rss
Arch Dis Child 96:927-931 doi:10.1136/adc.2009.177071
  • Original articles

Why children die: avoidable factors associated with child deaths

  1. D Kelly1
  1. 1Birmingham Children's Hospital, Birmingham, UK
  2. 2Newcastle University, Royal Victoria Infirmary, Newcastle upon Tyne, UK
  3. 3University of Oxford, Oxford, UK
  1. Correspondence to Dr G A Pearson, Paediatric Intensive Care, Birmingham Children's Hospital, Steelhouse Ln, Birmingham B4 6NH, UK; gale.pearson{at}bch.nhs.uk
  • Accepted 8 February 2010
  • Published Online First 8 June 2010

Abstract

Aim To describe the avoidable factors associated with child deaths identified by a confidential enquiry.

Method In the Centre for Maternal and Child Enquiries confidential enquiry, a sample (13%) of cases was subjected to case note review by multidisciplinary panels attempting to identify avoidable factors associated with the deaths. Cases were selected blindly but in equal numbers from predetermined age bands and participating regions. The anonymised records were reviewed in regions remote to where the child lived and died. Panel composition, conduct and reporting were standardised.

Results 119 of 126 cases reviewed by enquiry panels had sufficient information to determine avoidable factors. These cases were comparable with the whole dataset in terms of sex and causes of death. 31 (26%) of 119 had avoidable factors that were predominantly related to individuals or agencies with a direct responsibility to the child. 51 (43%) of 119 were defined as potentially avoidable. In all, 130 factors were considered in relation to these 82 cases, and 64% of the factors were healthcare related. Avoidable factors were more likely where life-limiting illness was not present. Recurring avoidable factors included failure to recognise serious illness at the point of presentation and death occurring in children who had been lost to follow-up.

Conclusion Child Death Overview Panels now have the responsibility to review child deaths using similar methods but relying upon data forms rather than the case record. Analysis of contributory factors on a national scale has the potential to improve understanding of why children die and indicate strategies to reduce child mortality.

More than 3000 children between the ages of 1 month and 18 years die each year in the UK. Child mortalities have declined in the developed world for the last century, but the rate of decline in Britain is slowing despite the fact that rates here are still worse than in other developed countries such as Australia and much of mainland Europe.1 Clearly, further reductions in child mortality could still be achieved.

Until now, child mortality has fallen largely because of significant improvements in housing, education, nutrition and access to healthcare. Immunisation and antibiotics have also led to the prevention and treatment of infectious disease from which childhood deaths were common. To determine the best next steps, potentially avoidable healthcare, social and environmental factors need to be identified to inform and direct government policy.

Local and regional child death enquiries have already been useful in highlighting issues in relation to child health2 3 and national confidential enquiries into maternity4 and perinatal health,5 and outcomes in paediatric surgery6 7 have identified avoidable factors leading to death in their study populations. The recent Centre for Maternal and Child Enquiries (CEMACH) confidential enquiry into child death8 looked at all child deaths in three regions of England, Wales and Northern Ireland. The dataset and processes involved in this confidential enquiry have helped inform the functions of Child Death Overview Panels convened by the Local Safeguarding Children Boards (LSCBs).

In this report, we present the examples of good practice highlighted by the panels in the enquiry. We also present all of the avoidable factors and consider the relevance of healthcare in relation to them.

Method

Local networks were established in South West, West Midlands and North East of England, in Wales and in Northern Ireland. Regional CEMACH managers established clinical contacts in local hospitals and other local contacts including coroners, police (including youth offending teams and prisons), other emergency services, health and safety executives, social workers and general practitioners. Local CEMACH staff also searched through local newspapers and other media sources in attempts to ensure comprehensive case ascertainment. Data were collected on all deaths of children (aged 28 days to 17 years 364 days) occurring in 2006. For each death, an initial notification dataset was used to assist completion of a comprehensive core dataset9 by the authorities/health and social care professionals involved in the case. The dataset itself was an adaptation of that used in the Arizona Child Fatality Review.10 A sample of cases evenly distributed across age groups and regions was blindly selected at the notification stage. Their anonymised records were then subjected to scrutiny by multidisciplinary panels convened by the regional CEMACH manager and chaired by the regional clinical lead (a volunteer who also shared responsibility for local liaison between CEMACH and LSCBs). The core panel members were an acute care (hospital-based) paediatrician, a community paediatrician (if the acute care paediatrician did not cover community), a pathologist (with paediatric expertise), a general practitioner, a nursing representative and two non-medical representatives. Specific expertise from medical subspecialties was recruited where necessary. Each panel reviewed the records of cases from regions remote to them. The conduct and reporting pattern of panels were standardised by the use of a reporting pro forma completed by consensus during the meeting. The principal author of the report observed as many panels as possible to ensure comparability and adherence to the agreed methods.

