Lucid Intervals - 2ndchairservices.com · Greenes D, Schultzman S.A. Occult intracranial injury in...

Lucid Intervals

Greenes D, Schultzman S.A. Occult intracranial injury in infants. Annals of Emergency Medicine 1998; 32(6):680-6.

The study looked at infants admitted to the emergency room of Children's Hospital Harvard (over a 6.5 year period). Occult (hidden) injuries (i.e. lucid intervals) were seen in fourteen of the 52 infants (27%) under the age of 6 months, 5 of 34 babies (15%) 6 months to a year and in none of the infants over one year old. 95% of the children had scalp contusions or hematomas, and 95% had fractures. None of the infants with occult injuries required medical assistance such as surgery, etc. to manage increased innercranial pressure.

6 8 0 A N N A L S O F E M E R G E N C Y M E D I C I N E 3 2 : 6 D E C E M B E R 1 9 9 8

Occult Intracranial Injury in Infants

David S Greenes, MDSara A Schutzman, MD

From the Division of EmergencyMedicine, Department of Pediatrics, Children’s Hospital,Harvard Medical School, Boston, MA.

Received for publication November 6, 1997. Revisionsreceived January 12 and May 14, 1998. Accepted forpublication June 5, 1998.

Presented at the Joint EmergencyMedicine Poster Symposium of theAmbulatory Pediatric Association and the Society for PediatricResearch, Washington DC,May 1997.

Address for reprints: David SGreenes, MD, EmergencyDepartment, Children’s Hospital, 300 Longwood Avenue, Boston, MA 02115; 617-355-6624,fax 617-355-6625,E-mail [email protected].

Copyright © 1998 by the AmericanCollege of Emergency Physicians.

0196-0644/98/$5.00 + 047/1/94147

P E D I A T R I C S / O R I G I N A L C O N T R I B U T I O N

Study objectives: The objectives of this study were as fol-lows: (1) to determine whether clinical symptoms and signs ofbrain injury are sensitive indicators of intracranial injury (ICI) ininfants admitted with head trauma, (2) to describe the clinicalcharacteristics of infants who have ICI in the absence of symp-toms and signs of brain injury, and (3) to determine the clinicalsignificance of those ICIs diagnosed in asymptomatic infants.

Methods: We conducted a retrospective analysis of all infantsyounger than 2 years of age admitted to a tertiary care pediatrichospital with acute ICI during a 61/2-year period. Infants wereconsidered symptomatic if they had loss of consciousness, his-tory of behavior change, seizures, vomiting, bulging fontanel,retinal hemorrhages, abnormal neurologic examination, depressedmental status, or irritability. All others were considered to haveoccult ICI.

Results: Of 101 infants studied, 19 (19%; 95% confidenceinterval [CI] 12%, 28%) had occult ICI. Fourteen of 52 (27%)infants younger than 6 months of age had occult ICI, comparedwith 5 of 34 (15%) infants 6 months to 1 year, and none of 15(0%) infants older than 1 year. Eighteen (95%) infants withoccult ICI had scalp contusion or hematoma, and 18 (95%) hadskull fracture. Nine (47%) infants with occult ICI received ther-apy for the ICI. No infants with occult ICI (0%) (95% CI 0, 14%)required surgery or medical management for increased intracra-nial pressure. Only 1 subject (5%) with occult ICI had any latesymptoms or complications: a brief, self-limited convulsion.

Conclusion: We found that 19 of 101 ICIs in infants admittedwith head trauma were clinically occult. All 19 occult ICIs occurredin infants younger than 12 months of age, and 18 of 19 hadskull fractures. None experienced serious neurologic deteriora-tion or required surgical intervention. Physicians cannot dependon the absence of clinical signs of brain injury to exclude ICI ininfants younger than 1 year of age.

O C C U L T I N T R A C R A N I A L I N J U R Y I N I N F A N T SGreenes & Schutzman

D E C E M B E R 1 9 9 8 3 2 : 6 A N N A L S O F E M E R G E N C Y M E D I C I N E 6 8 1

M A T E R I A L S A N D M E T H O D S

We performed a retrospective analysis of all patients youngerthan 2 years of age admitted to Children’s Hospital inBoston with an acute ICI between January 1, 1990, andJune 30, 1996. Patients were identified through a searchof the hospital’s medical records database. Initially, allpatients younger than 2 years of age with a discharge diag-nosis indicating any form of head injury were identified.The medical record was reviewed for each of these patients,and all patients who had an acute intracranial hemorrhage,cerebral contusion, or brain swelling identified on com-puted tomography of the head (head CT) were includedin the study population. Patients were excluded if they hadno identified acute ICI, or if they had a bleeding diathesis,prior intracranial lesion, ventricular shunt, or other med-ical problems that might make ICI more likely after headtrauma.

Data were collected from the patients’ charts regardingthe mechanism of injury, historical symptoms, physicalexamination findings, radiographic findings, hospitalcourse, discharge disposition, and condition on discharge.In those cases in which patients were transferred to ourhospital from an outside institution, data regarding thepatient’s initial presentation were abstracted from copiesof the referring institution’s records included in our hospital’schart. Information from follow-up visits (to NeurosurgeryClinic, the Emergency Department, or the hospital’s pri-mary care clinic), when available from the hospital chart,also was recorded.

Subjects were considered to have been victims of childabuse if the physicians’ or social workers’ notes in the chartindicated that this was the likely diagnosis on discharge.Those cases in which parents initially reported an acciden-tal injury but in which child abuse was ultimately suspectedwere reclassified as having child abuse as the mechanismof injury.

Symptoms and signs were recorded as “present” whenspecifically noted in the chart and “absent” when specificallydenied. When a finding was not mentioned in the chart, itwas recorded separately as “not documented.”

Subjects were considered to have a depressed mentalstatus when they were described as lethargic, listless, sleep-ingbut easily aroused, sleeping but responsive to verbal stim-uli, responsive to painful stimuli, or unresponsive. Subjectswere considered to have irritability when they weredescribed as irritable or crying inconsolably. Subjects wereconsidered to have a normal mental status if they did nothave a depressed mental status or irritability and weredescribedas alert, consolable, happy, playful, or appropriate.

[Greenes DS, Schutzman SA: Occult intracranial injury in infants.Ann Emerg Med December 1998;32:680-686.]

I N T R O D U C T I O N

Several authors recommend using clinical symptoms andsigns as a screen for determining which patients needradiographic imaging after head trauma.1-10 For severalreasons, however, many of these recommendations specif-ically exclude infants younger than 2 years of age.1,2,11

First, it has been recognized that infants are at increasedrisk of intracranial injury (ICI), and that infants may sufferICI even after more minor mechanisms of injury.1,12

Second, some authors express concern that clinical symp-toms and signs are less reliable in infants, because thesepatients are preverbal and have a more limited behavioralrepertoire.1,10 Third, because many of the studies of patientswith head injuries either specifically exclude infants orinclude only small numbers of infants in their samples,there is a paucity of data about the utility of clinical symp-toms and signs in assessing infants with head injuries.5,13

Because of these concerns, recommendations for theclinical management of infants with head injuries tend tobe more conservative than they are for older children andadults. For instance, Masters et al1 suggested that ageyounger than 2 years should in itself be considered a mod-erate risk factor for ICI after head trauma, and that clini-cians should have a very low threshold for radiographicimaging and hospital admission in such patients. Ros andCetta11 suggested that all infants presenting after headtrauma should have some form of radiographic imaging.Even more recent studies, which argue that clinical symp-toms and signs can detect the majority of ICIs in pediatricpatients, recommend that radiographic imaging be per-formed in infants even when symptoms of brain injury areabsent.2,10

These conservative recommendations are based, in part,on the limited availability of data about the clinical featuresof infants with ICI. Although previous authors differ as towhether or not they believe clinical features reliably identifythose infants with ICI, their data sets include small num-bers of infants with ICI.2,10,14 Our study was intended toaddress this issue by providing data to answer the follow-ing questions: (1) Are clinical symptoms and signs sensi-tive indicators of ICI in infants? (2) What are the clinicalfeatures of those infants in whom ICI is diagnosed in theabsence of symptoms or signs of brain injury? and (3)What is the clinical significance of those ICIs diagnosedin asymptomatic infants?



numeric coding for most study variables. When ambiguitiesarose in interpretation of the medical record, they wereresolved by discussion among the study authors. Chartsof all infants identified by the reviewing author as havingoccult ICI were then independently reviewed by the secondstudy author (SS), who verified the absence of the specificsigns and symptoms of ICI.

In addition, a subset of 10 randomly selected charts werereviewed by the second author, also a specialist in pediatricemergency medicine, to assess the reliability of the chartabstraction process. A κ statistic was calculated as a measureof interrater agreement.

R E S U L T S

Three hundred fifty-seven subjects younger than 2 yearsof age were admitted to Children’s Hospital with a dischargediagnosis related to head trauma over the study period.Of these, 182 subjects were excluded because they hadskull fracture but no intracranial injury. Another 74 sub-jects were excluded because they had no acute radio-graphic abnormalities noted, or because they had preexist-ing intracranial lesions or bleeding diathesis. The remaining101 subjects constituted the study sample (28% of alladmissions for head trauma in this age group). These sub-jects had a mean age of 7.1±5.8 months (range 3 days to23.5 months); 48 subjects (48%) were girls.

Overall, 82 (81%) of the 101 subjects had at least 1 ofthe symptoms or signs considered potential indicators of

All medical or surgical therapies related to the head injurywere recorded. Any late clinical symptoms or complications,requiring new or changed therapy after initial evaluationin the ED, were also noted.

From the group of all subjects with intracranial injury,2 subgroups were identified:

1. Symptomatic ICI Infants with 1 or more of a predeterminedset of symptoms or signs at the time of evaluation in the EDwere considered to have symptomatic ICI. The symptomsand signs that were considered indicators of potential ICIinclude: a history of loss of consciousness, history ofbehavior change, seizures, vomiting, depressed mentalstatus, irritability on examination, focal neurologic findings,asymmetric pupils, bulging fontanel, or retinal hemorrhages.

2. Occult ICI Infants with none of the symptoms or signsdescribed above were considered to have asymptomatic,or occult, injuries.

In determining whether patients had symptomatic oroccult ICI, only those historical features or physical exami-nation findings that were evident at the time of the EDevaluation were noted.

Results are expressed as mean±SD. Ninety-five percentconfidence intervals (95% CIs)were calculated by standardformulas and are included where relevant for frequency data.

All chart reviews were performed by 1 of the authors(DG), a specialist in pediatric emergency medicine, whowas not blinded to the purpose of the study. Reviews wereperformed with a standardized abstraction form, with

Table 1. Frequencies and documentation rates for symptoms and signs of head injury.

Symptomatic ICI (n=82) Occult ICI (n=19)Not Not

Symptom/Sign Present (%) Absent (%) Documented (%) Present (%) Absent (%) Documented (%)

Loss of consciousness* 21 48 32 0 95 5Behavior change 85 13 1 0 100 0Vomiting 37 38 26 0 100 0Seizures 21 32 48 0 32 68Depressed mental status 63 37 0 0 100 0Irritability 20 80 0 0 100 0Focal neurologic examination 30 70 0 0 100 0Abnormal pupils 15 85 0 0 100 0Bulging fontanel† 11 46 43 0 68 32Retinal hemorrhages‡ 12 16 72 0 37 63*In all cases in which loss of consciousness was not documented, there was either no history of trauma reported (as for the “unknown” mechanism category and many of the child abuse cases) or thetrauma was not witnessed.†The status of the fontanel was documented for 43 of 52 (83%) subjects younger than 6 months of age, and 17 of 49 (35%) subjects 6 months of age or older.‡The 71 subjects in the “not documented” category include 23 (32%) for whom it was documented that the fundi could not be adequately visualized.



victim of child abuse, 1 (5%) suffered a direct blow to thehead, and 1 (5%) was a passenger in a motor vehicle crash.

Local abnormalities on scalp examination (contusionor hematoma) were noted in 18 (95%) of the 19 subjectswith occult ICI. Skull radiographs were performed beforehead CT in 14 of 19 (74%) subjects with occult ICI. All 14of these skull radiographs indicated fracture. In all, skullfracture was noted among 18 of 19 (95%) subjects withoccult ICI.

ICIs were occult in 14 of 52 (27%) infants younger than6 months of age (95% CI 16%, 41%), 5 of 34 (15%) infantsbetween 6 months and 12 months of age (95% CI 6%,32%), and none of 15 (0%) infants 12 months of age orolder (95% CI 0%, 20%).

None of the subjects with occult ICI (0%) (95% CI 0,14%) had surgical procedures or required medical man-agement for increased intracranial pressure. In contrast,of the subjects with symptomatic ICI, 19 (23%) had surgical

intracranial injury and were considered to have symp-tomatic ICI (95% CI 72%, 88%]. Nineteen (19%) of thesubjects had none of these symptoms or signs and wereconsidered to have occult ICI (95% CI 12%, 28%). Thefrequencies and documentation rates for each of the individ-ual symptoms and signs of head injury are shown in Table 1.Specific descriptions of the age, mechanism of injury, headCT findings, mental status, and interventions performedfor each of the subjects with occult ICI are shown in Table 2.Overall, the 19 patients with occult ICI included 7 (37%)with subdural hematoma, 7 (37%) with cerebral contusion,6 (32%) with epidural hematoma, and 3 (16%) with sub-arachnoid hemorrhage (some subjects had more than 1type of ICI).

Of the 19 subjects with occult ICI, 6 (32%) had fallenfrom heights of greater than or equal to 3 feet, 4 (21%) hadfallen less than 3 feet, 4 (21%) fell down stairs, 2 (11%) hadunknown mechanisms of injury, 1 (5%) was a suspected

Table 2. Age, mechanism of injury, head CT findings, and description of mental status from the medical record for the 19 subjects with occult ICI.

Head CT Findings Mental Status Mechanism (as described by on Examination

Age of Injury attending radiologist) (as described in medical record) Interventions

3 days 5-foot fall “Small subdural and epidural hematomas. “Well-appearing, alert” NoneMild effacement of lateral ventricle.”

2 weeks 3-foot fall “Small subdural hematoma with maximum “Alert and active” Nonewidth 1 cm. Mild midline shift.Partial effacement of lateral ventricle.”

2 weeks 5-foot fall “Small subdural hematoma.” “Alert, active, feeding well” Anticonvulsants1 month Fall down 6 stairs “Small frontal hemorrhagic contusion.” “Alert, consolable, well-appearing” Anticonvulsants1 month 3-foot fall “Multiple small hemorrhagic contusions. “Alert, consolable” Anticonvulsants

Subarachnoid hemorrhage. Questionable midline shift.”2.5 months 5-foot fall “Interhemispheric subdural hematoma. “Alert, active, interactive” Anticonvulsants

Adjacent cerebral contusion versus shear injury.”3 months Fall down 9 stairs “Small subdural hematoma. Minimal midline shift.” “Alert and comfortable” None3 months Child abuse “Small contusion.” “Cheerful, alert, interactive” Anticonvulsants3 months 2.5-foot fall “Small epidural hematoma.” “Alert, active, smiling, appropriate” None3.5 months 3-foot fall “Small epidural hematoma.” “Alert, awake, vigorous” Anticonvulsants4 months 2.5-foot fall “Small hemorrhagic contusion.” “Alert and playful” None4.5 months 2.5-foot fall “Subarachnoid hemorrhage. Small contusion.” “Alert and happy” Anticonvulsants4.5 months Direct blow “Small epidural hematoma.” “Alert, playful, feeding well” Anticonvulsants5 months 2-foot fall “Small subdural hematoma.” “Alert and active” None6 months Motor vehicle crash “Multiple focal hemorrhagic contusions. “Awake, alert, consolable” None

Poor gray-white differentiation.”7 months Fall down 5 stairs “Subarachnoid hemorrhage.” “Alert, active” None8 months Fall down 7 stairs “Small epidural hematoma, maximum width 1 cm. “Alert and playful” None

Minimal midline shift. Slight effacement of lateral ventricle.”9 months Unknown “Epidural hematoma with maximum width 1 cm. “Alert, well-appearing” Anticonvulsants

Minimal midline shift.”10 months Unknown “Small subdural hematoma.” “Alert and playful” None



the neurologic examination, without the use of CT, andthey found that they missed no important injuries. Milleret al8 described a group of 1,382 adult patients with headinjuries, and noted that only 3% of patients with ICI had anormal neurologic status and that none of these neurolog-ically normal patients had lesions requiring any medicalor surgical intervention.

Conversely, many authors have suggested that abnor-malities on neurologic examination are not reliably presentin children with ICI. In several different studies, head-injured children with normal neurologic status at thetime of presentation to an ED had rates of ICI that rangefrom 3% to 6%.2,10,15 On the basis of these findings, Quayleet al10 and Dietrich et al15 recommend that all children withan abnormal neurologic examination after head injuryshould have head CT performed, but that head CT shouldalso be considered in children with other symptoms or signsof brain injury, such as loss of consciousness, vomiting,headache, drowsiness, or amnesia.

Our data suggest that even broader recommendationsfor radiographic imaging may be necessary to reliably detectICI in infants, because many infants with ICI have none ofthe typical symptoms of brain injury. These findings areconsistent with the data from Quayle et al,10 who foundthat 5 of 135 infants evaluated for head trauma had ICIwith scalp hematoma or contusion as their only abnormalfinding.

How, then, can these asymptomatic ICIs be detected?Previous investigators have suggested that assessing theforce of the mechanism of injury is not a sensitive meansof predicting which infants will have radiographicallyidentified injuries after head trauma. For example, Duhaimeet al12 found that skull fractures were as likely to occurfrom falls of less than 4 feet as they were from falls of greaterthan 4 feet. Similarly, we previously reported that 18% ofall skull fractures in infants resulted from falls of less than3 feet.16 Our current data suggest that ICIs, like skull frac-tures, may also result from low-impact injuries. In ourseries, 21% of infants with occult ICI had reported fallsfrom heights of less than 3 feet. This finding suggests that itwas not the force of the mechanism of injury that promptedclinicians to obtain radiographic imaging for many ofthese asymptomatic patients.

Although we cannot be certain from our retrospectivereview, it appears likely that it was the presence of abnormal-ities on physical examination of the scalp that motivatedradiographic imaging for many of the infants with occultICI. In our study 95% of the infants with occult ICI had localfindings of contusion or abrasion of the scalp. The fact that

procedures, and 23 (28%) had medical therapy for increasedintracranial pressure. Nine (47%) of the subjects with occultICI were treated with anticonvulsant medications comparedwith 32 (39%) of the subjects with symptomatic ICI.

Twenty-seven (33%) subjects with symptomatic ICIhad late symptoms or complications requiring new orchanged therapy after initial evaluation in the ED. In con-trast, only 1 subject (5%) (95% CI 1%, 28%) with occultICI had any late clinical symptoms or complications notedafter initial evaluation in the ED. This patient was a childwith a small cerebral contusion who had a brief generalizedseizure 48 hours after admission. The child subsequentlystarted anticonvulsant medication and remained wellthereafter.

All 19 (100%) (95% CI 86%, 100%) of the subjects withoccult ICI had normal neurologic examination results ondischarge from the hospital. Only 57 (70%) of the subjectswith symptomatic ICI had a normal neurologic status ondischarge from the hospital. Four (5%) of the children withsymptomatic ICI died, and another 21 (26%) had residualneurologic abnormalities.

Follow-up information was available for 13 (68%) of thesubjects with occult ICI and 56 (68%) of the subjects withsymptomatic ICI. All 13 of the subjects with occult ICI wereneurologically normal at the follow-up visit. Seventeen(30%) of the subjects with symptomatic ICI had persistentneurologic deficits at the time of the follow-up visit.

A random sample of 10 patient records was abstractedby a second investigator, and each of the symptoms and signsof head injury were recorded as “present,” “absent,” or “notdocumented.” The results of this review were comparedwith the data presented above, with an overall κ value of.94 for agreement between the 2 reviewers for all variablesstudied.

D I S C U S S I O N

We found clinical signs and symptoms to be insensitiveindicators of ICI in infants admitted after head injury; only81% of the infants in our sample had any of the symptomsor signs we studied. Occult ICIs were most common amongthe youngest infants, with 27% of the infants younger than6 months of age being asymptomatic.

These findings suggest that infants need to be evaluatedin a different manner than older children or adults after headtrauma. Increasingly in recent years, researchers havedetermined that neurologic examination reliably detectsmost clinically important ICIs in adults. Duus et al6 man-aged 2,204 patients with head injuries by relying only on



subdural hematoma, epidural hematoma, or intracerebralhematoma.18 We cannot know whether this anticonvulsanttherapy made any difference for the patients in our studysample. It is possible that some of these patients might havehad seizures, perhaps with complications, had they notbeen diagnosed and treated.

