Journal of the New Zealand Medical Association, 10-September-2004, Vol 117 No 1201
Infantile subdural haematoma in Auckland, New Zealand: 1988–1998
Patrick Kelly, Ian Hayes
In the 40 years since Henry Kempe1 first reminded the medical community of the fact of child abuse, subdural haematoma (SDH) in young infants has become increasingly recognised as a warning sign. There is now an extensive literature on the clinical presentation (often known as ‘shaken baby syndrome’), although debate continues as to the exact mechanisms of injury required.2–6
Auckland and Starship Children’s Hospitals provide general medical and surgical services to central Auckland. They also provide neurosurgery and intensive care for children from metropolitan Auckland (1996 population: 1,081,776) and elsewhere in New Zealand (1996 population: 3,618,300). This study was undertaken to characterise the infants we were seeing with SDH, and to identify features that might be helpful in diagnosis and management.
A retrospective study of those children under 2 years of age admitted to Auckland or Starship hospitals with subdural or retinal haemorrhage, from 1 January 1988 to 31 December 1998. We did not include infants that may have presented to National Women’s Hospital—the principal tertiary neonatal facility for metropolitan Auckland during that period. Nor did we include infants who may have presented in South Auckland, but were not transferred to Starship Children’s Hospital.
Cases were identified by ICD9 codes for SDH (8523, 8522, 8007, 8002, 8012, 8017) and retinal haemorrhage (RH) (3628.1). All six coding fields were included, so codes for child abuse were identified. However, these codes were not our primary search strategy. A trial run demonstrated clearly that the broader strategy identified more infants with SDH.
Medical records were analysed by standardised coding sheet, and data was entered onto a Microsoft Excel spreadsheet. Ethnicity was defined by enquiry from the infant’s family at the time of admission. Skull fractures were defined as ‘simple’ or ‘complex’—using Hobbs’ criteria of complexity (multiple or stellate fractures, non-parietal fractures, fractures involving multiple cranial bones, maximum fracture width >3 mm, growing fractures >5 mm, depressed fractures).7
Radiological diagnoses and age of SDH were taken from the final radiologist’s report. Infants were assigned to two diagnostic groups: NAI (non-accidental injury), and accident. If the information was available, they were assigned on the basis of the diagnosis made by the admitting clinicians. Where data was coded, equivalent data sets were compared using the Chi-squared test.
Eighty-four infants were identified. Twelve cases (coded for SDH) were excluded when review of the final computed tomography (CT) scan report found that SDH had not been confirmed. One case coded for SDH on the basis of the CT report was excluded because no SDH was found at postmortem. In three cases, medical records could not be found. Two infants with RH alone were excluded (one leukaemia, one penetrating trauma to the eye). Two postoperative SDH were excluded. There were no cases of SDH from bleeding disorder, inborn errors of metabolism, or hydrocephalus.
After these exclusions, there were 64 cases of traumatic SDH over 10 years. Twenty-three were accidental and 41 were not accidental.
In 18 of the 23 accidental cases, the history was of major trauma. In five of the cases, the history was less clear—and these cases received a skeletal survey, fundoscopy, and social work assessment before the diagnosis of accidental injury was accepted. All diagnoses of accidental injury were made during admission.
We assigned 39 of the 41 infants in the NAI group on the basis of the diagnosis made during admission. In 7 of the 41 cases, there was a confession from the child’s caregivers. In 32 cases, the diagnosis was made by experienced paediatricians. In 2 further cases there was no paediatric or social work assessment, no skeletal survey, and no fundoscopy. One of these infants fell off a sofa and the other from a stationary motor vehicle onto grass. On our review, the history did not explain the injuries, and we analysed them as NAI.
Of the 39 cases who were diagnosed as NAI during the original admission, only 32 were coded as ‘child abuse’ at discharge.
There were no differences in gender (24/41 were male in the NAI group, 13/23 in the accidental injury group), region of Auckland, or source of referral.
The average age at which the injury occurred was 30.6 weeks in the non-accidental injury group, and 61.8 weeks in the accidental injury group.
There were comparatively more Maori in the NAI group (p=0.008), see Table 1.
