The pathophysiology of infant subdural haematomas

Matthew A Howard, B A Bell, and David Utley

Atkinson Morley's Hospital, Cope Hill, Wimbledon, London, SW20 0NE, UK


The widely accepted theories concerning the pathophysiology of infant subdural) haematoma (SDH) were formulated in the pre-computed tomographic (CT) scan era. Violent shaking is considered to be a crucial cause of SDH in non-accidentally injured infants. This theory has been re-examined in a clinical and CT scan review, and our findings have been correlated with results of recent head injury research. A retrospective review was conducted of all head injured infants (up to 18 months old) treated at Atkinson Morley's Hospital over a recent 20 year period (n = 100). Twenty-eight infants with a SDH were identified. CT scans were reviewed and each SDH greater than 0.5 cm thick was morphometrically analysed. Seventeen infants were Caucasian, 10 were non-Caucasian and one was of mixed race. A race-dependent pattern of SDH pathophysiology was noted, with non-Caucasian infants with a head injury more likely to have a SDH than Caucasian infants (67 v 21%, p < 0.01 ) All had a history of, or radiographic and clinical findings most consistent with, an impact injury, and non-Caucasian infants were significantly more likely to have a normal scalp examination despite the impact injury (p < 0.05) and to have developed the SDH after a relatively trivial fall (p < 0.01). They were also more likely to have a large ( > 0.5cm thick) SDH (p < 0.05), and to suffer post-traumatic seizures (p < 0.05). Our findings do not support shaking as the only cause of infant SDH formation and also suggest that non-accidental injury is a less common cause of SDH than it is believed to be.


Traumatic SDH commonly occurs in infants with serious head injuries. Twenty years ago, Guthkelch expanded on earlier observations and proposed that in cases of non-accidental injury (NAI), vigorous shaking of the infant was often the mechanism underlying the formation of a SDH. The "shaken baby" hypothesis of SDH formation has subsequently gained wide acceptance. In western countries, a SDH and papilloedcma in an infant without external signs of head injury are considered findings highly suggestive of NAI by violent shaking.

Significant advances in knowledge about mechanisms of head injury and in clinical management have occurred since the shaken baby hypothesis was formulated. CT brain imaging, in conjunction with data derived from primate head injury research, have spawned new theories concerning the pathophysiology of traumatic SDH. Although these investigations have largely centred on adults, the methods utilized arc applicable to head-injured infants.

In view of the clinical and medicolegal importance of an infantile SDH, a clear understanding of the pathophysiology is essential. The current study is directed at re-examining this issue in a retrospective review of all infants with a diagnosis of traumatic SDH at Atkinson Morley's Hospital (AMH) over a 20-year period.

Materials and methods

The medical records and radiographic studies of all head injured infants (up to 18 months old at time of injury) admitted to AMH between July 1970 and July 1990 were reviewed. Infant's names and ages were obtained from the hospital register of discharge diagnoses. Infants in this series with a diagnosis of traumatic SDH are the subject of the current report.

Particular attention was directed at identifying the mechanism of injury that resulted in SDH formation. The history of injury in the medical record was supplemented by telephone enquiries and written questionnaires to parents, guardians, social workers, police officers, coroners and the injured infant's current general practitioner, to add any pertinent information that had come to light after discharge from AMH.

A range of pre-injury infant characteristics were examined in order to identify factors which could influence the formation of a SDH. These characteristics included race, nationality, prematurity, presence of a bleeding diathesis, history of seizures, hydrocephalus or other neurological disorder and developmental delay.

Head CT scans were reviewed and scans with a SDH greater than 0.5 cm thick were morphometrically analysed. The presence or absence of the following specific features were recorded: extracranial injuries, scalp soft tissue swelling, skull fracture, intracerebral haematoma or contusion, and intracerebral low density changes. Subdural haematoma volume, maximum haematoma thickness, and the amount of midline shift were quantified with a Colormorph Interactive Image Analyser (Perceptive Instruments Ltd., Halstead, Essex, UK) using a previously reported computerized technique.

The chi square test with Yates' correction for continuity was used to statistically analyse associations between infant subgroups and specific clinical and radiographic findings.


Infant selection

One-hundred-and-thirteen infants were identified with a discharge diagnosis of closed head injury. Six infants older than 18 months were excluded from the study. Two infants were misdiagnosed as head injuries and five infants' medical records could not be located. Of the l00 infants included in the study, 28 had a traumatic SDH. The diagnosis of SDH was confirmed by review of the head CT scan in 23 cases. In five infants, SDH was diagnosed by angiogram (n = 1 ) or exploratory burr holes (n = 4).