Our data indicate that most children diagnosed withoccult ICI will do quite well. Nonetheless, the data areinsufficient to prove that occult ICIs in infants are clini-cally unimportant. We believe it is reasonable, therefore,for clinicians to include strategies for detecting occult ICIsin their diagnostic evaluation of head-injured infants.

Some limitations of our study must be noted. First,because we limited our study to admitted patients, wemay have missed some patients who were diagnosed withICI and then discharged to home directly from the ED. Giventhat our usual clinical practice is to admit children withacute ICI to the hospital for observation, however, we believethat it is unlikely that many patients were missed by thisrestriction.

Another concern is that the retrospective nature of ourstudy may have limited the accuracy or completeness ofthe data. Our data on documentation rates, however, indicatethat documentation was generally thorough, especially forthose patients with occult ICI, in whom documentation wasvirtually complete. Documentation rates were somewhatlower for 3 of the features we studied—seizures, retinalhemorrhages, and bulging fontanel. It seems unlikely, how-ever, that a clinician would have discovered findings asdramatic as retinal hemorrhages, a bulging fontanel, or ahistory of seizures but then fail to document these facts inthe record. We believe, therefore, that our finding of manycases of occult ICI in infants is valid, and that it does notmerely represent poor or absent documentation.

It is important to note that there is no uniform protocolin our ED to guide the management of infants with headinjuries. Clinicians decide on an individual basis whethera head-injured child should or should not have radiographicimaging. This may introduce several biases into our data.Our finding that most cases of occult ICI are associatedwith scalp abnormalities, for instance, may merely reflectthe fact that it was only those patients with scalp abnormal-ities who had imaging performed. As another example, thefact that the youngest infants appear to be at the highestrisk for occult ICI may reflect the fact that clinicians aremore likely to perform imaging on asymptomatic patientsin this age group. Many infants with head injuries whowere seen in our ED during the study period were dischargedto home with no head imaging performed. Among this

74% of the occult ICI group had skull radiographs beforehead CT suggests that clinicians were concerned aboutpossible skull fracture in many of these patients. Clinicianslikely followed up an initial finding of skull fracture (notedin 100% of skull films performed in the occult ICI group)with a head CT to assess for associated ICI.

These data suggest that a significant fraction of ICIs ininfants may be detected by performing radiographic imag-ing on those infants who have signs indicative of possibleskull fracture. Quayle et al10 advocate such a screeningstrategy, recommending that head CT be performed on allinfants with any symptoms suggestive of brain injury, andthat all otherwise asymptomatic infants with scalp contusionor hematoma should have skull radiography performed(with head CT as follow-up for any child found to have askull fracture).

Other authors have suggested even broader indicationsfor radiographic screening for infants with head injuries,advocating radiographic imaging for all head-injured infantspresenting to an ED.2,11 However, our previous data indi-cate that skull fracture is uncommonly diagnosed in casesin which no scalp findings are present,16 as 97% of 101infants with isolated skull fracture were noted to have scalpabnormalitieson physical examination. Similarly, Kleinmanand Spevak17 found soft tissue swelling (as noted by headCT) to be virtually ubiquitous in cases of acute skull frac-ture. These findings—along with the observation that 95%of the subjects with occult ICI in our series had local scalpfindings—suggest that radiographic screening of head-injured infants without symptoms of brain injury shouldbe directed primarily toward those subjects with abnor-malities on scalp examination.

Although we have shown that ICI is commonly occultin infants, the clinical significance of occult ICIs in thesepatients is not entirely clear. If these patients have no signsor symptoms of brain injury, how important is it to make aradiographic diagnosis of intracranial injury? Our data indi-cate that clinical deterioration is rare among this group ofpatients, with only 1 of the subjects suffering a minor latecomplication, and none of the subjects requiring any sur-gical intervention or medical management of increased ICP.It should be noted, however, that our confidence intervalsare consistent with a rate of late clinical complications inthese patients that may be as great as 28%. Furthermore, itis important to recognize that clinicians provided anticon-vulsant therapy for 48% of the subjects with occult ICI.This practice was in keeping with published guidelineswhich consider prophylaxis for early posttraumatic seizuresto be a treatment option for patients with cortical contusion,

Jenny, C,. Hymel, K.P., Ritzen, A., Reinert, S.E. and Hay, T (1999) Analysis of missed cases of abusive head trauma. JAMA 281(7):621-626.

Interesting article that analyzes missed diagnosis of abusive head trauma, but in so doing, documents the fact that many head injuries go unnoticed only to result in further complications at a later date. This factor provides proof for the existence of a lucid interval and complicates prosecutor’s theories that the last person holding the baby was to blame. In this study 40% of the cases of head trauma resulted in complications from original injuries. The article sites the signs of pre-existing injuries. These conditions included seizure disorders, chronic vomiting and increasing head size because of increasing untreated subdural hematomas."

ORIGINALCONTRIBUTION

Analysis of Missed Casesof Abusive Head TraumaCarole Jenny, MD, MBALt Col Kent P. Hymel, MD, USAF, MCAlene Ritzen, MD, JDSteven E. Reinert, MSThomas C. Hay, DO

ABUSIVE HEAD TRAUMA (AHT) IS

a dangerous form of childabuse. More child abusedeaths occur from head inju-

ries than any other type of injury.1 In-fants and toddlers who survive AHT of-ten have serious neurologic sequelae.2,3

Head injury in infants and toddlers canbe difficult to diagnose because symp-toms are often nonspecific. Vomiting, fe-ver, irritability, and lethargy are com-mon symptoms of a variety of conditionsseen in children, including head trauma.When caretakers do not give a history ofinjury and the victim is preverbal, an abu-sive head injury can be mistakenly di-agnosed as a less-serious condition.

In March 1995, we evaluated a 14-month-old child who had sustained anabusive head injury 4 months previ-ously. Shortly after his initial injury, hehad been examined by his physician andhis new-onset seizures were attributedto his history of prematurity. During thenext 4 months, the child had 7 physi-cian visits and 2 cranial imaging stud-ies. At each visit, the diagnosis of AHTwas not recognized. When we exam-ined him 4 months later, he had mul-tiple old and new fractures and healingbrain injuries, including extensive brainatrophy and healing brain infarctions.This case encouraged us to review ourexperience with AHT cases to deter-mine if the appropriate diagnosis had

been previously missed. We also exam-ined factors that may have contributedto the unrecognized diagnosis of AHT.

METHODSWe studied cases of AHT in childrenyounger than3years evaluated at theChil-dren’s Hospital, Denver, Colo, from Janu-ary 1, 1990, through December 31, 1995.The Children’s Hospital is an academicmedical center affiliated with the Univer-sity of Colorado School of Medicine. It is

a referral center for Colorado, Wyoming,Montana, and western Nebraska.

The children in this study were evalu-ated by the hospital’s Child Advocacy andProtection Team (CAP Team). The CAPTeamis amultidisciplinarygroup that con-sults on cases of suspected child abuse andneglect. The team is led by pediatricianswhose clinical focus is child abuse. Socialworkers, nurses, psychologists, child psy-chiatrists, and attorneys also participate.The team routinely interviews caretakers

Author Affiliations: Department of Pediatrics, BrownUniversity School ofMedicine (Dr Jenny), and LifespanMedical Computing (Mr Reinert), Providence, RI; De-partment of Pediatrics, National NavalMedical Center,Bethesda,Md(DrHymel);DepartmentofPediatrics,Uni-versity of Oregon Health Sciences Center, Portland

Context Abusive head trauma (AHT) is a dangerous form of child abuse that can bedifficult to diagnose in young children.

Objectives To determine how frequently AHT was previously missed by physiciansin a group of abused children with head injuries and to determine factors associatedwith the unrecognized diagnosis.

Design Retrospective chart review of cases of head trauma presenting between Janu-ary 1, 1990, and December 31, 1995.

Setting Academic children’s hospital.

Patients One hundred seventy-three children younger than 3 years with head in-juries caused by abuse.

Main OutcomeMeasures Characteristics of head-injured children in whom diag-nosis of AHT was unrecognized and the consequences of the missed diagnoses.

Results Fifty-four (31.2%) of 173 abused children with head injuries had been seenby physicians after AHT and the diagnosis was not recognized. The mean time to cor-rect diagnosis among these children was 7 days (range, 0-189 days). Abusive headtrauma was more likely to be unrecognized in very young white children from intactfamilies and in children without respiratory compromise or seizures. In 7 of the chil-dren with unrecognized AHT, misinterpretation of radiological studies contributed tothe delay in diagnosis. Fifteen children (27.8%) were reinjured after the missed diag-nosis. Twenty-two (40.7%) experienced medical complications related to the misseddiagnosis. Four of 5 deaths in the group with unrecognized AHTmight have been pre-vented by earlier recognition of abuse.

Conclusion Although diagnosing head trauma can be difficult in the absence of ahistory, it is important to consider inflicted head trauma in infants and young childrenpresenting with nonspecific clinical signs.JAMA. 1999;281:621-626 www.jama.com

For editorial comment see p 657.

(Dr Ritzen); and the Department of Radiology, Univer-sity of Colorado School of Medicine, Denver (Dr Hay).Corresponding Author and Reprints: Carole Jenny,MD, MBA, Hasbro Children’s Hospital, MOC-140,593 Eddy St, Providence, RI 02903 (e-mail:[email protected]).

©1999 American Medical Association. All rights reserved. JAMA, February 17, 1999—Vol 282, No. 7 621

to document medical history and the his-tory of the acute injury, review previousmedical and social service records, re-view prior radiological studies, perform acareful physical examination, and orderappropriate new diagnostic studies. In allcases, organic illnesses thatmimicAHTareruled out. Confirmation that head traumawas inflicted requires multidisciplinaryteam consensus.

Head trauma cases were identifiedfrom the log records of the CAP Teamand charts were reviewed in depth. Toensure concurrence, study cases were re-viewed by at least 2 of the authors (in-cluding C.J.) and radiological imagingstudies were reviewed by a pediatric ra-diologist (T.C.H.). Permission for theanonymous chart review was granted bythe hospital’s human subjects commit-tee. Information gathered included de-mographics, social and family data, de-tails of the children’s injuries, presentingcomplaints, clinical course, and detailsof previous medical visits related to headtrauma, if applicable.

We limited the study to children withhead injuries who were younger than 3years for 2 reasons. First, children olderthan 3 years are not as likely to sustainsevere injury when struck in the head orshaken. Second, children older than 3years are more likely to be able to ar-ticulate their experiences. Hence, AHTis much less likely to be missed as theappropriate diagnosis.

Abusive head trauma was defined asinflicted cranial injury. Researchers de-bate whether shaking alone or shaking

and impact cause the signs and symp-toms commonly referred to as shakenbaby syndrome.4-6 The mechanism of in-jury cannot always be accurately deter-mined in child abuse cases. Because shak-ing, impact to the head, or both are allpotentially harmful to infants and tod-dlers, we grouped all head injuries causedby abuse into the single category of AHT.

Factors considered by the multidisci-plinary team in reaching the diagnosis ofAHT (rather than nonintentional head in-jury) included (1) confession of inten-tional injury by an adult caretaker; (2) in-consistent or inadequate histories givenbycaretakers (the history given did not ex-plain the nature and severity of the inju-ries); (3) associated unexplained inju-ries, such as fractures or intra-abdominalinjuries; and (4) delay in seeking care.

Cases of AHT were defined as missed ifreview of medical records and radiologi-cal studies confirmed the following pre-defined criteria: (1) Prior to the diagnosisof AHT, a physician evaluated the child(on �1 occasions) for nonspecific clini-cal sign(s) compatible with head trauma(ie, recurrent vomiting, irritability, facialand/or scalp injury, altered mental sta-tus, abnormal respiratory status, and/or sei-zures). (2) The medical evaluation(s) forthese nonspecific clinical sign(s) did notresult in a diagnosis of AHT. (3) Thereaf-ter, 1 or more of the following scenariosoccurred: (a) The child improved clini-cally, later experienced (repeat) acutetrauma confirmed as abusive, and under-went diagnostic imaging that revealed oldcranial injuries and other new injuries.(b)The child remained symptomatic or ex-perienced worsening clinical signs untilhead trauma was recognized, verified bycranial imaging studies, and confirmed asabusive. (c) The person who injured thechild later admitted to abusing the childshortly before the onset of the child’s non-specific clinical sign(s). In all cases, the es-timated age of the cranial injuries docu-mented by imaging studies was consistentwith the prior time of onset of the child’snonspecific clinical sign(s).

All remaining cases of AHT evalu-ated during the study period were con-sidered recognized. Children who sus-tained any new inflicted injuries during

the period of diagnostic delay were clas-sified as reinjured. Study patients whosemedical records after their inflicted headtrauma revealed abnormal head growth,recurrent seizures, psychomotor de-lays, chronic anemia, vomiting, weightloss, and/or sensory deficits were classi-fied as having medical complicationsof AHT.

We examined data to determine whatfactors were associated with a missed vsrecognized diagnosis. We used x2 test-ing to assess the independence of 10 vari-ables on the outcome variable of a cor-rect diagnosis of head trauma. Variablesresulting in x2 P�.25 or less were en-tered into an initial multivariate logisticregressionmodel.We thenusedWald andlikelihood ratio testing to iteratively re-move noncontributory variables from themodel.7 Analysis was performed usingStata software, Version 5.0 (Stata Corp,College Station, Tex).

RESULTSA total of 232 children with suspectedhead injuries were evaluated by the CAPTeam from January 1990 through De-cember 1995. Fifty-nine children did notmeet study criteria. Of these, 8 wereeliminated because they were aged 3years or older. It was determined that 38were not abused. The medical records of13 children could not be located. The re-maining study sample included 173abused children with head injuries.

The mean age of the 173 children was247 days (range, 10 days to 2.9 years).Ninety-five (55%) of the children weremale and78 (45%)were female. Theboys’ages at the time they were first seen forsymptoms of AHT were not significantlydifferent than the girls’ ages. In our studysample, minorities were overrepre-sented (33.5% minority) compared withthe racial distribution of the Denver met-ropolitan area (19.7% minority).8

The types of injuries noted in the chil-dren are shown in TABLE 1. Many of thechildren sustained more than 1 type ofinjury. Eighty-nine children (51.4%)were covered by Medicaid-funded in-surance programs. Twenty-seven chil-dren (15.6%) were uninsured. The re-mainder had private health insurance.

Table 1. Types of Injuries Sustainedby Study Population

Types of Injury No. (%)

Head injuries 173 (100)

Subdural hematoma 150 (86.7)

Diffuse parenchymal brain injury 77 (44.5)

Localized brain contusions orshearing injuries

64 (37.0)

Skull fracture 55 (31.8)

Epidural hemorrhages 4 (2.3)

Retinal hemorrhages 114 (65.9)

Facial or scalp trauma 98 (56.6)

Trauma to parts of body otherthan head or face

63 (36.4)

Fractures other than skull fractures 60 (34.7)

UNRECOGNIZED CASES OF ABUSIVE HEAD TRAUMA

622 JAMA, February 17, 1999—Vol 282, No. 7 ©1999 American Medical Association. All rights reserved.

Missed vs Recognized AHTIn the 173 children with AHT, 54 cases(31.2%) were classified as missed. Forchildren with missed AHT, the meannumber of physician visits before thetrauma was recognized was 2.8 (range,2-9 visits).

For children in whom the diagnosis ofAHT was missed, the mean length of timeto diagnosis of head trauma from the dayof the first visit was 7 days (range, 0-189 days). In 5 cases, the children wereseen twice in the same day and the di-agnosis was made on the second visit;hence, the designation of 0 days until di-agnosis in some cases of missed AHT.

When missed cases were comparedwith recognized cases, several factorswere found to be significantly different.

AgeChildren with missed AHT were muchyounger than those in whom the diagno-sis was recognized on the first physicianvisit. The mean age of missed AHT casesat the time of their first medical visit forhead injury symptomswas180days (95%confidence interval [CI], 125-236). Themean age of the recognized cases was 278days (95% CI, 228-328). The mean agesof children with missed and recognizedAHT were significantly different (inde-pendent samples t test, P = .02).

RaceAbusive head trauma was missed signifi-cantly more often in white children thanchildren of minority races. In white chil-dren,43 (37.4%)of115casesofAHTweremissed and inminority children, 11 (19%)of 58 were missed (Pearson x2, P = .01).

Family CompositionAbusive head trauma was more likely tobe missed in families in which both par-ents lived with the child. Thirty-seven(40.2%) of 92 cases were missed in in-tact families. In families in which themother and father of the child were notliving together, 14 (18.7%) of 75 caseswere missed (Pearson x2, P = .003).

Severity of Symptomsat Initial VisitNot surprisingly, the more severelysymptomatic children were more likely

to be recognized as having head traumaat first visit to the physician. TABLE 2summarizes the number and percent-age of children who were missed and rec-ognized as having AHT compared withtheir symptoms and signs. At the firstvisit, children who were comatose, whosebreathing was compromised, who werehaving seizures, or who had facial bruis-ing were more likely to be accurately di-agnosed. Conversely, children who pre-sented with irritability or vomiting at thefirst visit were less likely to be identi-fied as having AHT.

Factors Not Significantly DifferentSeveral factorswere foundnot to differ be-tween children with missed vs recog-nized AHT. These included whether theparents were employed, whether the par-ents had private insurance coverage, thesex of the child, the birth weight of thechild, andwhether thechildhadbeenbornprematurely (�37 weeks’ gestation).

Factors AssociatedWith Missed Diagnosis of AHTNine variables were found to be signifi-cantly associated with missing the diag-nosis of AHT by univariate analysis. These

were transformed to dichotomous vari-ables and entered into a logistic regres-sion model. They included age youngerthan 6 months, minority race, parents notliving together, and 6 signs and symp-toms noted at the first visit, including fa-cial injury, seizures, decreased mental sta-tus, abnormal respiratory status, vomiting,and irritability.Of these9variables, 4wereretained in themultivariate logisticmodel.These 4 independent variables predict-ing the correct diagnosis ofAHTat the firstvisit included (1) abnormal respiratory sta-tus (odds ratio [OR], 7.23; 95% CI, 2.4-21.3; P�.001); (2) seizures present (OR,6.67; 95% CI, 2.5-17.3; P�.001); (3) fa-cial and/or scalp injury present (OR, 4.81;95% CI, 2.1-11.0; P�.001); and (4) par-ents not living together (OR, 2.49; 95%CI, 1.1-5.7; P = .03).

Applying the logistic regression modelconstructed from the data, we found thatif none of these 4 factors were present,the probability that a physician wouldmake the correct diagnosis of AHT wasP = .20. That is, if a child had normal res-pirations, had no seizures, had no facialor scalp injury, and came from an intactfamily, the probability that AHT wouldbe recognized was less than 1 in 5.

Table 2.Missed and Recognized Abusive Head Trauma Cases: Severity of Presenting Symptoms

SymptomsNo. (%)

RecognizedNo. (%)Missed

�2

TestP

Value

Facial and/or scalp injuries 78/119 (65.5) 20/54 (37.0) 12.293 �.001

Other bodily trauma (not heador face trauma)

53/118 (44.9) 10/54 (18.9) 10.664 .001

Mental statusAwake and alert 35/119 (29.4) 35/54 (64.8)

Sleepy and/or lethargic 31/119 (26.1) 17/54 (31.5)31.397 �.001

Comatose and responsive to pain 21/119 (17.6) 1/54 (1.9)

Comatose and unresponsive to pain 32/119 (26.9) 1/54 (1.9)

Mental status by groupAwake and alert 35/119 (29.4) 35/54 (64.8)

19.326 �.001Depressed or comatose 84/119 (70.6) 19/54 (35.2)

Respiratory statusNormal breathing 45/119 (37.8) 44/54 (81.5)

Compromised 20/119 (16.8) 8/54 (14.8) 33.778 �.001

Requiring resuscitation or ventilation 54/119 (45.4) 2/54 (3.7)

Respiratory status by groupNormal 45/119 (37.8) 44/54 (81.5)

28.354 �.001Abnormal (compromised or requiring

resuscitation or ventilation)74/119 (62.2) 10/54 (18.5)

Seizures at first visit 55/119 (46.2) 8/54 (14.8) 15.820 �.001

Vomiting at first visit 42/111 (37.8) 30/54 (55.6) 4.637 .03

Irritable at first visit 53/111 (47.7) 34/52 (65.4) 4.426 .04



Misdiagnoses Appliedto Children With AHTThe 54 children with missed AHT re-ceived 98 diagnoses other than AHT dur-ing their 98 patient visits. TABLE 3 liststhe diagnoses applied to the children withmissed AHT. The most common diag-noses made were for viral gastroenteri-tis and accidental head injury. In somecases, the diagnoses were correct, eventhough coexistent head trauma was notrecognized. For example, in 1 case an in-fant was accurately assessed to have a ret-ropharyngeal abscess, but the accompa-nying subdural hematoma, retinalhemorrhages, and skull fracture were notrecognized. In other cases, the symp-toms of head trauma were attributed toconditions other than AHT. In 10 cases,the wrong diagnosis was applied morethan once to the same child. We did notcount these repeated diagnoses on ourfrequency table.

