Article Text

Download PDFPDF

Appearance and location of retinal haemorrhages in critically ill children
  1. G G W Adams1,2,
  2. Shruti Agrawal3,
  3. Rajnish Sekhri4,
  4. Mark J Peters5,6,
  5. Christine M Pierce5
  1. 1Department of Strabismus and Paediatrics, Moorfields Eye Hospital, London, UK
  2. 2Department of Paediatric Ophthalmology, Great Ormond Street Hospital for Children, London, UK
  3. 3Paediatric Intensive Care Unit, Addenbrooke's Hospital, Cambridge, UK
  4. 4Department of Paediatric Ophthalmology, Hull and East Yorkshire Eye Hospital, Hull, UK
  5. 5Paediatric and Neonatal Intensive Care Units, Great Ormond Street Hospital for Children, London, UK
  6. 6Critical Care Group—Portex Unit, Institute of Child Health, University College London, London, UK
  1. Correspondence to GGW Adams, Department of Strabismus and Paediatrics, Moorfields Eye Hospital NHS Trust, 162 City Road, London EC1V 2PD, UK; gill.adams{at}


Background There are few high-quality studies describing the appearance and location of retinal haemorrhages in critically ill children not due to birth or abusive head trauma.

Methods Prospective study from February 2008 to December 2009 of emergency admissions to a paediatric intensive care unit aged over 6 weeks. Children with a penetrating eye injury or suspected or proven abusive head injury were excluded. The children underwent either dilated funduscopy performed by a paediatric ophthalmologist or RetCam imaging.

Results Retinal haemorrhages were identified in 24/159 (15%) patients. 50% of the haemorrhages were bilateral. The severity was mild (<5 retinal haemorrhages) or moderate (5–20 retinal haemorrhages) in 75%. The location was in zone 1 in 45.8%, zones 1 and 2 in 33.3%, zone 2 alone in 8.3% and not described in 8.3%. Schisis cavities and perimacular folds were identified in two patients with one having a pseudohypopyon appearance; a further one patient had bilateral haemorrhagic retinal detachments. Three patients had exudates or scarring consistent with cytomegalovirus infection.

Conclusions Retinal haemorrhages are seen in a proportion of critically ill children, however most retinal bleeding is not extensive as indicated by location within the retina or layer of bleeding. Higher numbers and extent of retinal haemorrhages were only observed in the presence of severe coagulopathy, leukaemia, one victim of a road traffic accident, and one child who sustained a fatal witnessed fall down the stairs; all circumstances that would be readily distinguished by history and laboratory testing from abusive head injury.

  • Child health (paediatrics)
  • Retina
View Full Text

Statistics from


Retinal haemorrhages in a young child or infant in the absence of a history of trauma, raises the possibility of non-accidental injury. There are well-recognised non-traumatic causes of retinal bleeding in childhood, in particular haematological disorders and retinopathy of prematurity (ROP).1–7 In very young children, retinal haemorrhages may reflect a birth injury. Most birth-related retinal haemorrhages resolve by 4 weeks, although a single subretinal haemorrhage has been reported as persisting until 6 weeks after birth, and in another series one discrete haemorrhage was present at 58 days after delivery.8 ,9 There is a lack of high-quality ophthalmological studies on this topic in childhood. The majority of studies do not describe the examination technique, such as whether the pupils were dilated prior to funduscopy or whether an ophthalmologist using an indirect ophthalmoscope undertook the examination. Often it is not stated if child abuse was explicitly excluded as a cause of the retinal bleeding. The retinal findings may not be described in full detail. Currently there is no agreed standard to describe retinal bleeding but it has been proposed that it should include location, layer and severity of the haemorrhages.10

We have previously reported the medical details of a prospective study identifying the prevalence of retinal haemorrhages in critically ill children in whom abuse and birth-related retinal bleeding had been excluded, and we now describe the retinal findings in detail in relation to the laterality of bleeding, location in the retina, the layer within the retina, its severity and associated findings.11 To our knowledge, this is the first prospective study which has estimated the prevalence of retinal haemorrhages in a population of critically ill children and provided a detailed description of the retinal findings.


We undertook a prospective study of children over the age of 6 weeks who were admitted as an emergency to the 14-bedded tertiary centre paediatric intensive care unit at Great Ormond Street Hospital, London between February 2008 and December 2009. The study had local ethics committee approval and written, informed parental consent was obtained before recruitment. We excluded children under the age of 6 weeks because we wished to avoid including birth-related retinal haemorrhages in this study. Children with penetrating eye trauma where it would not be possible to perform RetCam imaging, and any child in whom non-accidental injury was a suspected or proven aetiology were excluded. Any child under 2 years of age with head injury had abusive head trauma actively excluded according to the hospital protocol in which a multidisciplinary team examined all the evidence before inclusion in the study. In our institution child abuse assessments are mandated in all children under 2 years of age with traumatic brain injury and in older children with any features that raise a concern of abuse.

