ArticlePDF Available

Exposure-Based, ‘Like-for-Like’ Assessment of Road Safety by Travel Mode Using Routine Health Data

Authors:

Abstract and Figures

Background: Official reports on modal risk have not chosen appropriate numerators and denominators to enable like-for-like comparisons. We report age- and sex-specific deaths and injury rates from equivalent incidents in England by travel mode, distance travelled and time spent travelling. Methods: Hospital admissions and deaths in England 2007-2009 were obtained for relevant ICD-10 external codes for pedestrians, cyclists, and car/van drivers, by age-group and sex. Distance travelled by age-group, sex and mode in England (National Travel Survey 2007-2009 data) was converted to time spent travelling using mean trip speeds. Fatality rates were compared with age-specific Netherlands data. Results: All-age fatalities per million hours' use (f/mhu) varied over the same factor-of-three range for both sexes (0.15-0.45 f/mhu by mode for men, 0.09-0.31 f/mhu for women). Risks were similar for men aged 21-49 y for all three modes and for female pedestrians and drivers aged 21-69 y. Most at risk were: males 17-20 y (1.3 f/mhu (95% CI 1.2-1.4)) for driving; males 70+ (2.2 f/mhu(1.6-3.0)) for cycling; and females 70+ (0.95 f/mhu (0.86-1.1)) for pedestrians. In general, fatality rates were substantially higher among males than females. Risks per hour for male drivers <30 y were similar or higher than for male cyclists; for males aged 17-20 y, the risk was higher for drivers (33/Bn km (30-36), 1.3 f/mhu (1.2-1.4)) than cyclists (20/Bn km (10-37), 0.24 f/mhu (0.12-0.45)) whether using distance or time. Similar age patterns occurred for cyclists and drivers in the Netherlands. Age-sex patterns for injuries resulting in hospital admission were similar for cyclists and pedestrians but lower for drivers. Conclusions: When all relevant ICD-10 codes are used, fatalities by time spent travelling vary within similar ranges for walking, cycling and driving. Risks for drivers were highest in youth and fell with age, while for pedestrians and cyclists, risks increased with age. For the young, especially males, cycling is safer than driving.
Content may be subject to copyright.
Exposure-Based, ‘Like-for-Like’ Assessment of Road
Safety by Travel Mode Using Routine Health Data
Jennifer S. Mindell
1
*, Deborah Leslie
2
, Malcolm Wardlaw
3
1Research Department of Epidemiology and Public Health, UCL (University College London), London, United Kingdom, 2Centre for Physical Activity and Nutrition
Research, Deakin University, Burwood, Australia, 3Edinburgh, United Kingdom
Abstract
Background:
Official reports on modal risk have not chosen appropriate numerators and denominators to enable like-for-
like comparisons. We report age- and sex-specific deaths and injury rates from equivalent incidents in England by travel
mode, distance travelled and time spent travelling.
Methods:
Hospital admissions and deaths in England 2007–2009 were obtained for relevant ICD-10 external codes for
pedestrians, cyclists, and car/van drivers, by age-group and sex. Distance travelled by age-group, sex and mode in England
(National Travel Survey 2007–2009 data) was converted to time spent travelling using mean trip speeds. Fatality rates were
compared with age-specific Netherlands data.
Results:
All-age fatalities per million hours’ use (f/mhu) varied over the same factor-of-three range for both sexes (0.15–
0.45 f/mhu by mode for men, 0.09–0.31 f/mhu for women). Risks were similar for men aged 21–49 y for all three modes and
for female pedestrians and drivers aged 21–69 y. Most at risk were: males 17–20 y (1.3 f/mhu (95% CI 1.2–1.4)) for driving;
males 70+(2.2 f/mhu(1.6–3.0)) for cycling; and females 70+(0.95 f/mhu (0.86–1.1)) for pedestrians. In general, fatality rates
were substantially higher among males than females. Risks per hour for male drivers ,30 y were similar or higher than for
male cyclists; for males aged 17–20 y, the risk was higher for drivers (33/Bn km (30–36), 1.3 f/mhu (1.2–1.4)) than cyclists (20/
Bn km (10–37), 0.24 f/mhu (0.12–0.45)) whether using distance or time. Similar age patterns occurred for cyclists and drivers
in the Netherlands. Age-sex patterns for injuries resulting in hospital admission were similar for cyclists and pedestrians but
lower for drivers.
Conclusions:
When all relevant ICD-10 codes are used, fatalities by time spent travelling vary within similar ranges for
walking, cycling and driving. Risks for drivers were highest in youth and fell with age, while for pedestrians and cyclists, risks
increased with age. For the young, especially males, cycling is safer than driving.
Citation: Mindell JS, Leslie D, Wardlaw M (2012) Exposure-Based, ‘Like-for-Like’ Assessment of Road Safety by Travel Mode Using Routine Health Data. PLoS
ONE 7(12): e50606. doi:10.1371/journal.pone.0050606
Editor: Hamid Reza Baradaran, Tehran University of Medical Sciences, Iran (Republic of Islamic)
Received June 27, 2012; Accepted October 24, 2012; Published December , 2012
Copyright: ß2012 Mindell et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This study received no funding.
Competing Interests: The authors have declared that no competing interests exist.
* E-mail: j.mindell@ucl.ac.uk
Introduction
Travel can provide many health benefits through access to
facilities, goods, and people. Active travel, i.e. walking and cycling,
can make additional large contributions to population health: daily
repetitive and necessary physical activity, such as commuting, can
have the greatest health benefits as these are more successful and
durable over long periods [1]. Cycling as a mode of commuting
has additional advantages for society, including reducing carbon
emissions, noise levels, and congestion on roads and other public
transport systems [2,3]. Increasing active travel in England and
Wales is estimated to save £17 bn in healthcare costs alone [4].
Despite these documented benefits and some increases in cycling
in several cities with specific interventions [5,6], the UK has no
nationwide cycling revival.
Perceived road danger is a strong disincentive to cycling [7];
many cyclists do not ride on the road due to safety concerns [8].
However, research regarding the safety of cycling tends to be
distorted by a number of substantial errors which are found
repeatedly in published papers and policy documents. These fall
into three main categories:
Nnot accounting for different types of journey undertaken in
each mode, notably long-distance car travel, which has no
comparison in walking or cycling, unless train travel is
included;
Nchoice of a misleading denominator, such as comparing
cycling fatality rates internationally using population size as the
denominator [9,10];
Nnot selecting comparable numerators, that is, failing to include
all transport casualties and exclude non-transport casualties.
Concerning the first two errors, the importance of using the
most appropriate measure of exposure has been demonstrated in
inter-country comparisons [10,11]. Risk by distance travelled does
not capture large differences in average speed, which enable
differential mobility for drivers, cyclists, and pedestrians. As the
speed differential between cars and bicycles is not great for local
PLOS ONE | www.plosone.org 1 December 2012 | Volume 7 | Issue 12 | e50606
5
journeys, time-based comparison minimises the distorting impact
of misleading comparisons with long distance car journeys. The
average ‘‘main driver’’ travels 7,034 miles, plus a further 1,254
miles as a passenger; the average cyclist rides about 830 miles
annually [12].
Surveys from many countries show that time spent travelling
has remained fairly constant over time at about one hour per day,
despite large changes in modal split [13]. This supports the use of
risk based on time as being most appropriate for comparing modes
with different average speeds.
The third category of error is to include off-highway falls and
children injured at play as cycling transport injuries, whilst
omitting all falls of pedestrians (whether on- or off-highway). This
substantially overstates cycling injuries and understates pedestrian
injuries (Figure 1).
In the UK, most drivers are middle-aged adults, with almost
half being female, while cycling is male-dominated and has
traditionally been considered predominantly an activity of youth:
24% of serious casualties among cyclists are ,20 y old, compared
with 12% of serious casualties among drivers [9]. The increased
risk-taking associated with both young age and male sex are well-
known and confound comparisons of cycling and driving.
The two sources of data on transport injuries, STATS 19 and
Hospital Episode Statistics (HES), include different categories of
incident but neither is totally comparable across different travel
modes (Figure 1). Other difficulties for like-with-like comparisons
are described in the methods section below. Not all these problems
are amenable to investigation using routine data, but some can be
addressed.
Because of these problems, we undertook a study based on
mortality data and HES, taking account of the inconsistencies in
the classification of the data with the aim of filling this important
gap. This paper provides age- and sex-specific deaths and injuries
in England by travel mode, using both distance and time as
measures of exposure. We considered both the ICD external codes
included in official travel risk data and those currently excluded
but required for comparative assessment, successfully providing
more accurate estimates on which to base comparison of travel
risks by mode. We also compared our findings with data from the
Netherlands.
