Objective. Head injury is the leading cause of death in abused children under 2 years of age. Evidence for establishing guidelines regarding screening for occult head injury in a neurologically asymptomatic child with other evidence of abuse is lacking. This is particularly important given that many children with acute inflicted head injury have evidence of old injury when they are diagnosed. The primary aim of this study was to estimate the prevalence of occult head injury in a high-risk sample of abused children with normal neurologic examinations. The secondary aim was to describe characteristics of this population.
Methods. Children under 2 years of age admitted to an urban children’s hospital between January 1998 and December 2001 with injuries suspicious for child abuse were eligible for this study if they had a normal neurologic examination on admission. Subjects were selected if they met 1 of the following “high-risk” criteria: rib fractures, multiple fractures, facial injury, or age <6 months. Subjects were excluded if they had a history of neurologic dysfunction, seizures, respiratory arrest, or if their initial physical examination revealed scalp injury.
Results. Of the 65 patients who met these criteria, 51 (78.5%) had a head computed tomography or magnetic resonance imaging in addition to skeletal survey. Of these 51 patients, 19 (37.3%, 95% confidence interval 24.2–50.4%) had an occult head injury. Injuries included scalp swelling (74%), skull fracture (74%), and intracranial injury (53%). All except 3 of the head-injured patients had at least a skull fracture or intracranial injury. Skeletal survey alone missed 26% (5/19) of the cases. Head-injured children were younger than non-head-injured children (median age 2.5 vs 5.1 months); all but 1 head-injured child was <1 year of age. Among the head-injured children, 72% came from single parent households, 37% had mothers whose age was <21 years, and 26% had a history of prior child welfare involvement in their families. Ophthalmologic examination was performed in 14 of the 19 cases; no retinal hemorrhages were noted.
Conclusions. Our results support a recommendation for universal screening in neurologically asymptomatic abused children with any of the high-risk criteria used in this study, particularly if that child is under 1 year of age. Ophthalmologic examination is a poor screening method for occult head injury, and one should proceed directly to computed tomography or magnetic resonance imaging. Given the high prevalence of occult head injury detected in this study, further study is warranted to estimate the prevalence of occult head injury in lower risk populations of abused children.
Head injury is the leading cause of death in abused children under 2 years of age,1,2 and early detection of this injury can limit its significant morbidity and mortality. Multiple investigators have shown that most children with inflicted head injury have evidence of other occult injuries, including fractures, at the time they present for medical care.3,4 As a result, physicians routinely screen head-injured infants for evidence of occult skeletal injury to ascertain the likelihood of an inflicted injury. Furthermore, because of the association of skeletal injury with inflicted head injury in infants, the American Academy of Pediatrics has recommended that all children under 2 years of age with injuries suspicious for child abuse receive a skeletal survey to detect occult skeletal injury.5
Although there are clear recommendations to screen for occult skeletal injury in the presence of acute head injury in infants, the recommendations regarding screening for occult head injury in the presence of suspicious skeletal or other injuries are less well-defined. A recent study examining 173 abused infants found that nearly a third of infants with inflicted head injury were “missed” on their initial presentation for medical care.6 Among this missed group of infants, 25% were reinjured before they were appropriately diagnosed. Nearly 10% of these missed cases suffered fatal injury after initial presentation. Similarly, 40% to 45% of abused children with acute head injury have evidence of prior head injury by computed tomography (CT) or magnetic resonance imaging (MRI) at the time of their diagnosis.7 Thus, some authors have recommended that head imaging be considered in high-risk cases of child abuse even when the child has no clinical evidence of head injury.5,8
To date, there has been no clinical research that addresses the prevalence of occult head injury in abused children or the utility of routine MRI or CT examination of the head to detect occult head injury in abused children. Such information might influence recommendations for screening for occult head injury in this population. The primary aim of this study was to estimate the prevalence of occult head injury in a cohort of abused children with high-risk skeletal and soft-tissue injuries. The secondary aim was to describe characteristics of the head-injured population and to explore risk factors that might be associated with the presence of an occult head injury.
