Injury, Int. J. Care Injured 45 (2014) 23–30

Epidemiology of trauma: A population-based study of geographicalrisk factors for injury deaths in the working-age population of Norway

Thomas Kristiansen a,b,c,*, Hans Morten Lossius a,d, Marius Rehn a,d,e, Petter Kristensen f,g,Hans Magne Gravseth f, Jo Røislien a,h, Kjetil Søreide i,j

a Department of Research and Development, The Norwegian Air Ambulance Foundation, PO Box 94, N-1441 Drøbak, Norwayb University of Oslo, Faculty Division Oslo University Hospital, Kirkeveien 166, N-0450 Oslo, Norwayc Diakonhjemmet Hospital, Department of Anaesthesiology, PO Box 23 Vinderen, N-0319 Oslo, Norwayd Field of Prehospital Critical Care, Network of Medical Sciences, University of Stavanger, Kjell Arholmsgate 41, N-4036 Stavanger, Norwaye Akershus University Hospital, Sykehusveien 25, N-1474 Nordbyhagen, Norwayf National Institute of Occupational Health, PO Box 8149 Dep., N-0033 Oslo, Norwayg Institute of Health and Society, University of Oslo, PO Box 1130 Blindern, N-0318 Oslo, Norwayh Department of Biostatistics, University of Oslo, PO Box 1122 Blindern, N-0318 Oslo, Norwayi Stavanger University Hospital, Department of Surgery, PO Box 8199, N-4068 Stavanger, Norwayj University of Stavanger, Institute of Health and Medicine, Rennebergstien 30, N-4021 Stavanger, Norway

A R T I C L E I N F O

Article history:

Accepted 6 July 2013

Keywords:

Trauma

Injury

Population-based

Epidemiology

Rural health

Norway

A B S T R A C T

Introduction: Trauma is a major global cause of morbidity and mortality. Population-based studies

identifying high-risk populations and regions may facilitate primary prevention and the development of

optimal trauma systems. This study describes the epidemiology of adult trauma deaths in Norway and

identifies high-risk areas by assessing different geographical measures of rurality.

Methods: All trauma-related deaths in Norway from 1998 to 2007 among individuals aged 16–66 years

were identified by accessing national registries. Mortality data were analysed by linkage to population

and geographical data at municipal, county and national levels. Three measures of rurality (centrality,

population density and settlement density) were compared based on their association with trauma

mortality rates.

Results: The study included 8466 deaths, of which 78% were males. The national annual trauma

mortality rate was 28.7 per 100,000. Population density was the best predictor of high-risk areas, and

there was a consistent inverse relationship between mortality rates and population density. The most

rural areas had 52% higher trauma mortality rates compared to the most urban areas. This difference was

largely due to deaths following transport-related injury. Seventy-eight per cent of all deaths occurred in

the prehospital phase. Rural areas and death following self-harm had higher proportion of prehospital

deaths.

Conclusion: Rural areas, as defined by population density, are at a higher risk of deaths following

traumatic injuries and have higher proportions of prehospital deaths and deaths following transport-

related injuries. The heterogeneous characteristics of trauma populations with respect to geography and

mode of injury should be recognised in the planning of preventive strategies and in the organisation of

trauma care.

� 2013 Elsevier Ltd. All rights reserved.

Contents lists available at SciVerse ScienceDirect

Injury

jo ur n al ho m epag e: ww w.els evier . c om / lo cat e/ in ju r y

Introduction

Trauma is a leading global cause of death and disability [1].Traumatic injuries largely affect young people, which increases the

* Corresponding author at: Norwegian Air Ambulance Foundation, PO Box 94,

1441 Drøbak, Norway. Tel.: +47 64 90 44 44; fax: +47 64 90 44 45.

E-mail addresses: thomas.kristiansen@norskluftambulanse.no,

thomas.kristiansen@snla.no (T. Kristiansen).

0020–1383/$ – see front matter � 2013 Elsevier Ltd. All rights reserved.

http://dx.doi.org/10.1016/j.injury.2013.07.007

societal burden of trauma [2]. Primary prevention has reducedmortality from trauma through the identification of effectiveprotective measures and the reduction in the incidence oftraumatic events [3]. The organisation of trauma systems [4,5]has become a major secondary preventive effort to reduce theimpact of traumatic injuries in many regions [6–11].

Population-based studies report higher trauma mortality ratesin rural compared with urban areas [12–15], with a majority ofdeaths occurring in the prehospital phase [16–19]. Despite this, themajority of trauma research has focused on hospital management

T. Kristiansen et al. / Injury, Int. J. Care Injured 45 (2014) 23–3024

within urban trauma systems. The benefit for rural areas of thecentralisation of trauma management is debated [18,20–22], andstudies indicate that rural populations are less likely to reachcentralised services [16,23].

Although a national trauma system has yet to be implementedin Norway, there is a tendency to centralise trauma care [24,25],increasing the distances to definite care for rural areas. Optimalinitial stabilisation by the emergency medical service (EMS) andlocal hospitals and efficient interhospital transfers are suggestedresources for rural trauma care [16,26,27]. Targeted primaryprevention and optimisation of trauma systems necessitate anincreased understanding of etiological, demographical and geo-graphical risk factors [3,28,29]. To identify high-risk areas,validated measures of rurality are needed. Several operationaldefinitions of rurality exist in trauma research. A Norwegian studyidentified large urban-rural differences in paediatric traumamortality using both municipal centrality and population densityas measures of rurality [30].