Methods
Definition of ‘‘risk’’
In this paper, ‘‘risk’’ is defined as the observed number of
casualties per unit of exposure. We considered two severities of
injury - fatal and hospital admission - and two units of exposure:
distance travelled (billion km) and time spent travelling (million
hours). Most emphasis was given to risk as ‘‘fatalities per million
hours’ use’’ (f/mhu). ‘‘Hospital admission’’ does not capture
differing severities of injury between modes of travel: mean length
of hospital stay after traffic collision was 2.9 d for cyclists and
drivers, but 4.7 d for pedestrians [17].
Denominator Data
The National Travel Survey (NTS), the primary source of data
on personal travel patterns in Great Britain, includes 15,000
households each year. The nationally-representative sample is
selected annually using a stratified two-stage random probability
sample of private households. Data collection comprises a face-to-
face interview and a one-week self-completed written travel diary
[12]. The Department for Transport also measures vehicle traffic
using on-road manual and automatic traffic counts. During the last
decade, the bicycle traffic estimates from these two independent
sources have agreed within 10% [18]. However, both data sources
exclude travel on routes not usable by motor vehicles, thus
underestimating travel by foot and by bicycle.
The Department for Transport NTS team provided the authors
with aggregated per capita distance travelled by age-group, sex
and mode in England from NTS data for each year 2007–2009,
weighted to be nationally-representative. Total distance travelled
by mode over the three years for each age- and sex-specific group
was obtained by multiplying the distance for each age/sex group
by the estimated population for that group and year.
Data for mean trip speeds (walking 4?2 km/h, cycling 12?2 km/
h, driving 39 km/h) were available from the DfT [19]. More
detailed data were not available, so the same speeds were used for
all age- and sex-specific groups.
Numerator Data
Defining ‘travel injuries/fatalities’. ICD-10 external
codes were grouped by type of incident; the fourth digit was used
to distinguish between different locations (e.g. on-highway, off-
highway, or unspecified) or other details. Table S1 shows the full
specification of the ICD-10 codes included in each group, and
identifies which groups are included/excluded from STATS19
statistics and NHS transport-related casualty data. Data were
sought for pedestrians, cyclists, and car/van drivers but not vehicle
passengers, using the age-groups for which NTS travel data were
provided. Non-transport casualties, such as pedestrian falls at work
or off-highway cyclist falls, were excluded as much as possible by
specifying the four-digit ICD-10 codes required (see table S1). This
enabled compatibility with denominator data for distance travelled
on highways only.
Casualty data. Numbers of hospital admissions, from Hos-
pital Episodes Statistics (HES) data [17], and deaths in England
aggregated over 2007–2009 were provided by the London Health
Observatory for each group of ICD codes specified in Table S1, by
age-group and sex. The three year period was used to improve
precision and data non-disclosiveness, especially for cycling.
Netherlands Data
Published age-specific fatality rates by distance travelled for car
occupants and cyclists from the Netherlands [20] were converted
to f/mhu using mean travel speeds produced by the Centraal
Bureau voor de Statistiek [21].These were: walking 5.5 km/h;
cycling 13.4 km/h; and driving 44.4 km/h.
Analysis
Age- and sex-specific fatality and injury rates for England were
calculated by dividing the casualty data (numbers aggregated over
the three years) for each age- and sex-specific group by the
distance travelled by that group over the three-year period. This
was also calculated for all persons by age-group. These rates per
unit distance were then converted into rates per unit time using
mean trip speed data for each mode, for both countries.
All analyses were conducted in Excel 2010. The 95%
confidence intervals for Poisson parameters were calculated for
the English data using the formulae for weighted sums [22].
Data Sharing Statement
The aggregated age- and sex-specific data can be obtained from
the data holders or, with the data holders’ permission, from the
authors. Anonymised National Travel Survey data are available
from the UK Data Service. Mortality data and Hospital Episodes
Statistics data are available from the Health and Social Care
Information Centre.
Like-for-Like Assessment of Road Safety by Mode
PLOS ONE | www.plosone.org 2 December 2012 | Volume 7 | Issue 12 | e50606
Figure 1. Sources of data on transport injuries and fatalities.
doi:10.1371/journal.pone.0050606.g001
Like-for-Like Assessment of Road Safety by Mode
PLOS ONE | www.plosone.org 3 December 2012 | Volume 7 | Issue 12 | e50606
Results
England
The ICD system allows for incomplete data to be coded in more
general categories such as ‘‘unspecified occupant in unspecified
transport accident’’ (eg V49.9). These are known as ‘‘dustbin’’
codes. All results were affected by dustbin coding. For drivers, an
unknown number of passengers will have been included as
‘‘unspecified occupants’’. For cyclists, an unknown number of non-
transport casualties will have been included in code V19.8. The
results for drivers and cyclists should thus be seen as pessimistic.
For pedestrians, an unknown number of on-highway falls will have
been excluded as ‘‘falls in unspecified location’’ (eg W19.9). The
results for walking are thus underestimates.
Caution is therefore advised when considering results for falls in
highway for walking and cycling, especially in older age groups.
The results apparently show more cyclist deaths due to falls than
traffic crashes in age groups 50–59 y and 70+y (Tables 1 and 2)
but it is not certain that all these casualties were transport-related.
In contrast, for walking, some of the 434 fatal falls in unspecified
location for the 70+y group will have been transport-related.
For both walking and cycling by under-17 s, only a very small
number of deaths due to falls occurred. It is thus difficult to draw
conclusions about fatal risk without a longer sample period. For
serious injuries, i.e. those resulting in hospital admission, the risk of
falls appears to be higher for cycling than walking (Tables 3 and 4).
While this may be the case, the large proportion of ‘‘unspecified
location’’ walking falls must be noted. In addition, for child
cyclists, distinguishing falls at play in the street from falls during
travel is particularly difficult. The risks for under-age drivers are
extremely high. Although unlicensed drivers face high risks - and
impose high risks on others [23] - the results in this study are over-
estimates due to under-recording of the denominator resulting
from the difficulty of measuring an illegal activity. These results
must be treated with caution.
The principal outcome, fatalities per million hours’ use, varied
substantially by age and sex. For most age groups, the rate was
within a similar range for all three modes. Disaggregated by age,
risks for all three modes were similar for men aged 21–49 and for
female pedestrians and drivers aged 21–69 y (Figure 2). The
change in risk with age followed different patterns by mode. For
drivers, the risk was highest in youth, falling by a factor of 20 into
middle age before rising again for 70+. For walking and cycling,
the variation in risk with age was much less, generally increasing
with age. The exception was female cyclists aged 17–20 y,
although this was a small group and the result could be chance
Table 1. Fatality numbers and rates per distance travelled, by travel mode, age, and type of incident, Males, England 2007–2009.
Mode Summary description Age-group
,17 17–20 21–29 30–39 40–49 50–59 60–69 70+ALL
Drive 3 yr distance (Mn km) 14 11,354 59,462 100,915 128,207 102,998 67,160 34,383 504,493
Driver Collision Fatality Drive-RTA 3114 164 108 91 71 66 122 739
Driver Single vehicle fatality Drive-RTA (single vehicle) 6158 221 122 76 51 28 35 697
Unspecified occupant unspecified
accident
a
Drive-RTA (unspecified) 36 101 122 65 51 47 24 48 494
On-highway fatality rate (per Bn km)
bc
3,214
d
33 8.5 2.9 1.7 1.6 1.8 6.0 3.8
95% CIs 2,3454,301 3036 7.89.3 2.63.3 1.5–.91.41.9 1.52.1 5.26.8 3.74.0
Cycle 3 yr distance (Mn km) 925 497 1,151 1,555 2,115 1,025 520 258 8,046
3 yr Cycle-RTA (n) Collision 36 6 20 33 26 23 20 21 185
3 yr Cycle-fall (n) On-highway
e
6 491710271026109
Off-highway
f
2 01110308
On-highway fatality rate (per Bn km)
e
45 20 25 32 17 49 58 182 37
95% CIs 3361 1037 1736 2442 1224 3664 3982 134243 3241
Walk 3 yr distance (Mn km) 5,132 1,464 2,951 2,827 2,638 2,342 2,115 1,561 21,030
3 yr Walk-RTA (n) Specified 70 50 75 73 62 54 60 185 629
Unspecified 31 21 44 39 47 47 30 96 355
3 yrn Walk-fall (n) On Highway 0 2 4 16 36 44 54 148 304
Other specified location
f
4 3119 13161745118
Unspecified Location
fg
2 1 5 4 11 32 42 434 531
On-highway fatality rate (per Bn km)
g
20 50 42 45 55 62 68 275 61
95% CIs 1624 3963 3550 3854 4665 5273 5780 249302 5865
a
This group may also contain passengers but probably notmany as ‘‘unspecified occupant’’ was rare for collisions.These have been included in the fatality rate estimate.
b
Estimates are likely to be a little too high due to the assumption that unspecified occupants are all drivers: an unknown proportion would have been passengers,
especially for females.
c
Note that these averages include both local roads and motorways/multi-lane divided roadways, where fatality rates are an order of magnitude lower than general
purpose roads, but data are not available by age and sex.
d
This figure is greatly exaggerated by under-measurement of under-aged driving.
e
These figures are too high as V19.8 is a dustbin code, including some off-highway falls.
f
Not included in the fatality rate.
g
Estimates are too low, as some ‘unspecified location’ deaths will have been on-highway.
doi:10.1371/journal.pone.0050606.t001
Like-for-Like Assessment of Road Safety by Mode
PLOS ONE | www.plosone.org 4 December 2012 | Volume 7 | Issue 12 | e50606
fluctuation, given the small casualty numbers and distances cycled.
