Accident Analysis and Prevention 82 (2015) 171–179

Contents lists available at ScienceDirect

Accident Analysis and Prevention

journal homepage: www.elsevier.com/locate/aap

Injury risk for matched front and rear seat car passengers by injury
severity and crash type: An exploratory study
R.J. Mitchella,* , M.R. Bambachb , Barbara Tosonc
a
Australian Institute of Health Innovation, Macquarie University, Australia
b
Transport and Road Safety (TARS) Research, University of New South Wales, Australia
c
Neuroscience Research Australia, University of New South Wales, Australia

A R T I C L E I N F O A B S T R A C T

Article history: Background: The risk of serious injury or death has been found to be reduced for some front compared to
Received 27 February 2015 rear seat car passengers in newer vehicles. However, differences in injury severity between car occupants
Received in revised form 28 May 2015 by seating position has not been examined. This study examines the injury severity risk for rear compared
Accepted 28 May 2015
to front seat car passengers.
Available online 15 June 2015
Method: A retrospective matched-cohort analysis was conducted of vehicle crashes involving injured rear
vs front seat car passengers identified in linked police-reported, hospitalisation and emergency
Keywords:
department (ED) presentation records during 2001–2011 in New South Wales (NSW), Australia. Odds
Injury severity
Rear seat passenger
ratios were estimated using an ordinal logistic mixed model and logistic mixed models.
Matched-cohort study Results: There were 5419 front and 4588 rear seat passengers in 3681 vehicles. There was a higher odds of
Road trauma sustaining a higher injury severity as a rear-compared to a front seat car passenger, with a higher odds of
rear seat passengers sustaining serious injuries compared to minimal injuries. Where the vehicle
occupant was older, travelling in a vehicle manufactured between 1990 and 1996 or after 1997, where the
airbag deployed, and where the vehicle was driven where the speed limit was
70 km/h there was a
higher odds of the rear passenger sustaining a higher injury severity then a front seated occupant.
Conclusion: Rear seat car passengers are sustaining injuries of a higher severity compared to front seat
passengers travelling in the same vehicle, as well as when travelling in newer vehicles and where the
front seat occupant is shielded by an airbag deployed in the crash. Rear seat occupant protective
mechanisms should be examined. Pre-hospital trauma management policies could influence whether an
individual is transported to a hospital ED, thus it would be beneficial to have an objective measure of
injury severity routinely available in ED records. Further examination of injury severity between rear and
front seat passengers is warranted to examine less severe non-fatal injuries by car seating position and
vehicle intrusion.
ã2015 Elsevier Ltd. All rights reserved.

1. Introduction estimated $27 billion annually (Australian Transport Council,

Worldwide, road trauma accounts for an estimated 1.3 million
deaths annually, with road trauma projected to become the third
leading cause of the burden of disease by 2030 (World Health
Organization, 2008). In Australia, fatal injury as a result of a vehicle
crash has declined over time (Australian Transport Safety Bureau,
2003; Bureau of Infrastructure Transport and Regional Economics,
2012), but still represents approximately 1400 deaths and 32,500
serious injuries each year (Henley and Harrison, 2011), costing an
* Corresponding author at: Australian Institute of Health Innovation, Macquarie
University, Level 6, 75 Talavera Road, Macquarie University, NSW 2109, Australia.
Fax: +61 2 8088 6234.
E-mail address: r.mitchell@mq.edu.au (R.J. Mitchell).
2011).
For car occupants, the risk of death and serious injury has
historically been found to be lower for rear compared to front seat
passengers (Smith et al., 2004; Mayrose and Priya, 2008), but
particularly for passenger cars without airbags or if the passengers
were not restrained (Smith and Cummings, 2006). There have been
improvements in passenger car crashworthiness, such as the
inclusion of frontal airbags and improved occupant restraint
mechanisms, that have decreased rear seat occupant vehicle
protection compared to front seat occupants in recent vehicle
models (Sahraei and Digges, 2009; Sahraei et al., 2009, 2010) and
that have resulted in reduced risk of injury among front compared
to rear seated car occupants (Bilston et al., 2010; Sahraei et al.,
2010, 2014). In the United States, a reduced injury risk was shown

