Article Text

Differences in incidence of injury between rural and urban children in Canada and the USA: a systematic review
1. Kyungsu Kim1,2,
2. Dejan Ozegovic3,
3. Donald C Voaklander1,3
1. 1Department of Public Health Sciences, University of Alberta, Edmonton, Alberta, Canada
2. 2Rural Development Administration (RDA), Suwon-si, Kyunggi-do, The Republic of Korea
3. 3The Alberta Centre for Injury Control & Research (ACICR), Edmonton, Alberta, Canada
1. Correspondence to Kyungsu Kim, 4075 Research Transition Facility, 8308-114 street, Edmonton, Alberta T6G 2E1, Canada; kyungsu1{at}ualberta.ca

## Abstract

Objective The goal of the study was to systematically review available evidence regarding differences in injury incidence between rural and urban paediatric populations in Canada and the USA.

Data source Eight electronic databases, institutional websites and reference lists of relevant studies including published and unpublished reports.

Selection criteria Population-based observational studies or surveys published from 1970 to February 2011 that compared injury incidence or injury-related healthcare outcomes between rural and urban children (<18) living in Canada or the USA.

Data collection and analysis Two reviewers independently applied selection criteria and assessed methodological quality of studies. Data were extracted by one author and independently verified by the second author. Injury rate ratios for rural and urban children were extracted or calculated. Data were synthesised descriptively due to substantial heterogeneity among studies.

Results A total of 41 studies were included for this review (seven surveys and 34 studies using administrative health databases). Internal validity of included studies was moderate. Rural children were at higher risk of overall injury, motor vehicle crash injury and suicide, whereas urban children in the USA experienced higher rates of firearm-related homicides. Greater rural–urban injury disparities were likely to be found between more extreme rural and urban areas. In particular, children in remote rural areas are at increased risk of severe injuries than urban counterparts. Overall, healthcare costs per child for injury were higher for rural children.

Conclusion These findings indicate the need of developing geographic area-specific injury-prevention strategies. Future research is required to investigate rural–urban disparity for less-studied injuries and related health outcomes (eg, disability).

Systematic review registration number CRD42011001244 (PROSPERO 2011).

• Wounds and injuries
• child
• rural health
• urban health
• review, interventions
• mechanism
• outcome of injury
• populations/contexts
• rural
• sports
• spinal cord
• falls
• older people

## 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.

## Introduction

Injury is the leading cause of death for children aged 1–19 years in Canada and the USA.1 ,2 WHO reported that on a daily basis, more than 2000 children die worldwide from unintentional injuries, and at least half of these deaths are preventable if appropriate prevention measures are adopted.3 Injury trends and problems vary between populations and communities, and an identification of specific injury problems for different populations is required to target injury control strategies.

Rural and urban children are surrounded by different physical and socioeconomic environments and risk factors of injury. Prior studies have demonstrated differences in injury rates and patterns of injury between these groups (eg, higher rates of overall injury and motor vehicle crash (MVC) injury for rural children than their rural counterparts).4 ,5 However, review or systematic examination of injury differences between these two populations has not been performed previously. Fahs et al6 reviewed literature on risk behaviours including injury experience among adolescents in rural, suburban and urban areas, but the study did not focus on comparisons of injury incidence between these geographic groups.

Canada and the USA share similarities in terms of geography, demographics, sociocultural environment, economic growth, infrastructure and so on.7 In both countries, rural areas comprise 93%–95% of the land mass, but rural populations make up 19%–21% of the total population with declining trends over time,8–11 and experience poor access to healthcare services and a shortage of healthcare providers (eg, physicians).10 ,12 Conversely, they have different health insurance systems (universal social system in Canada vs mixed public–private system in the USA)13 and firearm control laws (less restrictive regulations in the USA vs Canada).14

The objective of this study is to systematically identify and synthesise the existing evidence to assess whether there are differences in injury incidence between rural and urban children (aged under 18) in Canada and the USA. Understanding the injury-related health disparity between rural and urban children would help to develop effective injury prevention strategies specific to each setting and, in particular, to raise the profile of injury sustained by rural children for additional funding and intervention initiatives.