There is a scarcity of population-based trauma research, andepidemiological studies based on hospital data may underestimatethe true contribution of fatal trauma [19,31]. Differences inmortality rates and proportions of prehospital deaths in differentregions should be thoroughly described to ensure that both theEMS and hospital trauma resources are scaled in accordance withthe needs of each geographic area, while validated measures forrural trauma may serve as generalisable tools to identify high-riskareas.

The aim of this study is to describe the national epidemiologyand urban-rural distribution of fatal trauma. We analyse traumamortality rates and the proportion of prehospital deaths in ruralversus urban areas and explore which geographical measure bestidentifies the increased risk associated with rural trauma.

The study was approved by the Regional Committee for MedicalResearch Ethics (Ref. No. 2010/125) and informed consent fromwas not considered appropriate or necessary due the anonymousnature of the data and the results. The study is in compliance withthe Declaration of Helsinki of the World Medical Association [32].

Materials and methods

Data collection and study participants

This study is based on data from the Cause of Death Registry,Statistics Norway [33], which maintains consecutively collecteddata on all national deaths. Registry data sources include medicaldeath certificates, police reports and autopsy reports. Data qualityis ensured by follow-up inquiries on incomplete death certificatesand by linkage to other central health registries, StatisticsNorway’s transport accident registry and the population registryof the Norwegian Tax Administration [34]. Information on deathsfrom external causes (ICD-10 codes V01 to Y89) in a ten-yearperiod (1998-2007) was extracted. Deaths due to medical oriatrogenic causes (ICD-10 codes: X20 to X29, X40 to X49, X50 toX57, X60 to X69, X85 to X90, Y06, Y10 to Y19, Y40 to Y84, and Y88)were excluded.

A related article on paediatric (0-15 years of age) traumaepidemiology has previously been published [30]. The standardage of retirement in Norway is 67 years. We aimed to study traumamortality in the adult working-age population and thus includeddeaths in individuals aged 16–66 years of age.

Index year and setting

The number of municipalities was reduced during the studyperiod, necessitating a reference municipality-structure for data-base construction and aggregated analysis. The year 2002 was

chosen as an index year central to the study period (1998-2007)and served as a fixed year for descriptive purposes. Variablesregarding municipal characteristics and populations were thusbased on data from 2002.

In 2002, the population of Norway was 4.52 million, 2.95million of whom were between 16 and 66 years of age. During thestudy period, there was a 0.62% annual population growth, and themean total population during the period was 4.54 million (95%confidence interval (CI): 4.48, 4.60). The Statistics Norway’sStatBank [35] was used to ensure that the index year was themost representative year, and within the 95% CI of the mean, forthe entire study period; this was also ensured for rural and urbanpopulations when they were assessed separately.

Norway was divided into 19 counties and 434 municipalities,and the population density was 14.0 inhabitants per squarekilometre (inh/km2), in the index year. The county populationdensity ranged from 1.5 to 1129.5 inh/km2 [36].

Norwegian EMS is publically funded and consists of paramedic-manned ground ambulances and nationwide coverage of ananaesthesiologist-manned helicopter ambulance service (HEMS)[25,37]. In 2002, 51 hospitals were receiving trauma patients, andthe majority had multidisciplinary trauma teams in theiremergency departments (ED) [38,39]; however, there was noformal designation of trauma centres [25].

Variables and definitions

The mode of injury was divided into four categories based onICD-10 codes: transport (V01-V99), falls (W00- W19), self-harm(X70-X84), and assault (X91-Y09).

Data regarding place of death is collected from the deathcertificate and categorised in the Cause of Death Registry accordingto type of location. In-hospital deaths are defined as deaths thatoccurred within hospitals or other health care institutions. Deathsthat occurred outside hospitals or during transport to a hospital areconsidered prehospital deaths. The database does not containgeographical data locating the site of injury. Using place of death tolocate events, introduces bias in cases where deaths occur after orduring transport from the site of injury; e.g., overestimating risk forareas surrounding acute care hospitals. Place of residency wastherefore used to locate the events of the study.

Three different geographical measures of municipal rurality wereextracted from the Norwegian Social Science Data Services (NSD)data base [36]: centrality, population density, and settlement density.

Centrality describes the number of inhabitants in towns or othergeographically bounded areas within, or in the proximity of, amunicipality. The variable is based on the municipality classificationstandard defined by Statistics Norway for the index year [40]. Thevariable was aggregated to four levels. The most rural categoryincludes municipalities without, or with more than 45 min by roadto, areas with more than 5000 inh. The Rural municipalities include,or are within 45 min by road to, areas with 5-15,000 inh. Central

municipalities include, or are within 60 min by road to, areas with15-50,000 inh, and the most central municipalities include, or arewithin 75 min by road to, areas of more than 50,000 inh and/ortowns with a regionalised role for public services.