The group most at risk for each mode were: male drivers aged 17–
20 y (1.3 f/mhu (95% CI 1.2–1.4)); male cyclists aged 70+(2.2 f/
mhu (1.6–3.0)): and female pedestrians aged 70+(0.95 f/mhu
(0.86–1.1), Figure 2). In general, fatality rates were substantially
higher amongst males than females, except for drivers aged 60+.
Risks per hour for male drivers ,30 y were similar or higher than
for male cyclists; for those aged 17–20 y, the risk was higher for
drivers (33/Bn km (95% CI 30–36), 1?3 f/mhu (1.2–1.4)) than
cyclists (20/Bn km (10–37), 0.55 f/mhu (0.4–0.75)) using distance
or time.
Overall rates of hospital admissions per mhu were similar in
men and women, but varied more widely by mode than fatality
rates did (Table 5). A similar U-shaped pattern with age was seen
in both sexes and for all three modes. Males were at higher risk
than females as drivers aged 17–29 y, as cyclists aged 17–39 y, and
as pedestrians aged 17–59 y.
Comparison with the Netherlands
Comparing the fatality risk for cyclists and car users/drivers by
time between the Netherlands and England (Figure 3), a similar
pattern was seen of fatality rates for younger car users/drivers
exceeding those for cyclists for people under 30 y. Both countries
showed a marked increase in fatalities in those aged 70 y and over,
particularly for cyclists. This increase was of the same magnitude
in both countries. However, rates in the Netherlands for cyclists
,70 y were closer to those for drivers than the equivalent data for
England.
Discussion
Principal Findings in Relation to Existing Knowledge
Previous assessments showed that the all-ages risks per hour
vary by country for all travel modes, with the risk for UK cyclists
higher than for drivers or pedestrians [24]. Our study confirms this
is so overall. It also reveals the large variations by age and sex,
Table 2. Fatality numbers and rates per distance travelled, by travel mode, age, and type of incident, Females, England 2007–2009.
Mode Summary description Age-group
,17 17–20 21–29 30–39 40–49 50–59 60–69 70+ALL
Drive 3 yr distance (Mn km) 8 9,816 44,170 61,447 74,397 48,749 24,998 11,466 275,051
Driver Collision
Fatality
Drive-RTA 0 28 57 20 37 21 21 57 241
Driver Single
vehicle fatality
Drive-RTA (single vehicle) 0 30 36 20 16 16 11 17 146
Unspecified
occupant
unspecified
accident
a
Drive-RTA (unspecified) 27 47 35 24 29 25 25 40 252
On-highway fatality rate (per Bn km)
bc
3,375
d
11 2.9 1.0 1.1 1.3 2.3 9.9 2.3
95% CIs 2.2244,910913 2.43.4 0.81.3 0.91.4 1.01.6 1.73.0 8.212 2.12.5
Cycle 3 yr distance (Mn km) 286 74
e
391 399 440 288 118 60 2,056
3 yr Cycle-RTA Collision 5557842339
3 yr Cycle-fall On-highway
f
0013242012
Off-highway
g
000000000
On-highway fatality rate (per Bn km)
f
18 67 15 25 23 28 34 50 25
95% CIs 641 22157 633 1246 1142 1255 987 10146 1833
Walk 3 yr distance (Mn km) 5,022 1,300 3,184 3,568 3,425 2,483 2,024 1,674 22,680
3 yr Walk-RTA
(n)
Specified 47 15 15 19 29 24 31 159 339
Unspecified 19 6 15 12 19 11 11 103 196
3 yrn Walk-fall
(n)
OnHighway 0014857117142
Other specified location
g
001116238 49
Unspecified Location
gh
00015927634 676
On-highway fatality rate (per Bn km)
g13
16 10 10 16 16 24 226 30
95% CIs 1017 1025 714 714 1221 1222 1832 204250 2832
a
This group may also contain passengers but probably notmany as ‘‘unspecified occupant’’ was rare for collisions.These have been included in the fatality rate estimate.
b
Estimates are likely to be a little too high due to the assumption that unspecified occupants are all drivers: an unknown proportion would have been passengers,
especially for females.
c
Note that these averages include both local roads and motorways/multi-lane divided roadways, where fatality rates are an order of magnitude lower than general
purpose roads, but data are not available by age and sex.
d
This figure is greatly exaggerated by under-measurement of under-aged driving.
e
The base for this was much smaller than distances for other ages, sex, and travel modes.
f
These figures are too high as V19.8 is a dustbin code, including some off-highway falls.
g
Not included in the fatality rate.
h
Estimates are too low, as some ‘unspecified location’ deaths will have been on-highway.
doi:10.1371/journal.pone.0050606.t002
Like-for-Like Assessment of Road Safety by Mode
PLOS ONE | www.plosone.org 5 December 2012 | Volume 7 | Issue 12 | e50606
Table 3. Hospital admission numbers and rates per distance travelled, by travel mode, age, and type of incident, Males, England 2007–2009.
Mode Summary description Age-group All ages
,17 17–20 21–29 30–39 40–49 50–59 60–69 70+
Drive 3 yr distance (Mn km) 14 11,354 59,462 100,915 128,207 102,998 67,160 34,383 470,110
Driver Collision Drive-RTA 61 1,202 2,127 1,745 1,706 1,284 964 1,182 10,271
Driver Single vehicle Drive-RTA (single vehicle) 76 1,447 2,053 1,216 888 522 358 634 7,194
Unspecified occupant
unspecified accident
a
Drive-RTA (unspecified) 269 538 745 474 463 304 221 342 3,356
Hospital admission rate (per Bn km)
bc
29,000
d
281 82.8 34.0 23.8 20.5 23.0 62.8 41.3
95% CIs 26,247
31,963
271291 8185 3335 2325 2021 2224 6065 4445
Cycle 3 yr distance (Mn km) 925 497 1,151 1,555 2,115 1,025 520 258 8,046
3 yr Cycle-RTA (n) Collision 1,752 429 838 1,031 1,128 621 324 241 6,364
3 yr Cycle-fall (n) On-highway
e
5,212 883 1,357 1,729 1,589 1,014 636 484 12,904
Hospital admission rate (per Bn km)
e
7,529 2,642 1,906 1,775 1,285 1,595 1,845 2,815 2,395
95% CIs 7,3457,708 2,5012,789 1,8281,988 1,7091,842 1,2371,334 1,5191,674 1,7301,966 2,6133,027 2,3612,429
Walk 3 yr distance (Mn km) 5,132 1,464 2,951 2,827 2,638 2,342 2,115 1,561 21,030
3 yr Walk-RTA (n) Specified 4,692 1,283 2,001 1,595 1,477 1,095 883 1,550 14,576
Unspecified 411 128 223 177 216 167 104 224 1,650
3 yrn Walk-Fall (n) On-Highway 1,940 1,215 2,656 3,022 4,450 4,445 4,917 12,607 35,252
Unspecified Location
fg
16,923 4,571 10,173 10,425 13,046 12,824 14,311 58,613 140,886
Hospital admission rate (per Bn km)
g
1,372 1,794 1,654 1,696 2,329 2,437 2,791 9,213 2,448
95% CIs 1,3411,405 1,7261,864 1,6081,701 1,6481,744 2,2712,388 2,3742,501 2,7212,864 9,0639,365 2,4272,469
a
This group may also contain passengers but probably notmany as ‘‘unspecified occupant’’ was rare for collisions.These have been included in the fatality rate estimate.
b
Estimates are likely to be a little too high due to the assumption that unspecified occupants are all drivers: an unknown proportion would have been passengers.
c
Note that these averages include both local roads and motorways/multi-lane divided roadways, where fatality rates are an order of magnitude lower than general purpose roads, but data are not available by age and sex.
d
This figure is greatly exaggerated by under-measurement of under-aged driving.
e
These figures are too high as V19.8 is a dustbin code, including some off-highway falls.
f
Not included in the fatality rate.
g
Estimates are too low, as some ‘unspecified location’ deaths will have been on-highway.
doi:10.1371/journal.pone.0050606.t003
Like-for-Like Assessment of Road Safety by Mode
PLOS ONE | www.plosone.org 6 December 2012 | Volume 7 | Issue 12 | e50606
Table 4. Hospital admission numbers and rates per distance travelled, by travel mode, age, and type of incident, Females, England
2007–2009.
