Keenan HT, Runyan DK, Marshall SW, Nocera MA, Merten DF, Sinal SH. A Population-Based Study of Inflicted Traumatic Brain Injury in Young Children. JAMA. 2003;290(5):621–626. doi:10.1001/jama.290.5.621
Author Affiliations: Departments of Social Medicine (Drs Keenan and Runyan), Pediatrics (Dr Runyan), Epidemiology (Dr Marshall), Orthopedics (Dr Marshall), and Radiology (Dr Merten), University of North Carolina at Chapel Hill, and University of North Carolina Injury Prevention Research Center (Drs Keenan, Runyan, and Marshall and Ms Nocera), Chapel Hill; Department of Pediatrics, Brenner Children's Hospital, Wake Forest University School of Medicine, Winston-Salem, NC (Dr Sinal).
Context Physical abuse is a leading cause of serious head injury and death in
children aged 2 years or younger. The incidence of inflicted traumatic brain
injury (TBI) in US children is unknown.
Objective To determine the incidence of serious or fatal inflicted TBI in a defined
US population of approximately 230 000 children aged 2 years or younger.
Design, Setting, and Subjects All North Carolina children aged 2 years or younger who were admitted
to a pediatric intensive care unit or who died with a TBI in 2000 and 2001
were identified prospectively. Injuries were considered inflicted if accompanied
by a confession or a medical and social service agency determination of abuse.
Main Outcome Measure Incidence of inflicted TBI. Multivariate logistic regression models
were used to compare children with inflicted injuries with those with noninflicted
injuries and with the general state population aged 2 years or younger.
Results A total of 152 cases of serious or fatal TBI were identified, with 80
(53%) incurring inflicted TBI. The incidence of inflicted traumatic brain
injury in the first 2 years of life was 17.0 (95% confidence interval [CI],
13.3-20.7) per 100 000 person-years. Infants had a higher incidence than
children in the second year of life (29.7 [95% CI, 22.9-36.7] vs 3.8 [95%
CI, 1.3-6.4] per 100 000 person-years). Boys had a higher incidence than
girls (21.0 [95% CI, 15.1-26.6] vs 13.0 [95% CI, 8.4-17.7] per 100 000
person-years). Relative to the general population, children who incurred an
increased risk of inflicted injury were born to young mothers (≤21 years),
non–European American, or products of multiple births.
Conclusions In this population of North Carolina children, the incidence of inflicted
TBI varied by characteristics of the injured children and their mothers. These
data may be helpful for informing preventive interventions.
Child maltreatment involving physical abuse is the leading cause of
infant death from injury1 and serious head
injury in children.2 Most inflicted traumatic
brain injury (TBI) occurs in children younger than 2 years. Bruce and Zimmerman3 found in a clinical series that 80% of deaths from
head trauma in infants and children younger than 2 years were from inflicted
injuries. Developmental outcomes of children who survive inflicted TBI may
be worse than those of children with noninflicted TBI.4 Mortality
rate estimates for inflicted TBI range from 15%5 to
38%,6 and neurologic morbidity for surviving
children can be severe.7,8
One prospective, population-based study of inflicted TBI has been reported
in the literature.9 The study identified 19
cases in Scotland during an 18-month period and calculated the incidence of
inflicted TBI to be 24.6 per 100 000 infants.
The current study reports the first population-based estimates and demographics
of serious inflicted TBI in children aged 2 years or younger in the United
States. In a statewide population, we compared cases of inflicted TBI with
noninflicted TBI to assess the differences in risk factors between these 2
groups; we also compared prenatal risk factors for inflicted TBI with population
aggregates from the North Carolina vital statistics registry.
Hospitalization With TBI. We prospectively
collected data from all 9 hospitals in North Carolina with a pediatric intensive
care unit (PICU) or a monitored step-down unit. The study identified all North
Carolina resident children aged 2 years or younger in whom a serious or fatal
TBI occurred between January 1, 2000, and December 31, 2001. Subjects were
identified prospectively by contacting each PICU's charge nurse 3 times weekly
during the study period. Additionally, medical records at each center were
searched every 6 months by International Classification
of Diseases, Ninth Revision, Clinical Modification10 code
and matched to PICU admission logbooks to ensure that no eligible child had
been missed. Children were required to have been admitted to a PICU or monitored
step-down unit and have evidence of a nonpenetrating TBI on computed tomography,
magnetic resonance imaging, or pathologic findings. Most seriously injured
children in North Carolina would likely be transported to a North Carolina
PICU because of insurance or geography. However, to ensure complete case ascertainment,
the 3 closest out-of-state hospitals with PICUs to which children might be
transported (1 in Virginia and 2 in Georgia) were queried during the study
period about PICU admissions of North Carolina residents who were injured
in North Carolina with head trauma. This study was reviewed and approved by
the institutional review boards of all participating hospitals.
