Copyright 1999 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.1999
To determine the incidence of skull fracture (SF) and intracranial injury (ICA) among children younger than 2 years evaluated in a pediatric emergency department for head trauma; whether historical features and/or physical findings are predictive of injury type; and whether clinical criteria could allow a selective approach to radiographic imaging.
Retrospective medical record review.
Tertiary pediatric emergency department.
Case series of 278 children aged younger than 24 months evaluated for head injury.
Main Outcome Measures
Presence of SF and/or ICA.
Diagnoses at discharge included 227 minor head injuries, 39 isolated SF, 9 ICA with SF, and 3 isolated ICA. Children younger than 12 months had the highest incidence of SF/ICA (29%) vs 4% for children aged 13 to 24 months (P<.001). Seven percent of complications from SF/ICA resulted from falls greater than 3 ft (0.9 m). Incidence of behavioral change, loss of consciousness, emesis, and seizures did not differ significantly between those with minor head injuries and those with SF/ICA. Scalp abnormalities were more common in children with SF/ICA (P<.001). Sixty-two percent of children with isolated SF and 58% of children with ICA had no history of loss of consciousness, emesis, seizure, or behavioral change. Ninety-two percent of children with isolated SF and 75% of children with ICA had normal levels of consciousness and nonfocal neurologic examinations at diagnosis. Among children who fell 3 ft or more (≥0.9 m) and had no loss of consciousness, emesis, seizure, behavioral change, or scalp abnormality, none of 31 (95% confidence interval [CI], 0-0.10) children younger than 24 months and none of 20 (95% CI, 0-0.15) children younger than 12 months had SF/ICA.
Both SF and ICA are common in children younger than 2 years evaluated for head trauma. Children younger than 12 months are at highest risk. Injuries resulted from relatively minor falls and occurred in alert, neurologically normal children. Clinical signs and symptoms were insensitive predictors of SF/ICA; however, a grouping of features (fall ≤3 ft [0.9 m], no history of neurologic symptoms, and normal scalp physical examination results) identified a subset of children at low risk for complications.
HEAD TRAUMA is one of the most common childhood injuries, accounting for 600,000 visits to emergency departments and 95,000 hospital admissions per year.1- 3 The goal of clinicians is to identify the small subset of children at risk for complications of head trauma among the large number of patients who seek an evaluation. Skull radiographs and cranial computed tomography (CT) can accurately identify skull fractures (SFs) and intracranial injuries (ICAs), respectively; however, their indiscriminate use results in unnecessary costs and wasted resources. While guidelines for radiographic imaging have been developed for adults, most investigators have been hesitant to provide selective imaging guidelines for young children with head injuries.4- 8
Children younger than 2 years differ from older children and adults in several ways that would make a low threshold for imaging prudent. Clinical assessment is more difficult, the risk for abuse is higher, and SF from minor trauma are more common.2,4,5,9,10 Additionally, isolated SF that may not have clinical significance in adults or older children predisposes younger children to developing leptomeningeal cysts or "growing fractures"11- 13 and mandates a careful evaluation for intentional injury. As in older populations, the presence of an SF is associated with an increased risk of ICA,2- 4,9,14- 17 and ICA diagnosed prior to patient deterioration have better outcomes.18 These concerns have led to the current recommendations in the literature for the liberal imaging of all children with head injuries aged 2 years or younger (unless "trivial" injuries).3,5- 8 However, unlike adults, young children undergoing cranial CT imaging frequently require sedation, which carries potential risks.
Problems with the current recommendations for liberal use of imaging technology for children aged 2 years or younger following a head injury include the following: (1) there are no specific guidelines for which children require imaging (and "trivial" trauma is not defined); (2) recommendations are not clear as to what imaging modality is preferred; (3) these recommendations are based on little clinical data; and (4) these recommendations are not uniformly followed in practice. Certainly the child with a depressed level of consciousness or focal neurologic examination should undergo cranial CT to detect ICA. Less clear is which alert, neurologically intact children merit radiographic evaluation to identify complications.