## Methods

A protocol of this study was developed in advance to outline the objective and methods. The review protocol has been registered in PROSPERO (registration number: CRD42011001244).

### Search strategy

A research librarian and the first author developed search strategies, and the first author conducted searches of the following electronic databases to identify reports: Cochrane Database of Systematic Reviews, MEDLINE, EMBASE, Scopus, Web of Science, Health & Safety Science Abstracts, Safetylit, and ProQuest Dissertations and Theses Databases. We applied supplementary approaches by searching websites of institutions and government agencies in Canada and the USA and checking reference lists of relevant studies. There was no restriction by publication status, but the searches were restricted to studies conducted in Canada or the USA and published in English. We used all appropriate search terms that imply the concepts of ‘rural’, ‘urban’, ‘children’ and ‘injury’, while applying them for controlled vocabulary as well as free text terms, when available. The search strategy for MEDLINE is presented in Web Only file #1, and this search strategy was modified as appropriate to the specifications of other databases.

### Study selection

Eligible studies for review were population-based observational studies or surveys that compared injury incidence (primary outcome) or injury-related healthcare outcomes (secondary outcomes including hospital length of stay (LOS), hospital costs, etc) between rural and urban children within each study. Studies with a qualitative approach, reviews and case series were excluded. Populations considered were children under 18 years of age living in Canada or the USA. If the upper age limit exceeded 18, studies were included if the majority of participants were <18 or subgroup data for that age group were provided. We included studies that classify rural/urban status either by area of residence or by area of injury occurrence. Injury outcomes comprise any cause or intent of injury regardless of severity of injury including mortality and morbidity (eg, injury causing medical attention or restricted activity) excluding risk behaviours (eg, drug use experience or violence exposure).

To be included for review, the relative effect measures of injury for compared groups such as rate ratio (RR), OR or HR needed to be provided or be able to be calculated. Studies published from 1970 to February 2011 were included given that large-scale injury prevention efforts began to emerge in the 1960s in the USA.15

For study selection, two independent reviewers first scanned the titles and abstracts to remove obviously irrelevant reports being overinclusive with broad criteria and second, assessed full-text reports of remaining studies using a standardised form that listed predefined inclusion criteria. Disagreements between reviewers were resolved by consensus or third reviewer adjudication.

### Quality assessment

The internal validity of included studies was assessed using the Newcastle-Ottawa Scales (NOS)16 developed for cohort studies. The NOS evaluates selection bias, comparability and outcome assessment with eight items. It scores the validity by awarding a maximum of two stars for an item of comparability and one star each for other items. Scores are summed and range from zero to nine stars. We modified the NOS by removing three items (outcome status at start of study, follow-up period, and follow-up) that were assumed irrelevant to injury studies, leaving us with five items for a maximum possible score of six stars (Web Only file #3). As criteria to assess comparability, we chose age as well as socioeconomic status or medical service environment as important factors to be controlled for or matched. Two authors independently conducted critical appraisal of the included studies. Disagreements were resolved by consensus or by third reviewer adjudication.

### Data extraction

Data were extracted from reports using a standardised data collection form to gather information on the study design, population, data sources, urban/rural definition, outcomes, results and additional information for assessments of the risk of bias. In cases of multiple reports from one study, information was combined. Some data were extracted from graphs.17 ,18 Data extraction was performed by the first author, and independently verified by the second author. More information was obtained by contacting investigators, if necessary. Data were extracted into an Excel database.

