Article Text
Abstract
Objective: To examine the variation in serious injuries in children by the level of deprivation in an area and by types of settlement.
Methods: Hospital admission rates for serious injury to children aged 0–15 years in census lower super output areas in England during the 5-year period 1 April 1999 to 31 March 2004 were analysed.
Results: Rates of serious injury in children were higher in the most deprived areas than in the least deprived for pedestrians (rate ratio (RR) 4.1; 95% CI 2.8 to 6.0) and cyclists (RR 3.0; 95% CI 1.9 to 4.7). Rates of serious pedestrian injury were lower in towns and fringe areas (RR 0.67; 95% CI 0.53 to 0.86) and in village (RR 0.64; 95% CI 0.50 to 0.83) areas than in urban areas. The rate of serious injury to cyclists was lower in London than other urban areas (RR 0.78; 95% CI 0.62 to 0.98). The rate of serious injury to car occupants was higher in village than urban areas (RR 1.51; 95% CI 1.05 to 2.17). Rates of serious injury caused by falls were higher in London (RR 1.60; 95% CI 1.47 to 1.75) and lower in villages (RR 0.76; 95% CI 0.66 to 0.88) than in urban areas. Steeper socio-economic gradients in serious injury rates were identified in rural areas for cyclists and for children suffering falls.
Conclusions: Socio-economic inequalities in serious injury exist across the whole of England, particularly for child pedestrians. Rates of serious injury vary by settlement type, and inequalities vary by cause of injury between rural and urban settings.
Statistics from Altmetric.com
Child death rates from injuries in England and Wales have fallen over the past 20 years, but steep socio-economic gradients remain. Recent analysis of mortality data found that death rates of children with parents in routine occupations were over twice those of children with parents in higher managerial and professional occupations. Death rates of children in families with no adult in paid employment were 13 times higher than those with professional adults.1 There is also considerable morbidity from injury. Non-fatal injuries can result in disabilities that have a profound effect on the quality of life of the children as well as their parents.
Inequalities in injury morbidity rates are more difficult to investigate, in part because routine data sources on injury morbidity do not include information on socio-economic status. However, over the past decade a number of ecological studies have suggested that child-injury mortality gradients in the UK are reflected in indicators of injury morbidity such as Accident and Emergency (A&E) department attendance and hospital admissions. Socio-economic gradients have been identified for the most common injury mechanisms in several hospital A&E departments, 2–4 in English regions,5 and in Wales,6 but there has been some disagreement over whether, for instance, fractures are more common in deprived areas.7 8 We therefore aimed first to establish whether, on the basis of hospital admissions, socio-economic gradients in serious injuries in children exist across the whole of England and, if so, whether gradients vary by cause of injury.
Although there is strong evidence to suggest that socio-economic inequalities in child-injury mortality exist, and some evidence to suggest that these inequalities are reflected in morbidity, gaps remain in both our knowledge about other sources of variation in injury rates and about what mechanisms might link measures of socio-economic status with injury risk. There has been some interest in disentangling those effects of deprivation that relate to the area people live in (such as housing quality or road-traffic speeds) and effects that result from aggregating individual or household effects (such as lone parenthood or family size) at an area level.2 3 However, there has been less attention paid to other geographical variation that might result in different injury outcomes and rates. The risks to which children are exposed clearly relate to environmental factors such as road layouts, modes of transport available and housing stock, which vary considerably across the country.9 There is some evidence that these risk factors are associated with injury outcomes (for instance, that urban residents are more likely to present to A&E with head injury than those from mixed rural areas)10 but there is little evidence to explain how deprivation affects variations in injury rates across rural and urban areas. We therefore examined serious injuries in children stratified by rural and urban area classification,11 and we assessed whether deprivation gradients were similar across these settlement types.
METHODS
We obtained an extract of Hospital Episode Statistics (HES) from the Department of Health containing records of all child (0–15 years) admissions to NHS hospitals in England owing to external causes during a five-year period around the 2001 census. As HES data are based on financial years, we analysed records spanning the period 1 April 1999 to 31 March 2004. External cause was coded according to the external causes of morbidity and mortality chapter of the 10th revision of the International Classification of Diseases (ICD). Primary diagnosis was coded according to the ICD chapter on injury, poisoning and certain other consequences of external causes. Each record included year of admission, external cause code, primary diagnosis code and census lower super output area (SOA) code of residence (records with any of these data missing were removed before analysis).