Most previous studies of head injury in children aged 2 years and younger have studied populations of children selected for having undergone radiographic imaging or having been hospitalized, populations likely at higher risk for complications. We undertook a study of all patients younger than 2 years presenting to an emergency department for the evaluation of a head injury to determine (1) the incidence of SF and ICA in our emergency department's population, (2) whether historical features and/or physical findings predict complications of head trauma, and (3) whether clinical criteria could be developed to allow a selective approach to imaging for this population.
We reviewed medical records of previously healthy children aged 2 years and younger who had computer discharge diagnoses from the hospital or emergency department of head injury, SF, ICA, cerebral contusion, or cerebral edema and who were treated in the emergency department at Children's Hospital. Boston, Mass, from January 1992 to December 1992. Children with a primary diagnosis of scalp or facial laceration were not included without a secondary diagnosis of 1 of the above-listed types of head injury. Children with a previous history of blood dyscrasias, neurologic abnormalities, seizure disorders, or ventricular shunts were excluded. Institutional review board approval was obtained for this study.
Historical information, physical findings, radiological studies, surgical procedures, discharge diagnoses, and outcome were abstracted from the medical record. Age was rounded down to the nearest month. Heights for falls were estimated from historical information if not specified. Falls were estimated as follows: from sitting position, 1 ft (0.3 m); from a chair, another child's arms, standing position, or a children's bicycle, 2 ft (0.6 m); from a couch, bed, or baby swing, 3 ft (0.9 m); from an adult's arms, shopping cart, crib, table, or changing table, 4 ft (1.2 m); and from the top of a bunk bed, 6 ft (1.8 m). Falls down stairs were considered as a separate group with heights of 1 ft (0.3 m) per stair. Patients were considered to be referred if first examined by an outside physician who then sent the patient to our hospital's emergency department for further evaluation or management. Because application of the Glasgow Coma Score to young children remains unstandardized, level of consciousness was recorded as alert, responsive to verbal stimuli, responsive to painful stimuli, or unresponsive based on the nurses' and physicians' notes. Scalp abnormalities were considered present if a palpable step-off, ecchymosis, hematoma, laceration, soft-tissue swelling, or abrasion was noted. Children were given a diagnosis of minor head injury if they had a nonfocal neurologic examination, were alert at discharge, and had normal results of radiological studies (when obtained); SF if a fracture was identified on skull radiograph or cranial CT; basilar SF if clinical signs of a basilar SF were present (eg, hemotympanum), with or without radiological evidence of fracture; or ICA if an intracranial hemorrhage, cerebral contusion, or cerebral edema was identified on cranial CT.
Children were considered to have sustained inflicted injury if a referral was made to the department of social services and the ultimate diagnosis was felt to be an intentional injury. The group of children considered to have intentional injuries was excluded from all calculations involving historical information, as the validity of these histories is questionable.
Continuous and ordinal data were analyzed using the Student t test and Mann-Whitney analysis, respectively. Categorical data were analyzed by χ2 and the Fisher exact test. A stepwise logistic regression model was created using selected variables to determine their association with the dependent variable. Data were analyzed using the SPSS/PC (SPSS Inc, Chicago, Ill) and Epi-Info statistical packages. Values are reported as mean ± SD. P<.05 was considered significant. Sensitivity and specificity with 95% confidence intervals (CIs) were calculated using standard methods.
Two hundred ninety-one patients evaluated in the emergency department for head injuries during 1992 met criteria for entry into the study. Of these, 278 records (96%) were available for review; 223 patients came directly to the emergency department and 55 were initially evaluated at outside facilities. The mean age of patients was 11 ± 7 months, with a range of 0 to 24 months. There were 163 children (59%) younger than 12 months and 115 children (41%) 12 months or older. Diagnoses at discharge are summarized in Table 1. The overall incidence of SF/ICA was 18%, the incidence of SF/ICA for patients not referred was 8%, and the incidence of SF/ICA for patients referred was 60% (many referred patients were transferred after having SF/ICA diagnosed). Among the children initially presenting to the emergency department, 14 (6%) had isolated SF and 4 (2%) had ICA with or without SF.