Population density was defined as the number of inhabitants persquare kilometre based on the index year population from the NSDdatabase [36]. The variable was categorised according to quartiles ofpopulation density when weighted by the total population asfollows: most rural: <18.2 inh/km2, rural: 18.2–76.9 inh/km2, central:77.0–442.7 inh/km2, and most central: >442.7 inh/km2. Populationdensity was categorised according to population quartiles tocompare groups of equal proportions of the total population.

Settlement density is based on Statistics Norway’s municipalityclassification standard for the index year and describes the

T. Kristiansen et al. / Injury, Int. J. Care Injured 45 (2014) 23–30 25

percentage of inhabitants within a municipality that live in densesettlements [40], which are defined as areas with >200 inhabitantswith an inter-residential distance of �50 m. The variable wasstratified into five levels according to quintiles living in densesettlements.

Statistical analysis

Continuous data are presented as medians with interquartilerange (IQR), except for municipal, county and national averages,which are presented as means weighted by the population in therespective areas.

Mortality rates were calculated by linking mortality data topopulation data from the NSD database [36]. Rates per 100,000 inhper year for municipalities, counties and the nation, werecalculated using the number of deaths relative to the number ofinhabitants aged 16–66 years in the respective areas for the indexyear. Confidence intervals for rates were calculated as Agresti-Coull intervals for binomial proportions [41].

Associations between rurality measures and mortality rateswere assessed by univariate and multivariable linear regression.The rurality measures were included as categorical explanatoryvariables, with the most central category serving as the reference.For the multivariable regression analysis, possible multicolli-nearity was assessed by the calculation of variance inflationfactors (VIF), which did not exceed 4.5 for any of the includedexplanatory variables.

Two-tailed tests were used, and statistical significance wasassumed for P-values <0.05. SPSS v.19 (IBM, Chicago, IL) and thestatistical software R 2.12 [42] was used for analysis.

Results

A total of 8466 deaths were included in the study, yielding anational annual trauma mortality rate of 28.7 per 100,000 inh.

Age

444240383634323028262422201816

Rat

e/10

0 00

0/Y

ear

80

60

40

20

0

Fig. 1. Age-specific annual trauma mortality rate per 10

Males accounted for 6592 (78%) of the deaths, and theannual male and female mortality rates were 44.0 and 12.9per 100,000, respectively. The study included the ages 16–66years, and the median (IQR) age at the time of death was 40(27, 53) years. For men, there was a sharp increase in mortalityrates from 16 to 20 years, with an early adulthood peak of 68.7deaths per 100,000 per year at the age of 20 (Fig. 1). Forwomen, this peak was reached at age 18, with 19.2 deaths per100,000 per year (Fig. 1). Mortality rates increased in the sixthdecade of life, and in the seventh decade (ages 60–66 years),injury mortality rates reached levels equal to the earlyadulthood peaks (Fig. 1).

Rural vs. urban areas

The median (IQR) annual county mortality rate was 30.2 per100,000 (26.8, 33.3), and the county rates ranged from 23.0 inAkershus to 44.7 in Finnmark, the latter being the most rural andnorthern county (Fig. 2).

The municipal mortality rates increased with each ruralitylevel for all three geographical measures, and all measures weresignificantly associated with higher mortality rates in theunivariate linear regression models (Table 1). For centralityand population density there was a marked increase from therural to the most rural categories (Table 1). The squaredcorrelation (R2) between mortality rates and the explanatoryvariables in the univariate models were .205, .185 and .135 forpopulation density, centrality and settlement density, respec-tively. In the multivariable regression model, population densitywas the only measure that remained significantly associatedwith higher mortality rates for all levels (Table 1). Usingpopulation density as the measure for rurality, there was anincrease in the annual mortality rate from 23.3 per 100,000 inthe most central to 35.3 per 100,000 in the most rural

municipalities (Table 1).

6664626058565452504846

FemaleMale

0,000 Norwegian citizens in the period 1998–2007.

Fig. 2. Map of Norway depicting counties in shades of grey according to the adult trauma mortality rate per 100,000 per year. The reference bar indicate the mortality rate

values according to the grey scale.

T. Kristiansen et al. / Injury, Int. J. Care Injured 45 (2014) 23–3026

Mode of injury

The transport-related mortality rate was 8.6 per 100,000 peryear while the annual self-harm mortality rate was 11.3 per100,000 (Table 2). The mode of injury varied significantly withdifferent levels of population density (Table 2). The highestnumber of deaths in the most rural municipalities was found in thetransport category, while in the more urban areas, self-harmcaused the highest numbers of deaths. The transport injurymortality rate increased significantly with each category of rurality

and was almost three times higher in the most rural compared withthe most central municipalities. In the most central municipalities,self-harm caused more than twice the proportion of deathscompared to transport injuries. However, the self-harm mortalityrates still tended to be higher in the rural than in the centralmunicipalities. The assault mortality rates were significantlyhigher in the most central municipalities compared with the rural

and most rural municipalities (Table 2).The median age at the time of death varied according to the

mode of injury, with fall injuries occurring at a higher median (IQR)

Table 1Three measures of municipality rurality: centrality, population density and settlement density in linear regression analysis with mortality rates per 100,000 inhabitants per

year, weighted by study population. ‘‘Most Central’’ is the reference category in each variable.