Each child underwent dilated funduscopic examination, either photographic using a RetCam II (Clarity Medical Systems USA) with a 130° field lens or by a paediatric ophthalmologist (consultant or fellow) using a binocular indirect ophthalmoscope. The RetCam was used either by intensive care staff (SA, CMP or MJP) who have been trained to use the instrument or an ophthalmologist. The patients were recruited within the first 48 h of admission unless the intensive care staff considered that pupil dilatation was contraindicated because of neurological observations; in these patients the eye examination was done as soon as possible, usually within 72 h of admission. In children who were not fully sedated and paralysed the eyes were anaesthetised using proxymethocaine 0.5% drops prior to photography. Viscotears gel was used as an interface between the camera and the cornea. Indirect ophthalmoscopy by a paediatric ophthalmologist (consultant or fellow) was undertaken if RetCam imaging was not possible, for example in a very awake patient or those with lid oedema.

The RetCam images were initially reviewed by GGWA then the presence of retinal haemorrhages on RetCam imaging was agreed by GGWA and SA after reviewing the images together; neither was masked to the clinical scenarios. The findings following binocular indirect funduscopy were obtained from the hospital records; in two patients the findings were notified directly to one of the intensivists and in those two cases it was not possible to identify the zone in which the bleeding had been identified.

We noted whether the retinal bleeding occurred unilaterally or bilaterally. We described the location of retinal bleeding using the well-established ROP zone system; we could not use the more recently published grading systems which use photographic images to grade as not all patients in this study had RetCam imaging undertaken.12 ,13 RetCam imaging does not go out to the ora serrata, and only images zones 1 and 2 of the retina. Bleeding out to the ora can only be identified by binocular indirect ophthalmoscopy, probably with indentation, which was not used in this study. It is possible that some of the bleeding which was seen at the edge of the RetCam imaging extended into zone 3 but as we were not able to confirm this, it has been recorded as the maximum zone identified which would be zone 2 for RetCam imaging and indirect funduscopy without indentation.

Assessing the layer of retinal haemorrhage is recognised to be challenging, and achieving high agreement levels between observers has been reported to be difficult.14 In this study the retinal bleeding was reported as that identified by the examining paediatric ophthalmologist when photographs were not taken or assessed from the RetCam photographs. When assessing the RetCam photographs we described the bleeding in accordance with the classification described by Mulvihill et al.14 We described the bleeding as nerve fibre layer (superficial) where it was splinter or flame shaped, intraretinal where it was dot or blot shaped, and preretinal. In addition subhyaloid and subretinal were terms used clinically to describe the identified bleeding. We also noted whether schisis cavities (where there appeared to be splitting of the retina with elevation) or perimacular folds were present.

As there is no universally accepted system of extent of retinal bleeding we classified the retinal haemorrhage severity as mild when there were less than 5 retinal haemorrhages identified, moderate (5–20 haemorrhages) or severe (more than 20).

We noted where disc swelling was identified. If it was not reported upon directly and in the hospital records the discs were drawn with clear outlines we have assumed that the discs were not swollen. On the RetCam photographs the disc was deemed to be elevated or swollen if the rest of the retina was in focus and the disc margin or disc was blurred or elevated.

Coagulopathy was defined as prothrombin time more than 3 s above normal and/or twice the normal partial thromboplastin time and/or platelet count less than 50 000/mm3.


The ophthalmological findings, age and diagnosis are set out in table 1.

Table 1

Summary of patient details and ophthalmological findings

During the study period there were 1260 PICU admissions, of whom 650 were less than 6 weeks of age. Of the 610 patients aged 6 weeks or more, 254 were excluded as elective admissions (245) or ocular trauma (9). Three hundred and fifty-six patients met the inclusion criteria and a total of 161 consented to inclusion in the study. Final analyses were performed on 159 patients; two patients were excluded—one due to poor quality RetCam images which could not be graded and the other had recently undergone surgery for ROP. We were unable to recruit consecutive patients fulfilling the inclusion criteria due to consent refusal. The age, gender and admission diagnosis did not differ significantly between those who were recruited and those not recruited into the study.