Fatal TBIs. Children who died at a hospital
without a PICU or prior to hospital admission were identified through the
Office of the Chief Medical Examiner. North Carolina has a centralized medical
examiner system that reviews all cases of unexpected or violent deaths. The
North Carolina Office of the Chief Medical Examiner was queried to identify
all deaths among children aged 2 years or younger. Charts were manually reviewed
to identify children who died of head injuries. Data were abstracted, including
pathologic evidence of brain injury and the results of the medical examiner's
investigation of the cause of death. Autopsies are performed on all children
who do not have a self-evident cause of death. Double counting of deaths was
avoided by matching date of birth, date of death, race/ethnicity, sex, and
place and manner of death to children who died in the hospital.
Definitions. The outcome definition of serious
TBI included computed tomography, magnetic resonance imaging, or pathologic
evidence of intracranial injury, including any type of intracranial hemorrhage,
shear injury, lacerations, or contusions. Patients with skull fractures but
no evidence of intracranial injury were excluded. The International
Classification of Diseases, Ninth Revision, Clinical Modification10 codes used to query medical records included the
following: 800.1 to 801.49, 801.6 to 801.99, 803.1 to 803.49, 803.6 to 803.99,
804.1 to 804.99, 850.0 to 850.99, 851.0 to 851.99, 852.0 to 852.59, 853.0
to 853.19, 854.0 to 854.19, and brain injuries from 959.8 to 959.9.
Inflicted TBI required evidence of TBI as defined herein accompanied
by a confession or a medical and child protective services determination that
the injury was inflicted. The medical and social services evaluations were
performed by the treating team at each hospital and reviewed by 2 of the authors
Children aged 12 months or younger were defined as infants. Other children
were included up to and including their second birth date. Extended family
living in the home was defined as adults related to the child who were not
the child's parents, foster parents, or primary caretaker and spouse (if the
child was in the legal custody of a family member other than the biological
parents). Premature birth was defined as a birth occurring at 37 weeks or
less of completed gestational age.
Medical Record Abstraction and Data Collection. Two
medical abstractors reviewed the complete medical chart or medical examiner
record of each child. The patients' presentation, including details of the
history as told by the parent or guardian, the physical examination, hospital
course, and outcome, were abstracted. The results of all radiological, ophthalmologic,
and electroencephalographic examinations were reviewed, as were all surgical
procedures and discharge summaries. Involvement by the North Carolina Department
of Social Services and the child's posthospitalization disposition were documented
from the medical record. The Department of Social Services' involvement with
a family is always documented in the medical record to ensure legal clarity.
The initial head computed tomography and/or magnetic resonance imaging scan
and subsequent radiological examination results of each patient were read
by a single pediatric radiologist who was blinded to the mechanism of injury
to ensure that each subject met uniform entrance criteria. Children born in
North Carolina were matched to their birth records to ascertain data about
mothers' prenatal care, age, number of prior births, and education.
Unknown Etiology Classification. For cases
in which no social service or medical information was contained in the medical
record about injury etiology, the project coordinator prepared an abstract
of the case for review by 2 of the investigators (H.T.K., D.K.R.), adapting
the methods of Stier et al.11 Each abstract
included a summary of each case, a description of the cause of injury as presented
by the caretaker to the treating physicians, and a description of the injuries,
including specific findings such as metaphyseal fractures, retinal hemorrhages,
and old injuries found on long bone studies. Demographic information such
as race/ethnicity, sex, referring hospital, and socioeconomic status was not
included. Child age was included in the summary because the likelihood of
certain reported events is developmentally linked. Each reviewer independently
classified the case as inflicted or noninflicted TBI. If the 2 reviewers did
not agree, an additional 3 reviewers, including pediatric intensivists and
child abuse experts, were available to provide additional review and resolve
Incidence in person-years was calculated for infants using the number
of births in the state for each study year as the denominator. Incidence for
children aged 12 months to 2 years was calculated using the number of live
births minus the number of infant deaths for the preceding year as the denominator.
These numbers were added to obtain the denominator for the entire study population.
To assess a possible bias related to migration either in or out, population-based
data were obtained for infants and 2-year-old children from the 2000 census
and were found to be within 6% of the information from vital statistics records.12 Information on the racial/ethnic makeup of the population
was based on the North Carolina birth records for the comparable year.12
Three comparison groups were used for the analyses. The first comparison
group was composed of the children in the prospectively collected TBI cohort
who had noninflicted TBI. The second comparison group represented the population
at risk and was composed of aggregate statistics for all North Carolina births
in 2000. To perform multivariate logistic regression analyses of prenatal
risk factors, a third comparison group of 300 births was randomly selected
from the birth cohorts of 2000 and 2001 proportional to the number of injured
children born in each year.