Univariate linear regression models Multivariable linear regression modela

Rate Rate difference 95% CI P Adjusted rate difference 95% CI P

CentralityMost central 25.7

Central 29.3 3.58 (1.61; 5.55) <.001 1.46 (�0.71; 3.62) .187

Rural 32.7 6.98 (3.76; 10.20) <.001 1.77 (�2.07; 5.61) .365

Most rural 37.6 11.93 (9.44; 14.42) <.001 6.56 (3.11; 10.02) <.001

Population densityMost central 23.3

Central 27.0 3.65 (1.39; 5.90) .002 2.73 (0.28; 5.17) .029

Rural 29.3 5.99 (3.71; 8.28) <.001 3.82 (0.72; 6.92) .016

Most rural 35.3 11.98 (9.68; 14.28) <.001 7.16 (3.26; 11.06) <.001

Settlement densityMost central 25.6

Central 31.0 5.45 (3.24; 7.65) <.001 1.85 (�0.75; 4.45) .164

Intermediate 32.8 7.22 (4.64; 9.80) <.001 0.99 (�2.19; 4.18) .541

Rural 33.5 7.97 (4.85; 11.09) <.001 0.68 (�3.04; 4.39) .721

Most rural 34.8 9.27 (5.16; 13.37) <.001 1.11 (�3.48; 5.71) .634

a All three measures of rurality were included in the model.

T. Kristiansen et al. / Injury, Int. J. Care Injured 45 (2014) 23–30 27

age: 54 (43, 61) years, compared to 39 (27, 52) years for theremaining groups.

Prehospital deaths

Seventy-eight per cent of deaths occurred outside of hospitals.The proportion of prehospital deaths decreased for each level ofpopulation density (Table 2). Finnmark had a significantly higherproportion of prehospital deaths (86%) than the remainingcounties (P = .008). The proportion of prehospital deaths variedaccording to the mode of injury; 89% of deaths following self-harmoccurred in the prehospital phase, while 75 and 51% of deathsassociated with transport injuries and falls, respectively, occurredbefore reaching hospital.

Discussion

This population-based study assessed the deaths of patientsmanaged, or potentially managed, within a trauma system anddemonstrated an increased trauma related mortality rate in ruralareas; this difference was largely due to transport-related injuries.The proportion of prehospital deaths was high overall andincreased with rurality. Defining rural areas according to popula-tion density best predicted high-risk areas and also revealed anincremental relationship between rurality and risk.

The comparison of mortality rates between different regionsand different studies is limited by the non-uniformity of inclusioncriteria, definitions and level of study details available, and theneed for more uniform studies on trauma epidemiology is evident.An Australian study, using similar inclusion criteria, reportedcomparable mortality rates, except for the most remote areas [14].A US publication, based on national central registry data, includedall external causes of deaths and all age groups and reported anannual mortality rate of 55.7 per 100,000 [43]. After excluding thedeaths due to poisonings, which accounted for 16% of deaths in theUS study, the injury mortality rate for the comparable age groupstill ranged from 30% to 100% higher than that in our study;transport, fall and assault rates were higher in the US study, whileself-harm rates were more comparable. Our mortality rates arealso comparable to a previous Dutch national study [44], althoughthis study excluded all intentional injuries. Similarly, a study froma northeastern region of Italy reported comparable rates, but self-harm injuries were excluded [45]. Considering the relative

contribution of self-harm injuries to the mortality rate in ourstudy, these comparisons indicate a substantially lower non-self-harm rate in our results. In Canada, a group of researchers found alower overall mortality rate in a rural area compared with ourresults, but this study also excluded many of the injury typesincluded in our study [16]. Our reported mortality rates are lowerthan the global injury mortality rates estimated by the worldhealth organisation (WHO) [46]. For persons in high-incomecountries between the ages of 15 and 59 years, it is estimated thatroad traffic and self-inflicted injury mortality rates are 15.0 and14.9 per 100,000 per year, respectively. Our results for transportand self-harm injuries are approximately 43 and 24% lower thanthe WHO estimates, respectively [46]. However, the differences inage groups (16-66 vs. 15-59 years of age) and definitions arelimitation when comparing these results.

A Norwegian regional study reported a much lower traumamortality rate, but this study excluded several of the injurymechanisms included in our study (e.g., burns, hanging anddrowning) and included all age groups [47]. Our reported earlyadulthood peaks for men and women confirm results from anotherNorwegian study that analysed road traffic deaths in the 16- to 20-year age group [48]. The mortality rate increase described in thesixth and seventh decade of life, combined with the increased ageand the higher proportion of in-hospital deaths after falls,corresponds well to epidemiological patterns described previouslyfor trauma in older age groups [14,43,47,49].

For studies on rural trauma, a wide variety of rurality measuresare used [14,15,17,19,50], including population density [12,51]. Arelated previous study on paediatric trauma mortality assessed thesame rurality measures as this study and found that centrality bestpredicted the county mortality rates for children [30]. Our studydid not confirm this for adult trauma, but centrality was found tobe significantly associated with mortality rates for all ruralitylevels in the univariate regression model. In the multivariableregression analysis, the most rural level of centrality remainedsignificantly associated with trauma mortality, indicating poten-tial additional precision when combining this measure withpopulation density.