### Data analysis and synthesis

Data were summarised and synthesised by the category of injury based on intent and cause of injury. The principle summary measure of the effect of residential area (rural vs urban) on injury incidence was RR or OR. Meta-analysis was not conducted due to large heterogeneity across studies as well as a lack of SE in the majority of included studies as these studies mainly performed descriptive analysis with data for the whole population from administrative health databases. Therefore, data were synthesised qualitatively and displayed using a forest plot with point estimate of RR or OR without pooling of data. If a study classified rural/urban areas using more than two categories, RR for the most rural area versus the most urban area was used. In cases of multiple RRs for the same study outcome, an RR for the largest age group and higher level of severity was chosen for the forest plot; otherwise, multiple results are descriptively presented and not included in the forest plot. Effects of several factors (severity of injury, age and sex of participants, urban/rural categorisation methods, a level of data collection (ie, national or local), etc) on RRs were explored qualitatively. A test for publication bias was not performed because of an absence of SE in most studies as well as there being few studies for each injury outcome. Forest plots were drawn using RevMan software (version 5.1 for windows; The Cochrane IMS).

## Results

### Description of included studies

The flow of study retrieval and selection is shown in figure 1. Overall, 2922 reports were identified from literature searches. The full texts of 235 reports were examined and, of these, 189 studies were excluded. A list of excluded studies is available by contacting the first author. A total of 41 unique studies (46 reports) were included for the review, all of which were population-based cross-sectional studies (seven surveys19–25 and 34 studies using administrative health database). Three studies4 ,24 ,26 were reports from government agencies and the others were articles published in peer-review journals. The majority of studies (34 studies) were conducted for only a paediatric population, and seven studies4 ,17 ,24 ,27–30 were performed for all age groups providing subgroup data for the paediatric population. The majority of studies (30 studies) were carried out in the USA and 11 studies4 ,5 ,20 ,23 ,26 ,31–36 were undertaken in Canada. Eleven studies4 ,5 ,19–21 ,24 ,27 ,34 ,37–39 used national-level data, and 30 studies were conducted at the local level (28 studies at the state or province level, two studies23 ,40 at municipal level). Rural/urban areas were classified by participants' residential area (31 studies), school areas,20 ,22 ,25 location of injury occurrence (on MVC injury28 ,33 ,36 ,41 and on firearm injury37 ,42) or location of hospitals (sport-related injury39). Definitions and categories of rural/urban areas varied across studies: the number of rural/urban categories included two categories (26 studies), three–five categories4 ,17 ,20 ,21 ,23 ,25–27 ,29 ,34 ,38 ,42–45 and 10 categories37 (King et al43 applied two categories as well as four categories). Most of the study participants were children under 17–20 years of age (21 studies), children under 15,26 ,27 ,29 ,30 ,36 ,40 ,41 ,45–47 mostly teenagers,20–23 ,25 ,48 young children43 ,49 and children excluding young children.34 ,39 Key features of included studies were summarised in Web Only file #2.

Figure 1

Flow of study retrieval and selection.

### Methodological quality of included studies

The internal validity of studies was moderate. In total, 95% of studies (39 out of 41 studies) received four or more stars out of a possible six on the NOS. Stars given to studies using administrative databases included full stars,37 ,39 ,50 five stars4 ,26 ,30 ,31 ,33 ,36 and four stars (25 studies). Three surveys21 ,23 ,25 obtained five stars and the other four surveys19 ,20 ,22 ,24 were rated at three stars. The main reasons of lower star ranking were a lack of control for potential confounders and outcome assessment by self-report (Web Only file #3).

### Primary outcome (injury incidence)

Forty one studies compared injury incidence rates between rural and urban children. For effect measures, 32 out of 41 studies used unadjusted RRs, and the other studies reported adjusted RRs,33 ,37 ,50 adjusted ORs21 ,23 ,25 ,39 or standardised mortality ratios.4 ,30 Studies19–25 using survey methodology investigated only one category of injury, while studies with administrative data reported on one or several causes of injury within a single study.

### Overall injury

Twelve studies reported rates of overall injury (any cause/intent) varying in severity from any injury needing medical attention to fatal injury. Six studies4 ,19 ,20 ,24 ,27 ,38 using national-level data consistently demonstrated higher risk of overall injury for rural children in comparison with urban children ranging from RR (rural vs urban)=1.04 to 3.53, whereas five studies31 ,32 ,44 ,47 ,50 with state or province-level data presented variations in terms of effect size/direction ranging from RR=0.54 to 2.20 (figure 2). One study18 conducted in Ohio reported that rural children (mostly Caucasian) experienced higher rates of fatal injury than urban Caucasian children, but lower than urban non-Caucasian children (mostly African American). Two studies examined traumatic brain injury. Reid et al51 reported higher risk of death for rural children (RR=2.37), and Gabella et al17 also found higher incidence rates for rural children in all ages groups and both genders, with the exception of the young boys' group (<5 years).