Premorbid infant characteristics (Table i)

There were 18 boys and 10 girls with a mean age of 7.8 months (SE = 0.9, range 0.8-18 months). The premorbid histories were unremarkable with the exception of one premature infant, one infant with a cleft palate and one mentally retarded infant with severe gencralised epilepsy. Twenty-seven infants were developmentally normal at the time of injury.

Infant race and nationality

The race and nationality of all infants with a SDH was identified. There were 17 British Caucasian infants with a mean age of 5.9 months (SE = 1 .0, range 0.8-I5 months), and l0 non-Caucasians with a mean age of 9.4 months (SE = 1.4, range 2-18 months). The nationalities of the 10 non-Caucasian infants, all residents in the United Kingdom, were: British (n = 2), Nigerian (n = 2), Somalian (n = I ), Indian (n= 1 ), Egyptian (n = I ), Qatarian (n = 1), Sierra Leone (n = 1), and Iraqi (n = 1 ). One infant was of mixed racial origin (Caucasian father and non-Caucasian mother). In the group of head-injured infants who did not have a SDH (n = 72), 63 were Caucasians, five were non-Caucasians and two were of mixed racial origin. The race of two British infants in this category could not bc determined. The incidence of SDH in head injured Caucasians was 21%, end in head injured non-Caucasians was 67%. The increased incidence of SDH in non-Caucasian vs Caucasian infants is significant. (p < 0.01).

The possibility that this observed difference results from a biased referral pattern was examined. The 100 head injured infants included in the study were referred from 29 different hospitals and 87% of the non-Caucasian infants were referred from hospitals that had also referred Caucasian infants. In the group of head injured infants without a SDH, the severity of neurological injury assessed by the admission Glasgow coma score (GCS) was not significantly different [Caucasian mean GCS = 10.7 (SD 2.0), non-Caucasian mean GCS = 10.8 (SD 2.5)]. The time from injury to neurosurgical evaluation did not differ between the two racial groups and there is no evidence to suggest any racial difference in the criteria used by the infants' guardians to seek medical attention or in the referral practice of clinicians in the 29 referring hospitals.

Head injury history

Falls were the most frequently reported injury (n = 14), and occurred in 8 of the 10 (80%) non-Caucasian infants, and in 6 of the 17 (35"/0) Caucasian infants. All six Caucasian infants fell from a height of over 3 feet (from a high chair, window or the arms of an adult who was standing). Of the eight non Caucasian infants who fell, three were said to have lost their balance while in the sitting (n = 1 ) or standing (n = 2) position and sustained trivial head impact (multiple unbiased witnesses in two cases). Two others were said to have rolled off a low bed onto a carpeted floor and onlywhrce fell from a significant height.

Six infants presented with no history of trauma, three of these had scalp contusions noted on physical examination. One of the three infants without physical signs of an impact injury had a scalp contusion noted at autopsy. The duration of symptoms in the remaining two infants suggested that injury occurred more than a week prior to admission and there was no suspicion of NAI. It was presumed that these infants, sustained a minor accidental head impact resulting in chronic SDH formation. Four infants were unrestrained passengers in road traffic accidents and four infants were reportedly assaulted. The assaulted infants had been beaten or thrown against stationary surfaces and three of these infants had evidence of scalp contusion on physical examination. A history of shaking was obtained in three of the assaulted infants and they showed evidence of a violent impact injury. In each case the carer described shaking the baby with the intention of resuscitation after the infant had developed symptoms from the initial head impact.


Eleven infants (40%) immediately lost consciousness following their head injury. Ten additional infants were noted to have a delayed deterioration in their Ievel of consciousness ( > 1 hour from time of injury). Infants with a decreased level of consciousness were often noted to have abnormal breathing patterns. Vomiting (50%) and irritability (25%) were frequent presenting symptoms. Post-traumatic seizures (5 immediate and 11 delayed) were more common in non-Caucasian (90%) than in Caucasian (41%) infants (p < 0.05). The time interval to neurosurgical evaluation was highly variable (range 2.5 hours to 4 weeks), but 15 infants presented within 24 hours, four infants between 24 and 72 hours, and nine infants more than 72 hours after a head injury was reported or symptoms were first noted.