Mortality rates for the most rural areas were 52% higher thanthose of the most urban areas. The excess mortality in rural areaswas reported to be even greater in studies from both Australia andNorth America, but these studies used different criteria andcategorisations of rurality [12,14,15]. Several factors have been

Table 2Mode of injury and place of death, in total and stratified by quartiles of population density, given as number of events (n) and rate per 100,000 inhabitants per year with 95% CI.

Total Population density P=#

Most rural Rural Central Most central

n (%)a 8466(100) 2523 (100) 2155 (100) 2081 (100) 1707 (100)

Rate 28.7 35.3 29.3 27.0 23.3

95% CI (28.1, 29.3) (34.0, 36.7) (28.1, 30.6) (25.8, 28.2) (22.2, 24.5)

Mode of injury <.001

Transport

n (%)a 2548(30.1) 917 (36.3) 722 (33.5) 572 (27.5) 337(19.7)

Rate 8.6 12.8 9.8 7.4 4.6

95% CI (8.3, 9.0) (12.0, 13.7) (9.1, 10.6) (6.8, 8.0) (4.1, 5.1)

Falls

n (%)a 685(8.1) 181 (7.2) 184 (8.5) 138 (6.6) 182(10.7)

Rate 2.3 2.5 2.5 1.8 2.5

95% CI (2.2, 2.5) (2.2, 2.9) (2.2, 2.9) (1.5, 2.1) (2.2, 2.9)

Assaults

n (%)a 305 (3.6) 63 (2.5) 57 (2.6) 78 (3.7) 107 (6.3)

Rate 1.0 0.9 0.8 1.0 1.5

95% CI (0.9, 1.2) (0.7, 1.1) (0.6, 1.0) (0.8, 1.3) (1.2, 1.8)

Self-harm

n (%)a 3337(39.4) 896 (35.5) 812(37.7) 857 (41.2) 772(45.2)

Rate 11.3 12.5 11.1 11.1 10.6

95% CI (10.9, 11.7) (11.7, 13.4) (10.3, 11.8) (10.4, 11.9) (10.0, 11.3)

Other

n (%)a 1591(18.8) 466 (18.5) 380 (17.6) 436 (21.0) 309(18.1)

Rate 5.4 6.5 5.2 5.7 4.2

95% CI (5.1, 5.7) (6.0, 7.1) (4.7, 5.7) (5.1, 6.2) (3.8, 4.7)

Place of deathb <.001

In-hospital n (%)a 1819(21.5) 469 (18.6) 459 (21.3) 482 (23.2) 409(24.0)

Pre-hospital n (%)a 6589(77.8) 2046(81.1) 1692(78.5) 1585(76.2) 1266(74.2)

Study populationn 2.95 mill 714,677 735,149 771,608 731,929

a Percentages are for each column of population density and separately for mode of injury and place of death.b There were 58 cases with missing data on place of death: 8, 0, 14 and 32cases for population density quartiles most rural, rural, central and most central, respectively.# x2 test comparing each quartile of population density with five groups of type of injury and two groups of place of death; d.f: 12 and 3, respectively.

T. Kristiansen et al. / Injury, Int. J. Care Injured 45 (2014) 23–3028

proposed to explain the excess mortality in rural areas, such asincreased injury incidence, injury severity, discovery time,response time and distance of primary transport, compared withurban regions [15,17,52]. Furthermore, the access to specialisedservices is limited in rural areas, and factors such as rapid triageand effective interhospital transfer are thought to be important[16,20,23,26,53].

The overall proportion of prehospital deaths was higher in ourstudy compared with a study from the south-western region ofNorway [47] and several North American studies [16–19]. Thehighest proportion of prehospital deaths was found in the mostrural and northern county, and our findings correspond well withprevious studies from Finnmark [54,55]. In line with previousstudies, the proportion of prehospital deaths increased withincreasing rurality [16,17].

The place of death also varied with the mode of injury.While approximately half and one-quarter of deaths after fallsand transport injuries, respectively, occurred after reachinghospital, almost nine out of ten deaths caused by self-harmoccurred outside hospital. This has implications for primaryand secondary preventive strategies for these conditions andcircumstances [18].

The comparison of trauma mortality rates between regionsmay identify high-risk regions and populations, as well aseffective strategies for trauma care [3,29]. Our results pointtowards higher relative expenditure of resources on road trafficsafety measures in rural areas. Conversely, the most urbanareas have relatively more to gain from the prevention of

violent crimes. The very high rate of prehospital deaths fromself-harm injuries may indicate scenarios with very earlydeaths after non-survivable injuries. The majority of this groupof patients may only be reached through primary preventivestrategies. The high proportion of prehospital deaths anddeaths from transport injuries in rural areas indicate thatresources spent in the first phases of trauma care may benefitrural patients. Resources empowering local acute care hospi-tals and EMS providers to optimise initial stabilisation andinterhospital transfers represent the inclusiveness of a traumasystem.

There is a finite level of resources available in any health caresystem, and resources spent on preparedness in highly urbanareas may give higher cost–benefit ratios from a utilitarianperspective. Rural areas present a challenge to the equality oftrauma care, and there is little evidence to support that theseregions have the same needs and have experienced the samebenefits as urban areas with the development of modern traumasystems. A balance needs to be struck between the increasednumbers of trauma patients in urban areas versus the increasedrisk for sparsely populated areas. In this study an incrementalrisk for all levels of rurality has been demonstrated and the areasof highest risk represent one quartile of the adult population.Most regional trauma systems will include areas wheregeographical distance and rurality warrant special consider-ations. This study of Norwegian trauma epidemiology maytherefore be of relevance to all regions aiming to improve bothquality and equality in trauma care.