Outcome and ConsequencesTwenty-five (14.5%) of the 173 childrendied as a result of their head injuries. Ofthe recognized AHT cases, 20 (16.8%) of119 children died. In the missed AHT

cases, 5 (9.3%) of 54 children died. Thepercentage of children in the missed AHTgroup who died was not statistically dif-ferent than in the recognized AHT group(x2 = 1.712; P = .19). In our estimation,4 of the 5 deaths in the missed AHT groupmight have been prevented by earlier rec-ognition of abuse (TABLE 4).

Of the missed AHT cases, 15 (27.8%)of the 54 children were known to havebeenreinjuredbecauseof thedelay indiag-nosis. Twenty-two children (40.7%) hadmedical complications related to thedelayin diagnosis. These conditions includedseizure disorders, chronic vomiting, andincreasing head size because of increas-ing untreated subdural hematomas.

Radiological MisdiagnosisIn 7 of the children whose diagnosis ofAHT was missed, radiological errors con-tributed to the delay. These 7 childrenhad 8 studies in which trauma wasmissed, including 6 computed tomog-raphy scans of the head, 1 skeletal sur-vey, and 1 long-bone radiograph of thearm. The 2 longest delays in diagnosis(141 days and 174 days) and 6 of 25cases in which the diagnosis of AHT was

missed for longer than 7 days involvedradiological misreadings. TABLE 5 sum-marizes the nature of the errors made and

Table 3. Frequent Erroneous Diagnoses Madein Cases of Missed Abusive Head Trauma*

DiagnosisNo. of Times

Diagnosis Made

Viral gastroenteritis orinfluenza

14

Accidental head injury 10

Rule out sepsis 9

Increasing head size 6

Nonaccidental trauma(not head injury)

4

Otitis media 5

Seizure disorder 5

Reflux 3

Apnea 3

Upper respiratory tractinfection

2

Urinary tract infection orpyelonephritis

2

Bruising of unknown origin 2

Hydrocephalus 2

Meningitis 2

*Incorrect diagnoses made only once included anxiety,bronchiolitis, colic, complications of prematurity,constipation, failure to thrive, fever of unknown cause,hemiparesis, milk allergy, myositis, pneumonia,postmeningitic subdural effusion, retropharyngealabscess, rule out osteomyelitis, sudden infant deathsyndrome, torticollis, urticaria, viral encephalitis, andvomiting of unknown cause.

Table 4. Clinical Presentations of 4 Potentially Preventable Deaths With Missed AHT*

PatientAge, mo

Time BetweenVisits Documented Clinical Signs Evaluation Results Diagnosis

18 First visit Vomiting, sleepy, normal respirations, facialbruising

None Influenza

7 Days afterfirst visit

Vomiting, alert and responsive, normalrespiration, new bruising

None Otitis media


Vomiting, coma, unresponsive to pain,respiratory arrest

Retinal hemorrhages, subdural hemorrhage,focal brain injury, diffuse brain injury,noncranial trauma

AHT

2 First visit Failure to thrive, vomiting, alert and responsive,normal respiration, bruising to face andchest

Normal computed tomography result withmissed subdural hemorrhage and brainshearing tears

Apnea

7 141 Days afterfirst visit

Seizures, coma, unresponsive to pain,respiratory arrest

Retinal hemorrhages, skull fracture, subduralhemorrhage, diffuse brain injury, noncranialtrauma, old cranial trauma

AHT

5 First visit Vomiting, irritability, sleepiness, normalrespiration, “went limp”

None Anxiety secondaryto new day care


Vomiting, diarrhea, irritability, alert andresponsive, normal respiration

None Acute gastroenteritis


Vomiting, irritability, coma, unresponsive to pain,seizures, cardiorespiratory arrest

Retinal hemorrhages, subdural hemorrhages,diffuse brain injury

AHT

3 First visit Vomiting, irritability, alert and responsive, normalrespiration, dehydration

None Acute gastroenteritis


Coma, unresponsive to pain Retinal hemorrhage, subdural hemorrhage,diffuse brain injury, old brain injury, oldcranial trauma

AHT

*In all cases of missed abusive head trauma (AHT), the estimated age of cranial injuries documented by imaging studies was consistent with the time of onset of the child’snonspecific clinical sign(s) before his/her first physician visit.



the time in delay of diagnosis attributedto the radiological misreading.

COMMENTIt is difficult to study the cases of childabuse that clinicians do not recognize.In 1972, Jackson9 reviewed traumatic in-juries in children at King’s College Hos-pital in London, England, and found 18of 100 cases to have been missed casesof child abuse. O’Neill et al10 reported aseries of 110 battered children in 1973.Eighty percent of those children had signsof prior injury. Alexander et al11 foundphysical evidence of previous headtrauma in 8 of 24 children evaluated forhead injury due to shaking. Ewing-Cobbs et al12 discovered signs of preex-isting brain injury in 45% of childrenwith inflicted traumatic brain injury com-pared with none in children with acci-dental traumatic brain injury.

Incidental cases of missed child abusehave been published.13 In their study ofabusive head injuries, Benzel and Had-den mention that 9 of 23 abused chil-dren with head injuries “. . . were knownto have been seen by other physicians be-cause of similar problems or other inju-ries consistent with child abuse.”14 Sincethen, an increased awareness of childabuse has occurred, but similar studieshave not been reported.

We do not know how many cases ofAHT are never detected. Surely, the inju-ries occurring from impact or shaking rep-resent a range of severity, from no inju-ries to mild concussion or small subduralhemorrhage, severe brain damage, exten-sive intracranial bleeding, and cerebraledema. Caffey15 speculated in 1972 thatmany childrenwhoare found tohavemildneurologic abnormalities and learningdis-abilities may have been victims of AHT.

Parents who confess to shaking orhitting the heads of their children fre-quently report doing the same thingpreviously. In 1 study case, an infantwas hospitalized 3 times before some-one witnessed the child being shakenviolently. On 1 occasion, he was evalu-ated and treated for possible sepsis.The other 2 hospitalizations were forapnea and reflux, respectively. Thechild’s father admitted to multiple epi-

sodes of shaking that led to the infant’svarious illnesses.

In the current study, we found that31.2% of children who were clinicallysymptomatic after AHT were misdiag-nosed as having other conditions. Infantshave few ways to demonstrate illness orinjury. Nonspecific signs, such as vomit-ing, fever, and irritability, are seen in amyriad of conditions, including many be-nign, self-limited illnesses. The diffi-culty, then, is to be able to discern whenthese signs and symptoms indicate poten-tially serious or fatal pathology.

The possibility exists that in some ofthe visits we classified as missed, thechild had not yet been injured. How-ever, in another study by our group, wefound that patients became symptom-atic immediately after their injuries in37 cases in which perpetrators admit-ted to causing head injuries in infants.16

To guard against misclassification, weexamined the medical records ex-tremely carefully to correlate clinicaland radiological findings.

Not surprisingly, the infants and tod-dlers in our study whose head injurieswere misdiagnosed were overall less illthan those whose head injuries were rec-

ognized. The fact that they were not asill made the diagnosis of AHT difficult.Also, the children whose AHT wasmissed were, as a group, younger. Thedifficulty of diagnosing serious illness orinjury in young infants is complicated bythe limited range of their normal behav-ior. With less-sophisticated behavioraland neurologic signs to assess, thechanges in young infants with head in-juries are more difficult to detect.

Strikingdifferenceswere seen in the raceand family composition of infants withmissed and recognized injuries. Infantswith recognized AHT were more likely tobe minority children or children whosemothers and fathers were not living to-gether. We speculate that this may repre-sent a subtle bias indecisionmakingbasedon the physician’s assessment of risk. Aphysician examining a white child froman intact family may be less likely to thinkabout the possibility of child abuse. An-other hypothesis is that perhaps minor-ity and single-parent families were morelikely to obtain care from public clinics orhospital emergency departments, wherephysicians may be more attuned to abuseissues. In the current study, the childrenof intact, 2-parent households were much

Table 5. Radiological Errors in Cases of Missed Abusive Head Trauma*

CaseNo.

Visit No. in WhichRadiological Error

Was Made Nature of Misdiagnosis

Length of Delay inDiagnosis Due to

Radiological Error, d

1 First visit of 2 Result of CT of head read as normal; CTshowed subdural hemorrhage and shearingtears of the parenchyma

141

2 Third visit of 4 Result of CT of head read as consistent withinternal hydrocephalus; CT showed subduralhemorrhage

1

3 Second visit of 3 Result of CT of head read as normal; CTshowed subdural hemorrhage

4

4 First visit of 2 Result of skeletal survey read as normal; childhad a metaphyseal fracture of the tibia andunilateral periosteal elevation of the samebone

11

5 Second visit of 3 Result of CT of head read as normal; CTshowed subdural hemorrhage

4

6 First visit of 2 Result of CT of head read as normal; CTshowed subdural hemorrhage

51

7 Second visit of 9 Result of CT of head read as normal; CTshowed subdural hemorrhage and shearingtears of the parenchyma

174

Fifth visit of 9 Long-bone radiographs of both arms read asconsisent with myositis; x-ray film showedextensive periosteal reaction of both humeriand metaphyseal fractures of proximalhumeri bilaterally

74

*CT indicates computed tomography.





15. Dietrich AM, Bowman MJ, Ginn-Pease ME, et al: Pediatric head injuries: Can clinical factorsreliably predict an abnormality on computed tomography? Ann Emerg Med 1993;22:1535-1540.

16. Greenes DS, Schutzman SA: Infants with isolated skull fracture: What are their clinical char-acteristics, and do they require hospitalization? Ann Emerg Med 1997;30:253-259.

17. Kleinman PK, Spevak MR: Soft tissue swelling and acute skull fractures. J Pediatr 1992:737-739.

18. Brain Trauma Foundation: The role of antiseizure prophylaxis following head injury. JNeurotrauma 1996;13:731-734.

group, there were likely some patients who had occult ICIsthat were never diagnosed. Accordingly, it is uncertain howrepresentative our data are of the entire population of infantswith occult ICI.

Finally, our study is limited by the fact that we do not havedata on those infants presenting after head trauma who didnot have ICI. Evaluation of the costs and benefits of anystrategy for screening of head-injured infants would requiredata on all patients who would be subjected to the screeningprocess.

A prospective study of all infants with head injuriespresenting to an ED that uses a uniform protocol to deter-mine which infants receive radiographic imaging mightbe helpful to address some of these limitations.

In summary, we have found that 19 of 101 ICIs in infantsadmitted with head trauma were clinically occult. All 19infants were younger than 12 months of age, and 18 of 19had fractures on skull radiographs. None had serious neu-rologic deterioration or required surgical intervention.Physicians cannot depend on the absence of clinical signs ofbrain injury to exclude ICI in infants in the first year of life.

R E F E R E N C E S1. Masters SJ, McClean PM, Arcarese JS, et al: Skull x-ray examinations after head trauma:Recommendations by a multidisciplinary panel and validation study. N Engl J Med 1987;316:84-91.

2. Lloyd DA, Carty H, Patterson M, et al: Predictive value of skull radiography for intracranialinjury in children with blunt head injury. Lancet 1997;349:821-824.

3. Bonadio WA, Smith DS, Hillman S: Clinical indicators of intracranial lesion on computedtomographic scan in children with parietal skull fracture. Am J Dis Child 1989;143:194-196.

4. Chan KH, Yue CP, Mann KS: The risk of intracranial complications in pediatric head injury:Results of multivariate analysis. Childs Nerv Syst 1990;6:27-29.

5. Davis RL, Hughes M, Makela M, et al: Cranial computed tomography scans in children afterminimal head injury with loss of consciousness. Ann Emerg Med 1994;24:640-645.

6. Duus BR, Lind B, Christensen H, et al: The role of neuroimaging in the initial management ofpatients with minor head injury. Ann Emerg Med 1994;23:1279-1283.

7. Hahn YS, McGlone DG: Risk factors in the outcome of children with minor head injury.Pediatr Neurosurg 1993;19:135-142.

8. Miller EC, Derlet RW, Kinser D: Minor head trauma: Is computed tomography always neces-sary? Ann Emerg Med 1996;1996:290-294.

9. Mitchell KA, Fallat ME, Raque GH, et al: Evaluation of minor head injury in children. JPediatr Surg 1994;29:851-854.

10. Quayle KS, Jaffe DM, Kupperman N, et al: Diagnostic testing for acute head injury in chil-dren: When are head computed tomography and skull radiographs indicated? Pediatrics 1997;99.Available from: URL:http://www.pediatrics.org/cgi/content/full/99/5/e11.

11. Ros SP, Cetta F: Are skull radiographs useful in the evaluation of asymptomatic infants fol-lowing minor head injury? Pediatr Emerg Care 1992;8:328-330.

12. Duhaime AC, Alario J, Lewander J, et al: Head injury in very young children: Mechanisms,injury types, and ophthalmologic findings in 100 hospitalized patients younger than 2 years ofage. Pediatrics 1992;1992:179-185.

13. Godano U, Serracchioli A, Servadei F, et al: Intracranial lesions of surgical interest in minorhead injuries in paediatric patients. Childs Nerv Syst 1992;8:136-138.

14. Shane SA, Fuchs SM: Skull fractures in infants and predictors of associated intracranialinjury. Pediatr Emerg Care 1997;13:198-203.

more likely tohaveprivate insurance (Pear-son x2, 23.953; P�.001). In addition,white families were much more likely tohaveprivate insurance thanminority fami-lies (Pearson x2, 5.148; P = .02). How-ever, we did not collect data on the prac-tice setting inwhichmissedandrecognizeddiagnoses were made.

Are missed cases of AHT inevitable?If a child’s caretakers cannot or will notgive an accurate history, making the cor-rect diagnosis is extremely difficult. Phy-sicians cannot obtain cranial computedtomographic scans for every infant andtoddler who presents with vomiting, ir-ritability, and fever. Based on this studyand on our experience with these cases,we make the following suggestions to fa-cilitate the diagnosis of AHT.

1. Be alert for bruises or abrasions onthe faces or heads of children presentingwith nonspecific symptoms. In 20 of 54missed AHT cases in this study, facial orhead bruising was attributed to acciden-tal injury unrelated to the presenting ill-ness symptoms. One study of bruising inhealthy, nonabused children found nobruises on children who were not yetstrong enough to pull to standing.17 Thepresence of bruises in infants raises thepossibility of inflicted injury.

2. When evaluating infants and tod-dlers with nonspecific symptoms, suchas vomiting, fever, or irritability, con-sider head trauma in the differential di-

agnosis. Perform a head-to-toe physicalexamination, palpate the fontanelles,measure the head circumference, and bealert for signs of trauma.

3. When collecting spinal fluid incases of suspected infantile sepsis, ex-amine any bloody cerebrospinal fluid forxanthochromia. A supernatant of a spi-nal fluid contaminated by blood second-ary to a traumatic procedure should beclear in color if the specimen is exam-ined shortly after it is collected. Xantho-chromic spinal fluid can represent oldblood in the cerebrospinal fluid from pre-vious trauma, although it is not specificfor an intracranial bleed.18-20

4. Pediatrically trained radiologistsshould be consulted to interpret x-rayfilm and computed tomographic im-ages in cases of suspected child abuse.

In addition to these suggestions,other as yet unvalidated strategies todetect occult abuse could be consid-ered. Dilated retinal examinations ininfants and children with nonspecificsymptoms of illness could increase therecognition of retinal hemorrhages.Retinal hemorrhages have been re-ported in the majority of children whoare victims of AHT.21 Other possibili-ties need further research. Some mark-ers of brain trauma are known to crossthe blood-brain barrier, such as the BBfraction of creatine kinase. If rapid testswere available for such markers, a

simple blood test possibly could bedone to detect occult trauma. In a re-cent study by Hymel and colleagues,22

children with traumatic parenchymalbrain injury were frequently noted tohave prolonged prothrombin and par-tial thromboplastin times. These testsare generally available and inexpensiveto run. Their sensitivity and specificityas screening tests for head trauma in in-fants are not known.

There are other ways for AHT to pre-sent clinically that we did not see in thisgroup of patients. The list of signs andsymptoms we examined is not univer-sally inclusive. Another limitation of ourmethod is that the study was done ret-rospectively through record review.However, this seems to be the only op-tion we currently have for examining di-agnostic errors. Finally, information con-cerning the training, experience, orpractice setting of the physicians evalu-ating these patients was not obtained.

Although it is difficult to detect all se-rious AHT in the clinical setting, anawareness of the nonspecific nature of thesigns and symptoms of AHT, particu-larly in less-serious cases, could in-crease the likelihood that more cases willbe detected.

Disclaimer: The opinions and conclusions in this ar-ticle are those of the authors and are not intended torepresent the official positions of the US Air Force, USDepartment of Defense, or any other governmentalagency.

REFERENCES

1. Levitt CJ, Smith WL, Alexander RC. Abusive headtrauma. In: Reece RM, ed. Child Abuse: Medical Di-agnosis andManagement. Philadelphia, Pa: Lea & Fe-biger; 1994:1-22.2. HadleyMN, Sonntag VKH, Rekate HL, Murphy A.The infant whiplash-shake injury syndrome: a clinicaland pathological study.Neurosurgery. 1989;24:536-540.3. Sinal SH, Ball MR. Head trauma due to child abuse:serial computerized tomography in diagnosis andman-agement. South Med J. 1987;80:1505-1512.4. DuhaimeAC, Gennarelli TA, Thibault LE, Bruce DA,Marguilies SS, Wiser R. The shaken baby syndrome:a clinical, pathological, and biomechanical study. J Neu-rosurg. 1987;66:409-415.5. Alexander R, Sato Y, Smith W, Bennett T. Inci-dence of impact trauma with cranial injuries ascribedto shaking. AJDC. 1990;144:724-726.6. Duhaime AC, Christian CW, Rorke LB, Zimmer-man RA. Nonaccidental head injury in infants: the“shaken baby syndrome.” N Engl J Med. 1998;338:1822-1829.7. Hosmer DW Jr, Lemeshow S. Applied LogisticRegression. New York, NY: John Wiley & Sons Inc;1989.

8. US Bureau of the Census. 1990 Census of Popu-lation and Housing, Summary Tape File 1.Washing-ton, DC: US Government Printing Office; 1991.9. JacksonG.Child abuse syndrome: the caseswemiss.BMJ. 1972;2:756-757.10. O’Neill JA Jr, Meacham WF, Griffin JP, SawyersJL. Patterns of injury in battered child syndrome.J Trauma. 1973;13:332-339.11. Alexander A, Crabbe L, Sato Y, Smith W, Ben-nett T. Serial abuse in children who are shaken. AJDC.1990;144:58-60.12. Ewing-Cobbs L, Kramer L, Prasad M, et al.Neuroimaging, physical, and developmentalfindings after inflicted and noninflicted traumatic braininjury in young children. Pediatrics. 1998;102:300-307.13. Diamond P, Hansen CM, ChristofersenMR. Childabuse presenting as a thoracolumbar spinal fracturedislocation: a case report. Pediatr Emerg Care. 1994;10:83-86.14. Benzel EC, Hadden TA. Neurologic manifesta-tions of child abuse. South Med J. 1989;82:1347-1351.15. Caffey J. On the theory and practice of shakinginfants: its potential residual effects of permanent brain

damage and mental retardation. AJDC. 1972;124:161-169.16. Starling SP, Holden JR, Jenny C. Abusive headtrauma: the relationship of perpetrators to their vic-tims. Pediatrics. 1995;95:259-262.17. Wedgwood J. Childhood bruising. Practitioner.1990;234:598-601.18. Kortbeek LH, Booij AC. Bilirubin excess, eryth-rophages and siderophages in differentiation of bloodin cerebrospinal fluid. Clin Neurol Neurosurg. 1979;81:265-279.19. Resurreccion EC, Rosenblum JA. Common causesof spurious xanthochromia in cerebrospinal fluid.An-giology. 1972;23:105-110.20. Spear RM, Chadwick D, Peterson BM. Fatalitiesassociated with misinterpretation of bloody cerebro-spinal fluid in the “shaken baby syndrome” [letter].AJDC. 1992;146:1415-1417.21. SmithWL, Alexander RC, Judisch GF, Sato Y, KaoSC. Magnetic resonance imaging evaluation of neo-nates with retinal hemorrhages. Pediatrics. 1992;89:332-333.22. Hymel KP, Abshire TC, Luckey DW, Jenny C. Co-agulopathy in pediatric abusive head trauma. Pedi-atrics. 1997;99:371-375.



When considering patient selection criteria and care plan-ning, clinicians must further recognize the importance of hav-ing an experienced team. Naylor et al have spent at least 2 de-cades in this field. Comparing this study with their earlierwork,14 it appears that results improved when the interven-tion was lengthened from 2 to 4 weeks and home visits wereadded. Since it is hard to predict which patients will con-tinue to have an unstable course, one wonders whether con-tinued involvement would have additional value. More stud-ies are needed to address this question.