Mode Summary description Age-group All ages
,17 17–20 21–29 30–39 40–49 50–59 60–69 70+
Drive 3 yr distance (Mn km) 8 9,816 44,170 61,447 74,397 48,749 24,998 11,466 275,051
Driver Collision Drive-RTA 65 686 1,510 1,223 1,152 898 574 801 6,909
Driver Single vehicle Drive-RTA (single
vehicle)
18 524 732 465 351 285 168 381 2,924
Unspecified occupant
unspecified accident
a
Drive-RTA (unspecified) 270 337 629 420 243 229 197 485 2,810
Hospital admission rate (per Bn km)
bc
44,125
d
158 65 34 24 29 37.6 145 46
95% CIs 39,641
48,977
150166 6367 3336 2225 2731 3540 138153 4547
Cycle 3 yr distance (Mn km) 286 74
e
391 399 440 288 118 60 2,056
3 yr Cycle-RTA (n) Collision 296 85 247 225 209 161 93 75 1,391
3 yr Cycle-fall (n) On-highway
f
1,259 66 314 354 374 430 304 189 3,290
Hospital admission rate (per Bn km)
f
5,445 2,035 1,435 1,451 1,325 2,050 3,358 4,405 2,277
95% CIs 5,178
5,723
1,724
2,387
1,319
1,559
1,335
1,574
1,220
1,437
1,888
2,223
3,036
3,706
3,890
4,970
2,212
2,343
Walk 3 yr distance (Mn km) 5,022 1,300 3,184 3,568 3,425 2,483 2,024 1,674 22,680
3 yr Walk-RTA (n) Specified 2,658 576 931 702 688 698 663 2,096 9,012
Unspecified 248 43 65 54 68 55 65 253 851
3 yrn Walk-fall (n) On-highway 1,126 557 1,355 1,639 2,570 4,161 6,120 23,307 40,835
Unspecified
Location
gh
9,425 3,133 8,540 8,588 9,397 13,462 18,803 126,098 197,446
Hospital admission rate (per Bn km)
h
803 905 738 671 971 1,979 3,383 15,326 2,235
95% CIs 778828 854958 709769 645699 9381,005 1,924
2,035
3,304
3,465
15,139
15,515
2,216
2,255
a
This group may also contain passengers but probably notmany as ‘‘unspecified occupant’’ was rare for collisions.These have been included in the fatality rate estimate.
b
Estimates are likely to be a little too high due to the assumption that unspecified occupants are all drivers: an unknown proportion would have been passengers.
c
Note that these averages include both local roads and motorways/multi-lane divided roadways, where fatality rates are an order of magnitude lower than general
purpose roads, but data are not available by age and sex.
d
This figure is greatly exaggerated by under-measurement of under-aged driving.
e
The base for this was much smaller than distances for other ages, sex, and travel modes.
f
These figures are too high as V19.8 is a dustbin code, including some off-highway falls.
g
Not included in the fatality rate.
h
Estimates are too low, as some ‘unspecified location’ deaths will have been on-highway.
doi:10.1371/journal.pone.0050606.t004
Figure 2. Fatality rates per million hours’ use by travel mode, age, and sex. a. Males. b. Females. f/mhu: fatality rate per million hours use in
England, 2007–2009.
doi:10.1371/journal.pone.0050606.g002
Like-for-Like Assessment of Road Safety by Mode
PLOS ONE | www.plosone.org 7 December 2012 | Volume 7 | Issue 12 | e50606
especially for younger drivers and older pedestrians and cyclists.
However, the range of risks is similar for each mode. Comparisons
with the Netherlands revealed a similar pattern by age in both
countries for both cyclists and car users/drivers.
It has not been previously reported that males in England aged
17–20 y face higher risks as drivers than as cyclists, and do not
achieve better safety as drivers until their 20 s or 30 s, as has also
been shown in the Netherlands [20]. Due to constraints of
exposure data being provided by set age-groups, we were unable
to determine a precise age at which risk by mode crosses over.
However, in view of the risk implied to third parties, it is clearly
misleading for official publications to suggest that cycling is
relatively hazardous. An American study limited to road traffic
crashes involving motor vehicles found that the fatality rate per
Table 5. Hospital admission rates per million hours travel (mhu
a
) by travel mode, age, and sex, England 2007–2009.
Mode Age-group All ages
,17 17–20 21–29 30–39 40–49 50–59 60–69 70+
Men
Drive
b
1,131
c
11 3.2 1.3 0.9 0.8 0.9 2.5 1.6
95% CIs 1,0241,247 1111 3.1–.3.3 1.31.4 0.91.0 0.80.8 0.90.9 2.32.6 1.61.6
Cycle
d
92 32 23 22 16 20 23 34 29
95% CIs 90–94 31–34 22–24 21–22 15–16 19–20 21–24 32–37 29–30
Walk
e
5.8 7.5 6.9 7.1 9.8 10 12 39 10
95% CIs 5.6–5.9 7.2–7.8 6.8–7.1 6.9–7.3 9.5–10 10–11 11–12 38–39 10–10
Women
Drive
b
1,720
c
6.2 2.5 1.3 0.9 1.1 1.5 5.7 1.8
95% CIs 1,546–1,910 5.8–6.5 2.4–2.6 1.3–1.4 0.9–1.0 1.1–1.2 1.4–1.6 5.4–5.9 1.8–1.8
Cycle
d
67 25 18 18 16 25 41 54 28
95% CIs 63–70 21–29 16–19 16–19 15–18 23–27 37–45 47–61 27–29
Walk
e
3.4 3.8 3.1 2.8 4.1 8.3 14 64 9.4
95% CIs 3.3–3.5 3.6–4.0 3.0–3.2 2.7–2.9 3.9–4.2 8.1–8.5 14–15 64–65 9.3–9.5
a
mhu: million hours use, estimated using National Travel Survey average speed for all trips by this mode as not available by age and sex.
b
These averages include both local roads and motorways/multi-lane divided roadways, where fatality rates are an order of magnitude lower than general purpose roads,
but data are not available by age and sex.
c
This figure is greatly exaggerated by under-measurement of under-aged driving.
d
These figures are too high as V19.8 is a dustbin code, including some off-highway falls.
e
Estimates are too low, as some ‘unspecified location’ deaths will have been on-highway.
doi:10.1371/journal.pone.0050606.t005
Figure 3. Fatality rates per million hours’ use in the Netherlands and England, by age. a. The Netherlands, 2008. b. England, 2007–2009.
There are a number of limitations to these comparisons. First, data for the two countries use slightly different age-groups. Due to inaccuracies in
driving data for those below the legal limit for driving, data for the youngest groups are not shown, although they are included in the ‘all ages’
categories. Secondly, fatality rates for the Netherlands are for all car occupants whereas for England they are restricted to drivers (plus small numbers
of fatalities for unspecified car occupants), with passengers excluded from both the numerator and denominator. Thus the age-specific rates for the
Netherlands underestimate the variability in rate by age of driver. Thirdly, the data shown are the actual ‘all persons’ data from the two countries.
Therefore, the English data for cyclists are weighted to the figures for males. English males cycled four times the distance but had six times as many
fatalities as women, and spent twice as much time driving as women but had three times as many deaths. For the Netherlands, driving is similarly
dominated by males, but cycling distance is equally split between males and females.
doi:10.1371/journal.pone.0050606.g003
Like-for-Like Assessment of Road Safety by Mode
PLOS ONE | www.plosone.org 8 December 2012 | Volume 7 | Issue 12 | e50606
trip was higher for those aged 15–24 y as vehicle occupants (21.3
deaths per 100 million person-trips) than pedestrians (12.4);
although lower than for cyclists (30.9), the 95% CIs for vehicle
occupants and cyclists overlapped. It did not distinguish between
drivers and passengers, nor between males and females by age-
group [16].