T. Kristiansen et al. / Injury, Int. J. Care Injured 45 (2014) 23–30 29

Limitations

This study was based on consecutively gathered data fromcentral registries that were collected prior to the commencementof the study, but the study protocol was written prior to accessingthese data. Central registry data allow for large population-basedstudies, but the level of detailed anatomical information is limited.In contrast to hospital-based studies, the data do not allow for theevaluation of the severity of injuries or the preventability ofindividual deaths. However, with four out of five deaths occurringbefore reaching hospitals, epidemiological studies based solely onhospital data would severely underestimate the true contributionof fatal injuries. Central registry data are based on medical deathcertificates combined with numerous other sources, and incom-plete certificates with respect to cause and place of death areinvestigated [34]. A previous study indicated that sub-classifica-tions, such as whether an injury is work-related, may besuboptimal in the registry, but that the data captured with respectto injury deaths in general appear to be better [56]. It is thereforeunlikely that many trauma-related deaths in the study period weremissed, but the level of detail regarding mechanism of injury maynot have been optimal.

We based population data on an index-year central to the studyperiod. This could potentially bias the rate calculations if majordemographical changes occurred during an isolated part of thestudy period. There has been a gradual urbanisation of theNorwegian population since the 1970s [57]. The rate ofurbanisation was higher in the part of the study period after2002 compared with the first half of the study period [58]. Theimpact of this change in urbanisation rate could potentially bethat the actual urban-rural differences are larger than those foundin this study. To exclude a potential bias in either direction, weassessed the representativeness of the index year population forthe mean of the entire study period separately for rural and urbanpopulations.

The place of injury was not available in the dataset. The locationof events was based on residency rather than the place of death, asthe latter may result in bias (e.g., as some municipalities have acutecare hospitals while others do not). For primary preventivepurposes, this may be beneficial; e.g., for targeting high-riskpopulations with injury preventive measures, place of residencymay be the location of interest. For other purposes, the use of placeof residency as a proxy for site of injury, reduces the precision forlocating events. However, this lack of precision is equallydistributed among the subgroups of the study, thereby reducingthe risk of any systematic error invalidating the study results. OneUS study revealed that the majority of injury deaths occur in thevicinity of the residency and recommended place of residency tolocate events when using central registry data in trauma research[59]. Further studies are needed to validate these results also forother regions.

Conclusion

Rural areas present a challenge in the provision of optimaltrauma care. This population-based study confirms a higherrural mortality rate, due primarily to transport-related acci-dents. The increased risk in rural areas is best predicted bypopulation density. The proportion of prehospital deaths is highamong all trauma victims, and it is even higher in rural areas.Our findings point to high-risk areas and populations for trauma,and suggestions for optimising rural trauma care and targetingprimary preventive strategies are made. The comparison of ourfindings with related studies indicates that these suggestionsmay be generalisable to all trauma systems that include ruralregions.

Conflict of interest statement

The authors declare that they have no conflict of interest.

Authors’ contributions

TK, HML, MR, PK and HMG planned the study project. TK and PKaccessed the data and established the data file for injury deaths. TKaccessed population and geographical data and established aggre-gated data files. TK and JR conducted the analysis. TK, MR, PK and KSconducted literature searches. TK drafted early manuscripts. Allauthors contributed substantially to the specification of the studycontent. TK, HML, JR and KS drafted the final version of themanuscript and all authors contributed to the continuous revisionand approved the final version of the manuscript.

Funding

The study did not receive funding beyond the normal salariesfor the authors, provided by their respective institutions, specifiedby authors’ affiliations.

References

[1] Peden M, McGee K, Sharma G. The injury chartbook: a graphical overview ofthe global burden of injuries. Geneva: World Health Organization; 2002.

[2] WHO. The Global Burden of Disease: 2004 Update. Geneva; 2008.[3] WHO: World report on road traffic injury prevention. Geneva: World Health

Organization; 2004.[4] Gwinnutt CL, Driscoll PA, Whittaker J. Trauma systems – state of the art.

Resuscitation 2001;48(1):17–23.[5] American College of Surgeons. Resources for optimal care of the injured patient

2006. Chicago, IL: American College of Surgeons; 2006.[6] Cameron PA, Gabbe BJ, Cooper DJ, Walker T, Judson R, McNeil J. A statewide

system of trauma care in Victoria: effect on patient survival. Med J Aust2008;189(10):546–50.

[7] Celso B, Tepas J, Langland-Orban B, Pracht E, Papa L, Lottenberg L, et al. Asystematic review and meta-analysis comparing outcome of severely injuredpatients treated in trauma centers following the establishment of traumasystems. J Trauma 2006;60(2):371–8. discussion 378.

[8] Gabbe BJ, Biostat GD, Lecky FE, Bouamra O, Woodford M, Jenks T, et al. Theeffect of an organized trauma system on mortality in major trauma involvingserious head injury: a comparison of the United kingdom and Victoria,Australia. Ann Surg 2011;253(1):138–43.