The need for improved discharge planning and postacutecare management has been well recognized,6,15 yet limited medi-cal oversight and lack of active physician participation dur-ing transitional care and ongoing chronic care remain a seri-ous problem.16 Efforts must extend beyond simply creating

hospital discharge plans. Acute care must be truly linked withpostacute care. Naylor et al show that value is gained whenmedical decision makers, in this case advanced practice nurses,work closely with patients, caregivers, and other practition-ers during the transition, then visit patients’ homes and main-tain continuity. This solid case management connection alsopromotes efficiency by allowing many problems to be handledby telephone. Integration into the care process creates the op-portunities for effective intervention.

Finally, this study recalls important observations drawn fromlarge community care demonstrations in the 1970s and 1980s.To be most cost-effective, interventions must be targeted tospecific populations and efforts must be made to control theongoing cost of interventions.17 Those seeking to replicate thisapproach should heed these observations.

REFERENCES

1. Naylor MD, Brooten D, Campbell R, et al. Comprehensive discharge planningand home follow-up of hospitalized elders: a randomized clinical trial. JAMA. 1999;281:613-620.2. Garfinkel SA, Riley GF, Iannacchione VG. High-cost users ofmedical care.HealthCare Financing Rev. 1988;9:41-52.3. Alexandre LM. High-cost patients in a fee-for-service medical plan: the casefor earlier intervention. Med Care. 1990;28:112-123.4. Anderson GF, Steinberg EP. Hospital readmissions in the Medicare population.N Engl J Med. 1984;311:1349-1353.5. Zook CJ, Savickis SF, Moore FD. Repeated hospitalization for the same dis-ease: a multiplier of national health care costs. Milbank Q. 1980;58:454-471.6. Eggert GM, Friedman B. The need for special interventions for multiple hospi-tal admission patients. Health Care Financing Rev. 1988(suppl):57-67.7. Corrigan JM, Martin JB. Identification of factors associated with hospital read-mission anddevelopment of a predictivemodel.Health Serv Res.1992;27:81-101.8. Waite K, Oddone E, Weinberger M, Samsa G, Foy M, Henderson W. Lack ofassociation between patients’ measured burden of disease and risk for hospitalreadmission. J Clin Epidemiol. 1994;47:1229-1236.9. Chin MH, Goldman L. Correlates of early hospital readmission or death in pa-tients with congestive heart failure. Am J Cardiol. 1997;79:1640-1644.

10. Wasson JH, Sauvigne AE, Mogielnicki P, et al. Continuity of outpatient medi-cal care in elderly men. JAMA. 1984;252:2413-2417.11. Rich MW, Beckham V, Wittenberg C, Levin CL, Freedland KE, Carney RM. Amultidisciplinary intervention to prevent the readmission of elderly patients withcongestive heart failure. N Engl J Med. 1995;333:1190-1195.12. Fonarow GC, Stevenson LW, Walden JA, et al. Impact of a comprehensiveheart failure management program on hospital readmission and functional statusof patients with advanced heart failure. J Am Coll Cardiol. 1997;30:725-732.13. Stuck AE, Aronow HU, Steiner A, et al. A trial of in-home comprehensive dis-charge assessments for elderly people living in the community.N Engl JMed. 1995;333:1184-1189.14. Naylor M, Brooten D, Jones R, Lavisso-Mourey R, Mezey M, Pauly M. Com-prehensive discharge planning for the hospitalized elderly: a randomized clinicaltrial. Ann Intern Med. 1994;120:999-1006.15. Potthoff S, Kane RL, Franco SJ. Improving hospital discharge planning for el-derly patients. Health Care Financing Rev. 1997;19:47-72.16. Boling PA. The Physician’s Role in HomeHealth Care.NewYork, NY: SpringerPublishing Co Inc; 1997.17. Weissert WG. Seven reasons why it is so difficult to make community-basedlong-term care cost-effective. Health Serv Res. 1985;20:423-433.

The Challenges of Recognizing Child AbuseSeeing Is BelievingJohn M. Leventhal, MD

ALMOST 4 DECADES HAVE PASSED SINCE KEMPE AND COL-leagues1 published in THE JOURNAL their landmarkdescription of the battered child syndrome. Therewere 2 major findings in that study. The first was a

clinical description of children who had been physically abusedby their parents. Although the abuse and misuse of childrenhad been recognized for centuries2(pp3-28) and radiographic find-ings in children thought to be caused by deliberate injurieshad been described,3,4 publication of the article by Kempe etal1 in JAMA made it clear that injuries caused by physical abusewere clinical problems that required the attention of physi-cians. The second finding was the result of an epidemiologi-cal survey in which 749 abused children—many of whom ei-

ther had been killed or had sustained permanent braindamage—were identified by 71 hospitals and 77 district at-torneys in the United States. This large number of cases sug-gested that serious child abuse was unlikely to occur infre-quently. However, no one in 1962 would have predicted thatin the United States in 1997, almost 3.2 million reports of childmaltreatment would be made to child protective service agen-cies. Of these reports, approximately 1 million were con-firmed, including neglect (54%), physical abuse (22%), sexualabuse (8%), emotional abuse (4%), and other (12%).5

That parents couldphysically hurt their childrenwas a fright-ening notion for clinicians concerned with the health and wel-fare of children, yet the astute observations and clinical descrip-

Author Affiliation: Department of Pediatrics, Yale University School of Medicine,New Haven, Conn.Corresponding Author and Reprints: JohnM. Leventhal, MD, Department of Pe-diatrics, Yale University School ofMedicine, 333 Cedar St, NewHaven, CT 06520-8064 (e-mail: [email protected]).

See also p 621.

EDITORIALS


tions ofKempe et al and the research that followed2 changed theway injuries inchildrenwereviewed.Cliniciansno longerwouldreadilyacceptcertainhistoriesprovidedbyparentsas truthfuland,instead, would consider the possibility of child abuse.

There have been numerous challenges for the field of childabuse since 1962, including gaining an understanding of theextent of the problem and how abuse occurs in families, thedevelopment of statewide systems of protective services to evalu-ate suspected maltreatment and help ensure the safety of mal-treated children, the development of treatment programs forchildren and families and for adults who were maltreated as chil-dren, and the development of prevention programs to help chil-dren and families before maltreatment occurs. A cornerstonefor all these activities is the appropriate recognition of the abusedchild and, concomitantly, the appropriate recognition of inju-ries that are truly unintentional. Although recent media atten-tion6 and court cases7 have suggested that physicians are over-diagnosing child abuse, the true problem continues to be oneof underrecognition. For example, since the early 1970s, re-peated unexplained deaths of infants in a family usually werebelieved to be due to recurrent sudden infant death syndrome(SIDS); however, recent evidence has indicated that some casesof recurrent SIDS are, in fact, homicides.8,9

Although the recognition of abuse can be relatively straight-forward (eg, a young child with fresh bruises, healing frac-tures, and no history that explains the injuries), there are manytimes when recognition is difficult. Why is it so difficult torecognize an abused child? The first problem is the false ormisleading history that is often provided. In the usual clini-cal encounter, the physician is accustomed to a truthful (if some-times minimized or exaggerated) history. Because diagnosticreasoning is often shaped by the history provided, a mislead-ing history can misdirect the diagnostic process and result inan incorrect diagnosis. In a study of fractures in childrenyounger than 3 years,10 examination of the initial histories in52 abused children revealed that in only 1 instance did a par-ent indicate that the child had been hurt by an adult. Instead,the most common presenting histories were a report by a care-taker of an abnormality (eg, a seizure or decreased move-ment of a limb) in 52% of cases, a fall in 27%, being hit by anolder child in 10%, and a self-inflicted injury in 6%. Of course,young children cannot speak for themselves, but even olderchildren often learn very early the importance of keeping fam-ily secrets and telling physicians, teachers, social workers, andothers a false story (eg, they tripped and fell) to explain aninjury to the face that was caused by abuse.

A second set of factors that influence the likelihood of rec-ognition of abuse are personal biases related to the physician’seducation, experience, attitudes, andbeliefs.Unfortunately,mostmedical students receive little education—often only 1 or 2 lec-tures—about abuse or family violence. Physicians who pro-vide care to children may have had little formal education aboutchild abuse and limited clinical experience during their resi-dencies in evaluating suspected abuse. Attitudes and beliefs alsocan interfere with making a correct diagnosis. Physicians often

have difficulty believing that abuse can occur in families inwhichthe parents appear to be caring and interact well with the phy-sicians. In fact, some of the most difficult cases to diagnose canbe those in which parents have characteristics much like thephysicians who are conducting the evaluations. For instance,instead of considering possible child abuse to explain freshbruises on the arms of a 6-month-old infant, the physician asksonly about a family history of bleeding disorders. Sometimes,physicians do not ask about abuse because they do not want tooffend or falsely accuse the family or because they want to becertain about the diagnosis before discussing it. And some-times, physicians do not want to get involved, although in ev-ery state, laws mandate that physicians report suspected (notnecessarily confirmed) child maltreatment.

Progress has been made in helping physicians recognize in-juries that may indicate abuse. For example, studies on thebiomechanics of injuries11,12 and characteristics that distin-guish abusive injuries from those that are unintentional havebeen helplful.10,13,14 Also helpful has been the increasing num-ber of physicians who have pursued careers focusing on theproblem of abuse despite the limited opportunities for fund-ing and fellowship training, both of which fall far short of whatis needed, based on the extent of the problem.

Although learning from errors in diagnosis is always diffi-cult, physicians are accustomed to such an approach. In thisissue of THE JOURNAL, Jenny et al15 present an important studyon the failure of physicians to recognize head injuries that weredue to child abuse. During a 6-year period, 31% of 173 chil-dren younger than 3 years with the final diagnosis of a headinjury due to abuse had made at least 1 prior visit to a physi-cian at which the diagnosis of child abuse had been missed. Theauthors note substantial consequences to missing the diagno-sis: 28% of the children were reinjured because of a delay indiagnosis and 4 deaths might have been prevented by earlierdiagnosis.

The diagnoses made at these initial visits to physicians in-cluded the range of possibilities in young children who presentwith a false or misleading history and symptoms of central ner-vous system disease, including gastroenteritis, uninten-tional head injury, colic, or otitis media. Errors in the interpre-tation of computed tomographic scans of the head (ie, failure toidentify subdural hematomas) and radiographs of bones (ie, fail-ure to identify metaphyseal fractures) contributed to missing thecorrect diagnosis and are an important reminder that the cor-rect diagnosis of child abuse often relies on collaboration withspecialists, such as radiologists, orthopedic surgeons, and neu-rosurgeons.

Not surprisingly, at the initial presentation to the physi-cian, children who had milder symptoms (no seizures, nor-mal respiratory status, and no facial or scalp injury) were morelikely to have the diagnosis missed. Also, children whose par-ents were living together were more likely to have the correctdiagnosis missed, which suggests that in the absence of a truth-ful history, physicians’ attitudes and beliefs influenced the like-lihood of a correct diagnosis.

EDITORIALS


This study does not address a related and important typeof missed diagnosis, namely, labeling a child’s unintentionalinjury as abuse. Such an incorrect diagnosis can cause sub-stantial harm to the child and family, especially if the child isremoved from the home.

Can physicians do better at recognizing injuries caused byabuse and thereby reduce the frequency of delayed diag-noses? Jenny et al15 provide sound, specific recommenda-tions to improve the likelihood of making a correct diagno-sis, such as performing a complete examination on youngchildren with nonspecific symptoms (ie, vomiting or irrita-bility) and being suspicious of bruises or abrasions on the faceor head of infants. As part of the evaluation of an infant withsuch nonspecific symptoms and facial marks, the physicianshould ask the parent directly about how the facial injuriesoccurred and about the possibility that someone may have hurtthe child and should consider obtaining a computed tomog-raphy scan of the head and a skeletal survey.

To make the correct diagnosis of abuse, physicians need tobe suspicious for its possible occurrence. Injuries in youngchildren (except those that occur from normal activities, suchas bruises on the shins of toddlers, or those that occur fromcommon unintentional events, such as a short, linear, pari-etal skull fracture from a fall off a bed) need careful evalua-tion. When a clinician is concerned about the mechanism ofhow the injury occurred, the extent of the injury, or the tim-ing, child abuse should be considered. It is helpful to reviewthe history with the person(s) who actually witnessed the event;if concerns persist, the clinician should report the case to pro-tective services and obtain the appropriate diagnostic tests (eg,skeletal survey, computed tomography scan of the head, orophthalmologic evaluation). It may be necessary for the phy-sician to get help obtaining a more detailed social and familyhistory and hospitalize the child to complete the evaluation.

Two general recommendations deserve mention. First, allphysicians who see children, not just those in primary care,should receive education about child abuse. This educationshould begin in medical school, be incorporated into resi-dency training programs, and be extended through continu-

ing medical education courses. Second, collaboration amongphysicians and other professionals, including protective ser-vice workers, should be improved because it is necessary tosolve the diagnostic puzzle of child abuse. For instance, theradiologist who is made aware of the child’s bruises and thepediatrician’s suspicions would certainly be more alert to signsof abuse on the radiographs. Although making the correct di-agnosis of child abuse will continue to be a challenge and re-ducing missed or delayed diagnoses to zero is unlikely, phy-sicians can do better. The key, of course, is “seeing” clearly.When that happens, seeing will become believing, even if whatis seen correctly is painful to all—that sometimes adults canhurt children in serious and deadly ways.

REFERENCES

1. Kempe CH, Silverman FN, Steele BF, DroegemuellerW, Silver HK. The battered-child syndrome. JAMA. 1962;181:17-24.2. Helfer ME, Kempe RS, Krugman RD, eds. The Battered Child. 5th ed. Chicago,Ill: University of Chicago Press; 1997.3. Caffey J. Multiple fractures in the long bones of infants suffering from chronicsubdural hematoma. AJR Am J Roentgenol. 1946;56:163-173.4. Silverman FN. The roentgen manifestations of unrecognized skeletal trauma ininfants. AJR Am J Roentgenol. 1953;69:413-427.5. Wang CT, Daro D. Current Trends in Child Abuse Reporting and Fatalities:The Results of the 1997 Annual Fifty State Survey. Chicago, Ill: National Com-mittee to Prevent Child Abuse; 1998.6. Diagnosis murder. “20/20.” ABC television. December 23, 1998.7. Doherty WF. 2d doctor says boy had old injuries: testifies that death not dueto shaking. Boston Globe. October 21, 1997:B3.8. Southall DP, Plunkett MCB, Banks MW, Falkov AF, Samuels MF. Covert videorecordings of life-threatening child abuse: lessons for child protection. Pediatrics.1997;100:735-760.9. Firstman R, Talan J. The Death of Innocents. New York, NY: Bantam Books;1997.10. Leventhal JM, Thomas SA, Rosenfield NS, Markowitz RI. Fractures in youngchildren: distinguishing child abuse from unintentional injuries. AJDC. 1993;147:87-92.11. Hymel KP, Bandak FA, Partington MD,Winston KR. Abusive head trauma? abiomechanics-based approach. Child Maltreatment. 1998;3:116-128.12. Kleinman PK, Schlesinger AE. Mechanical factors associated with posterior ribfractures: laboratory and case studies. Pediatr Radiol. 1997;27:87-91.13. Duhaime AC, Alario AJ, Lewander WJ, et al. Head injuries in very young chil-dren: mechanisms, injury types, and ophthalmologic findings in 100 hospitalizedpatients younger than 2 years of age. Pediatrics. 1992;90:179-185.14. Hobbs CJ.When are burns not accidental?Arch Dis Child. 1986;61:357-361.15. Jenny C, Hymel KP, Ritzen A, Reinert SE, Hay TC. Analysis of missed cases ofabusive head trauma. JAMA. 1999;281:621-626.

EDITORIALS


Recognizing Abusive Head Trauma in Children

To the Editor: The article by Dr Jenny and colleagues1 raisesa critical question: how can practicing physicians improvetheir ability to recognize inflicted head trauma in young chil-dren? Unfortunately, their study does not provide enoughinformation to solve the practitioner’s constant question:what is the predictive value of the symptom or sign at hand?Jenny et al address the question of if a child has an inflictedhead trauma, then what is the chance the child will have facialbruising, nonspecific vomiting, fever, or irritability? In fact,these probabilities reported in their article were high enoughto be of interest.However, the practitioner needs to know if a child has fa-

cial bruising, nonspecific vomiting, fever, or irritability, thenwhat is the chance the child has sustained inflicted head trauma?The data presented by Jenny et al in no way answer this criti-cal question. For instance, if a child has vomiting that cannotbe fully explained, is the risk that the child has been abusedhigh enough to justify a formal evaluation for abuse, includ-ing head neuroimaging, ophthalmology consultation, and alert-ing the child protection authorities? The same question can beraised for the child presenting with facial bruising, irritability,and fever. The conclusion that the authors reach extends be-yond the data they presented.

Arthur Lavin, MDPediatric Partners of ClevelandBeachwood, Ohio

1. Jenny C, Hymel KP, Ritzen A, Reinert SE, Hay TC. Analysis of missed cases ofabusive head trauma. JAMA. 1999;281:621-626.

To the Editor: As Dr Jenny and colleagues1 point out in theiranalysis of missed cases of abusive head trauma in children,abusive head trauma can be difficult to diagnose because of thelack of external evidence of trauma. However, their article failedto emphasize that abusive head trauma often results in dra-matic ophthalmic manifestations.2,3

Examining small infants can be done easily in the hospital.Vision can be assessed by close observation of fixation and fol-lowing patterns. The pupillary reflexes should be examined care-fully for sluggish pupils or a defect in the afferent pupillary re-sponse. External signs of periorbital edema, ecchymosis, or lidlacerations may be seen. Palpation of the orbital rim may re-veal the characteristic step-off of a previous orbital floor frac-ture. A traumatic sixth-nerve palsy may result in esotropia. Avariety of anterior segment findings, including conjunctival lac-erations, corneal abrasions, traumatic cataract, or hyphemawithattendant intraocular pressure changes, may be noted.The hallmark ophthalmic finding in shaken-baby syndrome

is that of posterior segment hemorrhages.4,5 All infants withsuspected abusive head trauma should have a thoroughdilated funduscopic examination with an indirect ophthalmo-scope to view the entire posterior pole and the retinal periph-ery. Abusive head trauma characteristically produces multi-

layer hemorrhages, although vitreous, preretinal, intraretinal,or subretinal hemorrhages also may be present.6 Peripheraldome-shaped hemorrhagic lesions with white retinal bordersmay represent peripheral retinoschisis. The retina can even befolded in a circumferential manner around the macula. Whilemany retinal hemorrhages resolve without sequelae, involve-ment of the optic nerve or macula can produce profound life-long visual loss.While none of the reported ocular findings is pathogno-

monic, these eye findings along with the associated medicalhistory and other medical conditions give strong evidence forabusive head trauma. An ophthalmic examination is associ-ated with no morbidity, can be quickly and easily performed,and can often confirm this difficult diagnosis. We recommendthat clinicians have a low threshold for ordering ophthalmicconsultation in cases of suspected abusive head trauma.

Herbert Becker, MDBalaji K. Gupta, MDUniversity of Illinois at Chicago

1. Jenny C, Hymel KP, Ritzen A, Reinert SE, Hay TC. Analysis of missed cases ofabusive head trauma. JAMA. 1999;281:621-626.2. Altman RL, Kutscher ML, Brand DA. The “shaken-baby syndrome.” N Engl JMed. 1998;339:1329-1330.3. Spaide RF, Mikler WF, Williams GA. Shaken baby syndrome. Am Fam Physi-cian. 1990;41:1145-1152.4. Swensen J, Levitt C. Shaken baby syndrome: diagnosis and prevention. MinnMed. 1997;80:41-44.5. Han DP, Wilkinson WS. Late ophthalmic manifestations of the shaken babysyndrome. J Pediatr Ophthalmol Strabismus. 1990;27:299-303.6. Williams DF, Swengel RM, Scharre DW. Posterior segment manifestations ofocular trauma. Retina. 1990;10(suppl 1):S35-S44.