The risks calculated in this study are all very low. An individual
who cycles 1 hr/d for 40 y at a fatality rate of 34/Bn km would
cover about 180,000 km, whilst accumulating only a one in 150
chance of fatal injury. Several studies have shown health benefits
at least an order of magnitude greater than fatality risk [25,26],
which cannot be claimed for driving. Our study confirms that
pedestrians face higher fatality rates per km travelled (45 fatalities/
Bn km in 2007–2009) than cyclists (34/Bn km), yet walking has
never been regarded as unduly hazardous. Data from the USA,
Germany and the Netherlands have also found higher risks for
pedestrians than cyclists per distance travelled, despite the higher
risks for non-motor travel in the USA [27]. There is clearly
a discrepancy between the popular image of cycling as dangerous
and the reality that, in most age groups, it is safe by everyday
standards, and yields health benefits that, for the previously
sedentary middle-aged, are greater than cessation of cigarette
smoking [28].
The situations in other countries are often more favourable to
cyclists than shown by this study. Data from the French National
Travel Survey and road fatality records reveal that the all-ages
average risk per hour of cycling in France(0.3 f/mhu: 150
fatalities, 5.7 Bn km national distance, 13.3 kph, 2008 data) is
a little less than in the UK, despite the general absence of cycling
infrastructure and a worse safety record for drivers (Personal
communication to MW 19/7/2011 from the French Ministry of
Ecology, Sustainable Development and Energy).
Strengths
This is the first study in the UK to provide travel casualty rates
by distance travelled and per hour, by mode, by age-group and
sex, based on nationally-representative data, to enable unbiased
intermodal comparisons for population sub-groups. The main
strengths of this study lie in assessing fatal and serious travel
injuries from mortality statistics and hospital records respectively
rather than police reports; using data by mode, age and sex; and
by providing data both including and excluding relevant codes.
This shows that official studies based on NHS data exaggerate the
risks of cycling by including non-travel injuries (Cycle-fall (off-
highway)) and underestimate the risks of walking by excluding
pedestrian falls on the highway. These errors combine to sustain or
enhance the misleading stereotype that travelling by bicycle is
relatively hazardous. A study from Israel found that 63% of bicycle
injuries resulting in hospitalization in adults and 73% in children
did not involve a motor vehicle [29]. We did not include walking
casualties in unspecified locations or off-highway when calculating
rates, to avoid overestimation of pedestrian travel injuries.
Weaknesses
Dustbin coding hindered interpretation of casualty data for all
three modes. It was impossible to separate drivers from passengers
completely, or on-highway cyclist falls from non-transport falls;
nor were all pedestrian falls on-highway identified. Certain sex-
and age-specific groups have wide confidence intervals due to
small numbers of travellers, especially female fatality rates for
walking and cycling, and male fatality rates for cycling, especially
older cyclists. This is despite the three year sample period.
Although postcode of residence for the numerator data would
allow assessment of the area deprivation of those killed or injured,
the exposure data did not include socioeconomic data, precluding
assessment of socio-economic distribution of injury and fatality
rates. This would be important to explore in future studies because
of the well known social inequalities in traffic casualties [30,31].
It is important to recognise that additional bias occurs from not
distinguishing between motorway and local driving, with the latter
an order of magnitude more dangerous than the former [32].
Casualty rates should ideally compare either driving and cycling
on general-purpose roads only, or should compare all driving with,
for example, travelling by bicycle for short- and train for long-
distance journeys. Given the inclusion of long distance driving in
our data, and therefore an underestimation of the risk for driving
on local streets, it may be that the crossover for fatality risk
between male cyclists and drivers occurs at an older age.
Alternatively, cyclists should be compared with low-mileage
drivers, who face risks 15–100% greater than the average [33].
The variation in risk was substantial for hospital admissions per
billion km, with walking and cycling having rates more than 100
times greater than for driving in some age groups. However, this is
not a fair comparison, due to large differences in the mobility
achieved using different modes of travel. Differences in injury
severity by mode exacerbate the problems of making like-for-like
comparisons. Most cyclists ride at moderate speeds for distances
,8 km on urban or quieter rural roads [12]. The risks in this
paper should not be applied to sportive cycling, mountain biking
or BMX riding.
The comparison with published age-specific data from the
Netherlands, converted to f/mhu, had a number of limitations, as
detailed in the legend to Figure 2. First, De Hartog et al compared
car users (not drivers) with cyclists [20], thereby diluting the age
effect of drivers, as about 25% of car user km are by passengers.
Secondly, we were unable to compare sex-specific rates. Cycling in
England is dominated by males, raising the overall average risk,
whereas cycling distance is split equally between males and females
in the Netherlands. Because of the greater risk among males, this
increases all-ages average cyclists’ risk in England by one third.
Despite these limitations, we found a similar pattern for f/mhu by
age and mode of travel in both countries: a marked increase in
fatalities in the oldest age groups, particularly for cyclists, and
fatality rates for younger car users exceeding those for cyclists for
those under 30 y.
Whereas the increased injury and fatality rates among the
youngest adults are associated with increased risk of collision, the
biggest problem for older people is an increased case-fatality rate
[34]. However, although data are sparse for walking or cycling, at
least one study has reported increased risk of collisions for older
pedestrians and cyclists, which they attributed to a range of age-
related factors affecting behaviour [35].
This study did not account for the third parties killed or
seriously injured in collisions. This would have had an effect
almost exclusively on the driver risks, since pedestrians and cyclists
seldom impose risk on others. A previous study has shown that in
the UK, such inclusion made all-ages risks per hour similar for
cyclists and drivers [24]. In England during the three years of this
study, there were nearly 1,000 third party or passenger fatalities in
crashes that involved at least one male driver aged 17–20 y
(personal communication from Department for Transport to MW,
12/10/12). This means that there were almost two deaths of
others for every fatality of a male driver aged 17–20 y. Further
research is required to clarify the extent to which young males are
a direct cause of fatalities amongst other road users.
Like-for-Like Assessment of Road Safety by Mode
PLOS ONE | www.plosone.org 9 December 2012 | Volume 7 | Issue 12 | e50606
Conclusions
This paper examined the risks of walking, cycling and driving in
England using comparable categories of casualty from mortality
statistics and hospital records, for age- and sex-specific groups.
Cycling is no longer an activity predominantly of youth but it
remains very male-dominated, at 80% of distance cycled per year
in the UK. The risks per hour of the three modes vary within
similar ranges, but show different trends with age. Notably, males
aged 17–20 y faced almost five times greater risk per hour as
drivers than as cyclists; in contrast, male cyclists 70+were the most
at risk of any group in this study. Females of all ages faced low risks
relative to males, except for older drivers.
Comparisons with the Netherlands revealed a similar pattern by
age in both countries for both cyclists and car users/drivers.
Fatality rates by time travelled in the Netherlands were lower for
cyclists than for car users in age groups under 50 y, but for older
age groups cycling risk increased sharply, as in England.
Fatality rates by distance travelled were similar for cyclists and
pedestrians when collisions and falls on-highway were included,
but were generally an order of magnitude higher than for drivers.
The important exception was for males under 20 y, in whom the
fatality rates as drivers/car users were higher than as cyclists.
Results for driver risks did not include third parties killed or
seriously injured in collisions. Not making like-for-like comparisons
for numerator data and type of road combine to sustain the myth
that cycling is relatively hazardous.
Supporting Information
Table S1 ICD-10 coding groups used for transport
modal risk study
(XLS)
Acknowledgments
We thank: Gary Smith from the Department for Transport National
Travel Survey team for providing the travel data; Robel Feleke, Allan
Baker, and Ed Klodawski from the London Health Observatory for
providing the HES and mortality data; and Barbara Carter-Szatynska for
administrative support.
Author Contributions
Conceived and designed the experiments: MW JM. Analyzed the data: JM
DL MW. Wrote the paper: MW JM DL. Defined the data: MW. Obtained
the data: JM.
References
1. Oja P, Titze S, Bauman A, de Geus B, Krenn P, et al. (2011) Health benefits of
cycling: a systematic review. Scand J Med Sci Sports. 21: 496–509 .
2. Heinen E, van Wee B, Maat K (2010) Commuting by bicycle: An overview of
the literature. Transport Rev. 30: 59–96.
3. Rojas-Rueda D, de Nazelle A, Tanio M, Nieuwenhuijsen MJ (2011) The health
risks and benefits of cycling in urban environments compared with car use:
health assessment study. BMJ. 343: d4521.
4. Jarrett J, Woodcock J, Griffiths UK, Chalabi Z, Edwards P, et al. (2012) Effect of
increasing active travel in urban England and Wales on costs to the National
Health Service. Lancet. 379: 2198–205.
5. Cope A, Muller L, Kennedy A (2009) Cycling demonstration towns. Monitoring
report 2006 to 2009. London: Cycling England.