Mechanisms of head injury are summarized in Table 2. Most of the children in this study (85%) presented for evaluation after a fall. The height of falls ranged from 1 to 30 ft (0.3-9.0 m) with a mean height of 4.4 ± 3.8 ft (1.3 ± 1.1 m) (Figure 1). As expected, increasing height of fall was associated with a higher incidence of SF/ICA (P<.001); however, low height of fall did not exclude SF/ICA. Of the 138 children who fell 3 ft (0.9 m) or more, 7% had an SF and/or ICA. There was no significant difference in the incidence of SF/ICA among those children sustaining injuries from falls down stairs as compared with those children sustaining injuries from free falls.
Distribution of skull fractures (SF) and intracranial injuries (ICA) by age for the varying heights of falls (n=227). Data are presented as number of cases/total number (percentage) unless otherwise indicated.
The mean age of children with SF/ICA was lower than that for those with minor head injury (6.9 ± 5 months vs 11.6 ± 7 months, P<.001). Twenty-nine percent of children younger than 12 months had SF/ICA, compared with only 4% of children aged 12 months or older (odds ratio [OR]=11.24; 95% CI, 3.72-38.05; P<.001) (Table 3). Even when excluding referred patients, the incidence of SF/ICA was 14% for children younger than 12 months compared with only 1% for children aged 12 months or older (Table 3). Only 3 of 100 children older than 12 months sustained SF/ICA from falls, with only 1 child reported to have fallen less than 3 ft (0.9 m). All 4 patients who had ICA with or without SF resulting from falls of 3 ft (0.9 m) or less were younger than 12 months.
Incidence of loss of consciousness, seizure, emesis, and history of behavioral change did not differ significantly between the children with minor head injury and those with SF/ICA (Table 4). Sixty-two percent of children with isolated SF and 58% of children with ICA had no history of loss of consciousness, emesis, seizure, or behavioral change.
Physical examination findings differed between those with minor head injury and those with SF/ICA. There was a significantly higher incidence of scalp abnormalities in children with SF/ICA (P<.001) (Table 5). Overall, 94% of SF/ICA were associated with a scalp abnormality, yielding a high sensitivity (0.94) and negative predictive value (0.97) for SF/ICA. Of those with normal results of scalp examination, 0 of 44 (95% CI, 0-0.07) children aged 12 months or older had SF/ICA as compared with 3 of 64 children younger than 12 months (2 isolated SF, 1 ICA associated with SF).
As expected, there was a significant association between depressed level of consciousness and SF/ICA (P<.05, Table 5); however, 92% of children with isolated SF and 75% of children with ICA with or without SF had normal levels of consciousness and nonfocal neurologic examinations. Two patients had neurologic abnormalities other than decreased level of consciousness, were critically ill at presentation, and had presumed inflicted injuries.
A stepwise logistic regression model for the subset of children (n = 227) who had sustained falls and had nonfocal neurologic examinations (which included a normal level of consciousness) (n = 227) was created with SF/ICA as the dependent variable to include the variables of age younger or older than 12 months, height of fall less than or greater than 3 ft (0.9 m), loss of consciousness, emesis, seizure, history of behavioral change, and scalp abnormality. The symptoms of depressed level of consciousness and lateralizing neurologic examination were excluded as variables for the logistical regression model because it was assumed that these children would unquestionably meet criteria for radiological evaluation. The combination of variables age younger than 12 months, height of fall greater than 3 ft (0.9 m), and scalp abnormality were the group of variables best predictive of SF/ICA (R2 = 0.40, P<.005). Among the subset of children who had sustained falls and had nonfocal neurologic examinations, this combination of features would only identify 23 (66%) of 35 children younger than 2 years with SF/ICA and 23 (72%) of 32 children younger than 12 months with SF/ICA. Conversely, we tried to identify a group of predictors that would exclude SF/ICA among the same subset of patients. Our study found a rate of no SF/ICA (95% CI, 0-0.10) among 31 children younger than 2 years and no SF/ICA (95% CI, 0-0.15) among 20 children younger than 12 months who fell 3 ft (0.9 m) or less and had no loss of consciousness, emesis, seizure, behavioral change, or scalp abnormality.