http://dx.doi.org/10.1016/j.aap.2015.05.023
0001-4575/ ã 2015 Elsevier Ltd. All rights reserved.
172 R.J. Mitchell et al. / Accident Analysis and Prevention 82 (2015) 171–179
for front seat car occupants aged 16–50 years in newer vehicles (i.e.
1997–2007) compared to rear seat car passengers (Bilston et al.,
2010). However, the reduced injury risk for front seat occupants
was not evident for younger (9–15 years) or older (51+ years)
individuals, nor for all crash types (Bilston et al., 2010). Studies that
have specifically examined injury risk and young children’s seating
position in passenger vehicles have found that risk of death is
reduced for children if they are seated in rear passenger seats
(Braver et al., 1998; Lennon et al., 2008).
Studies of occupant injuries following vehicle crashes can be
affected by confounding factors that can hamper the identification
of associations of an outcome of interest with different crash or
injury risk factors (Cummings et al., 2003a,b). By matching front
and rear seat passengers in the same vehicle in the same crash, a
matched cohort study design is able to control for possible vehicle
and crash-related confounding characteristics such as vehicle
model and vehicle speed (Smith and Cummings, 2006). Many of
the vehicle matched-cohort studies have examined risk of death
between rear and front seat passengers (Cummings et al., 2003a,b;
Smith and Cummings, 2004, 2006) or have examined risk of
serious injury and mortality (Bilston et al., 2010). Further work is
needed to compare rear and front seat car passengers for finer
levels of injury severity (Brown and Bilston, 2014), such as by using
detailed injury severity categories (e.g. using six severity catego-
ries) or by using broad-level injury severity categories (e.g. using
three severity categories). Investigation of whether certain types of
injuries are more common in different types of crashes for rear
compared to front seat passengers is also warranted. This study
aims to examine the risk of injury or death for rear compared to
front seat car passengers using both fine and broad levels of injury
severity and to examine type of injuries sustained by crash type
using a matched-cohort study with linked police-reported road
crash, ED presentation and hospital admission data in New South
Wales (NSW), Australia.
2. Method
A retrospective matched-cohort analysis was conducted. The
cohort included fatalities identified in police-reported crash data
and non-fatal injuries identified in linked police-reported crash, ED
presentation and hospital admission records of passenger car
occupants during 1 January 2001 to 31 December 2011. Injury
severity of front- vs rear-seated car occupants in the same vehicle
were compared. Ethics approval was obtained from the NSW
Population and Health Services Research Ethics Committee (2010/
10/273).
2.1. Data collections
The Emergency Department Data Collection (EDDC) contains
information collected in public hospital EDs in NSW. There are
around 150 EDs in NSW and just under 100 (including all the
larger EDs) provide information to the EDDC, although numbers
have varied over time. During the study time period, 77 hospitals
provided data during 2005, 90 during 2006, 95 during 2007, 98
during 2008 and 2009 and 96 during 2010 and 2011. Data
collected by the EDs includes patient demographics, arrival and
departure dates/times, triage category, type of visit and clinical
procedures. A provisional diagnosis assigned by staff when a
patient presents to the ED is also included which could either
contain diagnostic or external cause information. The ED
diagnostic data were categorised using a number of different
ICD-based classification frameworks (World Health Organization,
1977, 1992; National Centre for Classification in Health, 2006;
National Center for Health Statistics, 2011) or using the
Systematized Nomenclature of Medicine—Clinical Terms
(Snomed—CT) (International Health Terminology Standards
Development Organization, 2011). Information from the EDDC
was only available from 1 January 2005.
The Admitted Patient Data Collection (APDC) includes informa-
tion on all inpatient admissions from all public and private
hospitals, private hospital day procedures, and public psychiatric
hospitals in NSW. The APDC contains information on patient
demographics, source of referral, diagnoses, external cause(s),
hospital separation type (e.g. discharge, death) and clinical
procedures. Diagnoses and external cause codes are classified
using the International Classification of Diseases, 10th Revision,
Australian Modification (ICD-10-AM) (National Centre for Classifi-
cation in Health, 2006).
The CrashLink data collection contains information on all
police-reported road traffic crashes on a public road in NSW where
a person was unintentionally fatally or non-fatally injured, or at
least one motor vehicle was towed away. Information pertaining to
the crash and conditions at the incident site, the traffic unit or
vehicle, and the vehicle controller and any casualties resulting
from the crash are recorded. Each individual is identified as being
non-injured, injured or killed (died within 30 days). No informa-
tion on injury severity is available. Information was not obtained
on individuals who were non-injured. Road users selected for this
research were limited to passenger car occupants only and were
identified using the traffic unit group (i.e. car/ car derivative driver,
including 4 wheel drives, panel and passenger vans, utilities, and
station wagons).
2.1.1. Data linkage
The EDDC and the APDC were probabilistically linked to the
police-reported crashes in CrashLink by the Centre for Health
Record Linkage (CHeReL) using ChoiceMaker (Choicemaker
Technologies, 2011). The CHeReL uses identifying information
(e.g. name, address, date of birth, gender) to create a person project
number (PPN), for each unique person identified in the linkage
process. A successful link was defined if the PPN matched in the
data collections, and the presentation date in the EDDC or the
admission date in the APDC was on the same day or the next day as
the crash date in CrashLink. Upper and lower probability cut-offs
started at 0.75 and 0.25 for a linkage and were adjusted for each
individual linkage to ensure false links were kept to a minimum.
Record groups with probabilities in between the cut-offs were
subject to clerical review. The linkage rates for road traffic-related
hospital admissions to police casualty records were 74.1% for
drivers and 55.7% for passengers and the linkage rates for ED
presentations to police casualty records were 62.8% for drivers and
46.6% for passengers. Road users were identified as died, were
injured and hospitalised, were injured and presented to ED (but
were not admitted) or were identified by police as injured and not
hospitalised.
2.1.2. Injury severity
Injury severity was calculated using the International
Classification of Disease Injury Severity Score (ICISS). The ICISS
is derived for each person by summing the probability of
survival for each injury diagnosis using survival risk ratios (SRR)
calculated for each injury diagnosis (Stephenson et al., 2004). In
a prior study of all land transport trauma, the diagnosis
classifications within hospital records and survival outcome
identified from mortality records for 109,843 individuals were
used to generate SRRs for all ICD-10 injury codes during 2001 to
2007 (Bambach et al., 2012a,b). For each ICD injury (ICDi) the
SRR was calculated from Eq. (1).
 R.J. Mitchell et al. / Accident Analysis and Prevention 82 (2015) 171–179 173