6. Transport for London (2008) Central London Congesting Charging Impacts
Monitoring. Sixth annual report, July 2008. London: TfL.
7. Sonkin B, Edwards P, Roberts I, Green J (2006) Walking, cycling and transport
safety: an analysis of child road deaths. J Roy Soc Med. 99: 402–5.
8. Rissel C, Campbell F, Ashley B, Jackson L (2002) Driver road rule knowledge
and attitudes towards cyclists. Aus J Primary Health. 8: 66–9.
9. Wegman F, Zhang F, Dijkstra A (2012) How to make more cycling good for
road safety? Accid Anal Prev. 44: 19–29.
10. Christie N, Cairns S, Towner E, Ward H (2007). How exposure information can
enhance our understanding of child traffic ‘‘death leagues’’. Injury Prev.13: 125–
129.
11. TRL, Transport Research Laboratory (2001) Cost-benefit analysis of measures for
vulnerable road users. PROMISING: Promotion of Measures for Vulnerable Road Users
Deliverable D5. www.swov.nl/rapport/promising/wp5final.pdf (accessed 10/
10/2012)
12. Department for Transport (2011) National Travel Survey 2010. London: DfT.
13. Shafer A (2000) Regularities in travel demand; an international perspecti ve.
Research and Innovative Technology Administration (RITA), US Department
of Transport. www.bts.gov/publications/journal_of_transportation_and_
statistics/volume_03_number_03/paper_01/(accessed 15/06/12).
14. Knowles J, Adams S, Cuerden R, Savill T, Reid S, et al. (2009) Collisions
involving pedal cyclists on Britain’s roads: establishing the causes. Transport
Research Laboratory report PPR445. Crowthorne, Berks: TRL. www.trl.co.uk/
online_store/reports_publications/trl_reports/cat_road_user_safety/report_
collisions_involving_pedal_cyclists_on_britain_s_roads_establishing_the_
causes_.htm (accessed 15/06/2012).
15. Tranter M (2008) The use of hospital data on road accidents. In: Road
Casualties Great Britain 2007. Department for Transport. London: TSO
Publications. 78 p.
16. Beck LF, Dellinger AM, O’Neil ME (2007) Motor vehicle crash injury rates by
mode of travel, United States: Using exposure-based methods to quantify
differences. Am J Epidemiol. 166: 212–8.
17. HES Online. www.hesonline.nhs.uk/Ease/servlet/
ContentServer?siteID = 1937&categoryID = 211 (accessed 05/02/2012).
18. Davis A, Cavill N, Wardlaw M, Mindell JS (2011) Physical activity, trends in
walking and cycling and the obesity epidemic. In: Mindell JS, Watkins SJ, Cohen
JM, editors. Health on the move 2. Policies for health promoting transport.
Stockport: Transport and Health Study Group. 2–12.
19. Bhagat A, Tranter M (2008) Comparative casualty rates by mode of travel:
2006. In: Road Casualties Great Britain 2007. Department for Transport.
London: TSO Publications. 82 p.
20. De Hartog JJ, Boogaard H, Nijland H, Hoek G (2010) Do the health benefits of
cycling outweigh the risks? Environmental Health Perspectives.118: 1109–16.
21. Centraal Bureau voor de Statistiek (2010) Mobiliteit in Nederland; mobiliteits-
kenmerken en vervoerwijzen, regio’s. http://statline.cbs.nl/StatWeb/
publication/?VW = T&DM = SLNL&PA = 81127NED&LA = NL (accessed
15/06/12).
22. Dobson AJ, Kuulasmaa K, Eberle E, Scherer J (1991) Confidence intervals for
weighted sums of Poisson parameters. Stat Med. 10: 457–62.
23. Blows S, Ivers RQ, Connor J, Ameratunga S, Woodward M, et al. (2005)
Unlicensed Drivers and Car Crash Injury. Traffic Injury Prev. 6: 230–4.
24. Wardlaw M (2003) Assessing the actual risks faced by cyclists. Traffic Engineer
Control. 43: 420–24.
25. Hillman M (1992) Cycling towards health and safety. British Medical
Association ISBN 0–19–286151–4.
26. Rabl A, de Nazzelle, A (2001) Benefits of shift from car to active transport.
Transport Pol. 19: 121–131.
27. Pucher J, Dijkstra L (2003) Promoting safe walking and cycling to improve
public health: lessons from the Netherlands and Germany. Am J Public Health.
93: 1509–16.
28. Byberg L, Melhus H, Gedeborg R, Sundstro¨m J, Ahlbom A, et al. (2009) Total
mortality after changes in leisure time physical activity in 50 year old men:
35 year follow-up of population based cohort. BMJ. 338: b688.
29. Siman-Tov M, Jaffe DH, Israel Trauma Group (2012) Bicycle injuries: A matter
of mechanism and age. Accid Anal Prev. 44: 135–9.
30. Christie N, Ward H, Kimberlee R, Lyons R, Towner E, et al. (2010) Road Traffic
Injury Risk in Disadvantaged Communities: Evaluation of the Neighbourhood Road Safety
Initiative, Publication No. 19.London: Department for Transport. Road Safety
Web.
31. Morency P, Gauvin L, Plante C, Fournier M, Morency C (2012) Neighborhood
Social Inequalities in Road Traffic Injuries: The Influence of Traffic Volume
and Road Design. Am J Public Health. 102: 1112–1119.
32. Department for Transport (2009) Road Casualties Great Britain 2008. London:
TSO.
33. Langford J, Koppel A, McCarthy D, Srinivasan S (2008) In defence of the ‘low
milage bias’. Accid Anal Prev. 40: 1996–1999.
34. Li G, Braver ER, Chen LH (2003) Fragility versus excessive crash involvement
as determinants of high death rates per vehicle-mile of travel among older
drivers. Accid Anal Prev. 35: 227–235.
35. Oxley J, Corben B, Fildes B, O’Hare M, Rothengatter T (2004) Older vulnerable
road users - measures to reduce crash and injury risk. Clayton, Victoria, Australia:
Monash University Accident Research Centre, Report no.218. 164pp. www.
monash.edu.au/miri/research/reports/muarc218.html (accessed 29/10/2012)
Like-for-Like Assessment of Road Safety by Mode
PLOS ONE | www.plosone.org 10 December 2012 | Volume 7 | Issue 12 | e50606

Supplementary resource (1)

... The major challenge for assessing relative injury risks by transport mode is that of identifying appropriate exposure measures, given that journeys by different mode are not comparable: they differ by average speed, length of journey and by the demographic profiles of those undertaking trips. Analyses using measures of distance travelled to characterise exposure may inappropriately compare the long journeys typically done by car with much shorter journeys by walking or cycling (Mindell et al., 2012). In the setting of our study (England), using time spent travelling may be preferable, given that the mean time spent on a cycling and driving trip is similar (Scholes et al., 2018) and that the overall time spent travelling tends to remain relatively constant across modes (Mindell et al., 2012). ...
... Analyses using measures of distance travelled to characterise exposure may inappropriately compare the long journeys typically done by car with much shorter journeys by walking or cycling (Mindell et al., 2012). In the setting of our study (England), using time spent travelling may be preferable, given that the mean time spent on a cycling and driving trip is similar (Scholes et al., 2018) and that the overall time spent travelling tends to remain relatively constant across modes (Mindell et al., 2012). However, data on travel time by mode is reliant largely on self-report measures, which are subject to various biases (Tenenboim and Shiftan, 2016), including changes in the social desirability of mode choice over time. ...
Article
Full-text available
Introduction Increasing levels of active travel in the population brings many public health benefits, but may also change the risks of road injury for different road users. We examined changes in rates of pedestrian injuries resulting from collisions with pedal cycles and motor vehicles in England during 2005–2015, a period of increased cycling activity, and described the gender, age distribution and locations of pedestrians injured in collisions with pedal cycles and motor vehicles. Methods Collisions data were obtained from police STATS19 datasets. We used two measures of cycle/motor vehicle use; miles per annum, and estimated average travel time, and assessed evidence for trends towards increase over time using Poisson regression analysis. Results There were 3414 pedestrians injured in collisions with one or more pedal cycles in England during 2005–2015, 763 of whom were killed or seriously injured (KSI). This accounted for 1.3% of the total pedestrians KSI from all vehicles. Of those KSI in collisions with cycles, 62% were female; 42% over the age of 60; 26% were on the footway or verge and 24% were on a pedestrian crossing. There was a 6% (IRR 1.056; 95% CI 1.032–1.080, p < 0.001) annual increase in the pedestrian KSI rate per billion vehicle miles cycled in England over the time span. This increase was disproportionate to the increase in cycle use measured by vehicle miles or time spent cycling. Conclusions Increases in cycling were associated with disproportionate increases in pedestrian injuries in collisions with pedal cycles in England, although these collisions remain a very small proportion of all road injury. Increased active travel is essential for meeting a range of public health goals, but needs to be planned for with consideration for potential impact on pedestrians, particularly older citizens.