[9] Oakley PA, MacKenzie G, Templeton J, Cook AL, Kirby RM. Longitudinal trendsin trauma mortality and survival in Stoke-on-Trent 1992–1998. Injury2004;35(4):379–85.

[10] Ruchholtz S, Lefering R, Paffrath T, Oestern HJ, Neugebauer E, Nast-Kolb D,et al. Reduction in mortality of severely injured patients in Germany. DtschArztebl Int 2008;105(13):225–31.

[11] Twijnstra MJ, Moons KG, Simmermacher RK, Leenen LP. Regional traumasystem reduces mortality and changes admission rates: a before and afterstudy. Ann Surg 2010;251(2):339–43.

[12] Baker SP, Whitfield RA, O’Neill B. Geographic variations in mortality frommotor vehicle crashes. N Engl J Med 1987;316(22):1384–7.

[13] Eberhardt MS, Pamuk ER. The importance of place of residence: examininghealth in rural and nonrural areas. Am J Public Health 2004;94(10):1682–6.

[14] Fatovich DM, Jacobs IG. The relationship between remoteness and traumadeaths in Western Australia. J Trauma 2009;67(5):910–4.

[15] Zwerling C, Peek-Asa C, Whitten PS, Choi SW, Sprince NL, Jones MP. Fatal motorvehicle crashes in rural and urban areas: decomposing rates into contributingfactors. Inj Prev 2005;11(1):24–8.

[16] Gomez D, Berube M, Xiong W, Ahmed N, Haas B, Schuurman N, et al. Identify-ing targets for potential interventions to reduce rural trauma deaths: apopulation-based analysis. J Trauma 2010;69(3):633–9.

[17] Gonzalez RP, Cummings G, Mulekar M, Rodning CB. Increased mortality inrural vehicular trauma: identifying contributing factors through data linkage. JTrauma 2006;61(2):404–9.

[18] Rogers FB, Madsen L, Shackford S, Crookes B, Charash W, Morrow P, et al. Aneeds assessment for regionalization of trauma care in a rural state. Am Surg2005;71(8):690–3.

[19] Rogers FB, Shackford SR, Hoyt DB, Camp L, Osler TM, Mackersie RC, et al.Trauma deaths in a mature urban vs rural trauma system. A comparison. ArchSurg 1997;132(4):376–81. discussion 372-1.

[20] Nutbeam T, Leaman A, Oakley P. Transporting major trauma patients from themargins of a UK trauma system. Emerg Med J 2012;29(3):182–3.

[21] Rogers FB, Osler TM, Shackford SR, Martin F, Healey M, Pilcher D. Population-based study of hospital trauma care in a rural state without a formal traumasystem. J Trauma 2001;50(3):409–13. discussion 414.

T. Kristiansen et al. / Injury, Int. J. Care Injured 45 (2014) 23–3030

[22] Clay Mann N, Mullins RJ, Hedges JR, Rowland D, Arthur M, Zechnich AD.Mortality among seriously injured patients treated in remote rural traumacenters before and after implementation of a statewide trauma system. MedCare 2001;39(7):643–53.

[23] Hsia RY, Wang E, Torres H, Saynina O, Wise PH. Disparities in trauma centeraccess despite increasing utilization: data from California, 1999–2006. JTrauma 2010;68(1):217–24.

[24] Kristiansen T, Ringdal KG, Skotheimsvik T, Salthammer HK, Gaarder C, NaessPA, et al. Implementation of recommended trauma system criteria in south-eastern Norway: a cross-sectional hospital survey. Scand J Trauma ResuscEmerg Med 2012;20:5.

[25] Kristiansen T, Soreide K, Ringdal KG, Rehn M, Kruger AJ, Reite A, et al. Traumasystems and early management of severe injuries in Scandinavia: review of thecurrent state. Injury 2010;41(5):444–52.

[26] Gomes E, Araujo R, Carneiro A, Dias C, Costa-Pereira A, Lecky FE. The impor-tance of pre-trauma centre treatment of life-threatening events on the mor-tality of patients transferred with severe trauma. Resuscitation 2010;81(4):440–5.

[27] Grossman DC, Hart LG, Rivara FP, Maier RV, Rosenblatt R. From roadside tobedside: the regionalization of trauma care in a remote rural county. J Trauma1995;38(1):14–21.

[28] Haddon Jr W. The changing approach to the epidemiology, prevention, andamelioration of trauma: the transition to approaches etiologically rather thandescriptively based. Am J Public Health Nations Health 1968;58(8):1431–8.

[29] Krug EG, Sharma GK, Lozano R. The global burden of injuries. Am J PublicHealth 2000;90(4):523–6.

[30] Kristiansen T, Rehn M, Gravseth HM, Lossius HM, Kristensen P. Paediatrictrauma mortality in Norway: a population-based study of injury character-istics and urban–rural differences. Injury 2011;43(11):1865–72.

[31] Vane DW, Shackford SR. Epidemiology of rural traumatic death in children: apopulation-based study. J Trauma 1995;38(6):867–70.

[32] World Medical Association: Declaration of Helsinki. Ethical principles formedical research involving human subjects. In: Amended by the 52nd WorldMedical Association General Assembly; 2000.