Figure 2

Rate ratios of any cause/intent of injury for rural children to urban children.

### Injury by intent

#### Unintentional injury

Two studies26 ,46 addressed unintentional injury demonstrating consistently higher injury rates for rural to urban children ranging from RR=1.30 to 1.45.

#### Intentional injury

Eleven studies4 ,21 ,29 ,35 ,44 ,46 ,48–50 ,52 ,53 reported on intentional injury and the main results are illustrated by subgroup (figure 3). As with overall intentional injury, Hammig and Weatherly46 described lower risk of injury for rural children (RR=0.5). For suicides/self-harm, four studies4 ,31 ,46 ,50 consistently demonstrated higher rates for rural children with RR ranging from 1.22 to 2.70. Thompson35 reported a higher suicide rate (RR=1.86) for rural boys to urban ones but no differences among girls. Regarding homicides/assaults, there was inconsistency in results among six studies29 ,31 ,44 ,46 ,48 ,50 ranging from RR=0.05 to 1.81. Child abuse was reported in three studies49 ,50 ,53 with inconsistency in results ranging from RR=0.56 to 1.82. Schnitzer et al49 also presented conflicting results of non-fatal child abuse by data source in the study with lower risk for rural children (RR=0.56) from medical data as well as inverse results (RR=1.49) from family services data. For fight-related injuries, one study21 reported a lower rate of injury for rural children with an adjusted OR of 0.78.

Figure 3

Rate ratios of intentional injury for rural children to urban children.

### Injury by cause

#### MVC injury

Injuries from traffic collisions were examined in 13 studies,4 ,5 ,28 ,30 ,31 ,33 ,34 ,36 ,41 ,43 ,44 ,46 ,50 all of which were based on administrative health records with severe injuries (deaths or hospitalisations). All studies on overall MVC injuries or ones as occupants/passengers/pedal cyclists consistently demonstrated higher risk of injury for rural than urban children. Relative risks (rural to urban) of overall MVC injuries, ones as occupants, as passengers and as pedal cyclists ranged from 1.40 to 2.75,4 ,5 ,31 ,46 from 1.20 to 5.40,5 ,33 ,41 ,44 ,46 ,50 1.9743 and from 1.06 to 2.30,34 ,46 respectively (figure 4). In addition to results in figure 4, for overall MVC injury, standardised mortality ratios were reported at >100 for rural children (137.62 in rural vs 73.87 in urban children).30 Also, for paediatric occupant injury,36 the same direction of effect but an extreme effect size for male subjects (RR=11.58) was reported. On the other hand, for MVC injuries as pedestrians, there was inconsistency in the direction of effects across three studies (two studies46 ,50 in figure 4 and one study28 reported RR of 1.0–1.5 by age subgroup). Oliver and Kohen5 reported RR of 0.90 for MVC injuries as pedestrian- and cyclist-related combined.

Figure 4

Rate ratios of motor vehicle crash-related injury for rural children to urban children.

#### Firearm-related injury

Seven studies (eight reports) in the USA37 ,42 ,45 ,50 ,54–56 using administrative health databases reported on firearm-related injuries with severe outcomes (deaths or hospitalisations) (figure 5). Overall, for firearm-related injuries, contrary results were found across three studies.37 ,42 ,57 Homicides using firearms were examined in five studies37 ,42 ,45 ,56 ,57 showing consistently lower rates for rural children with RR ranging from 0.02 to 0.42. As for unintentional firearm-related injuries, despite inconsistency in results among six studies, four37 ,45 ,55 ,57 out of six studies37 ,42 ,45 ,54 ,55 ,57 presented higher risk for rural children. For suicides, there were conflicting results from four studies.37 ,42 ,45 ,50

Figure 5

Rate ratios of firearm-related injury for rural children to urban children.