... Mindell and colleagues [57] also argued that the risk by distance travelled does not capture large differences in the average speed, which enable a comparison of different mobilities for drivers, cyclists, and pedestrians, and that a time-based comparison minimises the distorting impact of different comparisons of other modes with long distance car journeys. The travel time budget (TTB) concept argues that the amount of time that people spend travelling remains almost constant, at an average daily travel time of around one hour, and that individual travel choice is dictated by a tendency to stay within this TTB [58,59], with faster travel modes resulting in longer distances travelled. ...
... The travel time budget (TTB) concept argues that the amount of time that people spend travelling remains almost constant, at an average daily travel time of around one hour, and that individual travel choice is dictated by a tendency to stay within this TTB [58,59], with faster travel modes resulting in longer distances travelled. Mindell and colleagues [57] further argued that the TTB concept can be used to support the use of risk based on time as being the most appropriate for comparing modes with different average speeds. ...
Article
Full-text available
While absolute injury numbers are widely used as a road safety indicator, they do not fully account for the likelihood of an injury given a certain level of exposure. Adjusting crash and injury rates for travel exposure can measure the magnitude of travel activity leading to crash outcomes and provide a more comprehensive indicator of safety. Fatal and serious injury (FSI) numbers were adjusted by three measures of travel exposure to estimate crash and injury rates across nine travel modes in the Australian state of Victoria. While car drivers accounted for the highest number of injuries across the three modes, their likelihood of being killed or seriously injured was substantially lower than that of motorcyclists across all exposure measures. Cyclists accounted for fewer injuries than car passengers and pedestrians but had a higher risk per exposure. The results varied by both injury severity and exposure measure. The results of this study will assist with high level transport planning by allowing for the investigation of the changes in travel-related FSI resulting from proposed travel mode shifts driven by safety, environmental reasons or other reasons as part of the holistic goal of transforming the transport system to full compliance with Safe System principles.
... Exposure could include measuring the miles or minutes walked; however, determining how to precisely capture these data is important. Mindell and colleagues converted distance traveled to time spent traveling and showed the value of accounting for time when calculating exposure and determining risks for pedestrians and cyclists (Mindell et al., 2012). Future research is needed that account for exposure is determining risks to pedestrians. ...
Article
Full-text available
Modifying the built environment to make communities more walkable remains one strategy to promote physical activity. These modifications may have the added benefit of reducing the risk of pedestrian injury; however, there is a gap in the physical activity literature regarding how best to measure pedestrian injury. Examining the measures that have been used and related data sources can help inform the use of pedestrian injury data to evaluate whether safety is optimized as walking increases. We conducted a systematic review of the literature to identify studies that evaluated changes to the built environment that support walking and measures impacts on pedestrian injury as a measure of safety. We searched PubMed, PsycInfo, and Web of Science to identify peer-review studies and websites of fifteen organizations to document studies from the grey literature published in English between January 1, 2010 and December 31, 2018. Our search identified twelve studies that met the inclusion criteria. The few studies that measured changes in pedestrian injury used crash data from police reports. Injury frequency was often reported, but not injury severity, and no studies reported injury risk based on walking exposure. We conclude that few studies have measured pedestrian injury in the context of creating more walkable communities. Future research would benefit from using well-characterized measures from existing studies to support consistency in measurement, and from more longitudinal and evaluation research to strengthen the evidence on additional benefits of walkability. Increased collaborations with injury prevention professionals could bolster use of valid and reliable measures.
... Many studies have addressed sex or gender differences in population-based rates of road crash injury or death, with higher values usually observed in men [1][2][3][4][5][6], specifically in drivers' risk of road crash or crash-related injury or death. In fact, there is consistent evidence showing increased male-to-female ratios in this context [7][8][9][10][11][12], although this pattern seems to change with age, with higher risk values found in men, especially at younger ages [11,[13][14][15][16]. Furthermore, conflicting findings arise when specific issues, such as injury severity, are considered (i.e., depending on whether death or serious injury is considered as the outcome, the risk may be higher for men or for women in different studies) [3,8,17]. ...
Article
Full-text available
We designed a cross-sectional study in Spain, from 2014 to 2017. Our objective was to assess sex-related differences in the amount of driving exposure of car drivers, overall and stratified by the main environment-related driving conditions. We compared the sex distribution across three populations: (1) total number of person-years aged > 18 years; (2) total number of person-years aged > 18 years holding a valid car-driving license; and (3) total number of non-responsible car drivers involved in crashes with another offending driver, stratified by different environmental variables. The quasi-induced exposure approach was applied: the non-responsible drivers were considered as representative of the entire population of drivers on the road at the place and time at which the crash occurred. We calculated the female-to-male odds ratio (OR) by comparing population 2 versus 1, and population 3 versus 2. Finally, we performed separate regression models in population 3 for each environment-related variable as the dependent variable and driver’s age and sex as the independent variables. The female-to-male OR for the first comparison was 1.12, but values below 1 were found for extreme age groups. In the second comparison, an OR of 0.50 (0.49–0.51) was found, with progressively lower OR values as age increased. In population 3, women were found to drive less than men in environments known to be high risk (i.e., open roads, night-time, poor light conditions, and weekends). A significant gender gap exists in the amount and type of driving exposure. Although women obtain a driving license more frequently than men, they drive much less and tend to avoid high-risk environments. These results emphasize the need to incorporate a gender perspective in the development and implementation of road safety interventions.
Article
Although cycling has benefits to environment and physical health, bicyclists are vulnerable road users. Prior studies have identified the environment, traffic and road user factors that affect the risk of bicycle-related crashes. However, it is rare that difference in their effects on the risk amongst different bicycle crash types is investigated. For instance, there is possible correlation between counts of different crash types. In this study, multivariate Poisson-lognormal regression method is applied to examine the relationship between bicycle crash frequencies and possible explanatory factors, with which the correlation between bicycle-vehicle and bicycle-bicycle crashes is considered, using the crash data from London in 2018 and 2019. In addition, effects of road network characteristics on bicycle crash frequencies are also considered. Results indicate that proposed multivariate Poisson-lognormal model outperforms conventional univariate Poisson-lognormal models, in term of the metrics like deviance information criterion (DIC). In addition, factors including population socio-economics, land use, and road network characteristics that affect bicycle crash risk are identified. For instance, effects of traffic flow, residential area, network connectivity, and intersection density on the crash counts are different between bicycle-vehicle and bicycle-bicycle crashes. Findings are useful for the implementation of remedial measures that can improve overall bicycle safety in the long run.
Article
Children are experiencing an increase in so‐called ‘eco‐anxiety’. Along with others, we argue that children need more concrete, meaningful opportunities to feel they can make a difference in climate. In this article, we contend that bringing back the bicycle, and the opportunities for independent and active mobility that it affords children, is an important form of climate action for young people. We assert that as well as providing sustainability and wellbeing co‐benefits for children, cycling can help bolster the neighbourhood resilience and social cohesion that will be key resources for communities adapting to new climate risks.
Article
Introduction Previous studies have found a positive association between cycling injury risk and residential deprivation. However, most of these studies focused on serious and fatal injuries, children, and a specific point in time. This study explores i) inequalities in cycling injury risk by residential deprivation for all recorded casualties (slight, serious, and fatal) in England, ii) whether these inequalities vary by sex and age, and iii) how they have changed over time. Methods Using the STATS19 database of road traffic casualties in Britain, the English National Travel Survey, and population estimates for England over the six-year period 2014–2019, we estimated the ratio of slight, serious, and fatal cycling casualties per billion kilometres cycled by residential Index of Multiple Deprivation (IMD) quintile; by residential IMD quintile and sex and age group; and by residential IMD quintile and year. Results We found that the higher the level of residential deprivation, the higher the slight and serious cycling injury risk. The fatal cycling injury risk was also higher in individuals from the most deprived areas. Inequalities were particularly large for children, with slight and serious rates three times higher for children from the most deprived areas than for children from the least deprived areas. We also found that the linear trend lines of the slight and serious injury rates between 2014 and 2019 declined in the least deprived quintiles but not in the most deprived quintiles, which suggests that inequalities in slight and serious cycling injuries may have grown over the last years. Conclusion This study found that people from deprived areas are at higher risk of cycling injury for all types of severity; that children from deprived areas are most at risk; and that these inequalities may have recently increased.