To the Editor: Any diagnosis requires sound criteria to bedefensible. However, 3 of the 4 factors used by Dr Jenny andcolleagues1 to determine that intentional injury in a child hadbeen “missed” are questionable. A confession typically isobtained only after a suspect has been repeatedly confrontedwith the allegation that “you shook your child” or after thecaretaker has been offered a plea to a reduced charge with alimited jail sentence, given a deadline for accepting the offer,and threatened with an upward sentencing departure if thereis a conviction on the original charges. Such a confessionshould not be used as evidence that nonintentional injuryoccurred, and may lead to incarceration of an innocent per-son.2,3 Furthermore, the history may be inadequate if the care-taker does not know what happened, or may be consideredinconsistent if the team members evaluating child abuse areunfamiliar with the biomechanics of head injury. (Too often,any history other than “I did it” is considered inadequate.)Moreover, a delay in seeking care cannot be proved by assum-ing that an injury occurred at a certain time, and especially notby inferring that a child with a head injury never has a lucidinterval.If trained health care professionals have a difficult time rec-

ognizing severity of injury, why should a parent or caretakerbe expected to do any better? The diagnosis of abusive headtrauma is difficult at best. However, what special insight doesa multidisciplinary team, as suggested by Jenny et al, “led by

LETTERS

©1999 American Medical Association. All rights reserved. JAMA, October 20, 1999—Vol 282, No. 15 1421

pediatricians” and including “social workers, nurses, psycholo-gists, child psychiatrists, and attorneys” have to allow it to con-clude by the criteria given that a head injury in a child is in-flicted or unintentional?Where are those who understand thepsychology of a false confession? Where are the biophysi-cists? Where are those who know the limitations of injurydating based on biochemical, radiological, or physical obser-vation? Is “multidisciplinary team consensus” the “gold stan-dard” for “confirmation that head trauma was inflicted”? Thedanger is for such a group to become a de facto star chamber,unencumbered by accountability or self-doubt. The judicialweight of a medical conclusion of abuse is too great to allowthis to occur.4

The study by Jenny et al does not address the mislabeling ofa child’s unintentional injury as abuse, and, as Dr Leventhal5

pointed out, an “incorrect diagnosis can cause substantial harmto the child and family, especially if the child is removed fromthe home.” However, the greater harm, in my opinion, is forsomeone to be charged and convicted and spend many yearsin prison or be executed for a crime that never occurred.6 Thecriteria for a medical diagnosis of abusemust be based on stan-dards other than those used by Jenny et al. Physicians and otherhealth care professionals must humbly admit in some and, per-haps, many cases, and especially in children with an isolatedhead injury, that we simply do not know if the trauma was un-intentional or inflicted.

John Plunkett, MDRegina Medical CenterHastings, Minn

1. Jenny C, Hymel KP, Ritzen A, Reinert SE, Hay TC. Analysis of missed cases ofabusive head trauma. JAMA. 1999;281:621-626.2. State v Lehmer, State v Engberg, 24296 F Supp 24297 (D Ia 1997).3. Hoffman J. Police refine methods so potent, even the innocent have con-fessed. New York Times. March 30, 1998;sect A:1, 17-18.4. Wilkins B. Head injury: abuse or accident? Arch Dis Child. 1997;76:393-397.5. Leventhal JM. The challenges of recognizing child abuse: seeing is believing.JAMA. 1999;281:657-659.6. State v Burr, 21905 NC Super Ct Div, 21909 (1991).

In Reply: DrLavin points out that the practicing physician facesthe dilemma that effective “screening tests” to rule out headtrauma have not been tested. More research is needed to givephysicians the tools they need to make an accurate assess-ment of nonspecific symptoms such as vomiting and irritabil-ity in preverbal children.We agree with Drs Becker and Gupta that ophthalmologic

findings in cases of abusive head trauma are common and of-ten dramatic. More frequent use of ophthalmologic consulta-tion most likely would lead to more accurate diagnosis.Dr Plunkett takes issue with the criteria we used to define

abusive head trauma. First, he states that confession by a per-petrator is likely to be coerced. For many years, we have askedthe caretakers of injured young children a simple question:“What happened?” In response, some caretakers confess to in-juring their children by shaking, striking, or slamming them.These confessions of inflicted pediatric head trauma are uni-formly deeply emotional, guilt-ridden, and usually tearful. The

accounts of extreme stress, the infant’s prolonged crying, amo-mentary loss of control, and the sincere desire to take back themoment when the event occurred evoke our deepest sympa-thy for these caretakers, their families, and their victims. Plun-kett’s assumption that confessions of inflicted head trauma pre-dominantly follow threats and coercionminimizes the tangiblesincerity of these confessors. He is not familiar with the caseswe reported, and knows nothing of the circumstances of theconfessions.A conclusion that the caretaker’s history is inconsistent or

inadequate typically arises when severe or fatal pediatric cra-nial injuries are reported to have resulted from a simple fall.An extensive body of literature about injuries sustained in wit-nessed pediatric falls leads us to the conclusion that substan-tial force and distance are required to seriously injure chil-dren.1-3When presentedwith this information,many caretakersbegin to change their account of the circumstances of the pe-diatric head injury. Although not diagnostic, an evolving or fre-quently changing history ismore suggestive of inflicted trauma.Despite Plunkett’s assertion to the contrary, we recognize that

pediatric victims of head traumamay initially presentwithmini-mal symptoms. In contrast, most pediatric victims of fatal blunthead trauma reveal immediate and severe clinical signs of neu-rological deterioration.4,5

Finally, Plunkett does not seem to realize that the abuse ofchildren is a widespread problem. Our multidisciplinary teamworks froman assumption of innocence.Our decisions are basedon the best available research. To refer to a thoughtful, carefulmultidisciplinary team as a “de facto star chamber” does notadd to the scientific discourse on this extremely complicatedissue.

Carole Jenny, MD, MBABrown University School of MedicineProvidence, RILt Col Kent P. Hymel, MD, USAF, MCNational Naval Medical CenterBethesda, Md

Disclaimer: The opinions and conclusions in this letter are those of the authorsand are not intended to represent the official positions of the US Air Force, USDepartment of Defense, or any other governmental agency.

1. Williams RA. Injuries in infants and small children resulting fromwitnessed andcorroborated free falls. J Trauma. 1991;31:1350-1352.2. Chadwick DL, Chin S, Salerno C, Landsverk J, Kitchen L. Deaths from falls inchildren: how far is fatal? J Trauma. 1991;31:1353-1355.3. Lyons TJ, Oates K. Falling out of bed: a relatively benign occurrence. Pediat-rics. 1993;92:125-127.4. Willman KY, Bank DE, Senac M, Chadwick DL. Restricting the time of injury infatal inflicted head injuries. Child Abuse Negl. 1997;21:929-940.5. Starling SP, Holden JR, Jenny C. Abusive head trauma: the relationship of per-petrators to their victims. Pediatrics. 1995;95:259-262.

Electron Beam Computed Tomographyto Detect Coronary Artery Disease

To the Editor: Drs Siegel and Evens1 note that electron beamcomputed tomography (EBCT) is being used to diagnose ath-erosclerosis. However, their statement, “At best, however, there

LETTERS

1422 JAMA, October 20, 1999—Vol 282, No. 15 ©1999 American Medical Association. All rights reserved.

Nahelsky M, and Dix J. The time interval between lethal infant shaking and onset of symptoms: A review of the Shaken Baby Syndrome Literature. The American Journal of Forensic Medicine and Pathology 1995; 16(2):154-157.

The authors agree with the Bruce-Zimmerman and Duhaime theories which can be basically summarized as follows: you must have impact to create damages like those seen in SBS. This article comments on the paucity of studies done discussing the onset of symptoms for SBS. It looks at time between "shaking" and symptoms. This article also discusses three cases of "shaking" injury where children experienced lucid intervals of 3 hours, 3 days and 4 days. The last child had bilateral retinal hemorrhages. The article concludes that there is very little data available to suggest the actual time limits between fatal head injuries and death. This article shows that lucid intervals do exist and that perpetrators cannot be narrowed down to the last person holding the baby.

Dacey, R.G., Alves, W., Rimel, R., Winn, R., and Jane, J. (1986) Neurosurgical complications after apparently minor head injury. Neurosurgery 65:203-210.

Authors studied 610 patients at a Washington trauma center. Of 66 patients with skull fractures, 5 had intracranial hematomas, 13 had some type of Neurosurgical complications. Neurological complications and lucid intervals were more likely to be found in boys than girls and are more likely to occur in a fall rather than by some other mechanism. The increased ICP is found after about 50% of severe head injuries. Skull fractures increase likelihood of Neurosurgical procedures. Article documents the existence of lucid intervals. Authors found that 3% of minor head injury cases will deteriorate after experiencing a lucid interval.

Looney, Christopher; Smith, Keith; Merck, Lisa; Wolfe, Honor; Chescheir, Nancy; Hamer, Robert; Gilmore, John Intracranial Hemorrhage in Asymptomatic Neonates: Prevalence on MR Images and Relationship to Obstetric and Neonatal Rick Factors February 2007 Radiology 242(2): 535-541

The authors of this article did MRI’s on babies directly after birth and found that 26% of normal non-complicated vaginal births resulted in intracranial hemorrhages. These hemorrhages were asymptomatic.

Intracranial Hemorrhage inAsymptomatic Neonates:Prevalence on MR Images andRelationship to Obstetric and NeonatalRisk Factors1

Christopher B. Looney, BSJ. Keith Smith, MD, PhDLisa H. Merck, MD, MPHHonor M. Wolfe, MDNancy C. Chescheir, MDRobert M. Hamer, PhDJohn H. Gilmore, MD

Purpose: To retrospectively evaluate the prevalence of neonatalintracranial hemorrhage (ICH) and its relationship to ob-stetric and neonatal risk factors.

Materials andMethods:

Pregnant women were recruited for a prospective study ofneonatal brain development; the study was approved bythe institutional review board and complied with HIPAAregulations. After informed consent was obtained from aparent, neonates were imaged with 3.0-T magnetic reso-nance (MR) imaging without sedation. The images werereviewed by a neuroradiologist with 12 years of experi-ence for the presence of ICH. Medical records were pro-spectively and retrospectively reviewed for selected riskfactors, which included method of delivery, duration oflabor, and evidence of maternal or neonatal birth trauma.Risk factors were assessed for relationship to ICH by usingFisher exact test statistics.

Results: Ninety-seven neonates (mean age at MR imaging, 20.8days � 6.9 [standard deviation]) underwent MR imagingbetween the ages of 1 and 5 weeks. Eighty-eight (44 maleand 44 female) neonates (65 with vaginal delivery and 23with cesarean delivery) completed the MR imaging evalu-ation. Seventeen neonates with ICHs (16 subdural, twosubarachnoid, and six parenchymal hemorrhages) wereidentified. Seven infants had two or more types of hemor-rhages. All neonates with ICH were delivered vaginally,with a prevalence of 26% in vaginal births. ICH was signif-icantly associated with vaginal birth (P � .005) but notwith prolonged duration of labor or with traumatic orassisted vaginal birth.

Conclusion: Asymptomatic ICH following vaginal birth in full-term ne-onates appears to be common, with a prevalence of 26%in this study.

� RSNA, 2006

1 From the Department of Psychiatry, CB No. 7160,7025A Neurosciences Hospital, University of North Caro-lina School of Medicine, Chapel Hill, NC 27599-7160.From the 2005 RSNA Annual Meeting. Received January23, 2006; revision requested March 23; revision receivedJune 7; accepted June 21; final version accepted August21. J.H.G. supported by National Institute of MentalHealth grant 1 P50 MH064065. C.B.L. supported by aDistinguished Medical Scholarship from UNC School ofMedicine. Address correspondence to J.H.G. (e-mail:[email protected]).

� RSNA, 2006

ORIGINALRESEARCH

�PEDIATRIC

IMAGING

Radiology: Volume 242: Number 2—February 2007 535

Intracranial hemorrhage (ICH) in full-term neonates commonly is associ-ated with apnea, bradycardia, and

seizures (1–4). Subdural, subarachnoid,intraparenchymal, and intraventricularhemorrhages have been identified insymptomatic full-term neonates (5–8).Several factors have been reported toincrease the risk of symptomatic ICH infull-term newborns, and these factorsinclude assisted vaginal delivery (for-ceps or vacuum extraction), maternalparity, fetal weight, and prolonged du-ration of labor (9–14).

Imaging studies indicate that ICHcan occur in asymptomatic newborns(15–17), though precise incidence anddistribution of ICH in asymptomatic full-term neonates is not clear. Results ofone large prospective study with imag-ing (18), which was conducted by usinga 0.2-T magnetic resonance (MR) im-ager, indicated that there was an 8%prevalence of subdural hemorrhage innewborns; subdural hemorrhage wasassociated with vaginal delivery. In thestudy, all subdural hemorrhages re-solved at follow-up imaging 4 weekslater.

While we conducted a prospectivestudy of normal brain development byusing a 3.0-T MR imager, we notedseveral asymptomatic full-term neo-nates with ICH. We hypothesized thatasymptomatic ICH would be associ-ated with vaginal birth, traumatic vag-inal birth, and prolonged duration oflabor. Thus, the purpose of our studywas to retrospectively evaluate theprevalence of neonatal ICH and its re-lationship to obstetric and neonatalrisk factors.

Materials and Methods

PatientsPregnant mothers were recruited fromDecember 2002 to July 2005 as part ofan ongoing prospective study about theinvestigation of prenatal and neonatalbrain development. The cohort includedcontrol neonates and two groups of ne-onates at high risk for psychiatric orneurodevelopmental disorders: the off-spring of mothers with schizophreniaand neonates diagnosed with fetal iso-lated mild ventriculomegaly (MVM).Control mothers did not have a historyof psychotic illness, and their offspringdid not have MVM. Exclusion criteriafor all groups in the parent study in-cluded major maternal medical illnessesor major congenital abnormalities de-picted with ultrasonography at 18–20weeks gestation. Our ongoing Health In-surance Portability and AccountabilityAct–compliant study was approved byour institutional review board, and in-formed consent was obtained from theparents of the neonates. The analysisperformed in the retrospective study wereport here is considered within thescope of the initially approved research,as has been confirmed by our institu-tional review board.

Neonates who underwent MR imag-ing after 5 weeks of age were excludedfrom the nested case-control analysis,as Whitby et al (18) found that all hem-orrhages identified at birth had resolvedby that age. Both neonates at high riskfor psychiatric or neurodevelopmentaldisease and neonates without eitherwere included in the analysis, as there isno a priori evidence that such high-riskstatus confers risk of ICH.

ImagingNeonates were imaged by using a 3.0-TMR imager (Magnetom Allegra; Sie-mens Medical Systems, Malvern, Pa)without sedation (19). Neonates werefed, swaddled, fitted with ear protec-tion, and had their heads secured in avacuum-fixation device. T1-weightedstructural pulse sequences were either athree-dimensional magnetization-pre-pared rapid acquisition gradient-echo

sequence (repetition time msec/echotime msec/inversion time msec, 1820/4.38/400; flip angle, 7°) or a two-di-mensional spoiled gradient-echo fastlow-angle shot sequence (repetitiontime msec/echo time msec, 15/7; flipangle, 25°). Intermediate-weighted andT2-weighted images were obtained witha turbo spin-echo sequence (6200/20,119 or 7000/18, 108; flip angle, 150°).Spatial resolution was 1 � 1 � 1-mmvoxel for T1-weighted images and1.25 � 1.25 � 1.5-mm voxel with0.5-mm intersection gap (1 � 1 �3.0-mm voxel with 0.9-mm intersectiongap for earliest studies) for intermedi-ate-weighted and T2-weighted images.

A board-certified neuroradiologist(J.K.S.) with 12 years of experience inreading neonatal images retrospectivelyreviewed all MR images for ICH and wasblinded to neonatal and obstetric data.Abnormal areas of signal intensity withsignal characteristics compatible withblood products were identified (eg, sub-acute blood, which was bright on T1-weighted images and dark or isointenseon T2-weighted images). The location ofthis area of abnormality determined theclassification; for example, a collectionbetween the brain and skull or a duralreflection that did not enter the corticalsulci was considered subdural, one thatentered the cortical sulci or cerebrospi-nal fluid cisterns was considered sub-arachnoid, and a lesion that was sur-rounded by brain parenchyma was con-

Published online before print10.1148/radiol.2422060133

Radiology 2007; 242:535–541

Abbreviations:ICH � intracranial hemorrhageMVM � mild ventriculomegaly

Author contributions:Guarantors of integrity of entire study, C.B.L., J.K.S.,J.H.G.; study concepts/study design or data acquisition ordata analysis/interpretation, all authors; manuscript draft-ing or manuscript revision for important intellectual con-tent, all authors; manuscript final version approval, allauthors; literature research, C.B.L., J.K.S., L.H.M.,H.M.W., N.C.C., J.H.G.; clinical studies, J.K.S., L.H.M.,J.H.G; statistical analysis, C.B.L., J.K.S., R.M.H., J.H.G.;and manuscript editing, all authors

Authors stated no financial relationship to disclose.

Advances in Knowledge

� Intracranial hemorrhages arecommon in asymptomatic neo-nates after vaginal delivery, withan estimated prevalence of 26%.

� Intracranial hemorrhages inasymptomatic neonates deliveredvaginally are not associated withovert signs of trauma or with as-sisted delivery (forceps, vacuum).

PEDIATRIC IMAGING: Intracranial Hemorrhage in Neonates Looney et al

536 Radiology: Volume 242: Number 2—February 2007

sidered intraparenchymal. Admittedly,there could be some uncertainty as tothe exact location of some very smallhemorrhages. For example, a smallepidural hematoma may be indistin-guishable from a small subdural hema-toma. In such an instance, the locationwas assumed to be subdural. Patientswere classified as having ICH or nothaving it on the basis of this reading.Neonates without an adequate T1-weighted image were excluded, as allhemorrhages were identified on theT1-weighted image.

Data CollectionIn the protocol for the parent study, oneof two trained research associates pro-spectively collected the following obstet-ric data: maternal age and ethnicity,whether a cesarean or vaginal deliveryhad been performed, and whether as-sisted vaginal delivery (forceps or vac-uum) had been performed. Separatesets of medical records were retrospec-tively reviewed by a medical student(C.B.L.) and a neurosurgery resident(L.H.M.) to obtain information aboutthe duration of labor, duration of rup-tured membranes, and maternal trauma.Duration of labor was defined as thelength of time between the reported on-set of maternal contractions and deliv-ery. Duration of ruptured membraneswas defined as the length of time be-tween the rupture of maternal mem-branes and delivery. Maternal birthtrauma was defined as vaginal, labial, orperineal lacerations.

Two trained research associatesprospectively collected the followingneonatal data: Apgar scores; occur-rence of neonatal sepsis, neonatal as-phyxia, neonatal apnea, and neonatalseizures; duration of hospital stay; sex;neonatal head circumference; birthweight; gestational age at birth; and ges-tational age at MR imaging. Two indi-viduals (C.B.L., L.H.M.), who con-ducted a retrospective chart review byindependently reviewing individual setsof records, recorded external evidenceof neonatal birth trauma in the form ofcephalohematoma, scalp laceration, orbruising associated with the use of for-ceps.

Statistical AnalysisNeonates were classified by the neuro-radiologist (J.K.S.) who read the im-ages in two groups: those with ICH iden-tified on MR images and those withoutICH. Binary and categoric obstetric andneonatal variables were compared be-tween the groups by using the Fisherexact test. Group differences in contin-uous variables, such as duration of laborand duration of ruptured membranes,were compared with the Wilcoxon ranksum test. Data analyses were per-formed by using statistical software(SAS for Windows, release 9.1; SAS In-stitute, Cary, NC). A difference withP � .05 (two-tailed test) was consideredsignificant. Because MVM could theo-retically increase the risk of ICH, allanalyses were performed a second timewith exclusion of the cases with MVM.

Results

PatientsOne hundred eighteen neonates (meangestational age at birth, 39.2 weeks;range, 35.1–42.0 weeks) were imagedin the parent study; 97 had undergoneMR imaging between the ages of 1 and 5weeks after birth (mean age at MR im-aging, 20.8 days � 6.9 [standard devia-

tion]). Nine neonates were excludedfrom this analysis because a T1-weighted sequence was not performed.The final 88 (44 male and 44 female)neonates in our study included 69 con-trol neonates without risk for psychiat-ric or neurodevelopmental disorders(four from twin pregnancies), 12 neo-nates with prenatal MVM, and sevenoffspring of mothers with schizophre-nia. Among the 88 neonates, maternalethnicity was as follows: 69 (78%) werewhite, 16 (18%) were African Ameri-can, and three (4%) were Asian Ameri-can. Mean maternal age was 28.6years � 5.3. In 65 (74%) of neonates,delivery was vaginal; in 23 (26%), deliv-ery was cesarean.

Seventeen (19%)—15 control neo-nates without a high risk for psychiatricor neurodevelopmental disorders, oneneonate with MVM, and one offspringof a mother with schizophrenia—of 88neonates had ICHs that were clinicallysilent. The ICHs (Figs 1–3) were ob-served in 16 neonates with single ormultiple subdural hematomas and oneneonate with an isolated germinal ma-trix hemorrhage. Subdural hemorrhageoften coexisted with other types of ICH.Two neonates had additional subarach-noid hemorrhages, and five had coexist-ing intraparenchymal hemorrhages. The

Figure 1

Figure 1: Sagittal (left) and transverse (right) T1-weighted three-dimensional magnetization-preparedrapid gradient-echo MR images (1820/4.38/400; flip angle, 7°; section thickness, 1 mm) in a neonate showtypical size and location of subdural hemorrhage (arrow).



intraparenchymal hemorrhages wereperiventricular (hemorrhagic periven-tricular leukomalacia) in two neonates,were in the germinal matrix area in twoneonates, and were on the brain surface(contusions) in two neonates (Table 1).All subdural hematomas were infraten-torial or low in the occipital or temporalareas.