Retinal haemorrhages were found in 24/159 patients (15.1%, 95% CI 10% to 22%). In 12 cases (50%) the bleeding was unilateral, with 7/12 patients having right-sided bleeding. In 11 patients (45.8%), the bleeding was restricted to zone 1 with one patient noted as having peripapillary bleeding only. The location of bleeding was not described in two patients (8.3%) as the results of the ophthalmological examination, which had been undertaken by a paediatric ophthalmologist and not using RetCam imaging, were reported directly to one of the study intensivists and were not identifiable from the clinical records. In eight patients (33.3%) the bleeding was in zones 1 and 2 of the retina, and in zone 2 only in two patients (8.3%).

In patients with mild (<5) retinal haemorrhages, the layer of bleeding was superficial in one patient, intraretinal in nine patients and superficial and intraretinal in six patients. In the two patients with moderate (5–20) retinal haemorrhages, the layer of bleeding was superficial in one and superficial and intraretinal in the other patient. In patients with severe (>20) retinal haemorrhages, the bleeding was superficial and intraretinal in four patients, intraretinal and preretinal in one patient, and superficial, intraretinal and subretinal in one patient. One patient had bilateral haemorrhagic retinal detachments and subretinal bleeding.

In patients in whom the location was identified, superficial retinal haemorrhages only occurred in zone 1 of the retina. Intraretinal haemorrhages occurred in zones 1 and 2.

Severity of bleeding was mild (<5 retinal haemorrhages) or moderate (5–20 retinal haemorrhages) in 75% of the patients, and severe (>20 haemorrhages) in the other 25%. Most, but not all, of the mild retinal bleeding was restricted to zone 1 of the retina only. Three patients with severe retinal bleeding died (patients 21, 22 and 24). Two patients had subhyaloid haemorrhages with schisis cavities and perimacular folds. One of these patients had acute myeloid leukaemia and on one side there was a preretinal pseudohypopyon appearance to the schisis cavity with a white layer on the top of the blood meniscus (figure 1). The other patient with bilateral schisis and multilayered retinal bleeding suffered fatal traumatic brain injury from a witnessed fall while being carried which involved falling down 10 steps with multiple impacts (figure 2). The child was being carried by the mother when she slipped and the child fell out of her arms and rolled down two flights of stairs then hit a wall; this was witnessed by two unrelated people. This child had subarachanoid haemorrhage, a small left subdural haemorrhage, left uncal herniation and left posterior cerebral artery and right middle cerebral artery infarcts. The opening intracranial pressure was 25 mm Hg. The child did not have any other external injuries and developed bowel perforation related to necrotising enterocolitis and treatment was withdrawn. One patient presenting acutely unwell with acute myeloid leukaemia was noted to have bilateral haemorrhagic retinal detachments and subretinal bleeding.

Figure 1

Preretinal pseudohypopyon appearance of schisis cavity in patient with acute leukaemia. Access the article online to view this figure in colour.

Figure 2

Schisis and multilayered retinal bleeding following witnessed fall. Access the article online to view this figure in colour.

Only one patient had severe (>20 haemorrhages) unilateral retinal bleeding. This child suffered a low-level witnessed fall from a hammock and had a unilateral intracranial haemorrhage with ipsilateral retinal bleeding (figure 3). The child was found to have a severe coagulopathy due to late onset haemorrhagic disease of the newborn.

Figure 3

Unilateral retinal bleeding associated with severe coagulopathy and witnessed fall. Access the article online to view this figure in colour.

One child had a unilateral swollen disc in association with active cytomegalovirus (CMV) infection. One patient with a non-traumatic encephalopathy from sinus thrombosis with bilateral haemorrhagic temporoparietal infarcts and midline shift, had unilateral blurring of the disc outline although the disc was not elevated, suggestive of mild disc swelling; the other disc was normal.

Three patients had exudates or scarring consistent with active or past CMV infection. One patient (study number 17) had uniocular lid bruising suffered after a television set fell on the child. Two children had orbital fractures.


Because of the well-recognised association of retinal haemorrhages with abusive head trauma, an accurate description of retinal bleeding is extremely important. It has been reported that non-ophthalmologists may underdiagnose the presence of retinal haemorrhages.15 By having trained intensivists image the retina using the RetCam with the photographs being reviewed by a paediatric ophthalmologist, we hoped to reduce the risk of underdiagnosis of retinal bleeding, however it is also possible that photographic imaging may have allowed the identification of small discreet superficial haemorrhages which might have been missed on examination of an awake patient.