The child abuse team at Children’s Hospital of Philadelphia provides consults for children who present to a hospital setting with suspected maltreatment. Although some consultation occurs in the ambulatory and emergency department settings, most consults are performed in the inpatient setting. Over the last 4 years, the team has routinely, but not universally, screened young children under 2 years of age with head CT or MRI if they were suspected victims of child abuse with high-risk injuries and a normal neurologic examination. High-risk criteria have included the presence of multiple fractures, the presence of rib fractures, the presence of facial injury (bruises, abrasions, lacerations), and age <6 months. Therefore, the subjects for this study included all children under 2 years of age who were admitted to Children’s Hospital of Philadelphia between January 1, 1998 and December 31, 2001, with these high-risk criteria for child abuse, but with no overt symptoms of head injury, defined by a normal age-appropriate neurologic examination at admission to the hospital.
The source of patients for this study came from inpatient consultations during this time period. Subjects were identified by a retrospective review of the child abuse consultation team records over the time period of the study. Inclusion depended not only on the child meeting 1 of the above criteria, but also receiving a child abuse consult and report to social services. Following case ascertainment, subjects were excluded if they had a history of a seizure, history of respiratory arrest, history of an underlying neurologic condition, history of an abnormal neurologic examination, or evidence of scalp injury on physical examination. These findings would routinely necessitate brain imaging. Data were collected by abstraction of medical and radiology records. Permission was also granted by the child welfare agencies of New Jersey and Pennsylvania to gather data on abuse histories of the subjects and their siblings. The study was approved by the Institutional Review Board of Children’s Hospital of Philadelphia.
Study variables included demographics, such as the age, race, and gender of the child; age of the mother; history of prior child welfare reports on the child or his siblings; household composition; and insurance status of the family (private insurance versus Medicaid/none). Clinical data included weight (expressed as percent for age, adjusted for prematurity), head circumference (expressed as percent for age), presence of rib fractures, presence of multiple fractures, presence of facial injury, presence of retinal hemorrhages, and whether the child was screened with head CT or head MRI.
Because of our concern that bias of the initial radiologist reviewing the films might have led them to overestimate the presence of head injury for children with particular types of injuries, we chose to have 1 neuroradiologist review all of the studies that were initially interpreted as positive for head injury or for the presence of equivocal findings (eg, increased cerebrospinal fluid spaces, enhancement in cortex or white matter of brain). The neuroradiologist was blinded to the history of the child beyond knowing that the child was a suspected victim of child abuse. She was asked to denote the presence of scalp injury, skull fracture, or intracranial injury on a review of all available images, including skull radiographs, head CT, and/or head MRI.
Data analysis was conducted using Stata 7.0 software (Stata Press, College Station, TX). The study population was described using frequencies for categorical variables and means, standard deviations, and ranges for continuous variables. The prevalence of occult head injury was calculated with its associated 95% confidence interval (CI). Univariate χ2 statistics were used to explore the association of selected demographic and injury variables with the outcome of occult head injury.
During the study period between January 1998 and December 2001, the child abuse team provided consults on 474 children in the inpatient setting, of whom 314 were below the age of 2 years. Of the 314 children, 113 were immediately excluded because they presented with signs or symptoms to support the diagnosis of primary head injury before any diagnostic evaluation was conducted (18 with external head injury and/or primary skull fracture, 95 with brain injury). Of the remaining 201 children, 72 met the criteria for inclusion in this study. Seven children were subsequently excluded because of scalp injury noted on at least 1 physical examination on admission (3), history of respiratory arrest before admission (1), possible seizure activity before admission (2), or an abnormal neurologic examination noted by the referring physician (1). After exclusions, 65 children were eligible for the study.