Number of individuals with injury ICDi that survived

SRRICDi ¼
Total number of individuals with injury ICDi
(1)
Injury severity using ICISS was only calculated for individuals
who had been hospitalised using each individual’s injury diagnosis
classifications, using the previously developed SRRs. The ICISS is
equivalent or superior in assessing mortality risk compared to the
Abbreviated Injury Scale (Stephenson et al., 2004; Davie et al.,
2008; Willis et al., 2010). As there are no specific measures against
which different severity levels in the ICISS are compared, the
current study used the same ICISS levels as Dayal et al. (2008) to
define minor (ICISS:
0.99), moderate (ICISS: 0.941–0.99) and
serious (ICISS: 0.941) injury. An ICISS of  0.941 was used to
define serious injury in order to minimise threats to validity (Cryer
et al., 2004). This is equivalent to a survival probability of 94.1% or a
5.9% probability of death. No information regarding injury severity
could be obtained for ED presentations where the individual was
not admitted to hospital, termed ‘ED presentation, no hospital
admission’. Other minimal injuries were defined as a person who
was identified as injured by the police, but their police casualty
record did not link to an ED presentation or hospital admission
record.
2.2. Data management and analysis
All analyses were performed using SAS version 9.4 (SAS
Institute, 2014). Vehicles were included if there was at least one
injured rear and one injured front seat passenger in the same
vehicle that crashed. Vehicles of all ages were included so as to be
representative of the age of vehicles in Australia (Australian Bureau
of Statistics, 2013). Excluded from the analysis were children aged
less than nine years as car occupants of this age should be using
child restraints or child booster seats (Brown et al., 2006). As the
intent was to examine injury severity between rear and front seat
passengers in optimal conditions for injury prevention, car
occupants who were not wearing a restraint (n = 1203) or who
had unknown restraint use (n = 3045), and car occupants in other
and unknown seating positions and individuals who were in non-
passenger areas, such as the rear of vans (n = 2124) were excluded.
Where there was missing or unknown information for
passenger age (n = 1335), gender (n = 257), presence of airbag
(n = 2107) or the year of manufacture of the vehicle (n = 293), these
records were removed from the analysis. Due to small sample sizes,
vehicle impacts with animals, pedestrians, trains and unknown
impacts were excluded (n = 77). The removal from the analysis of
records where there was missing information resulted in the
exclusion of 1653 cars. Age of vehicle at time of crash was
calculated by subtracting the year the crash occurred from the year
of manufacture of the vehicle. For individuals who had been
hospitalised, identification of the injuries sustained was conducted
using up to 55 diagnosis variables.
The examination of injury severity was conducted in two stages.
Firstly, an ordinal logistic mixed model was conducted to examine
whether rear seat passengers had a higher odds of sustaining a
higher injury severity than a front seat passenger (i.e. driver or
front passenger) in the same vehicle. Secondly, logistic mixed
models were conducted using the same variables as the ordinal
logistic mixed model comparing each injury severity category to
the lowest injury severity category (i.e. other injury) to attempt to
tease out whether any differences in injury severity were evident.

Table 1
Demographic and injury type and severity characteristics for rear and front seat car passengers, linked police-report, ED presentation and hospital admission records in NSW,
2001–2011.

Front seat passenger (n = 5419) Rear seat passenger (n = 4588)

n % n %
Gender
Male 2671 49.3 2115 46.1
Female 2748 50.7 2473 53.9

Age group
9–15 years 319 5.9 1282 27.9
16–50 years 4226 78.0 2802 61.1
51+ Years 874 16.1 504 11.0

Injury type (hospitalised only)a

Head 532 39.4 419 36.5
Neck 244 18.1 239 20.8
Thorax 575 42.6 427 37.2
Abdomen, lower back, lumbar spine and pelvic 423 31.7 465 40.5
Upper extremities 474 35.1 347 30.3
Lower extremities 438 32.5 316 27.6

Injury severity
ED presentation (not admitted) 1343 24.8 1104 24.1
Minor 241 4.5 207 4.5
Moderate 720 13.3 600 13.1
Serious 455 8.4 401 8.7
Death 130 2.4 66 1.4
Other minimal injuryb 2530 46.7 2210 48.2

Seat position
Driver 2782 51.3 – –
Front passenger 2637 48.7 – –
Rear left – – 2416 52.7
Rear centre – – 452 9.9
Rear right – – 1720 37.5
a
Multiple injuries could be identified per person.
b
Identified by police as injured and no ED presentation or hospitalisation.
174 R.J. Mitchell et al. / Accident Analysis and Prevention 82 (2015) 171–179

These models were conducted using the GLIMMIX procedure, with
a multinominal distribution, cumulative link function and using
Kenward and Roger denominator degrees of freedom. A random
intercept for vehicle was used to model constant correlation within
vehicles (Brown and Prescott, 2006a). This method of analysis was
selected because injury severity was considered to be a ordered
categorical data and the use of a mixed-effects model was able to
accounted for the correlation among passengers in the same
vehicle. This model assumes proportional odds, implying that the
odds for cumulative logits among injury severity categories are
uniform.
Within the original ordinal logistic mixed model, variables and
variable interactions were trialled and retained if they improved
the model fit based on pseudo-likelihood values (Brown and
Prescott, 2006b) or had clinical or crash relevance. The variables
included in the models were based on the data available and were
consistent with the variables used in previous studies (e.g. Smith
and Cummings, 2006; Bilston et al., 2010). The following
independent variables were included in each model: rear seat,
age group, gender, year of vehicle manufacture, airbag deployment,
speed limit and interactions between rear seat and year of vehicle
manufacture. Each logistic mixed model retained the same
variables used in the ordinal logistic mixed model. Age and
vehicle model year groupings were selected to be consistent with
previous published research that examined injury and rear vs front
seat injuries (e.g. Bilston et al., 2010). Speed limit zone groupings
were those commonly used to describe speed zones in NSW. Crash
type was excluded from the ordinal logistic mixed model and the
logistic mixed models analyses to avoid collinearity with seating
position as information on site of crash or impact intrusion into the
vehicle was not available.
The ordinal logistic mixed models and then each logistic mixed
model were conducted for both detailed categories of injury
severity (i.e. six detailed categories) and three broad categories of
injury severity. Results generated using PROC GLIMMIX include
Type III Tests of Fixed Effects that include: F values, degrees of
freedom and p-values. Odds ratios and 95% confidence intervals
(95% CIs) were also requested.
To examine type of injury by crash type for rear compared to
front seat passengers who were hospitalised, adjusted risk ratios
(ARR) were estimated using regression analyses stratified by
vehicle. This was conducted using the GENMOD procedure using
generalized estimating equations (GEE), a binomial error distribu-
tion and log link function to generate relative risks and confidence
intervals. Lastly, to examine the distribution of type of injury by
injury severity for hospitalised injuries, the SURVEYFREQ proce-
dure was used to calculate Rao–Scott design-adjusted chi-square
tests. Only hospitalised individuals were able to be included in the
examination of type of injury as no (or limited) information is
available regarding injury type in police data or ED presentations.
between vehicles on an angle (37.1%) or vehicles into objects
(27.8%). Just over half (57.9%) of crashes occurred within designated
speed limits of 60 km/h. An airbag deployed in 19.6% of crashes
(Table 2).
Using the six detailed injury severity categories, there was a
higher odds of sustaining a higher injury severity as a rear seat
passenger compared to front seat passengers (F = 4.32; df = 1, 9990;
p = 0.04). Older age groups (F = 32.13; df = 2, 9990; p < 0.0001),
vehicles manufactured between 1990 and 1996 or after 1997
(F = 8.68; df = 2, 3439; p = 0.0002), crashes where an air bag
deployed (F = 31.95; df = 1, 9990; p < 0.0001) and crashes in speed
limits >70 km/h (F = 12.98; df = 3, 3224; p < 0.0001) all had a higher
odds of sustaining a higher injury severity. There was also a
significant interaction effect of seating position and year of vehicle
manufacture (F = 3.96; df = 2, 9990; p < 0.02) (Supplementary
Table 1). Passengers travelling in vehicles manufactured between
1990 and 1996 or after 1997, passengers travelling in a car where
the air bag deployed and passengers travelling in vehicles in higher
speed limits all had significantly higher odds of an ED presentation
(without admission), moderate and serious injuries compared to
other minimal injuries. Older age groups all had significantly
higher odds of moderate and serious injuries compared to other
minimal injuries. A rear seat passenger had one and a half times the
odds of sustaining serious injuries compared to other minimal
injuries compared to a front seat passenger (OR 1.42; 95%CI 1.15–
1.77) (Table 3).
Examining injury severity by the three broad severity categories
identified that there was a higher odds of sustaining a higher injury
severity as a rear seat passenger compared to front seated
passengers (F = 4.01; df = 1, 9993; p = 0.05). There was no interac-
tion between seating position and year of vehicle manufacturer
(p = 0.1). Older age groups (F = 37.46; df = 2, 9993; p < 0.0001),
vehicles manufactured between 1990 and 1996 or after 1997
(F = 4.29; df = 2, 2165; p = 0.01), crashes where an air bag deployed
(F = 5.22; df = 1, 8527; p = 0.02) and crashes in speed limits >70 km/
h (F = 23.99; df = 3, 2000; p < 0.0001) all had a higher odds of
Table 2
Vehicle and crash characteristics for injured rear and front seat car passengers,
linked police-report, ED presentation and hospital admission records in NSW,
2001–2011.
Vehicles (n = 3681)
n %
Vehicle model year
<1990 703 19.1
1990–1996 1223 33.2