Use of hospital admissions data to estimate injury rates can lead to confounding by extraneous factors, such as admission policies or distance to hospital (both of which are related to deprivation). To reduce possible effects of selection bias in the results owing to health-service supply and demand factors independent of the injuries, we selected a conservative selection of injuries known to be sufficiently serious that the majority of cases would be admitted to hospital (box 1).12
Box 1 Primary diagnoses defined as “serious” in this study
ICD10 Diagnosis
S72.0 Fracture of the neck of the femur
S06.1–.9 Intracranial injury (excluding concussion)
S14 Injuries of nerves and spinal cord at neck level
S22.4 Multiple fractures of ribs
T71 Asphyxiation
T68 Hypothermia
We obtained data at SOA level on the Index of Multiple Deprivation 2004 (IMD) score (recalculated to exclude the environment domain indicators for road-traffic injuries) and from the 2001 census on the following indicators: rural and urban area classification (four area types: (1) London – including all boroughs, (2) urban – excluding London, (3) town and fringe, and (4) village);11 the proportion of households classed as overcrowded; proportion of households without access to a car; proportion of households that contain lone-parent families; proportion of adults unemployed; and the percentage of the resident population that described its ethnicity as Black/Black British, Asian/Asian British, White, Mixed or Other. These variables were included a priori as being associated with injury risk and therefore potential confounders in our analysis. Research ethics approval was not required for this study. Use of patient-identifiable information was approved by the Patient Information Advisory Group, with support under section 60 of the Health and Social Care Act 2001.
Statistical methods
Admission rates were calculated for all external causes combined and for the leading causes of unintentional injury separately, using population data from the 2001 census. All admissions during the five-year period were combined to increase the accuracy of our estimates. We examined associations between injury rates, area deprivation and settlement type for the leading causes of injury. Negative binomial regression was used to examine associations between injury rates and the other SOA-level variables. Models were fitted using the backward-stepwise selection method, with a significance level of 20% for removal of variables from the model. We assessed evidence for deprivation gradients in injury rates using deciles and quintiles of SOAs, on the basis of IMD score: deciles were used to examine gradients across England as a whole; quintiles were used to examine gradients across settlement types. Interaction terms were fitted to assess whether deprivation gradients were similar across settlement types. Rate ratios with 95% confidence intervals were estimated, comparing injury rates in the most deprived SOAs with those in the least deprived. We used robust standard errors to allow for clustering of injury rates at SOA level within local authorities. All analyses were conducted using the Stata Statistical Software.13
RESULTS
Between April 1999 and March 2004 there were 663 542 hospital admissions for children aged 0–15 years in England owing to all external causes (table 1). Of these admissions, 7840 (1.2%) were for serious injuries (as defined in box 1). The average annual admission rates were 1340 (95% confidence interval 1337 to 1344) per 100 000 children for all injuries, and 15.8 (95% CI 15.5 to 16.2) per 100 000 children for serious injuries. Falls was the largest group, accounting for 36% of all admissions and 41% of admissions for serious injuries. The admission rate for serious injuries caused by falls was 6.4 per 100 000 children (95% CI 6.2 to 6.7 admissions). Transport injuries accounted for 10% of all admissions and 31% of admissions for serious injuries. Among admissions for serious transport injuries, the rate was higher for pedestrians (2.2 per 100 000 children; 95% CI 2.1 to 2.3 admissions) than for cyclists (1.5 per 100 000 children; 95% CI 1.4 to 1.6 admissions) and car occupants (0.6 per 100 000 children; 95% CI 0.5 to 0.6 admissions).