The characteristics of the study population appear in Table 1. The mean age of the children in the study was 6.4 months (standard deviation: 5.6 months) with 69% of the children below the age of 6 months (median: 4.7 months; range: 0.7–23.9 months). The population included 58% of children who were black, 31% who were white, and 11% who were of other racial background. The median weight was 35th percentile for age; 19% of the children were below the 5th percentile. Injury patterns included rib fractures in 34% of the children, multiple fractures in 59% of the children, and facial injury in 35% of the children. All of the children had a skeletal survey performed as part of their evaluation. Fifty-one of the 65 children (78.5%) had a head CT and/or MRI performed in addition to the skeletal survey. Of the variables included in the study, the only significant difference between the screened (n = 51) and unscreened children (n = 14) was that the unscreened children were older (median age: 5.8 vs 4.4 months; P = .03).
The neuroradiologist participating in the study (L.T.B.) reviewed brain imaging for more than half of the study subjects, in particular for all subjects whose initial head imaging revealed evidence of a head injury or equivocal findings. All head injuries noted on the initial radiology reports were also detected during this second review. None of the equivocal studies were found on second review to have clear evidence of head injury, and so these cases were added to the group of children without head injuries.
In total, 19 (37%; 95% CI: 24%–51%) of the 51 fully screened children had evidence of an occult head injury. If one makes a conservative assumption that the 14 children who were not screened fully did not have a head injury, then 29% (95% CI: 18%–40%) of the children were found to have an occult head injury. Occult head injuries included scalp injury not identified by physical examination (74%), skull fracture (74%), and intracranial injury (53%). Among the children with intracranial injuries were 9 children with subdural hemorrhage or hemorrhagic contusions and 3 children with subcortical white matter shear injuries. Of the 10 children with intracranial injuries, 7 were described by the radiologist as having both acute and subacute (within days) injuries; 3 had subacute or chronic intracranial injury only. Ophthalmologic consultation with a fundoscopic examination was provided in 74% of the children with occult head injuries (14 of 19); of those receiving fundoscopic examination, none were found to have retinal hemorrhages. None of the children with occult head injuries required operative intervention.
Of the 19 children with occult head injuries, 5 (9.8% of screened cohort; 95% CI: 1.2%–18.0%) would have missed detection if a skeletal survey had been performed alone. These 5 children included:
A 5-month-old boy with an unexplained femur fracture, found on CT and MRI scan to have left parietal scalp swelling.
A 3-month-old girl with a history of unexplained frenulum and lip lacerations, found to have multiple fractures and by MRI, high left parietal scalp swelling without a definitive skull fracture.
A 5-month-old boy referred to the hospital for evaluation of failure to thrive, diagnosed with an acromion and proximal humerus fracture on skeletal survey. CT and MRI scans both revealed high frontal scalp swelling.
A 1-month-old boy with unexplained abdominal bruising and a liver laceration, found by MRI to have subacute subdural hematomas along the tentorium of the posterior fossa.
A 3-month-old girl with history of colic and gastroesophageal reflux found on chest radiograph to have 2 healing rib fractures at the time of insertion of a esophageal probe for a reflux evaluation. MRI scan revealed bilateral cerebellar subdural hematomas.
The characteristics of the 51 children who were screened with head imaging in addition to a skeletal survey are summarized in Table 2. The characteristics are summarized according to head injury outcome. The crude association of risk factors with the presence of an occult head injury was explored despite the limitation that testing risk factors that were also selection criteria reduced the ability to evaluate the independent association of these factors with the head injury outcome. Despite this limitation, younger age was still associated with the presence of an occult head injury (median age of head-injured subjects: 2.5 months; median age of non-head-injured subjects: 5.1 months; P = .02). All but 1 of the children with a head injury were less than a year of age. Although trends were noted among other variables, sample size limited our ability to demonstrate statistical significance of these differences. Notably unrelated to the presence of a head injury was the race of the child; there were fairly equal proportions of black versus white children among the head-injured versus noninjured populations.
