12 months with 90% of children who completed the baseline assessment. However, only 57% of eligible children enrolled and completed baseline assessment. The characteristics of the children who did participate in the study compared with those who did not were similar except for the overrepresentation of severe TBI cases in the enrolled group. Therefore we may be overestimating novel psychiatric disorder. A second limitation in the study is the validity of baseline psychiatric findings. Any postinjury assessment of preinjury function is subject to inaccuracies. To minimize distortions in recollection of the child's preinjury behavior, we attempted to conduct these assessments as early as possible following the injury. The baseline assessment involving the parent occurred 14.8±13.1 (mean±SD) days following the injury. A third limitation is that although our results confirm that novel psychiatric disorders following TBI may be influenced by factors other than injury to brain tissue, a non-CNS trauma control group would be informative in the assessment of response to trauma and stress and in the documentation of naturally emerging comorbidities in children with preinjury psychiatric disorders. A control group was not essential in this study, since the question was not "How do TBI children differ from controls?," but rather, "Who among the TBI group develops a novel psychiatric disorder?" A fourth limitation is the sample size of our study. Replication of these findings with a larger sample would be necessary before these findings could be generally accepted. Nevertheless, the findings are remarkably similar to those of another prospective psychiatric study following TBI.2 A fifth limitation is that the psychiatric diagnoses were based on the psychiatric interviews alone. Data from other sources may have influenced diagnostic assessment, particularly with respect to externalizing disorders. A sixth limitation is that the psychiatric interviewer was not blind to severity of injury and was blind to only a portion of the family function and family psychiatric history data. Blind assessment would be useful, but injury-related physical stigmata are usually obvious. The good interrater reliability for the subjects' psychiatric diagnoses that was achieved with a second rater who was blind to all other data suggests that the lack of blindness did not cause a significant problem.
F1 illustrates the significant predictive models for the development of a novel psychiatric disorder in this cohort during this interval (6 to 12 months) as well as the first 6 months of follow-up.4 Lower family functioning and family psychiatric history were significant predictors throughout the first year. These findings are consistent with that of an earlier study2 with respect to the relationship between persistent new psychiatric disorders and "psychosocial adversity." This measure overlaps with our more differentiated evaluations of family function, family psychiatric history, and social class.
Our model "Socioeconomic Class and Preinjury Intellectual Function" was significant for the first 3 months and for months 6 to 12. Within this model, post hoc tests suggested that a lower teacher's rating of preinjury intellectual ability and preinjury school achievement were significant variables within the model for the 0—3-month and 6—12-month intervals, respectively. Our findings are consistent with research in children without brain injury, which suggests that intellectual skills constitute a protective factor in the avoidance of deviance despite chronic adversity.18
Our model, "Severity of Injury," only tended to significance in its prediction of novel psychiatric disorder in the second 6 months following TBI. This is in contrast to the first 6 months and suggests that at least in the interval studied, novel psychiatric disorder is relatively less strongly linked to injury severity and more to other factors such as psychosocial variables. The pattern following TBI in adults is similar.19 This is not to say that severity of injury is not associated with the presence of a novel psychiatric disorder during this interval. T2 reveals that 9 of 13 children with "severe" TBI, compared with 7 of 31 children with "mild" to "moderate" TBI, had a novel disorder during the second 6 months postinjury (Fisher's exact test, two-tailed; P=0.003). Earlier psychiatric studies2,3 underscored the importance of severity of injury as a predictor of psychiatric outcome 1 year following TBI.
Lifetime psychiatric disorder significantly predicted novel psychiatric disorder in the first 3 months only. This finding is consistent with the 1-year results of a mainly mild TBI cohort,3 where preinjury behavioral disturbance was not related to posttraumatic behavioral changes. Children with a preinjury lifetime psychiatric disorder may be especially vulnerable to the immediate disruptiveness of TBI in the first 3 months, but they are able to overcome the disturbance thereafter, provided the TBI was not "severe."
Another possible predictor of psychopathology, age at injury, was not significant in either the first or the second 6 months of follow-up. Nor was age at injury significant in other psychiatric studies of childhood TBI, which have generally excluded the 0—5-year age range for lack of a standardized diagnostic interview instrument. However, neurobehavioral function in children ages 0—5 years is more sensitive to disruption than in older children.20
Also frequently discussed is the effect of litigation status on psychopathology, which was not significant in either the first or the second 6 months of this study. Litigation is thought to be relevant to psychopathology because of the associated stresses as well as the potential monetary incentives for the maintenance of symptoms.
Other commonly discussed risk factors for psychopathology following TBI include the effect of gender and seizure activity/anticonvulsant medications. Neither of these was statistically significant in predicting novel psychiatric disorder in either the first4 or the second 6 months of follow-up.
In addition, none of the estimates of change in intellectual function (based on preinjury ITBS scores and observed postinjury verbal IQ scores) from before injury to immediately following resolution of PTA or to the 3-month or 12-month assessment were significantly correlated with novel psychiatric disorder in the second 6 months. This is in contrast to the findings at 6 months,4 where novel disorders were significantly associated with the change in intellectual function following PTA resolution, but not with change at the 3-month testing.
Risk factors for the onset of novel psychiatric disorder in children and adolescents in the second 6 months following TBI include family dysfunction, family psychiatric history, lower socioeconomic class, lower preinjury intellectual function, and lower preinjury behavior/adaptive function. The importance of severity of injury tends to diminish in the second 6 months compared with earlier follow-up. It is important from both clinical and public health perspectives that children and adolescents at risk for novel psychiatric disorders be identified and treated early following injury.
The authors acknowledge the helpful comments of Robert Robinson, M.D., and statistical advice from Stephan Arndt, Ph.D. This work was supported by a Young Investigator Award from the National Alliance for Research in Schizophrenia and Affective Disorders to Dr. Max.