1997 1755 47.7
Vehicle age at time of crash
0–5 years 1011 27.5
6–10 years 965 26.2

11 years 1705 46.3

3. Results Speed limit (km/h)

40–50 768 20.9
There were 10,007 passengers in 3681 vehicles where an 60 1365 37.1
occupant was identified by the police as either fatally or non- 70–90 820 22.3
100–110 730 19.8
fatally injured, 54.2% were front seat passengers and 45.9% were
rear seat passengers. There was a slightly higher proportion of Crash impact type
females as rear seat passengers compared to male passengers and Vehicle–vehicle—head-on 421 11.4
just over one-quarter of rear seat occupants were children aged 9– Vehicle–vehicle—angle 1365 37.1
Vehicle–vehicle—rear-end 606 16.5
15 years. The distribution of injury severity and injury type for
Vehicle–object 1022 27.8
hospitalised injuries was fairly similar for front and rear seat Vehicle–rollover 267 7.3
passengers, except for abdominal, lower back, lumbar spine and
pelvic injuries which were higher for rear seat passengers (Table 1). Airbag present
Of the 3681 vehicles, 47.7% were manufactured in
1997. Almost Not fitted 2132 57.9
Deployed 743 20.2
half (46.3%) of the vehicles were aged 11 years or older at the time Not deployed 806 21.9
of the crash. The most common crash impact types were crashes
 R.J. Mitchell et al. / Accident Analysis and Prevention 82 (2015) 171–179 175

Table 3
Ordinal logistic mixed model odds ratios rear seat compared to front seat passengers and 95% confidence intervals by detailed injury severity categories, linked police-report,
ED presentation and hospital admission records in NSW, 2001–2011.

Ordinal logistic Logistic mixed model: ED Logistic mixed model: Logistic mixed model: Logistic mixed model: Logistic mixed
mixed modela presentation vs other minor injury (hospital moderate injury serious injury (hospital model: death vs
injuryb (no hospital admission) vs other (hospital admission) vs admission) vs other other injuryb
admission) injuryb other injuryb injuryb

OR 95% CI OR 95% CI OR 95% CI OR 95% CI OR 95% CI OR 95% CI

Rear seat passengerc 1.10 1.01–1.21 1.07 0.92–1.23 1.07 0.81–1.42 1.15 0.97–1.36 1.42 1.15–1.77 na

Age groupd
16–50 years 1.19 1.04–1.37 0.92 0.74–1.13 1.00 0.68–1.48 1.10 0.86–1.41 1.62 1.17–2.23 na
51+ Years 2.02 1.68–2.43 1.05 0.79–1.41 1.46 0.86–2.49 1.81 1.30–2.53 2.89 1.91–4.35 na

Gendere
Female 1.06 0.97–1.16 1.12 0.97–1.30 1.21 0.92–1.60 1.10 0.93–1.30 1.07 0.87–1.31 na

Vehicle model yearf

1990–1996 1.16 0.98–1.38 1.34 1.04–1.73 0.89 0.56–1.41 1.37 1.03–1.82 1.11 0.79–1.58 na

1997 1.39 1.19–1.62 1.49 1.18–1.88 1.21 0.81–1.82 1.68 1.30–2.18 1.63 1.19–2.22 na

Air bagg
Airbag deployed 1.56 1.34–1.83 2.53 1.97–3.24 1.56 0.97–2.53 2.19 1.65–2.91 1.95 1.37–2.78 na

Speed limit (km/h)h

60 0.94 0.79–1.12 0.89 0.70–1.13 0.75 0.48–1.16 0.93 0.69–1.23 1.09 0.76–1.55 na
70–90 1.35 1.11–1.63 0.92 0.70–1.20 0.84 0.51–1.39 1.22 0.89–1.68 1.69 1.15–2.48 na
100–110 1.51 1.25–1.83 0.43 0.32–0.58 0.94 0.57–1.54 1.53 1.12–2.08 1.82 1.24–2.67 na

na: cell size not large enough for analysis.