Despite the universal recommendation of screening abused children less than 2 years of age with a skeletal survey for occult fractures, no comparable guidelines exist for detecting occult head injury in the same population. Previous recommendations have suggested screening with head CT or MRI in children with more serious and specific injuries for child abuse,5,8 but which injuries merit screening are unclear. Furthermore, it is unclear what sociodemographic factors might also be important in the consideration of screening this population.
To begin answering these questions, this study sought to estimate the prevalence of occult head injury in a high-risk cohort of abused children. In doing so, we found a high prevalence of occult head injury in over a third of a fully screened cohort of abused children under 2 years of age who were considered high-risk because of age (<6 months), rib fractures, multiple fractures, or facial injury. Although nearly three fourths of these children would have been detected by skeletal survey alone, 5 of 19 children (9.8% of the fully screened cohort) would have escaped detection without the addition of a head CT or MRI.
Given that the prevalence of occult head injury was high in this cohort, it is clear that we have identified potentially important risk factors associated with occult head injury in abused children. However, because these injuries served as selection criteria for this study, it is difficult to explore the independent association of these factors with head injury. Still, it is worth noting that only 1 child with a head injury was over 1 year of age (15 months), and if one excludes all children with skull fractures, the 5 children with isolated scalp and/or intracranial injuries were all under 6 months of age.
Although it is possible that this association between younger age and occult head injury may have resulted from a detection bias, such that the unscreened children were older than the fully screened children, there is prior literature to suggest that this finding may be important. A study of 173 children with symptomatic abusive head injury found that younger children were more likely to be “missed” on first contact with the medical system.6 Similarly, a recent review of head-injured infants younger than 2 years of age found that clinical signs of brain injury were insensitive as markers of intracranial injury in infants.9 Previous investigators have also found that the ability of physicians to perform neurologic screening examinations in young infants is often flawed,6,10 possibly providing 1 explanation of the poor utility of using clinical indicators to screen for occult head injury.
Another notable finding in this study was the significant number of children (14 of 19) with occult head injury who were found to have evidence of scalp injury by CT or MRI despite a normal physical examination. This finding is not surprising given that a previous investigation found that among 13 children with fatal abusive head injury, more than half had evidence of blunt trauma only after autopsy was performed.11 The data on scalp injury from this study merely support what many clinicians already know: that detecting scalp swelling in young children can be difficult.
In considering guidelines for screening abused children for occult head injury, one strategy might be to screen children with an ophthalmologic examination to detect retinal hemorrhages. This screening method would appear to have some merit because previous studies have estimated the prevalence of retinal hemorrhages in acute abusive head injury to be between 65% and 95%.1,7,12,13 However, this strategy is problematic because intuitively, one might not expect a child to have retinal hemorrhages in the setting of an occult head injury, particularly if the head injury had occurred several weeks earlier. Still, many physicians who regularly evaluate abused children will use this screening method in determining appropriate candidates for head CT or MRI. The finding in this study that none of 14 children with occult head injury had evidence of retinal hemorrhages confirms that retinal examination is a poor screening method for detecting occult head injury. Ophthalmologic examination remains important for the evaluation and care of the abused child with acute and/or symptomatic head injury.
One potential critique of developing screening recommendations for occult head injury in high-risk children is whether the identification of an additional injury in a child contributes added value when the child already has suspicious injuries and when, as in this study, the finding of additional injuries results in no further medical intervention. Although no child required surgical intervention for their occult head injuries, the detection of that injury, whether a significant intracranial injury or a simple scalp injury, both confirmed child abuse and allowed the child abuse team and regional authorities to better advocate for the safety of the child on discharge from medical care. In addition, more serious occult head injuries may dictate the need for follow-up rehabilitation or early intervention, as well as impact the ultimate neurodevelopmental outcome of the child.