Simple frequencies and proportions were calculated to describe the demographics. t Tests were used to compare continuous variables. Odds
ratios were calculated to compare children with inflicted injuries with children
with noninflicted injuries and to compare children with inflicted TBI with
the North Carolina aggregate data. Exact confidence intervals (CIs) were calculated
for expected cell sizes of 5 or less.
Two multivariate logistic regression analyses were fit to the data.
The first model estimated adjusted odds ratios comparing prenatal demographic
features of children with inflicted TBI to the North Carolina birth certificate
data. All variables available from both the vital statistics database and
the inflicted injury group were entered into the model. Indicator variables
were used to compare differences among racial/ethnic groups (European American,
African American, and all other minorities) and 3 maternal age groups (≤21
years, 22-26 years, and >26 years). The second model was constructed using
a forward model-building strategy. It compared demographic and family characteristics
of children having an inflicted TBI with children having a noninflicted TBI.
Analyses were computed with SAS software, version 8.2 (SAS Institute Inc,
A total of 152 cases of serious or fatal TBI were identified from a
population of approximately 230 000, with 80 (53%) incurring inflicted
TBI. There were 87 boys (57%) and 119 infants (78%). Twelve (7.9%) of the
hospitalized children were identified retrospectively through chart review.
Two cases had an undetermined etiology, both from the medical examiner's office,
which were adjudicated by the primary investigators as 1 inflicted TBI and
1 noninflicted TBI. All other cases were either injuries witnessed by a disinterested
person or had complete medical and social services evaluation. There were
a total of 71 non-Hispanic European American children, 53 African American
children, and 28 children from other minority groups. The "other" minority
group comprised 18 children of Hispanic ethnicity, 6 multiracial children,
2 Asian children, 1 Native American child, and 1 Pacific Islander. One case
of a North Carolina resident injured in North Carolina came from an out-of-state
hospital. There were 40 TBI fatalities (26.3%).
The incidence of inflicted TBI for all children aged 2 years or younger
was 17.0 (95% CI, 13.3-20.7) per 100 000 person-years (Table 1). Rates were markedly higher in infants (29.7 [95% CI, 22.9-36.7]
per 100 000 person-years) than children in the second 12 months of life
(3.8 [95% CI, 1.3-6.4] per 100 000 person-years). Boys were more likely
to incur inflicted injuries than girls, and non–European American children
were more likely to incur inflicted injuries than European American children.
There were 18 deaths from inflicted TBI (case-fatality rate, 22.5%).
The incidence of noninflicted TBI was 15.3 (95% CI, 11.8-18.8) per 100 000
person-years in the first 2 years of life. Unlike inflicted TBI, boys and
girls with noninflicted TBI had similar rates of injury. However, like children
with inflicted TBI, minority children had a higher incidence of noninflicted
TBI than European American children. The incidence rates of both inflicted
and noninflicted TBI decreased from the first to the second year of life;
however, the decrease was much more pronounced in the inflicted TBI group.
There were 22 noninflicted TBI deaths (case-fatality rate, 30.5%).
Demographic characteristics of the children are shown in Table 2. A younger median age at injury was observed in the group
with inflicted compared with noninflicted injuries (4.0 months vs 7.5 months; P<.001 by t test). Mean ages
at injury were 5.9 (95% CI, 4.7-7.1) months and 7.6 (95% CI, 7.8-11.5) months
in the inflicted and noninflicted TBI groups, respectively. Unadjusted odds
ratios indicated that risk factors for inflicted TBI compared with North Carolina
aggregate statistics included male sex, minority status, and multiple birth.
These risk factors were also present in the comparison of children with inflicted
vs noninflicted TBI; however, these estimates were less precise.
Maternal characteristics associated with increased risk of inflicted
TBI compared with North Carolina aggregate data were young maternal age (P<.001 by t test), unmarried
status at birth of child, prenatal care initiated after the first trimester,
and the index child being the first child (Table 2). Maternal education level higher than high school appeared
to be protective. The pattern was similar when comparing maternal characteristics
of children with inflicted TBI vs those with noninflicted TBI; however, again,
estimates were less precise.
Family and community characteristics (Table 3) associated with an increased risk of inflicted compared
with noninflicted injuries included presence of extended family in the home
and having a parent in the military.
Inflicted vs Noninflicted TBI. The adjusted
estimates largely confirmed the findings of the bivariate analysis. When comparing
inflicted with noninflicted TBI, younger maternal age, younger child age,
and having a parent in the military remained important risk factors, although
the estimates were imprecise (Table 4).