Child deaths were anticipated to be multifactorial complex events. The concept of avoidability/preventability was therefore deconstructed so that each factor identified by the panel could be independently assessed using the structure outlined in box 1.

Box 1

Grading avoidability

  1. Avoidable

    1. Where there were identifiable failures in the child's direct care by any agency, including parents, with direct responsibility for the child

    2. Where there were latent, organisational or other indirect failure(s) within one or more agency, including parents, with direct or indirect responsibility for the child

    3. Where there was a failure of design, dilapidation of barriers or inadequate maintenance by agencies with responsibility for public safety (eg, rail maintenance leading to Hatfield rail disaster)

  2. Potentially avoidable

    1. At a higher level than the agencies with direct or indirect responsibility for the child (eg, political violence, war, terrorism, crime and if the child is the victim of homicide)

    2. Where no agency, including parents, was involved directly or indirectly with the child

    3. Where intrinsic factors (eg, an acquired disease with a known high mortality such as meningococcaemia) were the principal factors leading to the death

    4. Where there were potentially modifiable factors extrinsic to the child

    5. Where the causal pathway leading to the death could reasonably be traced back to antepartum or intrapartum obstetric events

  3. Unavoidable

    1. Death caused by unmodifiable factors extrinsic to the child (eg, lightning strike, earthquake)

    2. Death due to undiagnosed, asymptomatic conditions presenting with a lethal event (eg, hypertrophic obstructive cardiomyopathy)

    3. Planned palliation for unpreventable, incurable disease or anomaly (eg, Leigh disease)

Results

Nine hundred and fifty-seven deaths from the calendar year 2006 were analysed in the enquiry. These are broken down by diagnosis in table 1. The data revealed higher suicide rates among children than previously reported. It also revealed regional differences among death rates in the 15- to 17-year group, ethnic differences (higher rates among Pakistani and black African children) and death rates increasing with deprivation. More detailed consideration of individual cases occurred during the panel enquiries that were conducted on a 13% sample of the deaths (n=126), which is 4% of all deaths in this age group in England, Wales and Northern Ireland. The case mix reviewed by panels matched that of the whole dataset in sex and diagnosis. This is shown in table 1, which compares the proportions of cases from different diagnostic groups in the whole dataset and the subset of cases that went to panel. In each case, the 95% confidence interval (CI) for the difference in these proportions includes zero. This fact about the CI makes an assertion that the groups are composed differently in this respect statistically untenable. An exception occurs in the case of deaths owing to sudden unexplained death in infancy where the panels, by chance, reviewed a slightly smaller proportion of cases.

Table 1

Diagnostic distribution of panel cases compared with all deaths

However, panels disagreed with the medical certificate of the cause of death in one third of those that they reviewed. They felt that only 119 cases had sufficient information to reliably determine avoidable factors, and they demonstrated a clear reticence to describe non-natural death as unavoidable. This is shown in table 2 where the cause of death is listed in column 1. The cases are then distributed and counted according to the grade of their most avoidable factor.

Table 2

Number of cases with avoidable factors by cause of death

Good practice and factors relating to the death are characterised in the online appendix table.

Appendix table

Aspects of good practice and avoidable factors

Eighty-three examples of good practice were found in 74 (62%) of panel cases, 20% of these examples related to care/practices after the death.

Thirty-one (26%) of 119 cases had avoidable factors, and 51 (43%) of 119 were defined as potentially avoidable. In all, 130 factors were considered in relation to these 82 cases, and 64% of the factors were healthcare related. Avoidable factors were more likely to be identified where life-limiting illness was not present. Terminal and bereavement care were often highlighted as areas of good practice, but across the whole cohort of deaths, this sort of care was infrequently delivered at home or in a hospice. Many cases were criticised for the lack of end-of-life planning even when the terminal phase of care was recognised and the decision to allow a natural death (do not resuscitate) had been discussed.

