Article Text

Download PDFPDF

Taking the long view: a systematic review reporting long-term perspectives on child unintentional injury
  1. Julie A Mytton1,
  2. Elizabeth M L Towner1,
  3. Jane Powell2,
  4. Paul A Pilkington2,
  5. Selena Gray2
  1. 1Centre for Child and Adolescent Health, University of the West of England, Bristol, UK
  2. 2Department of Health and Applied Social Sciences, Faculty of Health and Life Sciences, University of the West of England, Bristol, UK
  1. Correspondence to Dr Julie A Mytton, Centre for Child and Adolescent Health, University of the West of England, Oakfield House, Oakfield Grove, Clifton, Bristol BS8 2BN, UK; julie.mytton{at}


Objective The relative significance of child injury as a cause of preventable death has increased as mortality from infectious diseases has declined. Unintentional child injuries are now a major cause of death and disability across the world with the greatest burden falling on those who are most disadvantaged. A review of long-term data on child injury mortality was conducted to explore trends and inequalities and consider how data were used to inform policy, practice and research.

Methods The authors systematically collated and quality appraised data from publications and documents reporting unintentional child injury mortality over periods of 20 years or more. A critical narrative synthesis explored trends by country income group, injury type, age, gender, ethnicity and socioeconomic group.

Findings 31 studies meeting the inclusion criteria were identified of which 30 were included in the synthesis. Only six were from middle income countries and none were from low income countries. An overall trend in falling child injury mortality masked rising road traffic injury deaths, evidence of increasing vulnerability of adolescents and widening disparities within countries when analysed by ethnic group and socioeconomic status.

Conclusions Child injury mortality trend data from high and middle income countries has illustrated inequalities within generally falling trends. There is scope for greater use of existing trend data to inform policy and practice. Similar evidence from low income countries where the burden of injury is greatest is needed.

  • Child
  • mortality
  • epidemiology
  • advocacy
  • interventions
  • public health
  • mechanism
  • violence
  • methodology
  • systematic review
  • populations/contexts
  • school

Statistics from

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.


Each year injury and violence are responsible for 950 000 deaths in children aged 0–18 years throughout the world, with 90% of these considered unintentional subsequent to events such as road traffic collisions, drowning, burns, falls and poisoning.1 Tens of millions of children require hospital care for non-fatal injuries and many are left with some form of disability, often with life-long consequences. The burden of injury falls most heavily on the poor, both within and between countries.2 WHO has estimated that 95% of all child injury deaths worldwide occur in low and middle income countries.1 Rates of unintentional injury deaths are 3.4 times greater in low and middle income countries compared with high income countries. In May 2011, the World Health Assembly adopted a resolution on child injury prevention urging member states to promote the prevention of child injuries and implement a plan of action.3

The World Report on Child Injury Prevention highlighted substantial changes in the epidemiology of child deaths over the last century. As infectious diseases have declined, the relative importance of injury has increased in many countries. Studies of long-term trends in public health can identify patterns of morbidity and mortality that have the potential to inform policy and prioritisation. Systematic reviews of long-term trends in children's public health are uncommon in the published literature.4 To our knowledge, there are no systematic reviews of long-term trends in child injury. We therefore describe the findings of such a review and consider whether there are implications for policy, practice and research.


The aim of the systematic review was to collate and critically synthesise data on long-term trends in unintentional child injury mortality. We defined long-term as 20 years or more, acknowledging that authors will report data over variable time periods depending on the sources available. The review explored the following questions within this evidence base:

  1. Are injury mortality rates falling, rising, stable or variable for childhood unintentional injuries overall and for specific injury types?