a
Measuring higher risk of severity in this order: other injury, ED presentation (no hospital admission), minor injury (hospital admission), moderate injury (hospital
admission), serious injury (hospital admissions), death.
b
Identified by police as injured and no ED presentation or hospitalisation.
c
Front seat passenger is referent.
d
9–15 year age group is referent.
e
Male is referent.
f
990 vehicle model is referent.
g
No airbag, not deployed or fitted is referent.
h
40–50 km/h speed limit is referent.

sustaining a higher injury severity (Supplementary Table 2). There
was an increased odds of a rear seat passenger sustaining serious
injuries/death compared to other minimal injuries/ED presenta-
tions compared to a front seat passenger (OR 1.21; 95% CI 1.01–
1.46) (Table 4).
For all types of crashes, the risk of head injuries (ARR 0.44; 95%
CI 0.28–0.69) and injuries to both the upper (ARR 0.73; 0.62–0.87)
and lower (ARR 0.58; 0.38–0.90) extremities were lower for rear
seat compared to front seat passengers. The risk of neck injuries
(ARR 1.23; 95% CI 1.01–1.50) and injuries to the lower back, lumbar
spine and pelvis (ARR 2.81; 95% CI 1.82–4.34) were higher for rear
seat passengers compared to passengers in the front seat. For head-
on collisions, there was a higher risk of neck (ARR 2.12; 95% CI 1.28–
3.49) and abdominal, lower back, lumbar spine and pelvic (ARR
8.10; 95% CI 3.14–20.88) injuries for rear vs front seat passengers.
Abdominal, lower back, lumbar spine and pelvic injuries were also
higher for rear seat occupants injured in vehicle–object collisions
(ARR 3.14; 95% CI 1.52–6.47). For rear seat passengers, the risk of
injuries to the head, thorax and upper extremities were lower for
vehicle–object collisions (ARR 0.63; 95% CI 0.47–0.83), vehicles
colliding on angles (ARR 0.71; 95% CI 0.52–0.98), head-on collisions
(ARR 0.59; 95% CI 0.39–0.88) and vehicle-angle collisions (ARR
0.63; 95% CI 0.46–0.86) compared to front seat passengers
(Table 5).
For the hospitalised minor injuries, injuries to the neck
(x2 = 16.2, df = 1, p < 0.0001), abdomen, lower back, lumbar spine
and pelvis (x2 = 5.6, df = 1, p < 0.02), and hip and thigh (x2 = 39.0,
df = 1, p < 0.0001) were all significantly higher for rear vs front seat
passengers, with knee and lower leg injuries (x2 = 16.7, df = 1,
p < 0.0001) significantly higher for front seat passengers. For the
hospitalised moderate injuries, injuries to the thorax (x2 = 6.6,
df = 1, p < 0.01), elbow and forearm (x2 = 35.5, df = 1, p < 0.0001),
and wrist and hand (x2 = 8.4, df = 1, p < 0.004) were all significantly
higher for front vs rear seat passengers, with abdominal, lower
back, lumbar spine and pelvic injuries (x2 = 16.3, df = 1, p < 0.0001)
significantly higher for rear seat passengers. For the hospitalised
serious injuries, injuries to the head (x2 = 22.6, df = 1, p < 0.0001),
thorax (x2 = 10.9, df = 1, p < 0.0009), elbow and forearm (x2 = 18.8,
df = 1, p < 0.0001), wrist and hand (x2 = 4.9, df = 1, p < 0.03), knee
and lower leg (x2 = 43.1, df = 1, p < 0.0001), and ankle and foot
(x2 = 17.7, df = 1, p < 0.0001) were all significantly higher for front
vs rear seat passengers, with abdominal, lower back, lumbar spine
and pelvic injuries (x2 = 37.9, df = 1, p < 0.0001) significantly higher
for rear seat passengers (Fig. 1).
4. Discussion
Following improvements to roadways and vehicle crashwor-
thiness (e.g. airbags, stability control, ABS brakes, restrain
mechanisms) fatal injury following vehicle crashes has gradually
declined in Australia (Australian Transport Council, 2011). Evi-
dence suggests that there is a growing proportion of individuals
seriously injured following road trauma that has life-long
consequences for both the injured person and their family (Henley
and Harrison, 2011). However, there are also many individuals who
sustain relatively minor injuries following road trauma and are
also incapacitated for a period of time (Bambach et al., 2012a,b).
Using both the detailed and the broad injury severity categories,
this study identified that there was an increased odds of a rear seat
passenger sustaining an injury of a higher severity compared to a
front seated passenger. Overall, where the vehicle occupant was
older (aged 51+ years), travelling in a newer vehicle, where the
176 R.J. Mitchell et al. / Accident Analysis and Prevention 82 (2015) 171–179

Table 4
Ordinal logistic mixed model odds ratios rear seat compared to front seat passengers and 95% confidence intervals by broad injury severity categories, linked police-report, ED
presentation and hospital admission records in NSW, 2001–2011.

Ordinal logistic mixed Logistic mixed model: minor/moderate injury vs other Logistic mixed model: serious injury/death vs other
modela injuryb/ ED presentation (no hospital admission) injuryb/ ED presentation (no hospital admission)

OR 95% CI OR 95% CI OR 95% CI

Rear seat passengerc 1.12 1.00–1.25 1.10 0.95–1.26 1.21 1.01–1.46

Age groupd
16–50 years 1.36 1.15–1.61 1.14 0.93–1.40 1.71 1.27–2.29
51+ Years 2.57 2.06–3.22 1.86 1.41–2.44 3.62 2.51–5.21