The presence of a father in the home and maternal education level higher than
high school appeared to be protective. Race/ethnicity was not associated with
the odds of an inflicted TBI vs a noninflicted TBI given that the child had
incurred a TBI.
Inflicted TBI vs Birth Certificate Comparison Group. Adjusted estimates comparing the odds of incurring an inflicted TBI
with the North Carolina birth certificate comparison group showed an increased
risk associated with minority children, younger maternal age, male children,
and multiple birth, adjusted for all covariates in the model (Table 4). The odds of incurring an inflicted TBI increased with
decreasing maternal age in all racial/ethnic groups. Among European American
children, the adjusted odds of inflicted injury increased to 2.7 (95% CI,
1.2-6.0) for children born to mothers aged 22 to 26 years and increased further
to 4.6 (95% CI, 1.9-11.4) for children born to mothers aged 21 years or younger
compared with children born to the oldest maternal age group (>26 years).
Similarly, the odds of inflicted TBI for African American children increased
from 5.6 (95% CI, 1.9-16.2) to 9.8 (95% CI, 3.2-29.8) and for other minorities
from 16.0 (95% CI, 4.1-62.9) to 28.0 (95% CI, 6.5-120.2) with each decrease
in maternal age group compared with children of mothers older than 26 years.
This study is the first to report prospective, population-based estimates
of the incidence of serious inflicted TBI in young children in the United
States. It shows that inflicted TBI is a serious public health problem, with
approximately 30 per 100 000 infants experiencing a severe or fatal brain
Our estimate for the incidence of inflicted TBI in the first year of
life is within the confidence limits of the study by Barlow and Minns (24.6
[95% CI, 14.9-38.5] per 100 000 person-years).9 However,
our estimates have greater precision, possibly because we were able to calculate
incidence rates for a larger base population. The study by Barlow and Minns
differed from ours because they did not restrict their population to children
admitted to a monitored setting. It is unclear why they found no children
with inflicted TBI older than 12 months, but this may be related to the size
of their base population.
Sixty-two percent of children with inflicted TBI were boys, which is
consistent with other reports suggesting that male infants are at increased
risk for inflicted TBI.9,13,14 This
study found the mean age of inflicted TBI (5.9 months) to be consistent with
mean ages in previous studies (2.2-8.7 months).9,13,14 The
fact that our study restricted the age of children to 24 months or younger
may have lowered the mean age compared with studies that included children
as old as 3 years. Although previously not a predictor of inflicted TBI,15 in this study, race/ethnicity was highly predictive
of inflicted injury compared with the general North Carolina population. The
"other minority" group appeared to be at especially high risk; however, this
estimate was imprecise. Race/ethnicity was not a good predictor of inflicted
TBI vs noninflicted TBI; rather, minority children were at increased risk
of all types of TBI, suggesting that the role of race/ethnicity is a marker
for other social factors that put children at risk for injury. An association
between inflicted TBI and active military service has been previously described
(odds ratio, 3.5; 95% CI, 1.4-8.3).16 This
estimate is very close to our adjusted estimate for military dependents (odds
ratio, 4.7; 95% CI, 0.7-30.2); however, our estimate is imprecise because
of small numbers.
Young maternal age, prematurity, and multiple birth have been previously
reported as risk factors for child maltreatment.11,17 Overpeck
et al,1 in a national study linking infant
birth and death certificates, found that two thirds of infant homicide occurs
by the sixth month of life. Identified risk factors for infanticide included
maternal age younger than 19 years, less than 12 years of education, late
or no prenatal care, and premature birth.1 Our
study found that inflicted TBI appears to have many of the same risk factors
as those for child maltreatment and homicide. Because mothers are generally
not the perpetrator in cases of inflicted TBI,18 these
risk factors may be related to a decreased choice of suitable caregivers.
It is possible that this study may have systematically underascertained
cases of serious inflicted TBI. Clinicians vary in their practice of placing
children in a PICU for observation following a head injury. However, because
it is most common to watch infants in a closely monitored setting if they
have a new intracranial injury, and because of the medicolegal aspects of
inflicted TBI, this bias is not likely to be large. Ascertainment bias has
been reported for children with fractures screened for physical abuse in the
emergency department, with more minority children screened than nonminority
children19; however, this is less likely with
serious TBI because the children are hospitalized and a workup for inflicted
TBI is pursued after the child has been stabilized.
We studied only children who had injuries that were serious enough to
present to a PICU. However, many children may never present for medical care
after the first episode of inflicted TBI, and of those who do present for
medical care, many may be missed because of the nonspecificity of complaints
in infants with head injuries.20 It is known
that some children are found later in life to have had inflicted TBI when
they are evaluated with neuroimaging for other problems, such as developmental
delay or increasing head circumference.6 Thus,
our estimate is a lower-bound estimate of the overall problem of inflicted