Table 1. Ethnic breakdown of infants in the study
Mechanism (alleged) of injury
Explanations were given by caregivers in all 23 accidents and 30 NAI. In 11 NAI, there was no history of trauma. Five of the 11 cases had fractures and eight had RH (in 6 cases, bilateral). Three of these 11 cases died. Three of the 30 cases of NAI with a history also died. One caregiver blamed vigorous resuscitation following choking on a piece of bread, and the others blamed a minor fall and/or rough play.
Both groups often reported the mechanism of injury as a fall (14/23 accidents, 24/41 in NAI). However, falls less than 1 metre were far more common in NAI (p=0.0001), see Table 2.
Table 2. Height of reported fall
Of the 16 cases of NAI who fell <1 metre, 10 had RH, and 6 had at least one fracture. There were 2 accidental SDH resulting from a fall <1 metre. One was a corroborated fall in an adult’s arms, the other a fall from a wheelchair 1 month after surgical craniotomy. Neither had RH or fractures.
Of the 12 accidental falls >1 metre, 9 were >2 metres. Of the 3 falls between 1 and 2 metres, one was a fall downstairs in a walker, one a fall of 2 metres onto concrete, and one an infant thrown out of a runaway pram 1.5 metre onto rocks.
Of the remaining 9 accidental SDH, one was an infant onto whom an adult fell at a party, in an incident corroborated by multiple witnesses. Eight were either passengers in a motor vehicle accident (MVA) or struck in a car versus pedestrian accident.
Multiple explanations were given in the non-accidental injury group (usually minor trauma from every day activities)—use of a car seat (2), baby bouncer (1), minor accidental impact against walls when being carried around (4), and ‘rough handling’ during normal play (4). In 3 cases, cardiopulmonary resuscitation (CPR) of an unconscious infant was blamed for the injuries.
There was a confession in seven NAI cases. One did not describe the mechanism of injury. Four caregivers admitted to shaking, and two to hitting. The offenders were: three mothers, three fathers, and one 14-year-old babysitter. There was no difference in the rate of RH or fracture compared with the rest of the NAI group (five RH, and three fractures).
Lethargy (2/3), irritability, and poor feeding before admission (in 50% of cases) occurred in both groups. Severe impairment in state of consciousness was present in approximately 50% of patients in both groups. In accidents, deterioration was directly linked to the episode of trauma. In contrast, many NAI had a vague history of being unwell for a few days, or being found at night deeply moribund after going to bed well.
No infant in the accidental injury group presented with apnoea, whereas 9 of the 41 NAI did have apnoea (p=0.009). Conversely, those infants with accidental injuries were more likely to have seizures (15/23 compared to 16/41 NAI; p = 0.01).
There was no significant difference between the two groups in Glasgow Coma Score at presentation; need for resuscitation in hospital (18/41 vs 7/23); or neurosurgical intervention with tapping, burr holes, or craniotomy (18/41 vs 4/23). There was no difference in the presence of bruising to the head (10/41 vs14/23)—even in those with skull fractures (6/11 in NAI, 12/18 in the accidental injury group; p=0.67).
Bruising to the body was documented in 11 of the 41 infants with NAI, and in 1 of those with accidental injury.
Significantly more NAI infants were in shock (16/36 vs 4/23; p=0.04), and required ventilation (24/41 vs 7/23; p=0.02). Average time of ventilation in NAI was 1.6 days as against 0.8 days in the accidental injury group (not statistically significant).
All infants had skull X-rays. In infants with NAI, skeletal surveys were performed in 37 cases and bone scans in 16 cases. In the accidental injury group, skeletal surveys were performed in five cases and a bone scan in one case.
There were more skull fractures in the accidental injury group compared with the NAI group (18/23 vs 11/41; p=0.001). However, there was no difference in the ratio of complex fractures as defined above (10/18 in the accidental injury group, 6/11 in the NAI group; p=0.88).
In 5/6 of the complex skull fractures in NAI there was no history of trauma. In the accidental-injury group, the histories were as follows: 3 high-speed motor vehicle accident (MVA), 3 fall ≥3 m, 1 fall downstairs in a walker, 1 run over by a car, 1 propelled onto rocks from a runaway pram, and 1 fall in mother’s arms.