Gendere
Female 1.04 0.93–1.16 1.11 0.97–1.28 0.95 0.79–1.14

Vehicle model yearf

1990–1996 1.08 0.87–1.32 1.14 0.89–1.46 1.06 0.77–1.46

1997 1.30 1.08–1.57 1.35 1.08–1.69 1.35 1.01–1.80

Air bagg
Airbag deployed 1.24 1.03–1.50 1.39 1.10–1.75 1.26 0.94–1.70

Speed limit (km/h)h

60 0.98 0.80–1.21 0.93 0.73–1.19 1.10 0.79–1.54
70–90 1.49 1.18–1.87 1.18 0.90–1.55 1.88 1.31–2.69
100–110 2.17 1.73–2.74 1.79 1.36–2.35 2.86 2.00–4.09
a
Measuring higher risk of severity in this order: other injury/ED presentation (no hospital admission), minor injury (hospital admission)/moderate injury (hospital
admission), serious injury (hospital admissions)/death.
b
Identified by police as injured and no ED presentation or hospitalisation.
c
Front seat passenger is referent.
d
9–15 year age group is referent.
e
Male is referent.
f
<1990 vehicle model is referent.
g
No airbag, not deployed or fitted is referent.
h
40–50 km/h speed limit is referent.

Table 5
Adjusted risk ratios of rear seat compared to front seat passengers and 95% confidence intervals by type of injury for only hospitalised individuals (i.e. minor/moderate/serious
injuries), linked police-report and hospital admission records in NSW, 2001–2011.

Head Neck Thorax Abdomen, lower back, lumbar spine and Upper extremities Lower
pelvis extremities

ARR 95% CI ARR 95% CI ARR 95% CI ARR 95% CI ARR 95% CI ARR 95% CI
All crash types 0.44*** 0.28–0.69 1.23* 1.01–1.50 1.37 0.89–2.13 2.81*** 1.82-4.34 0.73** 0.62–0.87 0.58* 0.38–0.90

Type of crash
Vehicle–vehicle—head-on 0.44 0.13–1.50 2.12** 1.28–3.49 1.08 0.73–1.60 8.10*** 3.14–20.88 0.59** 0.39–0.88 0.08 0.03–0.25
Vehicle–vehicle—angle 1.21 0.89–1.65 1.16 0.84–1.61 0.71** 0.52–0.98 1.12 0.81–1.55 0.63** 0.46–0.86 1.01 0.74–1.37
Vehicle–vehicle—rear-end 1.13 0.56–2.28 1.13 0.61–2.10 0.83 0.37–1.87 1.27 0.60–2.70 1.01 0.48–2.10 0.74 0.31–1.75
Vehicle–rollover 0.77 0.47–1.29 0.73 0.36–1.48 1.66 0.92–3.01 1.71 0.76–3.83 1.01 0.58–1.74 1.53 0.87–2.68
Vehicle–object 0.63* 0.47–0.83 1.19 0.83–1.69 1.07 0.80–1.44 3.14** 1.52–6.47 0.87 0.66–1.13 0.78 0.59–1.04
*
p < 0.05.
**
p < 0.01.
***
p < 0.0001.

