Article Text

Trends in admission and death rates due to paediatric head injury in England, 2000–2011
  1. Robin Marlow1,2,
  2. Julie Mytton3,
  3. Ian K Maconochie4,
  4. Hazel Taylor5,
  5. Mark D Lyttle1,3
  1. 1Bristol Royal Hospital for Children, University Hospitals Bristol NHS Foundation Trust, Bristol, UK
  2. 2School of Clinical Sciences, University of Bristol, UK
  3. 3Faculty of Health and Applied Sciences, University of the West of England, Bristol, UK
  4. 4Emergency Department, St Mary's Hospital, Imperial College NHS Healthcare Trust, London, UK
  5. 5Research and Innovation, University Hospitals Bristol NHS Foundation Trust, Bristol, UK
  1. Correspondence to Dr Robin Marlow, Paediatric Emergency Department, Bristol Royal Hospital for Children, Upper Mauldin Street, Bristol BS2 8BJ, UK; robin.marlow{at}


Background The number of children admitted to hospital is increasing year on year, with very short-stay admissions doubling in the last decade. Childhood head injury accounts for half a million emergency department attendances in the UK every year. The National Institute for Health and Care Excellence (NICE) has issued three iterations of evidence-based national guidance for head injury since 2003.

Objectives To assess if any changes in the rates of admission, death or causes of head injury could be temporally associated with the introduction of sequential national guidelines by longitudinal analysis of the epidemiology of paediatric head injury admissions in England from 2000 to 2011.

Methods Retrospective analysis of English Hospital Episode Statistics data of children under the age of 16 years old admitted to hospital with the discharge diagnosis of head injury.

Results The number of hospital admissions with paediatric head injury in England rose by 10% from 34 150 in 2000 to 37 430 in 2011, with the proportion admitted for less than 1 day rising from 38% to 57%. The main cause of head injury was falls (42–47%). Deaths due to head injury decreased by 52% from 76 in 2000 to 40 in 2011. Road traffic accidents were the main cause of death in the year 2000 (67%) but fell to 40% by 2011. In 2000, children who were admitted or died from head injuries were more than twice as likely to come from the most deprived homes compared with least deprived homes. By 2011, the disparity for risk of admission had narrowed, but no change was seen for risk of death.

Conclusions Temporal relationships exist between implementation of NICE head injury guidance and increased admissions, shorter hospital stay and reduced mortality. The underlying cause of this association is likely to be multifactorial.

  • Accident & Emergency
  • Epidemiology
  • Injury Prevention

Statistics from

What is already known on this topic

  • Rates of attendances to emergency departments with minor medical problems have been steadily rising.

  • Head injury is the commonest presentation of moderate and major paediatric trauma.

What this study adds

  • Between 2000 and 2011, the number of children admitted with head injuries rose significantly and mortality halved.

  • These changes can be correlated with the introduction of NICE 2007 guidance, although alternatively may represent longitudinal shift due to other factors.

  • Falls were the predominant cause of admission but road traffic accidents the main cause of death.


Over the last decade, there has been an increase in childhood hospital admission rates in England, with a pronounced rise in very short-stay (less than 1 day) admissions for medical illnesses.1 The causes are not clear: reorganisation of primary care out-of-hours provision, the introduction of emergency department (ED) quality indicators and changes in health-seeking behaviour have all been suggested.2 Head injury is the commonest presentation of major and moderate paediatric trauma.3 Although most have no long-term consequences, a proportion suffer traumatic brain injury (TBI),4 the leading potentially avoidable cause of death and chronic neuro-disability in childhood.5

Prompt identification and early intervention in serious TBI may prevent life-changing detrimental outcomes. To detect such injuries, cranial CT is the gold standard investigation. It identifies those who require neurosurgical intervention or other intensive therapy, and aids in discharge decisions when normal. With technological advancement and increased availability, rates of CT for childhood head injuries approached 50% in North America by 2003.6 However, the recognition that radiation exposure from CT in childhood may reduce IQ7 or increase the risk of later malignancy8 ,9 has driven research to identify which children are at sufficiently high risk of TBI to warrant CT.4 ,6 ,10