The most significant recurrent avoidable factor between cases was a failure to recognise severe illness in children. This most often occurred at the point of first contact between the sick (and often febrile) child and the healthcare services. In some instances, there was a failure to understand the importance of the history, in others a failure to examine the patient or interpret physical signs correctly. There were also failures in anticipating or recognising complications of illness and failures in clinical supervision. In some cases, the impact was immediate, in others there resulted a critical delay in referral or treatment.

The largest cluster of cases with avoidable factors occurred in hospitals. In all cases (deaths in hospital or community), avoidable factors were concentrated where the children were previously relatively well (without pre-existent disability or life-limiting illness). There were disturbing cases where children had not been offered care or where death was due to the complications of a condition for which the children had been lost to follow-up. The cases also included deaths related to inadequate supervision by adults as well as assault by them. The lessons learnt from recurrent avoidable factors are listed in box 2.

Box 2

Outcomes of the confidential enquiry

What was learnt

  • National confidential enquiries in children are feasible. Common themes in child deaths include failure to recognise serious illness and fatal events occurring in the context of missed appointments and poor coordination of care.

  • Most deaths had some form of life-limiting illness, but most occur in hospital, few at home and 2.6% in hospices.

  • The contributory factors to deaths were complex, and death certificates were often inadequate or wrong.

What was recommended

  • All staff dealing with children should have key skills and competencies

  • Established national guidelines such as the National Service Framework and those produced by the National Institute of Clinical Excellence should be followed.

  • Parents should be encouraged to seek review if a child's situation deteriorates after medical advice has been given

  • There should be further use and evaluation of early warning scores in hospital and community-based initiatives.

  • Efforts should be made to improve the detection of children with mental health problems.

  • There should be proactive follow-up of children who do not attend appointments.

  • The recognition of life-limiting illness should prompt planning of terminal care.

  • There should be further epidemiological review of child deaths recognising that neonatal deaths require a different dataset to other child deaths.

Discussion

The sample of cases subjected to panel review by CEMACH was purposive but not in a way that was likely to introduce selection bias. We have demonstrated that the sample was similar in composition to the dataset from which it was derived and assert that it was representative of it. CEMACH panels were not unduly pejorative in their approach. As can be seen in the online appendix table, many avoidable factors that would be significant in a root cause analysis were ascribed “potentially avoidable” status. This may represent recognition both of the benefit of hindsight in making the judgement and/or the potential significance of a chance combination or juxtaposition of multiple factors. Furthermore, the proportion of cases found to contain avoidable factors in the CEMACH study concurs with rates found in other studies and is far lower than those found in the early stages of other enquiries.

The high suicide rates in the enquiry are not the result of case clustering but of the direct question used in the data collection tool. Death registration data (Office for National Statistics) include euphemistic terms and narrative verdicts that may disguise that the death was a suicide. These data like the other epidemiological data in the study as a whole deserve to be tracked over time to detect and confirm relevant trends.

Like the panels in the enquiry, Child Death Overview Panels have to report on the factors that contributed to the death and to classify the cause of death according to a system that we attempt to validate in the accompanying paper (“How Children Die”). The recent statistical release from the Department for Children, Schools and Families11 estimates that LSCBs have reviewed 41% of child deaths in 2008 to 2009. LSCBs have described only 5% of deaths as “preventable,” whereas CEMACH found 26% to have avoidable factors. There is wide regional variation in the Department for Children, Schools and Families data where the highest proportion of preventable cases comes from the South West region. In this region, many members of the Child Death Overview Panel also participated in the CEMACH study. Early evidence may therefore suggest that LSCBs are cautious in labelling a death “preventable” even when multiple avoidable factors are identified. Like the CEMACH panels' difficulty in labelling non-natural deaths as unavoidable, they may be more likely to identify non-natural deaths as “preventable.” In the CEMACH study, greater apparent clarity was achieved when panel cases were presented as though defined by their most avoidable factor. It is highly unlikely that the population of children dying in England in 2008 is remarkably different to that dying in England, Wales and Northern Ireland in 2006. However, there are several reasons why Child Death Overview Panels might be describing a lower proportion of deaths as “preventable” than the proportion identified as having avoidable factors in the CEMACH study.