#### Other causes of injury

Nine studies reported on several other causes of injury (figure 6). Drowning,46 ,50 burns,50 falls,31 ,50 poisoning34 ,50 and bicycle-related injury34 ,50 were described with higher risk for rural children, while risk of falls from windows40 was lower for rural children than urban children, although these are based on just one or two studies for each cause of injury. There were contrary results regarding work-related injury22 ,25 and sports-related injury,23 ,58 and no risk differences for suffocation were found from one study.46

Figure 6

Rate ratios of other cause of injury for rural children to urban children.

### Secondary outcomes (length of hospital stay, hospital costs, etc)

Three studies19 ,38 ,39 investigated secondary outcomes of injury. Two national surveys in the USA reported that healthcare costs for any injuries per injured child were higher for rural children compared with urban ones: for total cost $1200 for metropolitan children versus$1800 for non-metropolitan children19; $608 for urban children versus$661 for rural children in terms of emergency department expenditures.38 Yang et al39 examined secondary outcomes for sports injury, reporting 46.1% higher hospital charges per discharge and 5.61% longer LOS in urban hospitals than rural ones.

### Effect of other factors on injury differences

#### Severity of injury

Eleven studies reported multiple data for the same outcome by severity of injury within a study (Web Only file #2). Six19 ,20 ,22 ,33 ,36 ,51 out of eight studies19 ,20 ,22 ,26 ,31 ,33 ,36 ,51 that reported higher risk of injury for rural children consistently demonstrated increased RRs for more severe injuries and deaths. Three studies23 ,44 ,56 reporting lower risk for rural children showed inconsistency in results in terms of effects of severity on injury disparity.

#### Age of participants

Eight studies4 ,17 ,27 ,28 ,32 ,41 ,55 ,57 presented age-specific injury rates. Effect of age subgroups on RRs between urban and rural children were not consistent across other variables within a study or across studies for overall injury and MVC injury, but one study57 presented a positive linear relationship between age and RR (rural to urban) of firearm-related deaths (Web Only file #2).

#### Sex of participants

Seven studies17 ,18 ,33 ,36 ,47 ,57 reported sex-specific injury rates. Sex effects on RRs were not consistent across age groups within a study or across studies for any injury.4 ,17 ,18 ,47 However, relative risks (rural to urban) were higher for male than female subjects in three33 ,36 ,47 out of four studies4 ,33 ,36 ,47 on MVC injury and in one study on firearm-related injury57 (Web Only file #2).

#### Urban/rural categorisation

Fifteen studies4 ,17 ,20 ,21 ,23 ,25–27 ,29 ,34 ,37 ,38 ,42–44 provided rural–urban category-specific injury rates for more than two urban–rural categories (Web Only file #2). All except two studies23 ,29 demonstrated that rural–urban disparity was generally or consistently intensified when comparing more extreme rural and urban areas. For example, if RR of rural to urban children was larger than 1 (higher risk for rural children), the RR increased as the urban group was compared with a more distant rural group and vice versa.

## Discussion

### Summary of evidence

Injury is a major public health issue for paediatric populations.1 ,2 This systematic review summarises the available evidence regarding differences in risks of injury between rural and urban children in a qualitative fashion. A total of 41 population-based cross-sectional studies conducted in Canada and the USA were included for the review. Overall, our results show that rural children are at higher risk of overall injury, MVC injury and suicide, whereas urban children in the USA experience higher rates of homicide from firearms. Rural–urban disparities for injury tend to be intensified with injury severity (for injuries that rural children are at higher risk for) as well as when comparing more extreme rural and urban regions. Healthcare costs per child for overall injuries are higher for rural children. For other categories of injury, there was a lack of consistency in the results or an insufficient number of studies to review.