The National Institute for Health and Care Excellence (NICE) has produced three iterations of evidence-based guidelines for the early management of children with head injury in 2003, 2007 and 2014,11–13 replacing advice from the Royal College of Surgeons (RCS) (in 1984 and 1999).14 ,15 A crucial component in all versions has been guidance regarding imaging. The first NICE guideline was largely extrapolated from adult studies due to limited paediatric-specific evidence. Many clinicians felt this led to unacceptably high CT rates and used locally derived guidelines with higher thresholds and lower scan rates.16 ,17 The 2007 revision incorporated criteria from a paediatric UK study;4 with acceptable CT rates this became widely used, though with some centres making minor modifications.18 The most recent version (NICE 2014) introduced risk stratification, defining a cohort of children who may be actively observed rather than progressing to immediate CT scan. This practice for selected patients has been shown to be associated with approximately half the adjusted odds of performing CT and no apparent increase in adverse effects, leading to the suggestion that this approach could safely reduce CT rates.19

We aimed to determine whether mortality or admission rates from childhood head injury have changed with successive iterations of NICE guidance, and to explore the epidemiology to identify any potentially alternative influences.


Data sources

Hospital Episode Statistics (HES) is a centrally collected anonymised record of all admissions to English National Health Service hospitals. It uses nationally standardised coding systems to record diagnoses (International Classification of Diseases—ICD10), procedures and operations (Office of Population Censuses and Surveys—). We used this data set to identify all admissions during 2000–2011 of children under the age of 16 years old at the time of admission given a coded diagnosis of head injury (ICD10 codes S00–S09). We grouped the coded mechanism of injury into Road Traffic Injuries (V00–V99), Falls (W00–W19), Struck by animate object (W20–W49), Struck by inanimate object (W50–64), Assault (X85–Y09) and Unrecorded and Others (all other coded causes of injury). We identified episodes with neurosurgical interventions (OPCS4 codes A05, A40, A41, V03) or cranial imaging (OPCS4 code U051), although CTs were not routinely recorded before 2006. The length of stay was calculated as the difference between start and end dates of an admission.

To evaluate the association of socioeconomic status with the risk of avoidable injury, we identified the index of multiple deprivation (IMD) decile for the household location of each head injury admission. The IMD provides a relative measure of deprivation at small area level across England. Combining seven different dimensions of deprivation, it ranks England into 32 482 areas from least to most deprived.20

Data analysis

We used Office for National Statistics mid-year estimates of the English population as our denominator. To standardise between years, annual head injury admission, mortality and neurosurgery rates were calculated for the English population under the age of 16 years old as incidence rate ratios (IRRs) with 95% CIs, relative to the year 2000. For these outcomes, a negative binomial regression model was fitted and a test for a linear trend by year was carried out. We assessed guideline eras by assessing the last 3 years of their use to allow for implementation delays. Between these periods, admission rates, death rates and neurosurgery rates were compared by fitting a negative binomial regression model. Logistic regression models were fitted in order to test for a linear trend by year for the proportion of admissions being admitted for less than a day, 1 day or 2 days or more. To analyse the effects of socioeconomic status, we fitted negative binomial models for admission and for death rates over the 12-year period against IMD quintiles (using the least deprived as baseline) adjusting for quintile population density to calculate IRRs. Data extraction and analysis were completed using the statistical programs R21 and Stata.22


Epidemiological trends

Between 2000 and 2011, there was a statistically significant rise in admissions with head injury (table 1) (p=0.005) from 34 to 37 per 10 000 children, while the number dying due to head injury fell from 7.6 to 4.0 per million (p<0.001). There was no statistically significant trend in the neurosurgery rate (relative to the population) (p=0.220). Of those admitted, the proportion admitted for a very short duration (less than 1 day) rose 20% from 37.6% in 2000 to 57.4% in 2011 (p<0.001 for linear trend). Correspondingly, admissions for 1 day fell by 14% and for 2 days or more (≥2) dropped by 6% (both p<0.001 for a linear trend by year). Between 2006 and 2009, the proportion of admitted patients having a CT rose and plateaued thereafter.