Physical findings

Scalp Injury. Scalp contusion or bruising was noted on admission to hospital in fourteen infants (50%). In the 14 infants who had no evidence of scalp injury the reported mechanisms of injury were a fall (n = 9), assault (n = 1), road traffic accident (n = l) and no history of trauma (n = 3). Of these 14 infants, evidence supporting an impact mechanism of injury was present in 12 cases. This evidence consisted of unbiased witnesses of the injury (n = 6), skull fracture on skull X-ray (n = 2), scalp contusion on post-mortem examination (n1), and three infants who presented with a chronic SDH. In these three cases there was no suspicion of NAI, and a minor head impact was thought to have occurred more than 1 week prior to admission. An additional infant presented with a chronic SDH 3 weeks after a reported minor fall. In this case, however, the young single mother's history was deemed suspicious and the child was taken into protective custody. A case report of the only infant with a recent history of head impact ( < 72 hours), a normal scalp examination, and no supporting evidence to verify the presenting history is presented below. The incidence of a normal scalp physical examination was 29% for Caucasians and 80% for non-Caucasians. The association of non-Caucasian race with absence of external signs of head injury is significant (p < 0.05).

Retinal haemorrhages. Fundoscopic findings were recorded in 20 cases. Retinal haemorrhages were noted in 11 infants and no abnormalities were found in nine cases. Retinal haemorrhages were noted in infants with (n = 4) and without (n = 7) external signs of an impact injury, as well as infants suspected (n = 9) and not suspected (n = 2) of having a NAI. On further investigation, no evidence to support the suspicion of NAI could be found in two of the infants with retinal haemorrhages.

Extracranial injuries. On physical examination six Caucasian infants (33%) had evidence of extracranial injuries and none of the non Caucasian infants had evidence of such injuries.

Radiological findings: fractures

Eight of the Caucasian infants (47%) had a skull fracture diagnosed on plain skull X-ray or CT and nine (53%) had no skull fracture. Only two of the non-Caucasian infants (20%) had a skull fracture and eight (80%) had no skull fracture. Seven Caucasian infants (4l%) had skeletal survey examinations and an extracranial fracture was found in one. Nine non-Caucasian infants (90%) had skeletal survey examinations and no extracranial fractures were found.

Radiological findings: CT scan (Table ii)

Head CT scans were obtained in 23 infants (82%). A SDH was located over at least one cerebral hemisphere convexity in all 23 infants. Nine infants also had an interhemispheric SDH and six infants had bilateral convexity subdural haematomas. High (n = 14), mixed (n = 7), and low (n = 2) density SDH was observed. A low density SDH was seen exclusively in infants evaluated more than 72 hours from the onset of symptoms, whereas high and mixed density SDH was observed in infants falling within a wide range of time intervals. Maximum SDH thickness varied from less than 0.5-2.7 cm, with 13 infants having a maximum SDH thickness greater than 0.5 cm (mean = l .6cm, SE = 0.2, for infants with thickness > 0.5 cm). In this latter group, SDH volumes ranged from 26 to 269 ml (mean = 122.4 ml, SE = 18.4). A highly significant race dependent pattern in SDH size was observed with all non-Caucasian infants (n = 10) having a large ( > 0.5 cm thick) SDH compared to only three (23%) Caucasian infants (p < 0.01 ).

Three of 13 Caucasian infants with CT examinations (23%) had brain injuries in the form of visible parenchymal contusions, but none of the 10 non-Caucasian infants with CT examinations had contusions.

Treatment and outcome

Sixteen infants required surgical treatment. Thrce infants underwent craniotomy, and 13 infants had their SDH evacuated by burr hole drainage. Six infants died within 1 week of admission to hospital (21% early mortality). There were three late deaths: one infant who was normal prior to injury and was severely disabled as a result of her injury and died of pulmonary complications 8 years later, a second infant was severly disabled with pre-injury generalised epilepsy and died nine years after her injury of pulmonary complications, and one infant who had a good outcome from his head injury died of leukaemia 4 years later. Outcome data were available on 18 of the ninteen survivors. The distribution of Glasgow outcome scores was: nine good recovery, seven moderately disabled and two severely disabled (50, 39, and 11%, respectively, of survivors with available follow up data). Early deaths were more common (40%) in non-Caucasians (12% in Caucasians), and a good recovery occurred in 20% of non-Caucasians and 35% of Caucasians (Table iii).

Non-accidental injuries

Child abuse was initially suspected in 14 cases (50%). Legal action was taken in 10 cases (carer criminally charged or children taken away from their parents). In two cases legal action was not taken and in two cases the legal outcome could not be determined. Of the 10 cases where legal action was taken, three presented with a history of assault and two presented without a history of trauma, but had physical signs of a recent impact injury to the head.

Carers of the remaining five infants claimed their infants had fallen, but had no unbiased witnesses to verify their history. On physical examination all five infants had retinal haemorrhages and minimal (n = 1) or no (n = 4) evidence of a scalp injury. The accepted medical opinion in these cases was that the SDH could not have been caused by the simple falls described. In all five cases children were taken from the custody of their parents by court order. Three of these infants were non-Caucasian.