### Interpretation

Higher risk of overall injury for rural children can be explained in multiple aspects. First, it may be attributed to the greater distribution of populations at high risk of injury (eg, farm children59 ,60 or aboriginal children)61 and hazardous environmental conditions related with farming, mining, fishing and forestry62 in rural versus urban areas. Second, the disparity is more likely affected by apparent higher rates of MVC injury for rural children, which can be linked to several factors including higher driving speed in rural areas,63 more frequent alcohol use,64 lower restraint usage rates for rural children65 and delayed access to paediatric trauma care in rural areas.66 ,67 Third, the rural–urban injury disparity may be influenced by potential confounding factors including socioeconomic status and the medical environment that were not controlled for in most of the included studies. Lower income and education level of rural people10 ,12 ,68 as risk factors of injury32 may cause overestimation of relative risks of injury for rural versus urban children. Poorer access to health service in rural areas10 ,12 may confound the injury disparity in several ways: a possibility of underestimating minor injuries in health databases69 potentially due to distances required to receive care, an underestimation of hospitalised injuries because of greater prehospital deaths66 ,67 and transfers to urban hospitals,70 and increased cases of death due to delay in medical care.71

As for other injuries, studies were consistent in reporting higher risk of unintentional injury, traumatic brain injury, drowning, falls and bicycle-related injuries as well as those involving MVCs for rural children, although these were examined by only two studies for each injury. Regarding firearm-related injury, there were different patterns by intent of injury: higher rates of firearm homicides for urban children versus a tendency of higher rates of unintentional injuries for rural children, which is consistent with findings from previous studies that were conducted for all ages.72 ,73 One thing to be advised of is that all of seven studies of firearm-related injury were conducted in US populations. Results on firearm-related injuries and intentional injuries in the USA may not be generalised to Canadian populations when reflecting on differences in firearm control regulations between the two countries. Interpretation of child abuse injury findings require caution due to potential under-ascertainment of cases in hospital/ED discharge data49 as well as limited search of social care databases in this review. Longer LOS and higher hospital charge of urban hospitals for sport-related injury may partly be attributed by transfers of rural patients with severe injuries to urban hospitals.70

### Investigation of heterogeneity

To investigate heterogeneity, effects of some factors on rural–urban differences of injury were examined. First, rural–urban injury disparity tended to increase for more severe injuries, in cases of there being higher risk of injury for rural to urban children. This trend can be explained by more severe injuries for rural children,74 greater possibility of under-reporting of rural minor injuries69 and higher death rates due to delay in medical care.71 Underestimation of rural minor injuries may be more notable in the USA where rural people have lower coverage of health insurance unlike Canada's universal coverage.12

Second, greater rural–urban injury disparities were likely to be found between more extreme rural and urban areas indicating a linear relationship between levels of rurality and injury rates. Third, national-level data were more consistent across studies in general, while state or province-level data showed wider variations in effect size/direction as shown in the figures. This implies more stability of national-level data as well as the need to consider local-level data when prevention strategies target local regions. Last, the effects of age and sex on rural–urban injury differences were not consistent across studies except effects of sex on MVC injury (ie, greater RRs for male than female subjects) and combined effects on firearm injury (ie, greater RRs for older male subjects). Although injury death rates for children have declined since the late 1970s,15 whether the declines were comparable between rural and urban children over time was difficult to examine.

### Strengths and limitations

This systematic review was methodologically rigorous with a protocol and contained a comprehensive search strategy to discover the body of evidence of this topic. This review covers every cause/intent of injury outcome, providing a full spectrum of injury disparity patterns between rural and urban children. We included only population-based studies that provided injury rates so that the results from this review may be generalisable to paediatric populations in Canada or the USA.

There are some methodological limitations of the primary studies included. The risk of bias of primary studies may limit the validity of the results as the majority of the studies applied descriptive rather than analytical statistics by calculating injury rates without controlling for potential confounders. Descriptive statistics may be meaningful to show the real trends of incidence; however, they may not be sufficient to examine the effect of residential area on injury incidence due to potential confounding. In addition, some studies on non-fatal injuries made unit-of-analysis errors75 in calculation of injury rates by not adequately considering multiple injury events for an individual (ie, not using person-time at risk as a denominator). Although this can result in miscalculation of injury rates, it does not affect RR due to a ‘nullifying’ of denominators in calculations.