  2. Are socioeconomic and demographic inequalities decreasing, increasing or stable?

  3. Have analytical and descriptive analyses been undertaken to inform explanations of injury trends?

  4. What explanations are given by authors for reported trends in childhood unintentional injury?

  5. Have authors used their findings to make recommendations for policy and practice?

Identification of studies

Studies were included in the review if they met all of the inclusion criteria:

  1. long-term (20 years or more) unintentional injury mortality reported;

  2. children and young people (up to 19 years of age inclusive) included;

  3. data at a national or sub-national level reported;

  4. Studies published since 1990.

Studies were excluded if:

  1. data on injuries in children and adults could not be identified separately;

  2. data on unintentional injury deaths could not be separated from intentional deaths or all cause mortality;

  3. only infant mortality secondary to ‘birth injury’ reported;

  4. only crude numbers were reported, without denominators to estimate rates;

  5. only case series of injuries occurring in specific populations at increased risk of injury reported (eg, people with epilepsy);

  6. intentional injury data alone were presented;

  7. not written in English;

  8. data were only available for a state, county or city.

Search process

The following electronic databases were searched during February 2011: Medline, Embase, CINAHL, ASSIA, PsycINFO, the Cochrane Injuries Group Specialist Register and SafetyLit using a search strategy developed in Medline and adapted for application to other databases (strategies available from first author). Using reference management software where possible, one author screened the titles and abstracts of studies to exclude duplicate references and ineligible studies on the basis of their design, population and study outcomes. The full texts of remaining studies were obtained and further ineligible studies were excluded using the same criteria. Grey literature sources included the bibliographies of included studies, a Medline author search for further publications by the lead author of included studies and searches of the publication catalogues on the websites of WHO and Unicef.

Data extraction

A data extraction form was developed and piloted on six papers and modified accordingly. Data extraction was undertaken from all included studies on the number and description of the study participants, the study methods, outcomes and recommendations recorded. In order to categorise the explanations of injury trends reported by authors of included studies, a set of category headings were identified during the development of the data extraction form. These categories were: political, economic, social, technological (incorporating environmental), legislative, healthcare or other reasons for the trends seen. The country of each included study was categorised as high, middle or low income using World Bank definitions.5 Included studies were divided among all members of the review team for data extraction. The reviewers were all experienced in systematic review techniques. Reviewers were not blinded to the names of the journals, the authors, the institutions or the results when extracting data on study methods. Data extraction forms were checked by the lead author prior to collation of data into tables and uncertainty in data extraction was resolved through discussion.

To assess study quality, each paper meeting the inclusion criteria was quality assessed prior to inclusion in the synthesis, using a set of 10 questions developed by the review team from published critical appraisal tools.6 Studies were categorised into those with sound methodology and clear reporting (A), those with minor concerns regarding methodology or reporting, but not to the extent that the validity of the results was questioned (B) and those with serious concerns regarding the methodology or reporting leading to doubt regarding the validity of the results (C). Studies with a (C) rating were deemed poor quality and were excluded from the synthesis.

Data analysis

Data were organised into tables by setting, reporting interval, injury type, trends and author interpretation of results. Potential health inequalities were explored by extracting data on the age, gender, ethnicity and socioeconomic status of the child, where reported. Due to the heterogeneity of the studies identified and the risks of bias associated with unrecognised confounding within observational data,7 ,8 a formal meta-analysis was considered unsuitable. Therefore, a narrative synthesis using methods described by Petticrew and Roberts9 ,10 was used to provide an overview of the findings, exploring the similarities and differences in results between studies while taking into account the differences in design, setting and population.


In all, 27 eligible reports were identified from 1585 unduplicated citations found in the electronic databases searched. A further four reports were identified through grey literature searches, making a total of 31 studies meeting the inclusion criteria (figure 1, table 1).11–41 No unpublished reports were identified. Twenty-four studies were from high income countries (Canada, England, Finland, Japan, Norway, Scotland, Sweden, Switzerland and the USA) and six from middle income countries (Chile, Croatia, Lithuania and Taiwan). Three of the included studies were updated reports26 ,32 ,41 of earlier publications.25 ,31 ,40 No studies were identified from low income countries. One study provided international comparisons between 20 countries, including many from the former USSR, and therefore included both high and middle income countries.35 Time trends were reported from 20-year up to 98-year periods17 (table 1). The majority of studies presented data from national birth and death registries. Thirty of the studies were considered to be of satisfactory or good quality, and only one study meeting the inclusion criteria was excluded from the synthesis due to being given a poor quality rating.12