Deprivation gradients
Table 2 shows rate ratios for the leading causes of serious injury (pedestrian, cyclist, car occupant and falls) by deciles of SOAs ranked by IMD score. The first model includes deciles of SOA only and the second model includes census variables identified by the stepwise selection method. The rate of serious injury to pedestrians living in the most deprived SOAs in England was 4.1 times (95% CI 2.8 to 6.0) greater than that in the least deprived areas. Evidence for increased risk of injury with increased area deprivation was also found for cyclists (rate ratio (RR) comparing most deprived with least deprived: 2.6; 95% CI 1.7 to 4.0), car occupants (RR 2.0; 95% CI 1.0 to 3.9) and children who suffered falls (RR 2.1; 95% CI 1.4 to 3.2). After adjusting for the percentage of the population who were “Black” and the percentage of the population who were “White”, pedestrian injury rate ratios comparing most deprived with least deprived SOAs were unchanged. After similar adjustment for ethnic-group populations, rate ratios increased for serious injuries to cyclists (RR 3.0; 95% CI 1.9 to 4.7) and to car occupants (RR 4.8; 95% CI 1.6 to 13.9). In the model for car-occupant injury, car ownership was identified as a predictor. The association between area deprivation and injury rates from falls was lost after adjustment for ethnic-group populations, proportion of households without a car and proportion of households with lone-parent families.
Urban and rural effects
Table 3 shows admission rates by area type for the leading causes of serious injury. Rates of serious injury to pedestrians were lower in town and fringe (RR 0.67; 95% 0.53 to 0.86) and in village (RR 0.64; 95% 0.50 to 0.83) areas than in urban areas. The rate of serious injury to cyclists was lower in London than other urban areas (RR 0.78; 95% 0.62 to 0.98). The rate of serious injury to car occupants was higher in village than urban areas (RR 1.51; 95% 1.05 to 2.17). Rates of serious injury caused by falls were higher in London (RR 1.60; 95% 1.47 to 1.75), and lower in village (RR 0.76; 95% 0.66 to 0.88) than urban areas.
Urban and rural deprivation gradients
We tested whether the effect of deprivation on injury rates was the same for each settlement type (models including interaction terms are not shown). We found no evidence for differences between settlement types in the effect of deprivation on either pedestrian or car-occupant injury rates. For cyclists, there was evidence for steeper deprivation gradients in town and fringe and in village areas than in urban areas (Wald test of interaction term comparing most deprived quintiles with most deprived urban quintile p = 0.017 and p = 0.016, respectively). For serious injuries caused by falls, there was evidence for a lower gradient in London than in other urban areas (Wald test of interaction term comparing most deprived London and urban quintiles p = 0.026) and evidence for higher gradients in town and fringe (p = 0.012) and in village (p = 0.004) areas.
DISCUSSION
In a national data set, we have identified a positive relationship between serious injury admission rates and increasing area deprivation for the leading causes of unintentional injury. Our analysis used a selection of injuries judged so severe that differential self-referral patterns or differential admission policies would not account for the observed associations. When SOAs were classified by settlement type, we found that, compared with urban areas, serious injury rates were lower for pedestrians in rural areas, lower for cyclists in London and higher for child occupants of cars in rural areas. We also found some evidence that inequalities in cycling injury were greater in rural than in urban environments. The rate of serious injury from children falling was highest in London, where we found inequalities were lower than in urban areas, and rates were lowest in villages, where we found inequalities were higher. Before we consider policy implications of our results, we will address the methodological strengths and weaknesses of our study.
Strengths and weaknesses in relation to other studies
We used data covering England over a five-year period to investigate area-level associations between deprivation and rates of the leading causes of injury for four settlement types. Using a robust injury-severity measure, we identified similar rate ratios to those identified in Trent,5 and our results are consistent with findings from analyses of child-death rates by National Statistics Socio-economic Classification (which used individual-level data on the occupation of the children’s parents).1 We have now added to this evidence base on inequalities by reporting on variations by area settlement type. We did not examine associations between deprivation levels and admission rates for burns or poisonings, as the definition of serious injury we used in our study did not include these external causes.
The HES data set contained complete data on year of admission, external cause code and primary diagnosis code; data for SOA code of residence were missing in only 0.8% of records. Such high levels of completeness reduce the possibility of selection bias in our results. It is known, however, that completeness of the external cause of injury codes varies considerably between hospital trusts, and this may have introduced bias if the completeness of cause-coding by trusts is associated with average levels of deprivation among trust patients.
We chose to use IMD over other possible indicators of area deprivation. IMD was designed to be a robust measure of multiple deprivation at a small-area level, taking into account a range of distinct dimensions of deprivation that can be separately measured. IMD is the most up-to-date composite deprivation score currently available and has the advantages over census-based measures, such as Townsend and Carstairs scores, of being more frequently updated and based on a larger and more balanced set of deprivation indicators.