There was substantial heterogeneity among the included studies that originated from the clinical diversity (eg, cause/intent of injury, severity of injury, and characteristics of participants including age, race, and study setting) as well as from methodological variations (eg, study design, data sources, urban/rural classification/definition, and statistical methods). The large heterogeneity precluded a statistical pooling of effect estimates and hindered valid comparisons of results across the studies.

Most of the included studies classified rural–urban status by residential area, but several studies on MVC injuries and firearm injuries classified by a place of injury occurrence. Findings about MVC injury and firearm injury may be biased as we combined studies that applied the different rural–urban classification methods. However, this bias may not be significant as prior studies reported that place of fatal crash primarily matches with driver's residential areas76 and that for the majority of firearm deaths they occurred within a county of residence.37

Although the Newcastle-Ottawa quality assessment scale is recommended for a systematic review for non-randomised studies,75 ,77 ,78 reliability or validity of the tool was not provided. Further, we had to modify this scale without verification due to inapplicability to injury studies. A quality assessment tool considering unique characteristics of injury outcome should be developed in future studies. Even though this review was comprehensive, publication bias and selection bias (eg, not including publications in French language) may be possible.

### Implications for research and practice

To be more rigorous for future systematic reviews, primary studies on this topic are recommended to use a standard rural–urban definition to reduce heterogeneity among studies, to report injury rates with proper denominators, to perform analytical statistics controlling for potential confounders and offering SE of measure of comparison for meta-analysis, and to provide data with basic figures of variable-adjusted as well as variable-specific injury rates (eg, injury rates for age subgroups).

More efforts should be made to reduce the greater burden of overall injury for rural children, especially, of severe injury for children in remote rural areas. Interventions to reduce the higher risk of MVC injury for rural children, especially male subjects, are in need through increased use of child restraints, intensified driver-safety regulations and education. The clear pattern of higher suicide rates for rural children is alarming, and identification of risk factors focusing on this suicide disparity is required in future studies. Stricter firearm controls may help to decrease firearm-related injury for paediatric populations, in particular, older paediatric male subjects in the urban USA. Finally, more studies are needed for a clear understanding of geographical disparity on minor injuries such as burns, drowning, and poisoning and other health-related outcomes (eg, disability) that are likely to show rural–urban injury differences.

## Conclusions

There were rural–urban differences demonstrated in rates and patterns of injury in children based on 41 studies having a moderate risk of bias. Rural children sustain a higher rate of overall injury, particularly from MVC and suicide, while urban children in the USA suffer from a higher rate of firearm-related homicides. Greater injury disparities tend to be found between more extreme rural and urban regions. In particular, children in remote rural areas are at increased risk of severe injuries than urban counterparts, possibly due to more hazardous environment, lower socioeconomic status and delayed access to paediatric trauma care in remote rural areas. These findings indicate the need for developing preventive strategies specific to each setting. Design changes in primary studies could also increase the possible utility of future systematic reviews. Future research is required to investigate rural–urban disparity for less-studied injuries and related health outcomes as well as temporal trends of identified disparities.

### What is already known on the subject

• Injury is the leading cause of death for children.

• Although substantial studies have demonstrated injury differences for rural and urban children, this evidence has not been synthesised or systematically reviewed.

• Rural children are at higher risk of overall injury, motor vehicle crash injury and suicide, whereas urban children in the USA experience higher rates of firearm-related homicides.

• Rural–urban disparities for injury tend to be intensified when comparing more extreme rural and urban areas; children in remote rural areas are at increased risk of severe injuries than urban counterparts.

• There is the need of developing injury prevention strategies specific to rural and urban setting as well as local settings.

• ## 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.

Files in this Data Supplement:

## Footnotes

• Funding This research is supported by the Public Health and the Agricultural Rural Ecosystem (PHARE) and Canadian Institutes of Health Research and the Canadian Agricultural Safety Association (CASA) through the Alberta Centre for Injury Control and Research (ACICR).

• Competing interests None.

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