In NAI, there were 11 infants with rib fractures and 5 with metaphyseal fractures. In accidents, the only fracture outside the skull was a spiral fracture of one humerus in the infant whose uncle fell on him at a party in the presence of multiple witnesses.
All patients had a CT head scan. SDH was more often unilateral in the accidental injury group (p=0.04). Interhemispheric SDH did not differ between groups (3/23 accidental, 10/41 NAI). No accident had SDH of different ages. Twenty-two cases of NAI had SDH which were either non-acute or (when multiple) of greater than one age.
Table 3. Location of subdural haemorrhage
Relatively few infants had magnetic resonance imaging (MRI) (10/41 NAI and 1/23 accidental), and gradient echo sequences were performed in only 2 of these infants. We are therefore unable to comment on the presence or absence of haemosiderin deposits in the parenchyma, which may be a manifestation of previous injury.8
Cerebral oedema (8/23 versus 24/41) was more common in the NAI group (p=0.01). Massive global brain injury (for which diffuse axonal injury could be one possible explanation) was seen in 7 infants. All were in the NAI group. Five of these infants died prior to discharge, but the CT appearance could only be correlated with postmortem findings in two cases. In one there was global ischaemia, and in the other diffuse brain injury consistent with shaking.
There was no significant difference in subarachnoid (6 NAI, 2 accidental) or extradural haemorrhage (3 NAI, 4 accidental), nor in the overall frequency of all other cerebral injuries combined. However, the numbers in each subcategory of these other cerebral injuries were too small for meaningful statistical analysis. For example, evidence of focal or diffuse cerebral atrophy (suggestive of previous cerebral injury) was seen in 4 infants in the NAI group but in none of those with accidental injuries.
An ophthalmologist examined 39/41 of the infants suspected of NAI, but only 5/23 of those with accidental injury. 27 infants with NAI had RH (21 bilateral, 6 unilateral). Five of 6 infants with unilateral RH had bilateral SDH. No infant with bilateral RH had unilateral SDH.
All infants were alive on admission. There were no deaths in hospital in the accidental injury group, but there were 6 deaths in the NAI group (p=0.03)—see Table 4. Postmortem reports could be found in only 2 of the cases included in the study.
The average length of stay was 7 days in the accidental injury group, and 12.6 days in the NAI group. 24/41 infants with NAI were followed up for an average time of 3.6 months (range 1 month to 4 years), and 12/23 infants with accidental injury were followed up for an average time of 9.1 months (range: 1 month–6 years). In survivors, there was no difference in medical outcome at discharge from hospital or at last outpatient visit.
Table 4. Short-term neurological outcome
We defined impairment as follows:
For example, a child with a unilateral 3rd nerve palsy and some persistent upper motor neurone signs but essentially normal function was classified as mild; a child with a hemiplegia starting to resolve at 4 months as moderate; and a child with hemiplegia unable to walk, blindness in one eye, seizures and developmental delay as severe.
The limitations of a retrospective review did not permit a more detailed assessment of neurological outcome.
This is one of the larger published series of infantile SDH,9,10 and is subject to all the usual methodological flaws associated with retrospective studies based on clinical records. There was a marked disparity in the use of skeletal surveys, fundoscopy, and social work assessment between the two groups.
There is a further problem associated with all studies of suspected child abuse. That is, there is no ‘gold standard’ diagnostic test. Uncorroborated histories from the caregivers as to mechanism of injury cannot be accepted at face value. The criteria used to diagnose NAI must perforce be derived from the biological implausibility of trivial mechanisms causing major injuries, and from the scientific literature.
The findings of this study accord both with the literature on infantile SDH and with the known biomechanics of infantile injury:
The difference from the accidental injury group with respect to skull fracture supports the hypothesis that the mechanism of injury in abusive SDH does not necessarily require impact.12,13 That is, it is consistent with the hypothesis that many of our infants were in fact shaken. However, when skull fracture did occur, it was no more likely to be complex than in the accidental injury group. This is consistent with the obvious point that in both groups, major impact is required to produce the combination of skull fracture and SDH. It is in contrast to Hobbs’ classic paper 7, but his group of abused infants was much more severe (19 of his 29 cases died).