Risk FactorsAll 17 hemorrhages occurred in full-term neonates delivered through vagi-nal birth and yielded a prevalence of26% in vaginal births. Since all hemor-rhages occurred in vaginally deliveredneonates, the neonates born with cesar-ean delivery were excluded from thesubsequent analysis of other obstetric

and neonatal risk factors. Vaginal birthwas the only significant risk factor asso-ciated with ICH (P � .005). Mothers ofneonates with ICH were not more likelyto have had assisted vaginal delivery orvaginal, labial, or perineal lacerations.Newborns with ICH were not morelikely to have had evidence of birthtrauma (Table 2). There were no signif-icant differences in the duration of laboror duration of ruptured membranes(Table 3). There was no difference ingestational age at delivery, Apgar score,duration of hospital stay, birth weight,or head circumference between the ne-onates with vaginal birth who had ICHand those who did not have ICH; neo-nates with ICH had a significantlyyounger gestational age at MR imagingthan did neonates without ICH (Table3).

None of the neonates in this study(with or without ICH) had evidence ofasphyxia, sepsis, or a seizure. One neo-nate with ICH had apnea at birth. Aclinical MR image depicted an oropha-ryngeal mass that was clinically deter-mined to be the cause of respiratorydistress. With exclusion of MVM, find-ings in neonates did not significantlychange; vaginal birth was still signifi-cantly associated with ICH (P � .008,Fisher exact test).

Discussion

We found that 26% of asymptomaticneonates delivered vaginally had ICH atMR imaging, and this finding suggeststhat ICH is a fairly common conse-quence of a normal vaginal delivery.ICH has been thought to be unusual infull-term neonates (1,8,20–27), thoughthe results of this study and those of thestudy of Whitby et al (28) suggest other-wise.

ICH in full-term neonates often hasbeen associated with birth trauma(6,29–35), which is an association thatresults from the reporting of hemor-rhages identified with cranial imaging insymptomatic neonates in case reports,case series, and case-control studies. Inour study, neither assisted vaginal deliv-ery nor evidence of neonatal birthtrauma could be used to predict the

Figure 2

Figure 2: Sagittal (left) and transverse (right) T1-weighted three-dimensional magnetization-preparedrapid gradient-echo MR images (1820/4.38/400; flip angle, 7°; section thickness, 1 mm) in a neonate showthe largest infratentorial subdural hemorrhages (arrows) identified.

Figure 3

Figure 3: Sagittal (left) and transverse (right) T1-weighted three-dimensional magnetization-preparedrapid gradient-echo MR images (1820/4.38/400; flip angle, 7°; section thickness, 1 mm) in a neonate showintraparenchymal hemorrhage (arrow) in the temporal lobe.



presence of ICH; most (13 of 17, 76%)of the cases of ICH were in the setting ofnonassisted vaginal birth. This finding isin agreement with that of Whitby et al(28), who described nine neonates withasymptomatic hemorrhage; in six of thenine neonates, hemorrhage was associ-ated with assisted delivery; in only twoof nine neonates with subdural hemor-rhages, external birth trauma was anassociated finding. The authors con-cluded that a subdural hematoma wasnot necessarily associated with obviousbirth trauma. Holden et al (16) identi-fied four of 11 neonates with clinicallysilent ICH; in all, vaginal delivery wasuneventful.

The majority of the ICHs in ourstudy were subdural, and most were in-fratentorial. Subdural hematomas canresult from tears in the tentorium, falx,or bridging veins during labor (6,21).Subdural hematoma location may be animportant determinant of symptoms, asfindings of this study and those of thestudy of Whitby et al (28) suggest thatperitentorial subdural hematomas oc-cur frequently and without immediateclinical consequence. Alternately, evensmall amounts of subdural hemorrhagein the posterior fossa may lead to ob-structive hydrocephalus or neurologicdeficits (21).

Subarachnoid, intraparenchymal, andgerminal matrix hemorrhages in thefull-term neonate previously have beenassociated with birth trauma (6,36),though the neonates with subarach-noid and intraparenchymal hemor-rhages in our study did not have obvi-ous evidence of birth trauma. In areview of seven spontaneous paren-chymal and subarachnoid hemorrhagesin full-term neonates, Huang and Rob-ertson (7) suggest that open suturesand the compliance of the calvarialead to shifting of cranial bones duringvaginal delivery and, thus, to compres-sion of brain tissue, tearing of the falxor tentorium, or damage of bridgingveins, with resulting parenchymal orsubarachnoid hemorrhage. Of rele-vance to the neonate with bilateral an-terior temporal contusions in ourstudy, Huang and Robertson (7) notethat the pterion is a large “relatively

unprotected sutural confluence,” whichmakes it a site of injury.

It is important to note that the pat-tern of ICH in the neonates in our studyis different from that found in infantswith nonaccidental head injuries. Non-accidental head injuries typically causesubdural hematomas that are generallylocated in the interhemispheric fissureor over the cerebral convexities; thesehematomas are often, but not always, ofdiffering ages (37). In our study, thesubdural hemorrhages were of the sameage in all 16 neonates, and the majorityof the subdural hemorrhages were inthe posterior fossa or over the occipitallobes near the tentorium. The charac-

teristics of such hemorrhages are con-sistent with previously described birthinjuries (18). None of the subdural ICHswere interhemispheric. As MR imagingevaluation of neonates becomes morecommon in clinical practice, these typesof incidental hemorrhages are likely tobe identified more frequently. Radiolo-gists and pediatricians must become fa-miliar with the appearance of thesehemorrhages and be able to distinguishthem from nonaccidental head injuries.

Cranial MR imaging is superior tocomputed tomography for identificationof hemorrhages, especially extracere-bral hemorrhages and posterior fossasubdural hemorrhages, in neonates

Table 1

Description of the ICHs

Case No. Type Location No. of ICHs

1 Subdural, subarachnoid,intraventricular

Posterior fossa, occipital lobe Multiple

2 Subdural, parenchymal Posterior fossa, periventricular Multiple3 Parenchymal Germinal matrix Single4 Subdural Posterior fossa, occipital lobe Multiple5 Subdural Posterior fossa, occipital lobe Multiple6 Subdural Posterior fossa Single7 Subdural, parenchymal Posterior fossa, occipital lobe Multiple8 Subdural Posterior fossa Multiple9 Subdural Posterior fossa, occipital lobe Multiple

10 Subdural Posterior fossa Single11 Subdural Posterior fossa Single12 Subdural, parenchymal Posterior fossa, occipital lobe,

periventricularMultiple

13 Subdural Posterior fossa, occipital lobe Multiple14 Subdural, parenchymal Posterior fossa, occipital lobe,

temporal lobeMultiple

15 Subdural Posterior fossa, occipital lobe Multiple16 Subdural, parenchymal Temporal fossa, germinal matrix Multiple17 Subdural, subarachnoid Posterior fossa, occipital lobe Multiple

Table 2

Evidence of Traumatic Birth

TypeWith Hemorrhage(n � 17)

Without Hemorrhage(n � 71) P Value

Assisted delivery 4 (24) 9 (13) .28Neonatal trauma 1 (6) 8 (11) �.99Maternal lacerations 11 (65) 28 (39) .1

Note.—Data are numbers of deliveries for assisted delivery, numbers of neonates for neonatal trauma, and numbers ofmothers for maternal lacerations. Numbers in parentheses are percentages.



(38–40). The higher rate of ICH ob-served in our study compared with therate in the study of Whitby et al (18)may be related to the improved capabil-ity of 3.0-T MR imaging to depict hem-orrhages. Whitby et al used low-field-strength 0.2-T MR imaging within 48hours after birth. In the setting of hyper-acute hemorrhage, low-field-strength MRimaging may afford good detection be-cause of the T1 contrast (41). Images inour study were obtained 1–5 weeks af-ter delivery. The T1 contrast achievedby using a T1 technique at a higher fieldstrength also may be more sensitive todepiction of subacute hemorrhage, asthe increased time allows the hemoglo-bin to break down to methemoglobin,which is bright on T1-weighted images.All the hemorrhages in our study weredark, if they could be depicted at all, onT2-weighted images. Since the darkhemorrhage was usually adjacent to anormal dark structure, such as the duraor dural sinus, the hemorrhages werenot conspicuous on T2-weighted im-ages. Because 3.0-T imaging is becom-ing more common, our study findingssuggest that neonatal ICHs will becomeincreasingly recognized in clinical set-tings.

There were limitations to our study.The images were not obtained immedi-ately after birth but in weeks 1–5 after

birth, and, perhaps, we missed cases ofICH that had resolved. Neonates withICH had a younger mean gestational ageat MR imaging than did those withoutICH, and this finding suggests that im-ages obtained closer to birth may depictICH more frequently. Also, our imagingprotocol did not include T2*-weightedor magnetic susceptibility-weighted im-ages, which might be even more sensi-tive for depiction of hemorrhage. Nofollow-up images were obtained to de-termine imaging resolution of hemor-rhage, and no clinical follow-up was per-formed after the identification of ICH.The protocol in the parent study in-cluded neurodevelopmental follow-upand MR imaging follow-up at ages 1 and2 years, but, to date, follow-up has notbeen performed in any of the neonateswith hemorrhage. Ultimately, we hopeto be able to determine whether ICH isassociated with later neurodevelopmen-tal problems.

The high prevalence of ICH in ourasymptomatic population is importantfor several reasons. Our findings indi-cate that vaginal birth may be inherentlytraumatic to the neonatal brain and canresult in a spectrum of ICHs, which in-clude subdural hematomas and sub-arachnoid, intraparenchymal, and ger-minal matrix hemorrhages. Holden et al(16) pointed out that retinal hemor-

rhage also is observed in 20%–40% ofnewborns and that red blood cells oftenare found in the cerebrospinal fluid ofnewborns, and these findings indicatethat there is trauma after vaginal birth.The long-term consequences of thesehemorrhages are unknown at this time,though it is likely that small subduralhemorrhages resolve quickly withoutsubstantial consequence.

It is possible, however, that some ofthese incidental ICHs that occur aftervaginal birth may have long-term conse-quences for subsequent neurocognitivedevelopment or may contribute to thedevelopment of “idiopathic” epilepsy.These incidental hemorrhages also mayincrease the risk for complex multifac-torial neuropsychiatric disorders suchas schizophrenia, which has been asso-ciated with perinatal birth complica-tions (42). In addition, ICH may be amarker of traumatic forces that couldcause more subtle injury to the develop-ing brain that would not be apparent onMR images; examples of these trau-matic forces are transient ischemia orwhite matter tract damage, which wouldalso affect subsequent neurodevelop-ment. We plan longitudinal follow-up ofthis cohort, and the findings from thisfollow-up may suggest answers to theseimportant questions.

References1. Palmer TW, Donn SM. Symptomatic sub-

arachnoid hemorrhage in the term newborn.J Perinatol 1991;11:112–116.

2. Painter MJ, Bergman I, Crumrine P. Neona-tal seizures. Pediatr Clin North Am 1986;33:91–109.

3. Mercuri E, Cowan F, Rutherford M, AcoletD, Pennock J, Dubowitz L. Ischaemic andhaemorrhagic brain lesions in newbornswith seizures and normal Apgar scores. ArchDis Child Fetal Neonatal Ed 1995;73:F67–F74.

4. Chaplin ER Jr, Goldstein GW, Norman D.Neonatal seizures, intracerebral hematoma,and subarachnoid hemorrhage in full-terminfants. Pediatrics 1979;63:812–815.

5. Volpe JJ. Neonatal intracranial hemorrhage:pathophysiology, neuropathology, and clini-cal features. Clin Perinatol 1977;4:77–102.

6. Volpe JJ. Neurology of the newborn. Phila-delphia, Pa: Saunders, 2001.

Table 3

Newborn and Labor Variables

VariableWith Hemorrhage(n � 17)

Without Hemorrhage(n � 71) P Value

Gestational age at birth (wk) 39.0 (1.0) 39.3 (1.6) .44Gestational age at MR

imaging (wk) 41.4 (1.2) 42.4 (1.7) .02Apgar score

1 min 8.1 (1.0) 7.8 (1.5) .485 min 8.9 (0.4) 8.8 (0.7) .46

Duration of hospital stay (d) 2.5 (1.2) 2.38 (1.5) .82Data at birth

Head circumference (cm) 33.8 (2.1) 34.4 (1.6) .24Weight (g) 3267.0 (449.5) 3398.2 (501.5) .33

Duration of labor (h)* 5.8 (0.9–22.4) 7.5 (0–22.4) .74Duration of ruptured

membranes (h)* 4.8 (0.2–18.4) 3.5 (0–23.5) .18

Note.—Data are means. Numbers in parentheses are standard deviations except where indicated otherwise.

* Data are medians. Numbers in parentheses are ranges.



7. Huang AH, Robertson RL. Spontaneous su-perficial parenchymal and leptomeningealhemorrhage in term neonates. AJNR Am JNeuroradiol 2004;25:469–475.

8. Sandberg DI, Lamberti-Pasculli M, DrakeJM, Humphreys RP, Rutka JT. Spontaneousintraparenchymal hemorrhage in full-termneonates. Neurosurgery 2001;48:1042–1048.

9. Castillo M, Fordham LA. MR of neurologi-cally symptomatic newborns after vacuumextraction delivery. AJNR Am J Neuroradiol1995;16:816–818.

10. Towner D, Castro MA, Eby-Wilkens E, Gil-bert WM. Effect of mode of delivery in nul-liparous women on neonatal intracranial in-jury. N Engl J Med 1999;341:1709–1714.

11. Odita JC, Hebi S. CT and MRI characteris-tics of intracranial hemorrhage complicatingbreech and vacuum delivery. Pediatr Radiol1996;26:782–785.

12. Gardella C, Taylor M, Benedetti T, Hitti J,Critchlow C. The effect of sequential use ofvacuum and forceps for assisted vaginal de-livery on neonatal and maternal outcomes.Am J Obstet Gynecol 2001;185:896–902.

13. Hanigan WC, Morgan AM, Stahlberg LK,Hiller JL. Tentorial hemorrhage associatedwith vacuum extraction. Pediatrics 1990;85:534–539.

14. Wen SW, Liu S, Kramer MS, et al. Compar-ison of maternal and infant outcomes be-tween vacuum extraction and forceps deliv-eries. Am J Epidemiol 2001;153:103–107.

15. Heibel M, Heber R, Bechinger D, KornhuberHH. Early diagnosis of perinatal cerebral le-sions in apparently normal full-term new-borns by ultrasound of the brain. Neuroradi-ology 1993;35:85–91.

16. Holden KR, Titus MO, Van Tassel P. Cranialmagnetic resonance imaging examination ofnormal term neonates: a pilot study. J ChildNeurol 1999;14:708–710.

17. Tavani F, Zimmerman RA, Clancy RR, LichtDJ, Mahle WT. Incidental intracranial hem-orrhage after uncomplicated birth: MRI be-fore and after neonatal heart surgery. Neu-roradiology 2003;45:253–258.

18. Whitby EH, Griffiths PD, Rutter S, et al.Frequency and natural history of subdural

haemorrhages in babies and relation to ob-stetric factors. Lancet 2004;363:846–851.

19. Gilmore JH, Zhai G, Wilber K, Smith JK, LinW, Gerig G. 3 Tesla magnetic resonance im-aging of the brain in newborns. PsychiatryRes 2004;132:81–85.

20. Chamnanvanakij S, Rollins N, Perlman JM.Subdural hematoma in term infants. PediatrNeurol 2002;26:301–304.

21. Perlman JM. Brain injury in the term infant.Semin Perinatol 2004;28:415–424.

22. Bergman I, Bauer RE, Barmada MA, et al.Intracerebral hemorrhage in the full-termneonatal infant. Pediatrics 1985;75:488–496.

23. Roland EH, Flodmark O, Hill A. Thalamichemorrhage with intraventricular hemor-rhage in the full-term newborn. Pediatrics1990;85:737–742.

24. Hernansanz J, Munoz F, Rodriguez D, SolerC, Principe C. Subdural hematomas of theposterior fossa in normal-weight newborns:report of two cases. J Neurosurg 1984;61:972–974.

25. Sachs BP, Acker D, Tuomala R, Brown E.The incidence of symptomatic intracranialhemorrhage in term appropriate-for-gesta-tion-age infants. Clin Pediatr (Phila) 1987;26:355–358.

26. Pollina J, Dias MS, Li V, Kachurek D, Arbes-man M. Cranial birth injuries in term new-born infants. Pediatr Neurosurg 2001;35:113–119.

27. Fink S. Intraventricular hemorrhage in theterm infant. Neonatal Netw 2000;19:13–18.

28. Whitby EH, Paley MN, Smith MF, Sprigg A,Woodhouse N, Griffiths PD. Low fieldstrength magnetic resonance imaging of theneonatal brain. Arch Dis Child Fetal Neona-tal Ed 2003;88:F203–F208.

29. Scotti G, Flodmark O, Harwood-Nash DC,Humphries RP. Posterior fossa hemorrhagesin the newborn. J Comput Assist Tomogr1981;5:68–72.

30. Pierre-Kahn A, Renier D, Sainte-Rose C,Hirsch JF. Acute intracranial hematomas interm neonates. Childs Nerv Syst 1986;2:191–194.

31. Tanaka Y, Sakamoto K, Kobayashi S, Koba-

yashi N, Muraoka S. Biphasic ventricular di-latation following posterior fossa subduralhematoma in the full-term neonate. J Neuro-surg 1988;68:211–216.

32. Avrahami E, Frishman E, Minz M. CT dem-onstration of intracranial haemorrhage interm newborn following vacuum extractordelivery. Neuroradiology 1993;35:107–108.

33. Welch K, Strand R. Traumatic parturitionalintracranial hemorrhage. Dev Med ChildNeurol 1986;28:156–164.

34. Hayashi T, Hashimoto T, Fukuda S, OhshimaY, Moritaka K. Neonatal subdural hema-toma secondary to birth injury: clinical anal-ysis of 48 survivors. Childs Nerv Syst 1987;3:23–29.

35. Williamson WD, Percy AK, Fishman MA, etal. Cerebellar hemorrhage in the termneonate: developmental and neurologic out-come. Pediatr Neurol 1985;1:356–360.

36. Hayashi T, Harada K, Honda E, UtsunomiyaH, Hashimoto T. Rare neonatal intracerebralhemorrhage: two cases in full-term infants.Childs Nerv Syst 1987;3:161–164.

37. Lonergan GJ, Baker AM, Morey MK, BoosSC. From the archives of the AFIP. Childabuse: radiologic-pathologic correlation. Ra-dioGraphics 2003;23:811–845.

38. Keeney SE, Adcock EW, McArdle CB. Pro-spective observations of 100 high-risk neo-nates by high-field (1.5 Tesla) magneticresonance imaging of the central nervoussystem. II. Lesions associated with hypoxic-ischemic encephalopathy. Pediatrics 1991;87:431–438.

39. McArdle CB, Richardson CJ, Hayden CK,Nicholas DA, Crofford MJ, Amparo EG. Ab-normalities of the neonatal brain: MR imag-ing. I. Intracranial hemorrhage. Radiology1987;163:387–394.

40. Kidwell CS, Chalela JA, Saver JL, et al. Com-parison of MRI and CT for detection of acuteintracerebral hemorrhage. JAMA 2004;292:1823–1830.

41. Brooks RA, Di Chiro G, Patronas N. MRimaging of cerebral hematomas at differentfield strengths: theory and applications.J Comput Assist Tomogr 1989;13:194–206.

42. Gilmore JH, Murray RM. Prenatal and peri-natal factors. In: Lieberman JA, Perkins DO,Stroup TS, eds. Textbook of schizophrenia.Washington, DC: American PsychiatricPress, 2006; 55–67.



Rooks, V.J.; J.P. Eaton; L. Ruess; G.W. Petermann; J. Keck-Wherley; and R.C. Pedersen (2008) Prevalence and Evolution of Intracranial Hemorrhage in Asymptomatic Term Infants American Journal of Neuroradiology pg. 1-8

Article discusses study in which 101 infants were examined at 3-7 days, 2 weeks, 1 month, and 3 months after birth by using MRI’s of the brain and sonography to measure subdural hematomas in these asymptomatic infants. The study found that 46 of the neonates had SDH within 72 hours of birth though all were “normal” upon physical examination. SDHs found were in both cesarean and vaginal deliveries. Conclusions include understanding that SDH found in asymptomatic neonates is limited in size and location

ORIGINALRESEARCH

Prevalence and Evolution of IntracranialHemorrhage in Asymptomatic Term Infants

V.J. RooksJ.P. Eaton

L. RuessG.W. PetermannJ. Keck-Wherley

R.C. Pedersen

BACKGROUND AND PURPOSE: Subdural hemorrhage (SDH) is often associated with infants experienc-ing nonaccidental injury (NAI). A study of the appearance and natural evolution of these birth-relatedhemorrhages, particularly SDH, is important in the forensic evaluation of NAI. The purpose of this studywas to determine the normal incidence, size, distribution, and natural history of SDH in asymptomaticterm neonates as detected by sonography (US) and MR imaging within 72 hours of birth.