Figure 1

Flowchart of identified eligible studies. In all, 146 citations were excluded on review of the full text because they failed to meet all of the inclusion criteria or because they met one or more exclusion criteria. It is not possible to specify the number failing to meet each criterion as many papers failed to meet multiple criteria.

Table 1

Studies meeting the inclusion criteria

Trends in all cause unintentional injury rates and injury rates by type

Most studies reported falling rates for overall child unintentional injury mortality over time. Two studies reported variable rates19 ,28 and one rising rates.24 The study by Mujkic and colleagues28 reported on a period of conflict, and so is arguably atypical. Generally, mortality rates were also falling for specific types of injuries, where these were reported (particularly deaths from burns/fire, drowning and poisoning) (table 2). The notable exception was mortality from road traffic injury, where seven studies reported rising or variable road traffic injury mortality,19 ,23 ,24 ,30 ,35 ,40 ,41 particularly in older adolescents23 ,40 ,41 where this was specified. Greater variability in trends in injury mortality was seen in studies from the middle income countries compared with those from high income countries.

Table 2

Trends in child injury from included studies

Trends in inequalities

Studies were reviewed for their reporting of variables associated with inequality in injury mortality by age, gender, ethnic group and socioeconomic status. Analysis of injury rates by age and gender were generally well reported, with 23 studies providing rates by age groups and 18 studies providing rates by gender. Boys were reported to have a greater incidence of injury mortality than girls at all ages, a pattern that did not change with time. Two studies showed variable trends by gender over time, with decreasing rates for boys for falls34 and drowning27 while rates in girls remained broadly stable.

In contrast to reporting by age and gender variables, only two studies reported trends by ethnic group30 ,36 and only four reported trends by measures of socioeconomic status.11 ,15 ,30 ,39 Both of the two studies reporting injury mortality by ethnic group identified widening ethnic disparities with time across injury types. The study by Olson et al 30 found that childhood mortality rates for motor vehicle crashes and homicides increased for all three of the different ethnic groups studied (New Mexico's American Indian, Hispanic and non-Hispanic White populations), with the overall injury mortality rate remaining particularly high for the American Indian group. The study by Pressley et al 36 showed all cause injury mortality rates across 50 American states falling in general, but disparities widening by ethnic group. The four studies examining injury morality rates by socioeconomic status were from high income countries: Canada,11 the UK15 and the USA.30 ,39 All found widening social disparities, within the context of overall falling injury rates, with the rate of decrease among the most disadvantaged groups slower than that in those groups who were more advantaged.

Interpretation of trends provided by authors

All authors described the injuries identified, mostly using injury rates. There was more limited use of analytical techniques, with only 18 of the studies included in the synthesis using techniques such as regression modelling to explore statistically significant changes in mortality between subgroups or over time, and to consider associations between risk factors and injury outcomes.11 ,13 ,14 ,18 ,19 ,22 ,24–26 ,31 ,32 ,34–36 38–41 We explored whether authors considered any particular hypotheses to explain the trends reported either on the basis of their analyses or on the basis of their understanding of the context in which these trends had occurred (table 3). Authors were most likely to try to explain trends against social, technological, legislative or healthcare categories of change. Very few attempted to analyse the impact of specific legislative change or other interventions. Social changes such as greater uptake of safety equipment use, reductions in exposure to injury risk and increased parental awareness were cited as the most likely explanation for the change in trends, being reported by 25 (81%) studies. Improved healthcare was reported as a likely explanation in 16 (52%) studies (mostly advances in trauma care and improved access to trauma services). Six authors (19%) reported political explanations for trends: two relating to changing injuries secondary to war time injury risks;23 ,28 two relating to momentous social change following independence from USSR35 ,40 ,41 and one secondary to the establishment of a high profile child safety strategy.18