Conference Paper
This paper is concerned with the development of an intelligent nanoscopic safety model for cycling safety. The present models are primarily focused on motorists modelling at an aggregate level. In this work, a framework for safety analysis is proposed consisting of a) Data collection unit, b) Data storage unit, and c) Knowledge processing unit. The predictive safety model is developed in the knowledge-processing unit using supervised deep learning with neural network classifier and gradient descent backpropagation error function. This framework is applied to a case study in Tyne and Wear County in England's northeast using the crash database. An accurate safety model (88% accuracy) is developed with the output of the riskiest age and gender group, based upon the specific input variables. The most critical variables affecting the safety of an individual belonging to a particular age and gender groups are the journey purpose, traffic flow regime and variable environmental conditions it is subjected to. It is hoped that the proposed framework can help in better understanding of cycling safety, aid the transportation professional for the design and planning of intelligent road infrastructure network for the cyclists.
Book
Full-text available
Cahiers de l'Observatoire de la mobilité de la Région Bruxelles-Capitale, 2020 Ce septième Cahier vient compléter la collection des Cahiers de l’Observatoire de la mobilité de la Région de Bruxelles-Capitale (RBC). Après avoir traité de l’offre de transport, des pratiques de déplacement en général et de celles liées au travail et à l’école en particulier, de logistique et de transport de marchandises, ou encore de partage de l’espace public entre tous les modes, ce nouveau Cahier s’arrête pour la première fois sur un mode spécifique : le vélo. Cette publication comporte trois parties. La première offrira une brève histoire du vélo racontée depuis Bruxelles et évitera d’emblée toute naturalisation du phénomène : le lent déclin du vélo au cours de la seconde moitié du 20e siècle résulte d’évolutions structurelles et non d’explications selon lesquelles Bruxelles ne serait " pas faite pour le vélo ". Cette première partie comportera également une mise en contexte institutionnelle afin d’identifier qui sont les acteurs compétents en matière de politique cycliste et la place occupée par celle-ci dans les outils réglementaires et planologiques régionaux, ainsi que dans ses budgets. Elle se terminera par une définition et une typologie des vélos et autres engins de déplacement légers. La deuxième partie du Cahier abordera la pratique du vélo en RBC à travers une analyse approfondie du parc vélo et des déplacements à vélo. Enfin, la troisième partie analysera la cyclabilité de la Région : les aménagements pour le vélo en mouvement, la sécurité et l’insécurité des cyclistes, le stationnement des élos et les services liés au vélo. Une conclusion générale viendra clore ce vaste exercice de synthèse. À noter que les données mobilisées dans ce Cahier ont été arrêtées en juillet 2019. Il va de soi qu’une actualisation régulière de cette synthèse sera nécessaire pour suivre l’évolution de ce secteur en pleine ébullition.
Technical Report
Full-text available
Transportation has been linked to several adverse health impacts, with a large, but modifiable, burden of disease. In this work, researchers conceptualized and documented the linkages between transportation and health. Following that, the researchers quantified the impacts of transportation on health in a case study in Houston, Texas, that focused on premature mortality attributable to three pathways: air pollution, noise, and motor vehicle crashes. Researchers found that the pathways linking transportation to health include some that are beneficial, such as when transportation serves as means for social connectivity, independence, physical activity, and access. Some pathways link transportation to detrimental health outcomes from air pollution, road travel injuries, noise, stress, urban heat islands, contamination, climate change, community severance, and restricted green space, blue space, and aesthetics. Researchers defined each pathway and summarized its health outcomes as they occur in the literature and showed that transportation-related exposures and associated health outcomes, and their severity, can be influenced by inequity and intrinsic and extrinsic effect modifiers. In Houston, the researchers estimated 302 (95 percent confidence interval [CI]: 185–427) premature deaths were attributable to transportation-related noise, compared to 330 fatalities from motor vehicles, 631 (95 percent CI: 366–809) from PM2.5, and 159 (95 percent CI: 0–609) from NO2. Transportation-related noise and motor vehicle crashes were responsible for 1.7 percent and 1.9 percent of all-cause premature deaths in Houston, respectively. The estimated premature death rate attributable to transportation-related noise was comparable to the death rate caused by suicide, influenza, or pneumonia in the United States. PM2.5 was responsible for 7.3 percent of all-cause premature deaths, which is higher than the death rate associated with diabetes mellitus, Alzheimer’s disease, or motor vehicle crashes in the United States. Households with lower median income had a higher risk of adverse exposure and premature deaths. Researchers also showed a positive relationship between health impacts attributable to air pollution and road traffic passing through census tracts, which was more prominent for NO2. Although some of the pathways linking transportation and health are widely discussed in the literature, others are new or under-researched. This conceptual model can form the basis for future studies looking to explore the transportation-health nexus.
Article
Full-text available
We examined the public health consequences of unsafe and inconvenient walking and bicycling conditions in American cities to suggest improvements based on successful policies in The Netherlands and Germany. Secondary data from national travel and crash surveys were used to compute fatality trends from 1975 to 2001 and fatality and injury rates for pedestrians and cyclists in The Netherlands, Germany, and the United States in 2000. American pedestrians and cyclists were much more likely to be killed or injured than were Dutch and German pedestrians and cyclists, both on a per-trip and on a per-kilometer basis. A wide range of measures are available to improve the safety of walking and cycling in American cities, both to reduce fatalities and injuries and to encourage walking and cycling.
Article
Full-text available
Older pedestrians and cyclists are vulnerable road users, comprising a substantial proportion of all road fatalities world-wide. Pedestrian fatalities constitute between 20 and 30 percent of road fatalities, while cyclist deaths range
Article
Full-text available
Many potential cyclists do not cycle on the road because of safety concerns. Drivers' knowledge of road rules and attitudes towards cyclists on the road were assessed. A telephone survey of 105 randomly selected adults in Sydney, Australia, with a current driver's licence was conducted. Less than half the sample (43%) was aware of recent changes to the Australian road rules. The majority of respondents (76%) reported high perceptions of danger associated with cycling, although respondents who had recently cycled on the road were significantly less likely to report these concerns of danger.
Article
The following are key points: • Pedestrians bear a higher fatality rate than cyclists, by a factor of almost 1.5; • Cycling in Britain is safer than driving in many other countries, including France and Belgium; • Cycling is far safer than driving anywhere when the health benefits and reduced risk to third parties are included; • Cycling gets safer as it gets more popular; • There is no known example in recent decades when an increase in cycling led to an increase in cyclist deaths.
Article
There is a growing awareness that significant benefits for our health and environment could be achieved by reducing our use of cars and shifting instead to active transport, i.e. walking and bicycling. The present article presents an estimate of the health impacts due to a shift from car to bicycling or walking, by evaluating four effects: the change in exposure to ambient air pollution for the individuals who change their transportation mode, their health benefit, the health benefit for the general population due to reduced pollution and the risk of accidents. We consider only mortality in detail, but at the end of the paper we also cite costs for other impacts, especially noise and congestion. For the dispersion of air pollution from cars we use results of the Transport phase of the ExternE project series and derive general results that can be applied in different regions. We calculate the health benefits of bicycling and walking based on the most recent review by the World Health Organization. For a driver who switches to bicycling for a commute of 5 km (one way) 5 days/week 46 weeks/yr the health benefit from the physical activity is worth about 1300 €/yr, and in a large city (>500,000) the value of the associated reduction of air pollution is on the order of 30 €/yr. For the individual who makes the switch, the change in air pollution exposure and dose implies a loss of about 20 €/yr under our standard scenario but that is highly variable with details of the trajectories and could even have the opposite sign. The results for walking are similar. The increased accident risk for bicyclists is extremely dependent on the local context; data for Paris and Amsterdam imply that the loss due to fatal accidents is at least an order of magnitude smaller than the health benefit of the physical activity. An analysis of the uncertainties shows that the general conclusion about the order of magnitude of these effects is robust. The results can be used for cost-benefit analysis of programs or projects to increase active transport, provided one can estimate the number of individuals who make a mode shift.
Article
Commuting by bicycle has advantages over other modes of transport, both for the commuter and for society. Although cycling is an option for many commuters, a considerable number of them choose to use other forms of transport. In order to underpin policies that promote commuting by bicycle, this paper investigates the determinants for commuting to work. As many bicycle commuters do not cycle every day, we also examine people’s daily choices, in terms of frequency. We conducted a survey of the current literature in order to identify the determinants for commuting by bicycle. We found many determinants, not all of which are addressed by conventional mode choice studies and models. This suggests that predicting and influencing bicycle use needs to be grounded in other kinds of knowledge than those currently available for motorized forms of transport.