Thirty-nine percent (110/278) of children had a skull radiograph and/or cranial CT scan obtained. Twenty-nine percent (83/278) had skull radiographs, 25% (69/278) had cranial CT scan, and 14% (42/278) had both studies. The only significant difference in historical features or physical findings between those who did and did not undergo radiographic evaluation was the presence of a scalp abnormality (P<.001). Significantly more children younger than 12 months underwent radiological studies than those aged 12 months or older (86/163 vs 24/115, P<.001). Sixty-five percent (82/126) of radiographic studies obtained in children younger than 12 months were abnormal compared with 28% (8/29) for children aged 12 months or older (P<.05). Of children younger than 12 months, 51% (37/72) had positive findings on skull radiograph and 83% (45/54) had positive findings on cranial CT scan. Of children aged 12 months or older, 25% (3/12) had positive findings on skull radiograph and 29% (5/17) had positive findings on cranial CT scan.
Evaluation for possible nonaccidental injury was initiated in 75 children (27%). Nonaccidental trauma was considered likely in 14 (5%), of whom all had associated extracranial injuries consistent with abuse, admission of mechanism of injury, or concerns regarding previous inflicted injuries.
Forty-nine patients (18%) were admitted to the hospital: 7 of the patients with minor head injury, 31 of the patients with isolated SF, and 11 of the patients with ICA with or without SF. (The 1 patient discharged from the emergency department with an ICA presented to the emergency department more than 24 hours after injury.) One patient required intracranial pressure monitoring, no patients required other surgical procedures, and there was 1 known fatality. Four patients initially discharged from the emergency department were known to return for a second evaluation: 1 was admitted with the clinical diagnosis of a basilar SF (hemotympanum with a normal head CT scan) and the other 3 patients were all reevaluated and redischarged home with diagnoses of minor head injury.
We sought to study children younger than 2 years seen in a children's emergency department for head trauma based on the existing guidelines to strongly consider radiographic imaging of all children in this population. Although we found a high overall incidence of SF/ICA (18%) among children aged 2 years and younger, the incidence of complications of head injury was by far the greatest in the subset of children younger than 12 months. Even when excluding previously evaluated patients referred for known abnormalities, the incidence of SF/ICA was 14% among children younger than 12 months but only 1% among children between the ages of 12 and 24 months. These high incidences are in keeping with published data yet are even more striking, because most prior studies had selected for children who had undergone radiographic imaging or hospitalization, populations likely at higher risk for complications. Dietrich et al8 reported a 17% incidence of SF/ICA in 71 children younger than 2 years who had undergone cranial CT scanning for head injury and Ros and Cetta7 reported a 9% incidence of isolated SF in 35 asymptomatic children younger than 1 year who had undergone skull radiography. Additional reports of ICA (ie, isolated SF excluded) include Schunk et al,14 who found a 6% incidence of ICA among children younger than 2 years who had undergone cranial CT scanning; Shane and Fuchs,13 who found a 50% incidence of ICA among children younger than 13 months with SF who had undergone cranial CT scanning; and Hahn et al,10 who found an 11% incidence among 318 children younger than 36 months hospitalized for head injury. Our study's reaffirmation of the high incidence of complications in an outpatient population not selected for having undergone radiographic imaging should heighten the clinician's concern for SF/ICA, especially in the subset of children younger than 12 months.
ALSO STRIKING IS THE RELATIVE lack of signs and symptoms traditionally considered high risk for complications of head injury: 60% of children with SF/ICA resulting from accidental trauma had no history of loss of consciousness, emesis, seizure, or behavioral change and, of all patients studied, 92% of those with isolated SF and 75% of those with ICA with or without SF were alert with nonfocal neurologic examinations at diagnosis.