MATERIALS AND METHODS: Birth history, delivery method, duration of each stage of labor, pharma-ceutic augmentation, and complications during delivery as well as postnatal physical examination wererecorded. Brain MR imaging and US were performed on 101 asymptomatic term infants at 3–7 days,2 weeks, 1 month, and 3 months. Clinical follow-up at 24 months was recorded.

RESULTS: Forty-six neonates had SDH by MR imaging within 72 hours of delivery. SDH was seen inboth vaginal and cesarean deliveries. All neonates were asymptomatic, with normal findings onphysical examination. All 46 had supratentorial SDH seen in the posterior cranium. Twenty (43%) alsohad infratentorial SDH. US detected 11 of the 20 (55%) infratentorial SDHs and no supratentorial SDH.Most SDHs present at birth were �3 mm and had resolved by 1 month, and all resolved by 3 monthson MR imaging. Most children with SDHs had normal findings on developmental examinations at 24months.

CONCLUSION: SDH in asymptomatic term neonates after delivery is limited in size and location.

Subdural hemorrhage (SDH) is often associated with in-fants experiencing nonaccidental injury (NAI).1-13 Birth-

related trauma is used in the court of law as an explanation forSDH in infants with suspected NAI because a variety of hem-orrhages have been reported in term neonates. A study of theappearance and natural evolution of these birth-related hem-orrhages, particularly SDH, is important in the forensic eval-uation of NAI. A few published series report the finding ofhemorrhages in infants who were symptomatic in the neonatalperiod.14-18 Some reports suggest that the risk of SDH andother hemorrhages found on imaging of symptomatic infantsvaries with the method of delivery.19 Sonography (US) is stan-dard practice for detecting germinal matrix hemorrhage in thepreterm neonate and has also been proved to demonstrateposterior fossa SDH.14 MR imaging in general has a high sen-sitivity for intracranial hemorrhage, and, with its lack of ion-izing radiation, is a favorable technique for the evaluation ofbirth trauma over CT, especially for a neonate. Previous stud-ies conducted in an effort to determine the incidence and nat-ural history of asymptomatic SDH in the neonate have beenlimited by the use of low-field-strength (0.2T) MR imaging,

small patient numbers, or variable timing of imaging afterbirth.20-23

The purpose of this study was to determine the normalincidence, size, appearance, and distribution of SDH inasymptomatic term neonates as detected by US and 1.5T MRimaging within 72 hours of birth. In addition, we prospectivelystudied the natural history of these hemorrhages. This studycan then serve as a baseline for comparison with an abnormalpattern of SDH seen in abuse.

MethodsThe protocol was approved by the Scientific Review and Human Use

Committees of the hospital. Neonates of at least 37 weeks gestation,

with normal findings on neonate physical examination by a board-

certified physician were eligible for the study. The first 101 patients

whose parents gave written consent during the approved study period

were included. Birth history, delivery method, duration of each stage

of labor, pharmaceutic augmentation with oxytocin, and complica-

tions during delivery were recorded. All neonates had normal findings

on neurologic examination by a board-certified child neurologist be-

fore imaging. Ophthalmologic examination of the retina was not per-

formed on any neonates. The first MR imaging and US for each pa-

tient were performed at �72 hours of age.

US was performed on an Acuson Sequoia 512 (Siemens Medical

Solutions, Malvern, Pa) by using 8V5 and 15L8 transducers. Standard

coronal and sagittal images of the neonatal brain through the anterior

fontanelle and images of the posterior fossa via the mastoid fontanelle

were obtained. Color Doppler flow imaging was also used when the

findings of gray-scale imaging were positive for SDH. US was per-

formed within 1 hour of the MR imaging. SDH was defined as an

extracerebral curvilinear echogenicity subjacent to the calvaria with-

out evidence of central traversing vessels on color Doppler imaging.

Imaging was timed to occur after a morning feeding. Infants were

transported to the radiology department in a mobile bassinette,

placed on the MR imaging table in an 8-channel head coil, and se-

cured with a sheet, sponges, and tape to minimize motion. Pieces of

Received November 13, 2007; accepted after revision December 24.

From the Departments of Radiology (V.J.R., J.P.E., L.R., G.W.P.) and Pediatrics (J.K.-W.,R.C.P.), Tripler Army Medical Center Honolulu, Hawaii; the Departments of Radiology andRadiological Sciences (V.J.R., L.R., G.W.P.) and Pediatrics (L.R., R.C.P.), F. Edward HebertSchool of Medicine, Uniform Services University of the Health Sciences, Bethesda, Md; andWeed Army Community Hospital (J.P.E.), Fort Irwin, Calif.

Previously presented at: Annual Meeting of the American Society for Neuroradiology, April29 –May 5, 2006; San Diego, Calif.

The views expressed herein are those of the authors and do not reflect the official policyof the Department of the Army, Department of Defense, or the US Government.

Please address correspondence to Veronica J. Rooks, MD, Department of Radiology,MCHK-DR, Tripler Army Medical Center, Honolulu, HI 96859-5000; e-mail: [email protected]

indicates article with supplemental on-line table.

DOI 10.3174/ajnr.A1004

PEDIA

TRICSORIGIN

ALRESEARCH

AJNR Am J Neuroradiol ● :● � ● 2008 � www.ajnr.org 1

Published April 3, 2008 as 10.3174/ajnr.A1004

Copyright 2008 by American Society of Neuroradiology.

standard foam ear protection were taped in place, and a pacifier was

offered for comfort. No infants were given sedation medications.

With a Signa 1.5T MR imaging scanner (software 11.0_M4_0403a)

(GE Healthcare, Milwaukee, Wis), we used the following imaging

sequences: 1) 3-plane localizer; 2) sagittal T2 single-shot fast spin-

echo (SE) 2D pulse sequence imaging option with a TE of 90, TR of

3000, bandwidth of 31.25, FOV of 18, section thickness of 4, 0 skip,

matrix of 256 � 192, frequency signal intensity, NEX 1, phase FOV of

0.70; 3) axial multiplanar gradient recall (MPGR) pulse sequence gra-

dient-echo imaging option, flow comp, VBW, with a TE of 20, TR of

355, flip angle of 20°, bandwidth of 15.63, FOV of 18, section thick-

ness of 4, 0 skip, matrix of 256 � 192, frequency AP, NEX 1, phase

FOV of 0.75; 4) axial T1 conventional SE 2D pulse sequence imaging

option, VBW, TE min, TR of 377, bandwidth of 15.63, SAT I, FOV of

18, section thickness of 4, 0 skip, matrix 256 � 192, NEX 0.75, phase

FOV of 0.75, frequency AP; 5) coronal T1 (posterior fossa) 2D pulse

sequence SE imaging option, VBW, TE min, TR of 502, bandwidth of

15.63, SAT I, FOV of 18, section thickness of 4, 0 skip, matrix 256 �

192, frequency direction S/I, NEX 0.75, phase FOV of 0.75; 6) axial

fluid-attenuated inversion recovery (FLAIR) 2D pulse sequence IR

imaging option, tailored radio-frequency fast, zip of 512, TE of 120,

TR of 10,000, TI of 2200, bandwidth of 15.63, FOV of 18, section

thickness of 4, 0 skip, matrix 256 � 224, frequency direction A/P,

NEX 1; 7) axial diffusion-weighted echo-planar imaging (DWI EPI)

2D SE imaging option (DIFF), number of shots 1, TE min, TR of

10,000, DWI screen b-value of 1500, diffusion direction ALL, fre-

quency of 128/128, NEX 1, FOV of 18, section thickness of 44, 0 skip,

matrix 128 � 128. Conventional SE T1 was substituted for fast SE

after 42 patients were scanned.

MR and US images were independently reviewed on a PACS

(Centricity; GE Healthcare) by 2 board-certified radiologists each

with a Certificate of Added Qualification in neuroradiology or pedi-

atric radiology. The child neurologist discussed imaging results with

parents. Infants with SDH detected on initial imaging were scheduled

for follow-up MR imaging and US examinations at 3–7 days, 2 weeks,

1 month, and 3 months or until the MR imaging and US findings were

both negative. If the initial US findings were normal, no further US

images were obtained. Final interpretations regarding the presence of

SDH on MR imaging were determined by consensus of 2 of the radi-

ologists based on SDH seen on both the immediate postdelivery initial

MR imaging and the first follow-up at 3–7 days. SDH on MR imaging

was defined as an extracerebral curvilinear signal-intensity abnormal-

ity corresponding to blood products that did not extend into the sulci.

For US and MR imaging, SDH location and size were recorded, with

size measured as a maximal width in the axial plane by using elec-

tronic calipers. In infants with SDH in multiple locations, the size of

the largest SDH was recorded. The presence of cephalohematomas

was also recorded. Evaluation for coagulopathy was not routinely

performed.

Comparison of the incidence of SDH among the delivery groups

was made by using the Fisher exact test. The average labor times and

birth weights of those with SDH and those without were compared by

using a Student t test or the nonparametric Wilcoxon test if the vari-

ance in data was unequal between groups. The Fisher exact test was

used to compare prolonged duration of labor and incidence of cepha-

lohematoma in those with SDH versus those without. The compari-

son of the incidence of SDH in vaginal and cesarean deliveries aug-

mented with oxytocin administration was also performed by using

the Fisher exact test in addition to computation of the odds ratio of

increased SDH associated with giving oxytocin. Data are expressed as

mean � standard error of the mean and/or as a median within the

range of values obtained. For all tests, a value of P � .05 was consid-

ered significant. The first stage of labor was defined as the duration

from the onset of labor until the fetus was engaged in the birth canal.

The second stage of labor was defined as the duration of fetal descent

through the birth canal.

Clinical follow-up was performed in all patients who demon-

strated SDH on imaging. Patients were evaluated at their 24-month

well-child visit and assessed for developmental delay. A developmen-

tal delay (motor or speech) was defined as a delay in a particular

developmental domain compared with expected norms for age. De-

velopmental delay is used as a temporary diagnosis in young children

at risk for developmental disabilities, indicating a failure to achieve

age-appropriate neurodevelopmental milestones.24 At our institu-

tion, assessing for developmental delays is part of every well-child

visit, typically performed at 2, 4, 6, 12, 15, 18, and 24 months. The

Denver Developmental Screening Test II is applied to each child at

their well-child visit.25

ResultsOne hundred one patients were enrolled in the study betweenJanuary 2005 and March 2006. There were 58 male and 43female infants. Seventy-nine (78%) infants were born via vag-inal delivery with (80%) via spontaneous delivery (SVD), 10(12%) with vacuum assistance, and 6 (8%) with forceps assis-tance (supplemental on-line Table). Thirty-five vaginal deliv-eries were induced or augmented with oxytocin. Twenty-two(22%) infants were delivered via cesarean delivery: 13 electivecesarean deliveries and 9 for failure to progress and/or fetaldistress after a trial of labor. Four of the cesarean deliveries hada trial of labor augmented with oxytocin. One cesarean deliv-ery was assisted with forceps, and 1 was assisted with vacuumextraction. All neonates had normal findings on neurologicexaminations at birth.

All 101 initial MR imaging examinations were successful,without significant motion artifact. Most infants slept throughthe entire examination. Examination times required �10minutes to complete. Three MR imaging examinations werethought to be positive for SDH on initial sequences, but thefindings were normal at the first follow-up MR imaging by 3–7days of life. These were presumed to be false-positive findingsand were recategorized as negative findings. Forty-six (46%)infants had SDH on initial MR imaging that was confirmed onfollow-up studies (supplemental on-line Table). Forty-four of46 (95.9%) had SDH of �3 mm in thickness (range, 1.0 – 4.3mm; mean, 2.1 mm). SDH was best visualized on the initialMR imaging MPGR sequence performed before 72 hours oflife (Fig 1). All 46 patients with intracranial hemorrhage hadsupratentorial SDH confirmed on 2 imaging planes on fol-low-up imaging. All supratentorial SDHs identified within 72hours postdelivery were seen in the posterior half of the cra-nium. Twelve (26%) infants had SDH noted in only 1 location,whereas most infants had SDH in 2 or 3 locations. In all, SDHwas most commonly seen in the posterior interhemisphericfissure (parafalcine location) (30, 65%), with SDH also notedposteriorly along the occipital lobes in 29 (63%) and over thetentorium in 22 (48%) (supplemental on-line Table). AllSDHs were homogeneous in signal intensity on all sequences.

Twenty (43%) of the neonates with supratentorial SDHalso had posterior fossa SDH (Fig 2). No neonate had only

2 Rooks � AJNR ● � ● 2008 � www.ajnr.org

posterior fossa hemorrhage detected by MR imaging. No ne-onate had MR imaging evidence of subarachnoid, epidural, orintraparenchymal hemorrhage. No parenchymal contusionswere seen. Two neonates had grade I germinal matrix hemor-

rhages (1 unilateral, 1 bilateral) as well as SDHs. Twenty-twoneonates had a cephalohematoma noted at MR imaging. Eigh-teen (82%) of these neonates had SDH. Most (11/18, 61%)had posterior fossa SDH as well as supratentorial SDH. One

Fig 1. Posterior fossa SDH in a neonate delivered via SVD. A,Axial MPGR at �72 hours of life demonstrates lobularsymmetric low signal intensity with blooming in the posteriorfossa (arrows). B, Follow-up T1 images show high-signal-intensity SDH (arrowheads) by 7 days.

Fig 2. Neonate delivered via SVD with both supratentorial and infratentorial SDH. A and B, Initial examination shows the lobular occipital SDH to be very low signal intensity on MPGR(arrows, A) and isointense to gray matter and difficult to detect on the SE T1-weighted MR image (B). C and D, Five-day follow-up shows high T1 SDH (arrowheads) in 2 locations in 2planes, axial supratentorial (C) and coronal, both supra- and infratentorial (D). E and F, Two-week follow-up shows complete resolution of hemorrhage on T1 images.


had a 1.6-cm paraventricular mass incidentally detected onMR imaging, which was not seen on repeat US performed afterthe initial MR imaging. The mass, thought to be a hamartoma,was observed with expectant management. It remainedasymptomatic and unchanged in size on 4-month follow-up atour institution before the patient’s family moved from ourarea.

Posterior fossa SDH was seen at US in 11 (11%) neonates,and all SDHs were confirmed on MR imaging (Fig 3). Thus,only 55% of the 20 posterior fossa SDHs seen on MR imagingwere identified independently on US examination. US wasfocused along the lateral aspects through the mastoid fonta-nelle. Sensitivity of US detection of posterior fossa SDH im-proved when the 3 infants with posterior fossa SDH isolated tomidline were excluded on MR imaging3; thus, 11/17 (65%)lateral posterior fossa SDHs were detected on US. All SDHsseen on US were also seen on MR imaging. No supratentorialhemorrhages were detected at US.

The incidence of SDH versus mode of delivery is shown inTable 1. All 4 neonates with SDH delivered by cesarean birthhad supratentorial SDH only. One of the neonates with SDHand delivered by cesarean birth was born via elective cesareandelivery for macrosomia, whereas 3 of 4 (75%) neonates withSDH and delivered by cesarean birth had failed a trial of oxy-tocin-augmented labor before cesarean delivery. One of thesecesarean deliveries required vacuum assistance. In compari-son with the neonates delivered via cesarean delivery, rates ofSDH were significantly higher in all the vaginal deliverygroups (Table 1). There was no statistically significant differ-ence in the presence of SDH in each of the vaginal deliverygroups.

The duration of the first and second stages of labor wasrecorded for all neonates delivered vaginally. For neonateswith SDH, the mean duration of the first stage of labor was notsignificantly different from that in those without SDH (Table

2). The second stage of labor was significantly longer in neo-nates with SDH than in those without SDH. A prolonged sec-ond stage of labor (�2 hours) was also significantly longer inthe group with SDH, compared with the group without SDH.The incidence of cephalohematoma was greater in neonateswith SDH than in those without SDH. There was no differencein average second-stage labor duration in those with a cepha-lohematoma compared with those without. The mean birthweight of neonates with SDH on MR imaging was higher thanthat of those with normal findings on MR imaging (Table 2).

The overall incidence of SDH in the 39 patients who re-ceived oxytocin was not different from the incidence of SDHin the 62 patients who did not receive oxytocin (Table 3). Thiswas also true for the subgroup of vaginal deliveries. However,closer examination of cesarean delivery revealed that the inci-dence of SDH when oxytocin was given before cesarean deliv-ery was much higher (Table 3).

Follow-up imaging was completed in 18/46 (39.1%) pa-tients with SDH. All 18 patients demonstrated resolution by 3months. Two patients were only imaged at birth and at 3months due to scheduling conflicts. Both of these patients hadnormal MR imaging findings at 3 months. Fifteen of 16 pa-tients (93.8%) whose follow-up imaging included a 1-monthMR imaging had interval resolution of their SDHs. One pa-tient had a new frontal SDH on the 2-week MR imaging fol-low-up examination (Fig 4). This patient had bilateral occip-ital and posterior fossa SDH on initial imaging at birth,confirmed on the 7-day follow-up MR imaging. He was alsonoted to have extra-axial collections of infancy. At 26-dayspostnatal age, the MR imaging demonstrated left frontal sub-dural collections that did not conform to CSF signal intensity.Of the 46 infants with SDH, 43 children had records of 2 yearsof well-baby examinations at our institution. One child wasonly followed to 2 months, 1 child’s family had moved out ofthe area, and 1 child was not eligible for continued care in oursystem. None of the 43 infants had gross motor delay. Six(14%) children were noted to have speech delay, and 1 (2%) iscurrently being evaluated for an autistic spectrum disorder.

DiscussionWe confirmed reports that SDH occurs in the asymptomaticneonate after delivery.20-22 The incidence of SDH (46%) issignificantly higher in our study than in previous reports. Our

Fig 3. Neonate delivered via SVD with posterior fossa SDHseen on US and confirmed on MR imaging. A, Axial sonogramof the posterior fossa through the mastoid fontanel demon-strates initial curvilinear echogenic focus adjacent to thetransverse sinus (arrow). B, Axial T1-weighted MR imageconfirms high-signal-intensity posterior fossa SDH (arrow-head) on day 7 of life.

Table 1: SDH versus mode of delivery

SVD Vacuum Forceps C/STotal deliveries 63 10 6 22SDH 32 6 4 4

Percentage 51 60 67 18P value* �.05 �.05 �.05

Note:—C/S indicates cesarean delivery; SVD, spontaneous vaginal delivery.* P values represent significance of comparisons with the C/S group.


higher incidence may be related to improved detection andincreased sensitivity with a higher magnetic-field-strength1.5T MR imaging scanner. Whitby et al,20 by using a low-field-strength 0.2T magnet, reported an SDH incidence of 8% over-all and 10.5% in vaginal deliveries when they imaged withinthe first 48 hours of life. Our reported incidence is most likethat of Holden et al,21 who, in a pilot study also using 1.5T MRimaging in 1999, saw SDH in 4 of 8 (50%) asymptomatic ne-onates in the first 4 days of life. These results suggest that SDHafter uncomplicated vaginal delivery is a common finding onMR imaging.

Patient age at the time of MR imaging is an importantfactor in determining the incidence of SDH in neonates. Weimaged neonates within the first 72 hours of life and foundSDH most readily detectable on a gradient-echo sequence,confirmed on follow-up T1 sequences at 3–7 days of life. Mostof the SDHs resolved by 4 weeks. Whitby et al20 also found thattheir 9 patients with SDH first seen within 48 hours of life hadresolution of hemorrhage on MR imaging at 4-week follow-up. Recently, Looney et al,22 by using 3T MR imaging, re-ported SDH in 26% of neonates delivered vaginally. Infants inthis study were scanned between 1 and 5 weeks of age. Weagree that the true incidence in the population of Looney et almay have been higher than the prevalence reported becausethey may have missed SDHs that were present earlier in lifeand had resolved by the time of first imaging. Patient age at thetime of MR imaging may also be important in determining anetiology for neonate SDH. In our patients, not only were mostSDHs resolved by 1 month but SDHs had resolved by 3months in all patients. This information may be useful to theradiologist asked to comment on the etiology of SDH in aninfant. Our study suggests that SDH in an infant older than 3months of age is unlikely to be birth-related regardless of themode of delivery.