In this study we described the location of the retinal bleeding by zone as described in the international classification of ROP. As not all patients in this study had RetCam imaging undertaken but also clinical examination we had to use a well-established description for identifying the location of the bleeding within the retina which would encompass those who had undergone a clinical examination and those who had photo documentation. The ROP zone system is well established and understood by paediatric ophthalmologists. It has been used by others when developing a standardised tool for describing retinal findings in abusive head trauma.10

Our study would support the suggestion that most retinal bleeding in seriously ill children admitted to paediatric intensive care is not severe, multilayered or extensive. Severe extensive retinal bleeding is uncommon but can be seen in children with severe coagulopathy, and severe accidental traumatic head injury. These are also the most severely ill children and if retinal haemorrhages are not seen in this population then they are highly unlikely to be seen in children presenting with milder severity of the same illnesses.

A recent systematic review has concluded that retinal haemorrhages are more common after abusive head trauma than after non-abusive head trauma and those seen in accidental head trauma are more frequently unilateral, sparse and restricted to the posterior pole.16

Although no specific particular retinal finding is pathognomonic for abusive head injury, studies consistently show an extremely high association of severe haemorrhagic retinopathy with abuse.17–20 In keeping with the medical literature to date, our study shows that extensive haemorrhagic retinopathy (ie, ‘too numerous to count’, confluent haemorrhages extending to the ora, with or without macular retinoschisis and perimacular folds) is not seen after accidental head injury except for fatal head crush, 11 m fall onto concrete and fatal motor-vehicle accidents, all situations that would be readily discernible based on the history. We add an additional case of a witnessed severe fatal fall to these reports.

Our study reports that schisis and perimacular folds can occur in accidental head trauma in children. It also highlights the necessity of investigating any child found to have retinal haemorrhages for an underlying medical condition, in particular a coagulopathy. In many conditions examination by an ophthalmologist would identify other findings, which would suggest an underlying problem other than possible abusive head trauma for example, scarring or exudates consistent with the diagnosis of CMV retinitis or a retinal pseudohypopyon appearance suggestive of leukaemia.

The strengths of this study are that we excluded children under 6 weeks of age to ensure that birth-related retinal haemorrhages were not present. Only two children were between 6 weeks and 7 weeks of age with all the other children being older; therefore birth-related retinal bleeding would not be a confounding factor. We excluded children with suspected or proven abusive head trauma, by active investigation and with the involvement of a multidisciplinary team. We used RetCam imaging after pupil dilatation or dilated binocular indirect funduscopic examination by paediatric ophthalmologists.

The limitations of the study are that we could not recruit all eligible patients being admitted to the unit. In any study where informed consent has to be requested from parents of children who have been admitted to intensive care, it is understandable that many parents do not wish any, even minor, additional investigations done on their child. It is possible that peripheral retinal haemorrhages could have been missed by RetCam imaging but equally some retinal haemorrhages identified on retinal photography may not have been identified by indirect ophthalmoscopy undertaken by the paediatric ophthalmologist. It has been reported that RetCam imaging may enhance accurate documentation, however classification of retinal haemorrhages using the RetCam has been reported to be difficult by some but not all authors.13 ,14 ,16 ,21


In this prospective study we found a prevalence of 15.1% of retinal haemorrhages in critically ill children admitted to a paediatric intensive care unit in whom birth-related retinal haemorrhages and physical abuse had been excluded. In the majority of children these retinal haemorrhages were not extensive in terms of area and layers of retina involved or severity of bleeding. Higher numbers and extent of retinal haemorrhage were only observed in the presence of severe coagulopathy, leukaemia, one victim of a road traffic accident, and one child who sustained a significant fatal witnessed fall down the stairs; all circumstances that would be readily distinguished by history and laboratory testing from abusive head injury. Dilated retinal examination can be undertaken either by an ophthalmologist using indirect ophthalmoscopy or trained non-ophthalmological personnel can undertake wide-angle retinal photography and the images reviewed by a paediatric ophthalmologist. Our study emphasises the importance of a detailed paediatric ophthalmological assessment of children with retinal bleeding.


View Abstract


  • Contributors GGWA, RS, SA, MJP and CMP contributed to conception and design of the project, and its conduct. All were involved in manuscript preparation and review. GGWA is the guarantor.

  • Funding There was no funding for this study. This work was undertaken at Great Ormond Street Hospital/UCL Institute of Child Health, which does receive a proportion of funding from the Department of Health's NIHR Biomedical Research Centres funding scheme.

  • Competing interests GGWA has provided expert witness and court reports in cases of possible abusive head trauma in relation to the eye findings in these cases, has been paid a fee for these reports, and has been instructed in civil and criminal cases. MJP has received fees from prosecution and defence teams for providing expert opinion to Family and Criminal courts in cases of suspected non-accidental head injury. SA,  CMP and RS have no financial disclosures to report.

  • Ethics approval National Research Ethics Committee UK, ref 07/H0713/65.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.