Table 1

Description of admissions and deaths due to head injury in children aged 0–15 years, 2000–2011

Guideline eras

There was a statistically significant difference in admission rates between guideline eras (p=0.0022) (table 2). Admission rates were higher in NICE 2003 compared with RCS. Admission rates were also higher in NICE 2007 compared with either NICE 2003 or RCS. There was a statistically significant difference in death rates between guideline eras (p=0.0006), with the death rates statistically significantly lower in NICE 2007 compared with either RCS or NICE 2003. Neurosurgery rates did not differ between guideline eras (p=0.1647). For those admitted, the proportion of less than 1 day admissions rose significantly (p<0.001) between the guideline periods, and the proportion of longer admissions (1 day or 2 days or more) dropped significantly (p<0.001).

Table 2

Comparisons between guideline eras*

Aetiology of admissions and deaths

The leading cause of head injuries requiring admission was falls (42–47%) with other causes remaining constant (figure 1). The predominant cause of death (figure 2) was road traffic accidents, reducing from 67% of all head injury-related deaths in 2000 to 40% in 2011 (p=0.01). Of fatal road traffic accidents, 65–100% were pedestrians or cyclists.

Figure 1

Causes of head injury admission/year.

Figure 2

Causes of head injury death (averaged over 2-year periods to smooth random fluctuations).


Differences in admission and death rates exist between children from different socioeconomic quintiles. One per cent of records were missing socioeconomic data and were excluded from the analysis. In 2000, twice as many children from the most deprived quintiles were admitted (IRR 2.06 (95% CI 1.99 to 2.13)) and more than twice as many died (2.31 (95% CI 1.77 to 3.02)) compared with those from the least deprived quintiles. Less marked, although still significant, differences were seen comparing other quintiles (see online supplementary tables S3 and S4) Over the period studied, the disparity in rates of admissions significantly reduced (p<0.001) across all levels of socioeconomic status. By 2011, when compared with the highest quintile the IRR of admission in the lowest quintile had fallen to 1.47 (95% CI 1.43 to 1.52). With comparatively small numbers of deaths, there was no statistically significant interaction detected between year and deprivation quintile, suggesting that the relationship between death rates and deprivation quintile remained similar across the years.


Between 2000 and 2011, rates of admission for childhood head injury rose. However within this overall rise, we have demonstrated an increasing proportion of admissions lasting less than 1 day. Over the same period, there was a reduction in mortality due to head injury, while the number of children requiring neurosurgical intervention remained constant.

Previous analysis of HES data23 demonstrated that after the introduction of the NICE 2003 guideline, admission rates for adults increased but children were unaffected. We have shown that following implementation of the NICE 2007 guideline there was a statistically significant rise in the number of admissions and fall in the number of deaths in the paediatric population. The clinical decision rule on which NICE 2007 was based predicted a rise in rates of CT scanning from a baseline of 3.3% to just over 14%. It would be expected that a full implementation of NICE 2007 guidance would have resulted in increased CT rates, perhaps with an associated reduction in admissions and length of stay. CT scanning and discharge direct from ED compared with admission for observation is a cost-saving strategy.24 Using the NHS 2013 reference payment tariffs,25 an ED attendance where the patient has a CT scan and is discharged (VB03Z £242) costs one-third the combined tariffs of an ED attendance and admission for observation (£730=(VB04Z £228)+(PR07B £502)). The effects of new guidance being implemented need to be monitored; if clinicians observe children for a short period of time prior to a decision on the need for imaging (which many already were, but is now supported by NICE 2014), we may see yet further increasing admission rates and resulting costs.