Given the importance of confirming child abuse and influencing safety recommendations before medical discharge, we believe the finding of such a high prevalence of occult head injury in this study should influence guidelines regarding screening of this population. Based on our results, we would recommend screening with CT or MRI all children under 6 months of age with any evidence of physical maltreatment, and children under 1 year of age with high-risk injuries, including rib fractures, multiple fractures, and facial injury. Skeletal survey alone is not an adequate screening measure, as nearly 10% of the screened children with occult head injury escaped detection by skeletal survey alone. The choice of CT or MRI is also a debatable issue. Although some investigators have suggested that MRI is more sensitive in detecting the sequelae of abusive head injury,5,14,15 CT may be a practical reality in areas where MRI is not readily available.
Future investigation should seek to estimate the prevalence of occult head injury in a broader population of abused children, and in doing so, identify risk factors for occult head injury. This would help establish evidence for whether universal or targeting screening would be preferential in a lower risk population.
There were several limitations in interpreting the results of this study. The sample size was limited, and thus the analysis of the data were limited to describing characteristics of the population. The ability to study the independent association of risk factors with injury was limited not only by sample size, but also by the design of the study, where potential risk factors of interest were used as selection criteria for the population.
It is also possible that some children who were symptomatic were included in the cohort, thus increasing the prevalence we reported. However, we relied on the admission neurologic examinations by the emergency department and inpatient attending physicians, which typically would function as the “gold standard” evaluators at most institutions.
Finally, detection bias may have occurred in the CT/MRI interpretation by our radiologists, such that the radiologist who knows the specific injuries of a child may overreport subtle findings on the scans. We believe we minimized this bias by reviewing all positive and equivocal scans with a radiologist who was blinded to the clinical history of the children, besides knowing that we were trying to detect evidence of head injury. Although simply knowing that these children were abused might have contributed to bias in determination of the outcome, our results are strengthened by the complete agreement of the second reviewer with the initial radiologists.
This study is the first to estimate the prevalence of occult head injury in a high-risk cohort of abused children. Our finding of a relatively high prevalence of occult head injury in this cohort suggests the need for universal screening of similar high-risk abused children. These risk factors include the presence of rib fractures, multiple fractures, facial injury, and younger age (<6 months). Retinal examination was a poor screening method for detecting occult head injury, and should not be relied on as a screening tool. One should proceed directly to CT or MRI in high-risk populations. Further study is needed to estimate the prevalence of occult head injury in broader populations of abused children and refine risk factors associated with occult head injury.
This study was supported in part by a grant from the National Institutes of Health (NIH RO1NS 39679).
The corresponding author also wishes to thank the Center for Clinical Epidemiology and Biostatistics at the University of Pennsylvania School of Medicine for their support in developing the project.
- Received May 7, 2002.
- Accepted October 15, 2002.
- Reprint requests to (D.M.R.) Division of General Pediatrics, Children’s Hospital of Philadelphia, Room 2423, 34th St and Civic Center Blvd, Philadelphia, PA 19104. E-mail:
- ↵Billmire ME, Myers PA. Serious head injury in infants: accident or abuse? Pediatrics.1985;75 :340– 342
- ↵O’Neill JA, Meacham WF, Griffin JP, Sawyers JL. Patterns of injury in the battered child syndrome. J Trauma Inj Infect Crit Care.1973;13 :332– 339
- ↵American Academy of Pediatrics, Section on Radiology. Diagnostic imaging of child abuse. Pediatrics.1991;87 :262– 264
- ↵Ewing-Cobbs L, Kramer L, Prasad M, et al. Neuroimaging, physical, and developmental findings after inflicted and noninflicted traumatic brain injury in young children. Pediatrics.1998;102 :300– 307
- ↵Kleinman PK, Barnes PD. Head trauma. In: Kleinman PK, ed. Diagnostic Imaging of Child Abuse. St Louis, MO: Mosby, Inc; 1998:285–342
- ↵Greenes DS, Schutzman SA. Clinical indicators of intracranial injury in head-injured infants. Pediatrics.1999;104 :861– 867
- Copyright © 2003 by the American Academy of Pediatrics