First, it is unnecessary and wrong to assume that correction of the most avoidable factor in a child's death would, on its own, have necessarily prevented the death. Although within the CEMACH system, the likelihood of such increases with avoidable factors of higher grades. Hence, one can describe a death as being associated with a highly avoidable factor but then not classify the death as being “preventable” without necessarily demonstrating conflicting logic. Second, CEMACH panels considered anonymised records and were convened in regions remote to where the care was delivered. Neither condition applies to Child Death Overview Panels. Hence, the proximity of the Child Death Overview Panel to the patient and carers may affect their behaviour, and the lack of anonymity may raise concerns over the potential response of the family (or the law) to the panel's determination. Third, Child Death Overview Panels are reviewing large numbers of cases from reporting forms without access to the case notes.12 It is not known whether a comparable depth of consideration is being applied in comparison with the CEMACH study panels.

The nature of the “good practice” identified by both CEMACH and Child Death Overview Panels will inevitably be selective because it is in the nature of a death enquiry that interventions successful in preventing death will not come to the panel's attention. Hence, it is not surprising that the numerically dominant examples of good practice relate to documentation, counselling and bereavement. Nevertheless, learning points can be found in deaths at all levels of avoidability, and this coincides with the observation within the CEMACH report that avoidable factors were not restricted to deaths that were “unexpected” using the definitions in “Working Together” (HM Gov).

We found that avoidable factors (and therefore preventability) could only really be determined by multidisciplinary panel case note review. Other authors have used different methods—for example, by assigning “avoidability” to different causes of death as described by their International Statistical Classification of Diseases and Related Health Problems, 10th Revision codes and even distinguishing those “amenable to medical care” and those “mainly responsive to health policy” from the same list.13 However, this approach would not work with our data. We experienced difficulty in some cases defining the most appropriate cause of death and attempting to group deaths by organ system–specific categories such as International Statistical Classification of Diseases and Related Health Problems 10th Revision codes left a large miscellaneous group that confounds further analysis and even forms the dominant cause of death in some age groups (table 3). Although other systems of classifying deaths have been reported, they have not been applied to the entire age range covered in this study.14 This issue is explored further in the accompanying article.

Table 3

Cause of death by age and ANZPIC15 group

The CEMACH study demonstrated that using local networks to identify cases can be a very successful method for capturing child deaths and that lessons can be learnt from nationally aggregated data. However, small localities on their own do not have sufficient numbers of deaths to reflect national trends of avoidable factors. Although there is clearly the potential for LSCBs to learn lessons from individual or groups of cases, it is nevertheless the analysis of nationally aggregated data that shows the greatest promise of contributing to improvements in child health.

Conclusion

In this paper, we have presented avoidable factors and good practice in greater detail than was used in the CEMACH report. We consider that the proportion of cases defined in this way is justified. Confidential enquiries on a national scale provide epidemiological and qualitative evidence that will improve understanding of why children die and hence indicate strategies that are likely to be successful in reducing child mortality.

Acknowledgments

In addition to the authors, the steering group for this project included Richard Congdon (Chief Exec CEMACH) and Michael Weindling (Chair of CEMACH). The working group consisted of Angela Bell*, Sue Dunstall, Peter Fleming, James Fraser, Barbara Fulton, Anthony Harnden, Deidre Kelly, Julie Maddocks*, Richard Mayon-White, Gale Pearson, Marjorie Renwick*, Fiona Reynolds, Dawn Roberts*, Moira Stewart, Rosie Thompson*, Allan Wardhaugh, Martin Ward-Platt and Michael Weindling. The data were gathered by local networks organised by CEMACH regional managers* with the assistance of Jo Coffee, Terry Falconer, Judith Hopkins, Lisa Hydes, Ann-Marie McCann, Mandy Neish and Sue Tierney. A list of panel contributors is included in Pearson.8

Footnotes

  • Funding Funding for the national confidential enquiry on which this article is based was provided by the National Patient Safety Agency and by the Department of Health, Social Services and Public Safety of Northern Ireland. The views of the authors are their own and do not necessarily represent those of the funding bodies.

  • Competing interests None.

  • Ethics approval This study was conducted with the approval of the North West Multicentre Research Ethics Committee Ref: 05/MRE08/51.

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

References