Although the incidence of complications was higher with increasing height of fall, SF/ICA occurred even with short falls (≤3 ft [0.9 m]); therefore, a "minor" mechanism did not exclude complications. Duhaime et al,19 in their study of 100 hospitalized patients with head injury aged younger than 2 years, concluded that simple falls from a low height rarely resulted in significant primary brain injury; however, they reported that SF were as likely to occur from a fall less than 4 ft (1.2 m) as from a fall greater than 4 ft (1.2 m) and that the 3 patients in their study with epidural hematomas all fell less than 4 ft (1.2 m). Ros and Cetta7 likewise reported that 3 of 35 children younger than 12 months who had fallen less than 3 ft (0.9 m) had isolated SF, and Greenes and Schutzman20 reported that 18 of 90 children with isolated SF resulting from a fall fell less than 3 ft (0.9 m) (17 of the 18 were children younger than 12 months and 1 of these injuries was a suspected nonaccidental trauma).
Given that complications of head injury may result from relatively minor falls, the clinician is faced with the question of which clinical parameters might identify those children at higher risk for SF/ICA. We found a significantly higher incidence of complications in children younger than 12 months than in those aged between 12 and 24 months. There was only one child older than 12 months who was not suspected of being abused and who fell less than 3 ft (0.9 m) who sustained an isolated SF (positive scalp abnormality but no historical symptoms). Although Leonidas et al4 reported a similarly higher incidence of SF in children younger than 1 year (14% vs 3% of older children), most studies group all children younger than 2 years together with recommendations for liberal imaging.3,5,6,8
Other than focal neurologic abnormalities and depressed level of consciousness (accounting for only a small minority of patients), no single symptom, group of symptoms, or physical finding could reliably predict SF/ICA. Factors traditionally considered high risk for complications of head trauma, such as loss of consciousness, emesis, seizure, and behavioral change, were not present in 62% of children with isolated SF and 58% of children with ICA (children with presumed intentional injuries excluded). Other studies have similarly reported insensitivity for clinical signs to reliably predict complications of head trauma.8,10,13,20- 22 Unlike historical features, the presence of a scalp abnormality was sensitive (although incompletely) for SF/ICA. Other investigators have also found an association between scalp abnormalities and SF/ICA.4,7,22 In children without scalp abnormality, our study found no SF/ICA in those aged 12 to 24 months (0/44; 95% CI, 0-0.07) and 3 cases of SF/ICA in those younger than 12 months. Greenes and Schutzman20 found that 4 (4%) of 101 children younger than 2 years with isolated SF had normal scalp examination results,20 and Shane and Fuchs13 found 4 (4%) of 102 children younger than 13 months with SF had no evidence of overlying scalp trauma. Therefore, scalp abnormalities had a high negative predictive value and the lack of a scalp abnormality markedly reduced the risk for SF/ICA, particularly in the older age group.
If one cannot reliably predict which patients will have SF/ICA, can we predict those patients without SF/ICA? Our study found that no child who fell less than 3 ft (0.9 m) and had no loss of consciousness, vomiting, seizure, behavioral change, or scalp abnormality had SF/ICA (95% CI, 0-0.10); however, the numbers are relatively small (n = 31). In a study of 322 subjects younger than 21 years with head trauma who received CT scans, Dietrich et al8 found that of those younger than 2 years (n = 71), 12 (17%) had SF/ICA; of these, all patients had a history of at least 1 suspicious feature (loss of consciousness, vomiting, seizure, depressed level of consciousness, or focal neurologic examination). Despite small sample sizes, it seems that children sustaining falls of 3 ft (0.9 m) or less without historical symptoms or scalp abnormalities (and no concerns of inflicted injury) may not require imaging studies. This may be especially true for the subset of children 12 months or older, who seem to be at lower risk for SF/ICA. Conversely, imaging studies should be considered for any child not fitting the above criteria and clearly for any child with suspicious symptoms.