Table 3

Author interpretation of trends and recommendations

Policy and practice recommendations

The authors of eight studies made no specific policy or practice recommendations. The remaining studies frequently made non-specific recommendations relating to the need to develop new or more effective injury prevention strategies, or to target groups that remained the most vulnerable, with adolescents specified as vulnerable in six studies19 ,20 ,24 ,27 ,40 ,41 (table 3). Two studies by Nakahara and Wakai29 and Jansson et al 18 prioritised recommendations relating to changing the environment of children to reduce injury occurrence, while two studies emphasised the need for adequate surveillance to monitor changing trends.35 ,36


The papers identified in this study largely provide a welcome illustration of falling unintentional child injury mortality rates in high and middle income countries around the world, albeit with substantial differences in absolute rates between high and middle income countries. However, we did not identify any papers that reported long-term trends in low income countries, where the greatest burden of injury exists, and only six from middle income countries. The paucity of such studies may be due to a number of reasons. We applied an English language restriction to included studies which may have biased the identification papers from low income countries. Additional grey literature searching, including contacting researchers and organisations working with low income countries, may have led us to further studies. The lack of studies may reflect lack of availability of timely, complete and high quality mortality data over a long time period. To monitor child deaths effectively, countries require adequate systems to collate information on births and deaths to produce robust national estimates of child populations, and schema, such as the International Classification of Disease, to categorise the cause of death.

While we attempted to conduct this review in a robust manner, limitations of our methods are acknowledged. A more extensive list of electronic databases could have been included in the search strategy. Contacting authors of included studies for clarification of methods, and additional data or further studies may have improved the quantity and quality of studies. Furthermore, restricting data extraction to two reviewers would have reduced the risk of inter-rater variability in data extraction, and facilitated reporting of inter-rater reliability.

Despite an overall trend of falling injury rates, disparities in child injury mortality persist within countries and are widening for some groups, the most disadvantaged in society remaining the most at risk of injury-related deaths.11 ,15 ,30 Similarly, some long-term studies have shown widening ethnic disparities within populations.30 ,36 While few of the studies explored socioeconomic or ethnic disparities, the consistent evidence of widening disparities suggests that this should be a major focus for child injury researchers in future. This will be challenging, as the national statistics upon which surveillance of child injury is based may not include data to routinely identify socioeconomic or ethnic groups.

There were some long-term changes in relation to gender and age groups. A time trend study of childhood injuries in Scotland investigated gender patterning. Over time, previous male excess in injuries had disappeared in some age groups and cause categories, for example, in pedestrian injuries.34 The trend towards falling injury rates was not consistent across age groups in all countries, with rising rates for 15–19-year-olds for road traffic injuries,23 ,41 falls24 and burns,14 suggesting adolescence may be a further emerging area for targeted injury prevention.

Sweden has the lowest child injury mortality rates in the world. These rates are potentially attainable by other countries. It was notable that studies from Sweden, such as that by Jansson et al,18 emphasised the impact that improving the safety of the environment can have on child injury mortality rates. It may be that the child and family friendly policies together with an emphasis on the environment, as opposed to behavioural approaches, are what have made the Swedish so successful in this area. In contrast, the authors reporting data from middle income countries gave few suggestions explaining the injury trends reported, with apparently little reflection on the impact of the wider environment on injury rates.