airbag deployed, and where the vehicle was driven where the
speed limit was
70 km/h there was a higher odds of sustaining a
higher injury severity. In particular, there were higher odds of rear
seat passengers sustaining serious injuries or serious injuries/
death compared to more minor injuries for front seat passengers
for both the detailed and broad levels of injury severity,
respectively. This finding highlights that rear seat occupant
protective mechanisms should be examined with a view to
improvement (Bilston et al., 2010, Brown and Bilston, 2014).
The odds of serious injury compared to a minimal injury for
older passengers (i.e. 16–50 and 51+ years) was higher compared to
individuals aged 9–15 years. Similarly, Bilston et al. (2010)
identified higher risk ratios for older individuals who sustained
serious injuries. They also found that individuals aged 51+ years
had better protection as a front than a rear seat passenger,
particularly in vehicles manufactured between 1997 and 2007
compared to earlier model vehicles, largely due to safety
improvements found in more modern vehicles (Bilston et al.,
2010).
The current study found an increased injury risk overall and for
serious injury compared to other minimal injuries for rear
compared to front seat passengers travelling in more recently
manufactured vehicles. Similarly, Bilston et al. (2010) found a
higher risk of serious injury for individuals who were sitting in the
rear passenger seats in newer vehicles (i.e. manufactured between
1997 and 2007) compared to older vehicles (i.e. manufactured
between 1990 and 1996) that crashed. In the current study, the
increased risk of serious injury for rear seat passengers is likely to
be due to the introduction of safety features aimed at front seat
protection, such as frontal airbags. On 1 July 1995, the Australian
 R.J. Mitchell et al. / Accident Analysis and Prevention 82 (2015) 171–179
Fig. 1. Proportion of type of injury for rear and front seat passengers for only hospitalised individuals (i.e. minor/moderate/serious injuries), linked police-report and hospital
admission records in NSW, 2001–20111.
1
Includes multiple injuries per person.
Design Rule (ADR) Full Frontal Vehicle Protection No. 69
commenced. This ADR required vehicle manufacturers to design
to meet specific injury level criteria for vehicle crashes. The exact
design specification choices were left to manufacturers, with
manufacturers appearing to prefer to include airbags to meet the
ADR (Fitzharris et al., 2006). In the current study, airbags did not
deploy in 21.9% of crashes. In these cases, either the crash was
relatively minor or the type of collision did not trigger the airbag.
Where there was an airbag present, the most common types of
crashes where the airbag did not deploy were rear-end collisions
(86.7%), rollovers (66.7%), and vehicle–vehicle angle collisions
(48.6%).
Using the detailed injury severity categories, this study did not
find a significant increased odds of an ED presentation compared to
other minimal injury for rear vs front seat passengers. This finding
could be influenced by the current pre-hospital management of
major trauma protocol in NSW (Ambulance Service of New South
Wales, 2008). The protocol uses the MIST trauma criteria (i.e.
Mechanism of injury, Injuries, Signs and Symptoms, and Transport)
to guide patient transport decision-making. The mechanism of
injury criteria states that a patient should be transported to a
trauma centre if there was a death in the same vehicle, intrusion
into occupant compartment >30 cm, steering wheel deformity,
patient side impact, or any rapid deceleration mechanism that
results in a large inertia change at impact. This could partly explain
why individuals with relatively minor or no apparent injury were
transported to a hospital ED. Therefore, it is possible that ED
presentations may be influenced by the pre-hospital trauma
management protocol, making it likely that there is little to
177
discriminate other minimal injury from ED presentations without
hospital admission. Having an objective measure of injury severity
routinely available in the ED would assist in this discrimination. It
is also possible that examinations of crash and injury data may also
be less likely to uncover significant differences in injury risk
between rear and front seat car passengers for relatively minor
injuries. Overall, using the broad injury severity categories
identified a similar pattern of higher odds for key characteristics
as using the detailed injury severity categories for rear compared to
front seat passengers.
The differences identified for injury type for minor, moderate
and serious injury showed that abdominal, lower back, lumbar
spine and pelvic injuries were significantly higher for rear
compared to front seat car passengers for each severity category.
Abdominal injuries have previously been identified as common for
rear seat passengers (Brown and Bilston, 2014). For front seat
passengers, injury type varied by injury severity, with knee and
lower leg injuries more common for front seat passengers with
minor and serious injuries and injuries to the thorax, elbow and
forearm, and wrist and hand more common for front seat
passengers with moderate and serious injuries. Further research
needs to be undertaken with a larger sample size in order to
indicate differences in injuries sustained and their severity by
seating position and crash type.
There are several limitations of the current study. The analysis
only included road crashes that occurred on a public roadway in
NSW that were reported to police and where the individual
presented or was admitted to hospital, so will under-enumerate
the total number of passenger vehicle crashes. Data completeness
178 R.J. Mitchell et al. / Accident Analysis and Prevention 82 (2015) 171–179
for passenger age and gender, presence of airbag and year of
vehicle manufacture was an issue with some records excluded due
to missing values and this might have resulted in some selection
bias. Data validity was not able to be assessed and it is possible that
there could be some misclassification in records. Detailed
information regarding type of restraint, such as whether the seat
belt was a shoulder and lap or lap-only, were not recorded. No
information was available regarding intrusion into the vehicle
passenger compartment, nor was any information available
regarding the crash that would allow investigation of a vehicle
drivers and their potential self-protection tendency.
The identification of other minimal injury was reliant upon
police identification of injured passengers. It is possible that other
minimal injury represented passengers with minimal injuries or
these may have been passengers with more serious injuries and
their police casualty and health records did not link. The
identification of injuries by police has been shown to involve
misclassification of injury severity (Cryer et al., 2001; Amoros et al.,
2007; UK Department for Transport, 2011). Injury severity for a
matched-cohort analysis with multiple rear/front seat passengers
with a vehicle strata would best be analysed using multinominal
methods with a repeated function for vehicle. This type of analysis
method is not currently supported by main statistical software.
Although not ideal, using GLIMMIX circumvented the problem by
using the random effects statement to control for correlation
within vehicle. For some analyses, there were relatively small
sample sizes, particularly for deaths. The ED presentation data
used for this exploratory study was not ideal. There was variation
in the number of EDs participating in data collection over time, and
data was only available from 2005 for data linkage purposes.
Using the matched-pair analysis for this study enabled the
control of any possible vehicle and crash-related confounding
characteristics. However, using the matched-pair regression
methodology only produces estimates where the matched-pairs
are discordant on the outcome of interest (Cummings et al., 2003a,
b). Absolute relative risks were not able to be estimated, as
information was only available regarding injured individuals. This
research did not consider young child passengers and there are
likely to be specific injury risk issues related to passenger seating