Possible mechanisms and implications for policymakers
We have reported ecological analyses based on hypothetical subgroups of the population of children in England, defined by area-level deprivation and settlement type. Any inferences we make about the relative safety of children living in these areas are thus based on averages of population data, aggregated at SOA level. It is not possible for us to distinguish contextual factors (relating to the area itself) and those that result from aggregates of individual risk factors.
We identified gradients in serious transport injury rates. For transport-related injury deaths, cycling and walking present greater risks than car occupancy, both per mile travelled 14 and per hour journey time.15 These modal differences are likely to contribute to differential injury rates, as children from households without access to vehicles walk more than their counterparts in car-owning families do.14 There is also evidence that cycling risk varies by deprivation, with a small study in Nottingham finding that children in deprived areas cycled more frequently and were less likely to own helmets.16 Variations in transport mode may help to explain some of the rural–urban differences observed in this study. We found that children in rural areas have higher injury rates as car occupants. Car access is higher in rural areas, with lower rates in urban areas, and the lowest in London.9 With lower levels of access, there are likely to be fewer injuries to car occupants.
Child-injury mortality has declined over the past 20 years, from about 11 deaths per 100 000 children aged 0–15 years in 1981, to four deaths per 100 000 in 2001; however, socio-economic inequalities remain.1 This study has demonstrated that a child’s risk of non-fatal injury is associated with a measure of area deprivation. Persisting inequalities in childhood injury are the subject of current national policy in England in the areas of health, education and transport.17–19 The National Service Framework for Children, Young People and Maternity Services provides guidance for local authorities and Primary Care Trusts to address injuries in areas with marked inequalities. Policy makers and practitioners, however, face this task with little evidence about which interventions are most likely to reduce inequalities in injury rates,20 as there are still large gaps in our understanding of the mechanisms that link deprivation (or area deprivation) and injury risk, and how those mechanisms might operate in different settings.
For transport-related injury, our findings provide further support for focusing on the level of exposure as a key factor in any attempt to explain how children living in the most deprived areas of England face increased risks of serious injury. To reduce both incidence and inequalities in non-fatal injury, the major policy need is for road environments to be safer for young pedestrians and cyclists. There is evidence that reducing the speed and volume of car traffic would achieve this,21 reducing the overall number of injuries, and – given that greater exposure as a pedestrian partly accounts for the inequalities – the higher injury rates in more deprived areas. In the UK, cycling and walking appear less safe as modes of transport. However, given evidence for a “critical mass” effect, whereby walking and cycling become progressively safer as more people walk and cycle,22 in the longer term encouraging more walking and cycling may well reduce the overall risk of these modes and thus the added “exposure” risk for those in the most deprived areas.
Falls account for a larger number of serious injuries than do transport accidents. Although the majority of fatal falls in childhood result from falls from buildings or other structures (eg, from roofs or through windows),23 hospital admissions also include other settings, such as falls on the same level (eg, slipping, tripping and stumbling) and falls on or from stairs and steps. We found rates of serious injury in children from falling to be highest in London and lowest in villages. Exposure to risk is likely to be one determinant, but more detailed research is needed to establish how area-based deprivation measures might be associated with, for instance, housing stock, hazards in the built environment or leisure facilities, which would put children in some areas at increased risk of these injuries.
What is already known on this topic
Socio-economic inequalities in child-injury deaths have persisted in England for several decades.
Regional studies in the UK have identified inequalities in non-fatal injuries in childhood.
What this study adds
Compared with urban areas, rates of serious injury in children are lower for pedestrians in rural areas, lower for cyclists in London and higher for car occupants in rural areas.
Inequalities in serious injuries caused by falls and when cycling are greater among children living in the most deprived rural areas.
Acknowledgments
We would like to thank Susan Alpay at the Health and Social Care Information Centre for her assistance with accessing HES data. We would also like to thank Richard Reading, Tinneke Beirens and a third reviewer whose comments helped improve this paper.
REFERENCES
Footnotes
Funding: This work was undertaken by the London School of Hygiene and Tropical Medicine who received funding from the Department of Health. The views expressed in the publication are those of the authors and not necessarily those of the Department of Health.
Competing interests: None declared.