Few papers have attempted to analyse systematically why changes have occurred. A notable exception was a paper from the USA which examined road traffic injury mortality for the period 1910–1994, using the age–period–cohort analysis technique. In this report, Li et al 22 reported significant period effects in the temporal trends of motor vehicle mortality, and linked these to changes in the socioeconomic environment, such as periods of war or fuel shortage. The most widely suggested explanations for falling road traffic injuries were technological advances in vehicle design and safety enhancement of the road environment. Around a half of the studies suggested legislative explanations for improvements in injury mortality, although there was a limited attempt to demonstrate the effectiveness of legislation through formal time series analyses. The majority of studies that highlighted legislation as an explanation for falling trends in injury mortality rates identified reductions in traffic speed, the use of vehicle safety devices, for example, seatbelts, or building regulations, for example, fitting of smoke detectors. Improved trauma care and the development of poisons information centres were the two most common healthcare suggestions provided for reductions in injury mortality rate trends.

Papers identified in our systematic review demonstrated how in high income countries the process of responding to high rates of child injury took place over decades.18 This review was able to identify studies in medium income countries showing that injury rates can be reduced much more quickly by implementing and adapting evidence based approaches. For example, in Taiwan, unintentional injury rates increased from 1965 to 1994,24 followed by a steep decline to at least 2006, after a concerted programme of road safety interventions.13 Many low income countries, however, will need to respond more rapidly to a faster pace of economic and social development than that experienced in medium income countries. Analysis of long-term trends can provide the intelligence to facilitate this response.


This systematic review synthesised evidence on child injury mortality rates across countries, explored whether inequalities in mortality persist, and whether trend data are being used to inform policy, practice and research. We identified 31 studies indicating generally falling all cause injury mortality but masking variations by injury type; for example, there was clear evidence of rising trends in children and young people dying in road traffic incidents. We failed to identify any studies from low income countries and due to the global distribution of child injury deaths, such evidence needs to be identified or support should be provided to enable it to be collected.

There is some evidence that inequalities in child injury mortality are persisting and widening in some cases. Adolescent age groups have been identified in a number of countries as increasingly vulnerable, with evidence of rising trends in deaths due to road traffic injuries, falls and burns. Recent publications have emphasised the need to devote more attention and resources to adolescents to consolidate gains achieved in health improvements in early and middle childhood.4 ,42 The increased incidence of injury deaths in boys compared with girls is well documented. This review identified that for some types of injury rates may be falling in boys while remaining static for girls. The paucity of reporting of child injury mortality by ethnic group or socioeconomic status may suggest that these inequalities are not fully recognised, and should be a target for further research, especially in light of the fact that in the two cases where data were available by ethnicity, disparities appeared to be widening.

This systematic review has illustrated the limited use of trend data for the development of recommendations for policy, practice and research. There is some evidence that studies of long-term injury trend data can be a powerful advocacy tool. The paper by Edwards and colleagues15 about widening inequalities related to deprivation has been used to advocate a targeted approach to child injury prevention in more deprived areas and has an impact on policy. In England, it has been cited in the National Institute for Health and Clinical Excellence guidance on the prevention of home injuries43 and in WHO European region to promote policies relating to the reduction of inequalities in child injuries.2 There appears to be considerable scope for a more analytical approach to studying child injury mortality trends that could be invaluable for informing both policy and practice, especially in those countries where the future burden of unintentional injury threatens to be greatest.32 ,41

What is already known on the subject

  • Reduction in child mortality from infectious diseases has resulted in mortality from injury assuming increasing importance, especially in high income countries.

  • The greatest burden of global child injury mortality lies in low and middle income countries.

What this study adds

  • There is an overall trend of falling unintentional child injury mortality rates in high and middle income countries around the world, albeit with substantial differences in absolute rates between countries. This trend masks disparities by injury type, with child road traffic deaths rising in several countries.

  • Inequalities in child injury mortality persist within countries and are widening for some age groups (eg, adolescents). Data on disparities by ethnic group and socioeconomic status are inadequate and should be the focus for further study.

  • There is scope for greater use of trend analysis to inform policy and practice particularly in those countries where the burden is greatest and where inequalities in child injury mortality persist.


The authors would like to acknowledge the support of the library staff at the University of the West of England for their assistance in obtaining publications for consideration in the review.



  • Competing interests None.

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