This is the first publication on infantile SDH in New Zealand—a comparable Australian study identified 21 cases of NAI over a 10-year period.9 Our study identified 41 cases, suggesting that the extent of our problem is certainly no less. The incidence figure for ‘shaken impact syndrome’ in infants less than 1 year of age in Wales was estimated at 21 per 100,000,11 and in Scotland: 24.6 per 100,000.19 Our study cannot provide a reliable incidence but we are addressing this with a prospective study.
Our data raise the possibility that Maori may be at greater risk of abusive SDH. Our search strategy should have avoided the risk of ethnically biased coding.20 However, the data in the clinical records was inadequate to control for the many risk factors for child abuse that may confound this result. These include the age and educational status of the mother, the number of children in the household, the presence or absence of the biological father, poverty, domestic violence, mental illness, and drug and alcohol abuse.21 Although a Japanese author suggested a predisposition to SDH after minor trauma in Japanese infants,22 and one English retrospective study hypothesised that non-Caucasian infants were more susceptible to SDH after trivial trauma,23 these articles appeared uncritically to accept uncorroborated histories.
One recent US study found no predictive effect of ethnicity.24 There is other evidence that Maori are disproportionately represented in statistics for child abuse.25 The reasons for this are debated, and could not be adequately addressed within the scope of this study.
The international literature shows that abusive infantile SDH has a worse prognosis than accidental injury.26 It is concerning, therefore, that our study suggested a relatively short duration of follow-up for abused infants. We were unable to assess the reasons for this short follow-up, although it may be connected with the same factors that contributed to the risk of abuse in the first place—complicated by the addition of multiple changes of residence and social worker for infants in alternative care. Our group would be suitable for a long-term follow-up study, which could address these questions and assess outcome in greater detail.
The trauma we have described is a major cause of mortality and morbidity. The scope of this problem cannot be determined from hospital discharge data. Our study shows clearly that such data are inaccurate. In addition, the literature suggests that the diagnosis of abusive head injury is often missed in infants who present to doctors,27 and the true incidence of infantile SDH presenting in Auckland in the years of this study may well have been higher.
It is important that health practitioners who see young children are aware of the possibility of head injury presenting with non-specific symptoms, absent or minimal external signs of injury, and a false or misleading history. Physical signs that are reliable indicators of risk for head injury in general, become unreliable when the risk is posed not only by the direct effects of any previous trauma, but by the risk that any earlier trauma will be repeated. Increasingly, the literature suggests that in infancy, a CT scan and ophthalmoscopy should be as much a part of the work-up of possible child abuse as a skeletal survey.28 Although not available for many of the infants in this study, MRI offers a modality far more sensitive in the detection of small SDH, and should be considered whenever there is a high degree of clinical suspicion.8
Finally, there is an urgent need not only to provide better diagnosis and follow-up for those who have been injured, but also to research and apply preventive strategies for those who have so far escaped injury.29
Author information: Patrick Kelly, Paediatrician, Clinical Director, Te Puaruruhau (Child Abuse Assessment Unit), Starship Children’s Hospital, Auckland; Ian Hayes, Fellow in Clinical Genetics, Royal Childrens Hospital, Melbourne, Australia
Acknowledgements: We thank Dr Teuila Percival (who first began to audit these cases in 1993); Sue Guthrie, Susan Grieve and Miriam Rea from Clinical Records (who gave invaluable assistance with coding and tracking files); and John Thompson from the Department of Paediatrics, University of Auckland (who kindly provided his statistical expertise).
This data was first presented at the Paediatric Society of New Zealand Conference in Napier, New Zealand in November 2000.
Correspondence: Patrick Kelly, Starship Children’s Hospital, Park Road, Private Bag 92024, Auckland 1. Fax: (09) 307 4930; email: firstname.lastname@example.org
issue | Search journal |
Archived issues | Classifieds
| Hotline (free ads)
Subscribe | Contribute | Advertise | Contact Us | Copyright | Other Journals