Like older children and adults, any young child with features particularly suspicious for ICA should undergo CT scan. Although debatable, these features would include loss of consciousness, significant change in behavior, seizure, ongoing emesis, persistently depressed level of consciousness, focal neurologic examination, hemotympanum, suspected inflicted injury, or significant mechanism of injury (particularly if mechanism other than a fall). Because data suggest that most young children with ICA are either symptomatic or have an SF, one reasonable approach to imaging would be to use skull radiographs (which do not require sedation) as a screening tool in the child with a scalp abnormality who does not have features highly suspicious for ICA. Children with SF identified on skull radiographs would then undergo cranial CT scanning because these children are at increased risk for associated ICA.2- 4,9,14- 17 Additionally, children identified with SF could be referred for follow-up for potential growing fractures and, if deemed appropriate, more closely scrutinized with a social service evaluation. Using this strategy, patients with ICA would be missed only if there was no associated SF and the child had a normal physical examination result and no suspicious symptoms. In our study, this imaging approach would have identified all children with complications. Of the 3 patients with isolated ICA, 2 were symptomatic and the third was injured by a mechanism other than a fall (motor vehicle crash). Among other outpatient studies that report SF associated with ICA for children younger than 2 years, Schunk et al14 found no ICA not associated with SF; however, Dietrich et al8 found 1 isolated ICA among 71 children younger than 2 years who had undergone cranial CT scanning for head injury (need for surgical intervention and mechanism of injury unknown).
There are several limitations to our study, many inherent to its retrospective nature. The study design assumed that signs or symptoms not documented in the medical record were absent; however, it is possible that findings were present and incompletely documented. Likewise, because only 39% of children underwent radiographic imaging, the true incidence of complications (SF/ICA) may be underestimated and the incidence of findings associated with cranial complications may therefore be skewed, especially in the group of children aged 12 months or younger. There are no follow-up data available on most patients; thus, injuries not identified during the initial visit may have been missed. We screened for return visits to our institution but return visits to other institutions would have been missed. Another potential limitation is that, given the relatively small number of serious complications, the sample size may not be large enough to identify predictors of significant injury. Data obtained from a population presenting to a children's hospital emergency department (especially a referral center) may not be generalizable to patients in other outpatient settings. Likewise, given the large numbers of children sustaining injuries from falls, extrapolating these data to patients sustaining injuries from other mechanisms may be difficult. Larger, prospective studies are needed to further refine a selective imaging strategy of children younger than 2 years with head injury and to determine the long-term outcome of children with complications.
In summary, we found a high incidence of SF/ICA in a population of children younger than 2 years presenting to an emergency department after head injury. These complications were significantly more common in children younger than 12 months. Many complications resulted from minor falls and most children with complications were asymptomatic at presentation. Except for scalp hematomas, clinical signs and symptoms were insensitive predictors of SF/ICA; however, a grouping of features (fall ≤3 ft [0.9 m], no historical symptoms, and normal scalp examination result) identified a subset of children at low risk for complications who seem not to need radiographic evaluation.
Clinicians should have a high suspicion for SF/ICA in any child aged 2 years or younger sustaining a head injury; however, the threshold for obtaining an imaging study should be particularly low for the subset of children younger than 12 months. Cranial CT scanning would be appropriate for any symptomatic child but the skull radiograph may be an inexpensive adequate screening tool that does not require patient sedation for asymptomatic children who sustain falls and have no suspicion for inflicted injury.
Accepted for publication July 21, 1998.
Presented as a poster at the Ambulatory Pediatric Association and Society for Pediatric Research meeting, San Diego, Calif, May 7-11, 1995.
We thank Michael Shannon, MD, for his statistical support and Gary Fleisher, MD, for his valuable suggestions.
Corresponding author: Karen D. Gruskin, MD, Division of Emergency Medicine, Children's Hospital, 300 Longwood Ave, Boston, MA 02115.
Editor's Note: While this study does not provide crystal indications of low risk for complications, it begins to clear the murky waters.—Catherine D. DeAngelis, MD
Gruskin KD, Schutzman SA. Head Trauma in Children Younger Than 2 YearsAre There Predictors for Complications?. Arch Pediatr Adolesc Med. 1999;153(1):15-20. doi:10.1001/archpedi.153.1.15