Case report

A 1 year old Somalian boy was reported by his mother to have fallen from a bed onto a carpeted floor. The child did not immediately lose consciousness, but had a generalised seizure less than a minute after the haed impact. The infant was brought to a local hospital, observed to be neurologically normal and discharged home. A fundoscopic examination was not recorded. At home the infant became irritable, vomited and had another generalised seizure. At the time of neurosurgical evaluation, 48 hours after head injury, the infant was unconscious with signs of uncal herniation and retinal haemorrhages. A large SDH was diagnosed and surgically evacuated, but the infant died soon after. The prevailing medical opinion was that the child had been the victim of a non-accidental shaking injury. The mother was charged with murder and the infant's two surviving siblings were taken from their parents' custody. An exhaustive investigation of both parents and the two siblings revealed no evidence to support a history of abuse or poor care. There was insufficient evidence to proceed with criminal charges and on judicial appeal 5 months later the siblings were returned to their parents.


Head CT scan data and the results of recent primate head injury research demonstrate that in adults the formation of a SDH is highly dependent on two factors: the mechanics of the injury sustained and the cerebral anatomy of the victim. A clear relationship has been shown between mechanism of injry and the type of cerebral pathology that results. The strain rate or duration of head deceleration is a critical variable in this respect. High strain rate injuries are characterised by very abrupt deceleration. A head impacting on an immovable surface alter a fall is an example of high strain rate injury. Low strain rate injuries are those where the head decelerates over a longer period of time. Non-impact shaking of the head is a low strain rate mechanical insult. Cerebral bridging veins are particularly susceptible to disruption following high strain rate head injuries; low energy, high strain rate impact injuries can cause vascular disruption without causing significant direct damage to the brain parenchyma. In contrast, low strain rate injuries are more likely to result in direct injury to the brain and begin to cause significant vascular disruption only at high levels.

Age-dependent changes in cerebral anatomy strongly influence the pathophysiology of a SDH. The combination of cortical atrophy and degeneration of the walls of bridging veins in older adults is an ideal setting for the formation of a SDH. Thus, older adults who sustain high strain rate injuries from a simple fall can develop a large, lethal acute SDH.

Falls are the most frequently reported cause of infant and toddler closed head injury, Hendrick and his colleagues report falls as the mechanism of injury in 52.8% of head injured Canadian infants, while Ivan and his colleagues reported that 66% of childhood head injuries in Canada result from falls. Although the specific relationship between falls and a SDH was not addressed in these two large studies, they concluded that simple falls infrequently caused serious neurological injury in the Canadian populations studied.

In a smaller American study of infants with acute SDH, falling from a height was a common mechanism of injury." Another American report on infants with acute SDH elicits falls as the presenting cause in 33% of the infants, but the details of the falls (whether from a height or not) are not provided. In America it is a widely held view that simple falls from the sitting or standing position cannot cause an acute SDH in an infant. This view is consistent with our findings in Caucasian infants, where none fell from a height of less than 3 feet, and the 33% incidence of falls from a greater height in this series is identical to the incidence of falls in an earlier American infant SDH study.

In striking contrast, simple falls from the sitting or standing position are often reported as the cause of acute SDH in Japanese infants. This pattern of low energy, but high strain rate injury was also frequently found in the non-Caucasian infants in our study. The possibility that these infants did not sustain a simple fall, but were deliberately shaken is unlikely for several reasons. The circumstances of the simple nature of a number of the falls were corroborated by unbiased witnesses and the 80% incidence of falls found in our non-Caucasian infants with SDH is at variance with the 30% incidence of falls reported in children known to have developed SDH as a result of child abuse. The mechanical forces of a fall are more effective in causing a SDH than shaking, and subdural haematomas resulting from accidental simple falls and from child abuse in Japanese infants are distinctly different.

If the carers of the non-Caucasian infants in our series had caused non-accidental injuries, it is unlikely that they would, as a group, show such a difference in the injury history pattern from that reported in earlier non-accidental infant head injury series. This is particularly unlikely when simple falls are so over represented, as this is a mechanism of injury that would be viewed with considerable scepticism by western medical attendants.

Head-injured infants represent a unique population in terms of their cerebral anatomy and the types of injuries they sustain. In the current study, the clinical history, physical examination or radiographic findings in infants with SDH were most consistent with a direct impact injury to the head. Fourteen vf these infants (50%), however, showed no external signs of head injury, a finding consistent with earlier reports. Thirteen of the fourteen infants with a normal scalp examination either sustained a fall observed by multiple unbiased witnesses (n = 6), had an underlying skull fracture (n = 2) or scalp contusion on post-mortem examination (n = 1 ) or presented mth a chronic SDH which is most likely to have been caused by an impact injury sustained more than 1 week prior to evaluation (n = 4).