Proposed mechanisms for SDH have included tears of thefalx and tentorium or bridging cortical veins secondary tostretching,11 difficult delivery,26,27 or abnormal labor.19 One

suggested mechanism of hemorrhage after vaginal delivery isthat increased circumferential pressure and squeezing of thehead in the birthing canal result in overlap at the sutures, me-chanical compression, and shearing of the bridging veins dur-ing delivery, resulting in SDH.28 The true etiology remainsunknown because there is a paucity of evidence-based litera-ture on this subject. Most reports of SDH in the neonate ap-pear in the larger body of literature on infants who presentwith symptomatic SDHs.

The forensic literature suggests that SDH can result fromrupture of bridging cerebral veins; however, it is difficult todemonstrate rupture of bridging cerebral vessels at autop-sy.29,30 Towner et al19 suggested that abnormal labor was acommon risk factor for hemorrhage in infants, after a retro-spective review of deliveries in nulliparous women demon-strated a low incidence of intracranial hemorrhage. Pollina etal27 suggested that the method of assisted delivery rather thanthe urgency of the delivery or dysfunctional labor is a moreimportant variable in cranial birth injuries. Although all typesof intracranial hemorrhage were more common in vacuumextraction, not all term neonate SDHs can be explained bycircumferential head squeeze and overlapping sutures. Thisfinding is particularly true because we found SDH after cesar-ean delivery as well. Perhaps additional forces during parturi-tion are at work contributing to the rupturing of veins and orcapillaries.

In our study, the first and second stages of labor werelonger in the infants with SDH than in those without SDH.Perhaps compressive force from the uterus during the firststage, which propels the infant into the birth canal, is a caus-ative factor. A prolonged first stage in combination with aprolonged second stage of labor may be causative in that theremay not only be increased prolonged propulsive forces butalso increased molding and overlapping of sutures, which maylead to failure of tensile strength of the stretched vessels. In-creased pressure during the labor process may augment theintracranial venous pressures, which also may be an additionalfactor leading to SDH. The incidence of SDH in our study wasgreater in neonates with cephalohematoma and was also asso-ciated with a longer second stage of labor. The overall birthweight of neonates with SDH was also significantly higher,which may have resulted in increased circumferential pressureforces from the birth canal. Although all of these factors or acombination of these factors is plausible for the mechanism,SDH as a product of parturition has now been documented inasymptomatic neonates in multiple studies.14-23,26-28

Although most of the asymptomatic SDHs seen at MR im-aging and US were in neonates delivered vaginally, 18% (4 of

Table 2: Vaginally delivered neonates with and without SDH: mean length of each stage of labor, birth weight, and incidence ofcephalohematoma

1st stage (min) 2nd stage (min) 2nd stage �120 minutes Cephalohematoma Birth Weight (g)No SDH 414, 50, 3 5 3404,(n � 55) range, 37–1439; range, 11–554; 5% 9% range, 2842–4379

median, 357 median, 19SDH 448, 96, 12 18 3589,(n � 46) range, 75–1397; range, 1–593; 26% 39% range, 2867–4583

median, 380 median, 27P value* �.01 �.01 �.01 �.01

* P values represent significance of comparisons between no SDH and SDH.

Table 3: SDH and use of oxytocin in vaginal and cesareandeliveries

Vaginal Cesarean DeliveryOxytocin No Yes No Yes

No. 44 35 18 4No SDH 18 19 17 1

Percentage 41 54 94.4 25SDH 26 16 1 3

Percentage 59 46 5.6 75P Value* �.01 �.01 �.01 �.01

* P values represent significance of comparisons between no SDH and SDH.


22) of our neonates delivered by cesarean birth also had SDH.Most infants with SDH delivered by cesarean birth (75%) hada trial of labor with oxytocin administration before the cesar-ean delivery. This supports the proposal that SDH may berelated to labor. Presumably, the neonate experienced laborduring oxytocin administration before the decision for cesar-ean delivery.

All previous reports of SDH associated with cesarean deliv-eries have been in symptomatic infants. Welch and Strand31

reported a series of neonates with a variety of intraparturi-tional intracranial hemorrhages, including 3 who had SDHand complicated cesarean deliveries either for failure to de-scend, forceps failure, or fetal distress. Studies reporting theincidence or prevalence of SDH in asymptomatic neonateshave not reported hemorrhages in association with cesareandeliveries. The series by Whitby et al,20 using low fieldstrength, did not report SDH after cesarean delivery evenwhen vacuum-assisted delivery was attempted. Most recently,Looney et al22 reported no SDH in 23 cesarean deliveries. Thedelayed initial imaging at 1–5 weeks could account for the lowincidence of SDH detection in that study because most SDHsin our patients resolved by 4 weeks.

The only hemorrhages detected were SDH. The locationand size of the SDHs were limited. Most SDHs in our neonateswere �3 mm. There were 2 neonates with an initial SDH �3mm. One of these neonates had a presumed hamartoma withan occipital SDH measuring 3.3 mm. The other infant hadincreased extra-axial spaces and an initial occipital SDH of 4.3mm. We believe that these infants had factors that may havepredisposed them to a larger initial SDH. Like other investiga-

tors,20,22 we found most SDHs were in the posterior half of thecalvaria.

In our patients, supratentorial hemorrhage was more com-mon, with 39% also having infratentorial posterior fossa hem-orrhage. Both Looney et al22 and Whitby et al20 reported in-fratentorial hemorrhage alone being significantly morecommon. We believe that confirmatory coronal imaging washelpful in assessing supratentorial-versus-infratentorial hem-orrhage. Only if we saw the blood products below the tento-rium on the coronal view, would we assess the hemorrhage asinfratentorial, which is depicted in Fig 2D. This finding wasdifficult to assess on the initial imaging series obtained withinthe first 72 hours of life but was confirmed on subsequentcoronal T1 imaging. Also very small 1- to 2-mm supratentorialhemorrhages, which were raised as possible SDHs on initialgradient-echo sequences, were not confirmed to be SDH un-less found as hyperintense on the T1 follow-up imaging. Thisfinding on 2 subsequent imaging studies may have increasedthe number of overall supratentorial SDHs that were detectedin comparison with that of other investigators.

In our patients, both the infratentorial and supratentorialhemorrhages were posterior in the cranium except for 1 SDHnot present on the initial MR imaging (�72 hours postdeliv-ery) but found at a follow-up study. Initially, this patient hadbilateral posterior occipital SDHs, which were being followedfor resolution. At 26 days of life, the patient returned for thefollow-up MR imaging and was noted to have a 1-cm extra-axial left frontal collection that did not conform to CSF atten-uation, consistent with a spontaneous SDH. The patient wasadmitted for full evaluation for nonaccidental injury to in-

Fig 4. Images obtained at 7 and 26 days postnatal age for follow-up of bilateral occipital SDH in a neonate with extra-axial collections. Axial T2, T1, gradient-refocused echo (GRE), andFLAIR images (left to right, top row) show CSF-intensity frontal subarachnoid collections that were present since birth. Also note a thin linear T1 hyperintense GRE hypointense bilateralposterior occipital SDH. At 26 days postnatal age (bottom row), left frontal subdural collections that do not conform to CSF signal intensity are present, consistent with spontaneous SDH.The patient had no history of trauma and had a negative evaluation for NAI.


clude skeletal survey, ophthalmologic examination, coagula-tion panel, metabolic studies, as well as social work enquiries.These investigations did not reveal any additional injuries orfindings to support NAI as an etiology of the spontaneousfrontal SDH (Fig 4). At a 5-month follow-up MR imaging, theleft frontal SDH resolved; however, the subarachnoid spaceremained prominent in this patient. This finding suggests thatthough not typical in a neonate, prominent extra-axial space isa predisposing factor for SDHs as has been reported by otherauthors.32-35

Although SDH along the interhemispheric fissure, parafal-cine in location, is widely associated with NAI, we would sug-gest that the pattern and location of SDH alone should not beused to make a distinction between SDH due to NAI or birthinjury. In the pilot study of Holden et al, 21 there is a descrip-tion and illustration of an interhemispheric SDH in an asymp-tomatic neonate. The posterior location of the SDH is gener-ally common to reports of asymptomatic SDH, including ourstudy. Interhemispheric SDHs have been previously reportedin accidental trauma as well as in birth trauma and are nolonger considered specific for the type or mechanisms of inju-ry.36-38 We noted that the SDH was in a more dependent po-sition on follow-up imaging regardless of the location of theinitial hemorrhage and propose that this is likely due to therecommended practice of the American Academy of Pediat-rics of placing infants on their backs for sleep.39 When lyingsupine, gravity may account for the posterior locations of theSDH, indicating communication of the subdural space.

Although US could detect approximately half of the SDHs,the area imaged was limited to the lateral posterior fossa viathe mastoid fontanelle. Midline imaging of the posterior fossawas not routinely performed and thus the utility of US fordetection of SDH may have been underestimated. Still, nosupratentorial SDHs were detected on US. Clearly, MR imag-ing is more sensitive than US for the detection of SDH.

The 2-year follow-up of the infants with SDH was reassur-ing because all (100%) of the 43 children with documentedfollow-up had no gross motor delay. In our study population,6 (14%) of the children were noted to have speech delay, whichis similar to the known incidence in the general population.40

The 1 boy being evaluated for a possible autistic spectrumdisorder is not unexpected because autism is currently re-ported to have a prevalence of 1:150, with the prevalence inboys reported as high as 1:80.41-43 Normal findings on clinicalfollow-up are reassuring but are limited because there is nobaseline for comparison in the study design. We comparednormal development with that in children who met the crite-ria for the Denver Developmental Screening Test, which listsexpected milestones at each chronological age through 5 years.This expected development is our norm when assessing chil-dren in our clinic. Children not meeting expectations aremarked as having a delay and are referred for further evalua-tion to a subspecialty clinic.

One limitation of our study was the evaluation for rebleed-ing of SDH, which has been reported in the literature. Re-bleeding may present either with or without clinical symp-toms.44 Although none of our infants re-presented clinicallywith an SDH rebleed, the subclinical incidence of rebleedingin our population was not studied because none of the infantswere reimaged after 3 months of age. Normal development on

clinical examination is reassuring, indicating that major re-bleeding did not take place.

Another limitation in our study included the need tochange MR imaging sequences and timing. We found early onthat SDH was isointense to gray matter and intermittentlydifficult to see on initial imaging within the first 72 hours oflife. The SDH was seen as lobular low signal intensity withblooming on the gradient-echo imaging. The MR imagingfindings were considered positive for SDH if the positive gra-dient-echo sequence was confirmed on subsequent T1 imag-ing with hyperintense signal intensity by 3–7 days. To improveMR imaging for maximal sensitivity, we changed the originalT1-weighted fast spin-echo imaging sequence, spoiled gradi-ent-recalled (SPGR), to a spin-echo T1 sequence. The initialstudy performed with fast spin-echo SPGR was recategorizedfrom positive for extra-axial blood products to negative forextra-axial blood products if findings of the follow-up studyperformed at 3–7 days were negative. The recategorizationfrom positive to negative for SDH may have underestimatedthe actual number of SDHs in our neonate population. TheSDH on initial MR imaging may have been very small andresolved by the second MR imaging between 3–7 days of life.Therefore, the true incidence of SDH may be slightly higherthan that reported in this study. The initial follow-up timeinterval was also changed from 3 days to 5–7 days to accountfor the signal-intensity conversion changes. The initial fol-low-up at 3 days did not consistently demonstrate T1 hyper-intense signal intensity. We, therefore, lengthened the intervalto 5–7 days to allow blood products to change to T1 hyperin-tense, consistent with methemoglobin. This variability of thetime-interval imaging may have masked the actual timeframein which fetal hemoglobin changes signal-intensitycharacteristics.

Another limitation to the study is the lack of follow-upimaging in some patients. Follow-up imaging was only com-pleted in 18 of 46 infants with SDH. We were surprised to findthat, despite the parent knowing that their infant had SDH,follow-up appointments were often missed after the first2-week follow-up MR imaging and US. Selection bias of thepatient population is also a potential limitation. We relied ona random selection process limited by our ability to obtainwritten consent from the parents for our sample population.

ConclusionSDH is a common result of parturition and may be seen aftervaginal and cesarean delivery. MR imaging is more sensitivethan US for the detection of SDH. The hemorrhages seen inasymptomatic term neonates are limited in size and location.SDH after 1 month of age is unlikely to be birth-related.

AcknowledgmentsWe thank Dr. Adam Huillet for recruitment assistance, TessSchmidt for early morning sonography support, Paul Inuukand Cindy Lopes for early morning MR imaging expertise, Dr.Kristen Liddell for obstetric knowledge, Drs. Catherine Ure-thra and John Claybaugh for statistical support, and Dr. Mi-chael V. Krasnokutsky for manuscript review and comments.


References1. Barkovich AJ. Pediatric Neuroimaging. 3rd ed. New York: Lippincott Williams &

Wilkins; 20002. Kleinman PK. Diagnostic Imaging of Child Abuse. 2nd ed. Toronto, Ontario,

Canada: Mosby; 19983. Billmire ME, Myers PA. Serious head injury in infants: accident or abuse?

Pediatrics 1985;75:340 – 424. Hoskote A, Richards P, Anslow P, et al. Subdural haematoma and non-acci-

dental head injury in children. Childs Nerv Syst 2002;18:311–17. Epub 2002 Jun26

5. Jayawant S, Rawlinson A, Gibbon F, et al. Subdural haemorrhages in infants:population-based study. BMJ 1998;317:1558 – 61

6. Caffey J. Multiple fractures in the long bones of infants suffering from sub-dural hematoma. Am J Roentgenol 1946;56:163–73

7. Caffey J. On the theory and practice of shaking infants: its potential residualeffects of permanent brain damage and mental retardation. Am J Dis Child1972;124:161– 69

8. Caffey J. The whiplash shaken infant syndrome: manual shaking by the ex-tremities with whiplash-induced intracranial and intraocular bleedings,linked with residual permanent brain damage and mental retardation. Pedi-atrics 1974;54:396 – 403

9. Hadley MN, Sonntag VK, Rekate HL, et al. The infant whiplash-shake injurysyndrome: a clinical and pathological study. Neurosurgery 1989;24:536 – 40

10. Kempe CH, Silverman FN, Steele BF, et al. The battered-child syndrome. JAMA1962;181:17–24

11. Duhaime AC, Christian CW, Rorke LB, et al. Nonaccidental head injury ininfants: the “shaken-baby syndrome.” N Engl J Med 1998;338:1822–29

12. Barnes PD, Krasnokutsky M. Imaging of the central nervous system in sus-pected or alleged nonaccidental injury, including the mimics. Top Magn ResonImaging 2007;18:53–74

13. Feldman KW, Bethel R, Shugerman RP, et al. The cause of infant and toddlersubdural hemorrhage: a prospective study. Pediatrics 2001;108:636

14. Huang CC, Shen EY. Tentorial subdural hemorrhage in term newborns: ul-trasonographic diagnosis and clinical correlates. Pediatr Neurol 1991;7:171–77

15. Hayashi T, Hashimoto T, Fukuda S, et al. Neonatal subdural hematoma sec-ondary to birth injury: clinical analysis of 48 survivors. Childs Nerv Syst1987;3:23–29

16. Hanigan WC, Powell FC, Miller TC, et al. Symptomatic intracranial hemor-rhage in full-term infants. Child Nerv Syst 1995;11:698 –707

17. Huang AH, Robertson RL. Spontaneous superficial parenchymal and lepto-meningeal hemorrhage in term neonates. AJNR Am J Neuroradiol2004;25:469 –75

18. Miall LS, Cornette LG, Tanner SF, et al. Posterior fossa abnormalities seen onmagnetic resonance brain imaging in a cohort of newborn infants. J Perinatol2003;23:396 – 40

19. Towner D, Castro MA, Eby-Wilkens E, et al. Effect of mode of delivery innulliparous women on neonatal intracranial injury. N Engl J Med1999;341:1709 –14

20. Whitby EH, Griffiths PD, Rutter S, et al. Frequency and natural history ofsubdural haemorrhages in babies and relation to obstetric factors. Lancet2004;363:846 –51

21. Holden KR, Titus MO, Van Tassel P. Cranial magnetic resonance imagingexamination of normal term neonates: a pilot study. J Child Neurol1999;14:708 –10

22. Looney CB, Smith JK, Merck LH, et al. Intracranial hemorrhage in asymptom-atic neonates: prevalence on MR images and relationship to obstetric andneonatal risk factors. Radiology 2006;242:535– 41. Epub 2006 Dec 19

23. Tavani F, Zimmerman RA, Clancy RR, et al. Incidental intracranial hemor-rhage after uncomplicated birth: MRI before and after neonatal heart surgery.Neuroradiology 2003;45:253–58. Epub 2003 Mar 15

24. Batshaw ML, Pellegrino L, Roizen NJ, eds. Children with Disabilities. 5th ed.Baltimore: Paul H. Brooks Publishing Co; 2002

25. Frankenburg WK, Dodds J, Archer P, et al. The Denver II: a major revision andrestandardization of the Denver Developmental Screening Test. Pediatrics1992;89:91–97

26. Doward W, Sgouros S. Acute subdural haematomas following ventouse-as-sisted delivery. Pediatr Neurosurg 2001;35:335

27. Pollina J, Dias MS, Li V, et al. Cranial birth injuries in term newborn infants.Pediatr Neurosurg 2001;35:113–19

28. Menezes AH, Smith DE, Bell WE Posterior fossa hemorrhage in the term neo-nate. Neurosurg 1983;13:452–56

29. Maxeiner H. Detection of ruptured cerebral bridging veins at autopsy. Foren-sic Sci Int 1997;89:103–10

30. Stein KM, Ruf K, et al. Representation of cerebral bridging veins in infants bypostmortem computed tomography. Forensic Sci Int 2006;163:93–101. Epub2005 Dec 20

31. Welch K, Strand R. Traumatic parturition intracranial hemorrhage. Dev MedChild Neurol 1986;28:156 – 64

32. McNeely PD, Atkinson JD, Saigal G, et al. Subdural hematomas in infants withbenign enlargement of the subarachnoid spaces are not pathognomonic forchild abuse. AJNR Am J Neuroradiol 2006;27:1725–28

33. Amodio J, Spektor V, Pramanik B, et al. Spontaneous development of bilateralsubdural hematomas in an infant with benign infantile hydrocephalus: colorDoppler assessment of vessels traversing extra-axial spaces. Pediatr Radiol2005; 35:1113–17. Epub 2005 May 19

34. Pittman T. Significance of a subdural hematoma in a child with external hy-drocephalus. Pediatr Neurosurg 2003;39:57–59

35. Ravid S, Maytal J. External hydrocephalus: a probable cause for subdural he-matomas in infancy. Pediatr Neurol 2003;28:139 – 42

36. Tung GA, Kumar M, Richardson RC, et al. Comparison of accidental and non-accidental traumatic head injury in children on noncontrast computed to-mography. Pediatrics 2006;118:626 –33

37. Vinchon M, Noule N, Tchofo PJ, et al. Imaging of head injuries in infants:temporal correlates and forensic implications for the diagnosis of child abuse.J Neurosurg 2004;101(1 suppl):44 –52

38. Steinbok P, Singhal A, Poskitt K, et al. Early hypodensity on computed tomo-graphic scan of the brain in an accidental pediatric head injury. Neurosurgery2007;60:689 –94, discussion 694 –95

39. American Academy Of Pediatrics Task Force on Sudden Infant Death Syn-drome. The changing concept of sudden infant death syndrome: diagnosticcoding shifts, controversies regarding the sleeping environment, and newvariables to consider in reducing risk. Pediatrics November 10,2005;116:1245–55

40. Feldman HM. Evaluation and management of language and speech disordersin preschool children. Pediatr Rev 2005;26:131– 42

41. Autism and Developmental Disabilities Monitoring Network Surveillance Year2000 Principal Investigators; Centers for Disease Control and Prevention. Prev-alence of autism spectrum disorders: autism and developmental disabilitiesmonitoring network, six sites, United States, 2000. MMWR Surveill Summ2007;56:1–11

42. Autism and Developmental Disabilities Monitoring Network Surveillance Year2002 Principal Investigators; Centers for Disease Control and Prevention. Prev-alence of autism spectrum disorders: autism and developmental disabilitiesmonitoring network, 14 sites, United States, 2002. MMWR Surveill Summ2007;56:12–28

43. Autism and Developmental Disabilities Monitoring Network Surveillance Year2002 Principal Investigators; Centers for Disease Control and Prevention. Eval-uation of a methodology for a collaborative multiple source surveillance net-work for autism spectrum disorders: autism and developmental disabilitiesmonitoring network, 14 sites, United States, 2002. MMWR Surveill Summ2007;56:12–28.

44. Hymel KP, Jenny C, Block RW. Intracranial hemorrhage and rebleeding insuspected victims of abusive head trauma: addressing the forensic controver-sies. Child Maltreat 2002;7:329 – 48


Lucid Intervals - 2ndchairservices.com · Greenes D, Schultzman S.A. Occult intracranial injury in...

Documents

Transcript of Lucid Intervals - 2ndchairservices.com · Greenes D, Schultzman S.A. Occult intracranial injury in...