for this group (Braver et al., 1998; Lennon et al., 2008). Lastly, when
using record linkage there is likely to be some degree of error in the
data linkage process. In the current study, the CHeReL estimates
the false positive rate for this linkage to be 0.4% (i.e. the proportion
of false matches) and estimated the rate of false negatives at 0.5%
(i.e. failure to identify matches) (Centre for Health Record Linkage,
2012).
5. Conclusion
Rear seat car passengers are sustaining injuries of a higher
severity compared to front seat passengers travelling in the same
vehicle, as well as when travelling in newer vehicles and when the
front seat occupant is shielded by an airbag deployed in the crash.
Rear seat occupant protective mechanisms should be examined
with a view to improvement. Pre-hospital trauma management
policies could influence if an individual, whether injured or not, is
transported to a hospital ED for either observation or treatment.
Therefore, it would be beneficial to have an objective measure of
injury severity routinely available in ED records. Further examina-
tion of injury severity categories is warranted to examine less
severe non-fatal injuries by car seating position and vehicle
intrusion.
6. Conflict of interest
Nil.
Acknowledgements
The authors wish to thank the NSW Ministry of Health for
providing access to the EDDC, the APDC, the CHeReL for conducting
the record linkage, particularly Katie Irvine, Michael Smith, Reinier
de Vos and Cho Phyu, and the Centre for Road Safety at Transport
for NSW for providing access to CrashLink.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.aap.2015.05.023.
References
Ambulance Service of New South Wales, 2008. Pre-hospital Management of Major
Trauma. Protocol T1. Trauma Triage Tool. Major Trauma Criteria (MIST).
Ambulance Service of New South Wales, Rozelle.
Amoros, E., et al., 2007. Road crash casualties: characteristics of police injury
severity misclassification. J. Trauma 62 (2), 482–490.
Australian Bureau of Statistics, 2013. Motor Vehicle Census. Catalogue 9309.0.
Australian Bureau of Statistics, Canberra.
Australian Transport Council, 2011. The National Road Safety Strategy, 2011–2020.
Australian Transport Council, Canberra.
Australian Transport Safety Bureau, 2003. Road Crash Data and Rates, Australian
States and Territories, 1925 to 2002. ATSB, Canberra.
Bambach, M., et al., 2012a. Road Trauma and Determinants of Injury Outcome for
Road Users. Transport and Road Safety (TARS) Research, University of New
South Wales, Kensington.
Bambach, M., et al., 2012b. Injury Severity Indicators and Their Impact on Data
Linkage. Transport and Road Safety (TARS) Research, University of New South
Wales, Kensington.
Bilston, L.E., et al., 2010. A matched-cohort analysis of belted front and rear seat
occupants in newer and older model vehicles shows that gains in front occupant
safety have outpaced gains for rear seat occupants. Accid. Anal. Prev. 42, 1974–
1977.
Braver, E., et al., 1998. Seating positions and children’s risk of dying in motor vehicle
crashes. Inj. Prev. 4, 181–187.
Brown, H., Prescott, R., 2006a. Generalised linear mixed models. Applied Mixed
Models in Medicine. West Sussex, John Wiley & Sons, pp. 107–152.
Brown, H., Prescott, R., 2006b. Mixed models for categorical data, In: Brown, H.,
Prescott, R. (Eds.), Applied Mixed Models in Medicine. 2nd ed. John Wiley &
Sons, West Sussex, pp. 153–181.
Brown, J., Bilston, L.E., 2014. The scope and nature of injuries to rear seat passengers
in NSW using linked hospital admission and police data. Traffic Inj. Prev. 15,
462–469.
Brown, J., et al., 2006. Serious injury is associated with suboptimal restraint use in
child motor vehicle occupants. J. Paediatr. Child Health 42 (6), 345–349.
Bureau of Infrastructure Transport Regional and Economics, 2012. Road Deaths
Australia, 2011 Statistical Summary. Canberra, Bureau of Infrastructure
Transport and Regional Economics.
Centre for Health Record Linkage, CHeReL: Quality Assurance. Retrieved 14/7/2012,
2012. from http://www.cherel.org.au/quality-assurance.
Choicemaker Technologies, Open Source Choicemaker Technology. Retrieved 10/7/
2011, 2011. Available from http://oscmt.sourceforge.net/.
Cryer, C., et al., 2004. Developing Valid Injury Outcome Indicators. A Report for the
New Zealand Injury Prevention Strategy. Injury Prevention Research Unit,
University of Otago, Dunedin.
Cryer, C., et al., 2001. Investigation of bias after data linkage of hospital admissions
data to police road traffic crash reports. Inj. Prev. 7, 234–241.
Cummings, P., et al., 2003a. Matched-pair cohort methods in traffic crash research.
Accid. Anal. Prev. 35, 131–141.
Cummings, P., et al., 2003b. Estimating seat belt effectiveness using matched-pair
cohort methods. Accid. Anal. Prev. 35, 143–149.
Davie, G., et al., 2008. Improving the predictive ability of the ICD-based injury
severity score. Inj. Prev. 14, 250–255.
Dayal, S., et al., 2008. Mapping Injury Severity Scores Against Hospitalisation Day
Stays for Injury Priority Areas (Excluding Workplace Injury). Public Health
Intelligence, Health and Disability Systems Strategy Directorate, Ministry of
Health, Wellington.
Fitzharris, M., et al., 2006. Crash-based Evaluation of Australian Design Rule 69 (Full
Frontal Impact Occupant Protection). Australian Transport Safety Bureau,
Canberra.
Henley, G., Harrison, J., 2011. Trends in Serious Injury Due to Land Transport
Accidents, Australia 2000–01 to 2007–08. Injury Research and Statistics Series.
Australian Institute of Health and Welfare, Canberra.
International Health Terminology Standards Development Organization, Snomed-
CT. Retrieved 10/7/2011, 2011. Available from http://www.ihtsdo.org/.
Lennon, A., et al., 2008. Rear seat safer: seating position, restraint use and injuries in
children in traffic crashes in Victoria, Australia. Accid. Anal. Prev. 40, 829–834.
 R.J. Mitchell et al. / Accident Analysis and Prevention 82 (2015) 171–179
Mayrose, J., Priya, A., 2008. The safest seat: effect of seating position on occupant
mortality. J. Saf. Res. 39, 433–436.
National Center for Health Statistics, International Classification of Diseases, Ninth
Revision, Clinical Modification (ICD-9-CM), 2011. Available from http://www.
cdc.gov/nchs/icd/icd9cm.htm
National Centre for Classification in Health, 2006. ICD-10-AM. National Centre for
Classification in Health, Sydney.
Sahraei, E., Digges, K., 2009. Trend of Rear Occupant Protection in Frontal Crashes
over Model Years of Vehicles. SAE World Congress, Detroit, SAE.
Sahraei, E., et al., 2009. Protection of rear seat occupants in frontal crashes,
controlling for occupant and crash characteristics. Strapp Car Crash J. 53, 75–
91.
Sahraei, E., et al., 2010. Reduced protection for belted occupants in rear seats relative
to front seats of new model year vehicles. Ann. Adv. Autom. Med. 54, 149–158.
Sahraei, E., et al., 2014. High strength steels, stiffness of vehicle front-end structure,
and risk of injury to rear seat occupants. Accid. Anal. Prev. 66, 43–54.
SAS Institute, 2014. SAS: statistical software, version 9.4. SAS Institute, Cary, North
Carolina.
179
Smith, K.M., Cummings, P., 2004. Passenger seating position and the risk of
passenger death or injury in traffic crashes. Accid. Anal. Prev. 36 (2), 257–260.
Smith, K.M., Cummings, P., 2006. Passenger seating position and the risk of
passenger death in traffic crashes: a matched cohort study. Inj. Prev. 12, 83–86.
Stephenson, S., et al., 2004. Diagnosis based injury severity scaling: investigation of
a method using Australian and New Zealand hospitalisations. Inj. Prev. 10, 379–
383.
UK Department for Transport, 2011. Linking Police and Hospital Data on Road
Accidents in England: 1999 to 2009 results. UK Department for Transport,
London.
Willis, C., et al., 2010. Predicting trauma patient mortality: ICD [or ICD 10 AM] versus
AIS based approaches. ANZ J. Surg. 80 (11), 802.
World Health Organization, 1977. ICD-9 International Classification of Diseases, 9th
revision World Health Organization, Geneva.
World Health Organization, 1992. ICD-10 International Classification of Diseases,
10th revision World Health Organization, Geneva.
World Health Organization, 2008. The Global Burden of Disease 2004 update. World
Health Organization, Geneva.