The limitations of our study are mainly due to the nature of the routinely collected data set. Our choice of outcomes as ‘admission or death due to head injury’ is pragmatic; we were not able to assess how many children survived but had adverse neurological outcomes. Despite HES data being rigorously collected using a strictly defined data set, the introduction of new fields or changes to coding guidance may generate changes in data that are at risk of misinterpretation as a result of epidemiological or clinical change. This can be seen in the anomalous rise in numbers of CT scans following their introduction to the data set in 2006. HES is the gold standard NHS activity data set, extracted directly from hospitals’ reports to their commissioners to claim payment; and estimated to capture 99.8% of all admissions.26 It has been postulated that the introduction of ED quality indicators (especially the drive to spend no more than 4 h in an ED, first introduced in 2001) has caused the rising number of short-term admissions. However, admission rates for children with medical complaints were already rising2 prior to their introduction. Neurosurgery rates have not significantly changed despite increasing admissions, suggesting that the overall incidence of severe head injury in childhood has remained relatively constant. We believe that these findings may therefore represent a change in clinician behaviour, with lower severity thresholds for admission combined with shorter stays before discharge. An alternative hypothesis is that the proportion of children being admitted has remained constant, but more children are being taken to ED and so more being admitted. HES data did not include information on ED attendances until 2007 and we have therefore not been able to examine this further.

Our rates of inpatient mortality due to head injury in England combined with those for Wales over the same time period (personal communication—NHS Wales Informatics Service) closely approximate the Office for National Statistics figures for total childhood deaths due to head injury in England and Wales.27 This suggests that most children who die from head injuries do so in hospital rather than at the scene of the incident. The reduction in deaths due to road traffic accidents is likely to be due to a combination of factors. Over this period, there have been improvements in safety for car occupants (eg, the use of booster seats, airbags). However, as the greatest reduction in road traffic deaths has been in pedestrians/cyclists, it may be that the focus on hard-hitting road safety campaigns has been beneficial, or that this reflects a reduction in exposure to risk with less walking and cycling in children.

Although there has been improvement over time, the effects of deprivation are still alarming; with children dying of head injuries twice as likely to come from the most deprived neighbourhoods. The recent RCPCH report ‘Why do Children Die?’5 identified injuries and social inequality as two of the modifiable targets to improve the health of children in the UK.

While we cannot directly link implementation of national guidance with reduction in mortality and bed occupancy, we have highlighted significant change in outcomes and resource use over time. This also represents the same time period over which the underpinning evidence in several key areas has expanded. These trends should continue to be monitored and explored in greater detail to determine where the key priorities for primary and secondary prevention lie, and guide future research.


These data provide an important snapshot of NHS healthcare use in England for the commonest childhood injury. Admission rates have changed since the introduction of national guidelines with more frequent, shorter admissions for the same rate of neurosurgical events, indicating that admission rates may not provide a consistent proxy for severity of injury. They provide a useful measure against which to compare the effects of the NICE 2014 head injury guidance and benchmark any future changes of emergency care provision. They also highlight the utility of acquiring a more complete data set from ED presentations, especially those due to the most common conditions.


Thanks to Dora Wood for comments on the manuscript and Anna Morris (NHS Wales Informatics Service) for tabulating Welsh paediatric head injury mortality data. HES data were provided to the University of Bristol by the Health and Social Care Information Centre under data reuse agreement IC Ref: NIC-164132-C45WP, IG Ref: RU919. Copyright 2013, reused with the permission of The Health and Social Care Information Centre. All rights reserved. ONS: Adapted from data from the Office for National Statistics licensed under the Open Government Licence V.1.0.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • Contributors RM and MDL conceived and designed the study, and JM contributed to study design. RM obtained the data, and RM and HT undertook the analyses. RM wrote the first draft of the study, and JM, IKM, HT and MDL critically reviewed and edited the manuscript to develop the final version. MDL acts as guarantor.

  • Funding RM is funded by a University Hospitals Bristol NHS Foundation Trust Clinical PhD studentship.

  • Competing interests None declared.

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

  • Data sharing statement Summary data and analysis scripts are available from the corresponding author. Source HES data are used under agreement between University of Bristol and Health and Social Care Information Centre and cannot be reused.

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.