[33] Statistics Norway: Causes of death registry; 2013, http://statbank.ssb.no/statistikkbanken/Default_FR.asp?Productid=03.01&PXSid=0&nvl=true&PLan-guage=1&tilside=selecttable/MenuSelP.asp&SubjectCode=03 [accessed17.04.13.].

[34] Gjertsen F. The cause of death registry and research. Tidsskr Nor Laegeforen2000;120(6):723–5.

[35] Statistics Norway: StatBank; 2013, http://statbank.ssb.no/statistikkbanken/[accessed 17.04.13.].

[36] Norwegian Social Science Data Services (NSD); 2013, Regional database:http://www.nsd.uib.no/nsd/english/index.html [accessed 17.04.12.].

[37] Langhelle A, Lossius HM, Silfvast T, Bjornsson HM, Lippert FK, Ersson A, et al.International EMS Systems: the Nordic countries. Resuscitation 2004;61(1):9–21.

[38] Brattebo G, Wisborg T, Hoylo T. Organization of trauma admissions at Nor-wegian hospitals. Tidsskr Nor Laegeforen 2001;121(20):2364–7.

[39] Isaksen MI, Wisborg T, Brattebo G. Organisation of trauma services – majorimprovements over four years. Tidsskr Nor Laegeforen 2006;126(2):145–7.

[40] Statistics Norway. Standard Classification of Municipalities Oslo-Kongsvinger;1994.

[41] Agresti A, Coull BA. Approximate is better than ‘‘exact’’ for interval estimationof binomial proportions. American Statistician 1998;52:119–26.

[42] R Development Core Team. R: A language and environment for statisticalcomputing. Vienna, Austria: R Foundation for Statistical Computing; 2012.

[43] Minino AM, Anderson RN, Fingerhut LA, Boudreault MA, Warner M. Deaths:injuries, 2002. Natl Vital Stat Rep 2006;54(10):1–124.

[44] Polinder S, Haagsma JA, Toet H, van Beeck EF. Epidemiological burden ofminor, major and fatal trauma in a national injury pyramid. Br J Surg2012;99(Suppl. 1):114–21.

[45] Di Bartolomeo S, Sanson G, Michelutto V, Nardi G, Burba I, Francescutti C, et al.Epidemiology of major injury in the population of Friuli Venezia Giulia, Italy.Injury 2004;35(4):391–400.

[46] WHO. Injury: a leading cause of the global burden of disease. Geneva: WHO;2000.

[47] Soreide K, Kruger AJ, Vardal AL, Ellingsen CL, Soreide E, Lossius HM. Epidemi-ology and contemporary patterns of trauma deaths: changing place, similarpace, older face. World J Surg 2007;31(11):2092–103.

[48] Kristensen P, Kristiansen T, Rehn M, Gravseth HM, Bjerkedal T. Social inequal-ities in road traffic deaths at age 16–20 years among all 611,654 Norwegiansborn between 1967 and 1976: a multilevel analysis. Inj Prev 2012;18(1):3–9.

[49] Meisler R, Thomsen AB, Theilade P, Abildstrom H, Borge P, Treschow M, et al.Age-related differences in mechanism, cause, and location of trauma deaths.Minerva Anestesiol 2011;77(6):592–7.

[50] Rogers FB, Osler TM, Shackford SR, Cohen M, Camp L, Lesage M. Study of theoutcome of patients transferred to a level I hospital after stabilization at anoutlying hospital in a rural setting. J Trauma 1999;46(2):328–33.

[51] McGuffie AC, Graham CA, Beard D, Henry JM, Fitzpatrick MO, Wilkie SC, et al.Scottish urban versus rural trauma outcome study. J Trauma 2005;59(3):632–8.

[52] Carr BG, Caplan JM, Pryor JP, Branas CC. A meta-analysis of prehospital caretimes for trauma. Prehosp Emerg Care 2006;10(2):198–206.

[53] Galvagno SM, Haut ER, Zafar SN, Millin MG, Efron DT, Koenig GJ, et al.Association between helicopter vs ground emergency medical services andsurvival for adults with major trauma. J Am Med Assoc 2012;307(15):1602–10.

[54] Bakke HK, Wisborg T. Rural high north: a high rate of fatal injury andprehospital death. World J Surg 2011;35(7):1615–20.

[55] Wisborg T, Hoylo T, Siem G. Death after injury in rural Norway: high rate ofmortality and prehospital death. Acta Anaesthesiol Scand 2003;47(2):153–6.

[56] Wergeland E, Gjertsen F, Lund J. Mortality due to occupational injury isunderreported. Tidsskr Nor Laegeforen 2009;129(10):981–6.

[57] Carling J. Centralizing tendencies in domestic migration between economicregions. The 1977–1998 experience and assumptions about the 1999–2050projection period. Statistics Norway; 1999, 67.

[58] Karlstad S, Lie I. Tendency of centralization in Norwegian regions. Alta:Northern Research Institute.; 2008. 12.

[59] Myers SR, Branas CC, Kallan MJ, Wiebe DJ, Nance ML, Carr BG. The use of homelocation to proxy injury location and implications for regionalized traumasystem planning. J Trauma 2011;71(5):1428–34.