Deprivation Child Pedestrian Road Casualties

DOENI
Deprivation and
Child Pedestrian
Road Casualties January 2010
Final Report
colinbuchanan.com
Deprivation and Child Pedestrian Road
Casualties
Final Report
Project No: 16216-01-1

January 2010
46 Hill Street,
Belfast,
BT1 2LB
Telephone: 028 9023 4110
Fax: 028 9031 4182
Email : Belfast@cbuchanan.co.uk
Prepared by: Approved by:
____________________________________________ ____________________________________________
David McGuigan Stephen Wood
Status: Final Issue no: 2 Date: 12 January 2010
final_report_120110.doc
(C) Copyright Colin Buchanan and Partners Limited. All rights reserved.
This report has been prepared for the exclusive use of the commissioning party and unless otherwise agreed in writing by Colin
Buchanan and Partners Limited, no other party may copy, reproduce, distribute, make use of, or rely on the contents of the report.
No liability is accepted by Colin Buchanan and Partners Limited for any use of this report, other than for the purposes for which it
was originally prepared and provided.
Opinions and information provided in this report are on the basis of Colin Buchanan and Partners Limited using due skill, care and
diligence in the preparation of the same and no explicit warranty is provided as to their accuracy. It should be noted and is expressly
stated that no independent verification of any of the documents or information supplied to Colin Buchanan and Partners Limited has
been made
Deprivation and Child Pedestrian Road Casualties
Final Report
Contents
Summary 1
1 Introduction 3
1.1 Objectives of the study 3
1.2 Report structure 3
2 Literature Review 4
2.1 Purpose of the review 4
2.2 Context 4
2.3 Literature search 5
2.4 Literature review 6
3 Data sources 22
3.1 Road collision casualty data 22
3.2 Population data 22
3.3 Deprivation data 23
3.4 Urban Rural Classification 23
3.5 Road infrastructure data 23
3.6 People in employment 24
3.7 Workplace population 24
3.8 School pupil data 24
3.9 Car ownership data 25
3.10 Household tenure data 25
3.11 Weather data 25
3.12 Creation of composite variables 25
4 Ten Year Analysis 28
4.1 Datasets 28
4.2 Deprivation dataset 29
4.3 Population dataset 29
4.4 Casualty rate analysis 29
4.5 Location of collision as a proxy for location of residence 33
4.6 Analysis of collision factors 35
5 Statistical Modelling 39
5.1 Overview 39
5.2 Statistical package 39
5.3 Four stages 39
5.4 Stage 1: intercorrelation of dependent variables 40
5.5 Stage 2: Level 1 Modelling 44
5.6 Stage 3: Level 2 modelling 48
5.7 Stage 4: Level 3 modelling 56
6 Conclusions 62
6.1 Link between deprivation and child pedestrian casualties 62
6.2 Statistical modelling 62
6.3 Implications for road safety planning 63
Appendix 1
Appendix 2
Appendix 3
Appendix 4
Final Report
Summary
The study was commissioned by the Northern Ireland Department of the Environment to review child
pedestrian road casualty data for the ten year period 1999-2008 and, in particular, to determine if
there is evidence that suggests a link between child pedestrian casualties and deprivation. The study
also included a comprehensive literature review of UK and Irish sources relevant to the study.
The study was based on demographic and infrastructure data for each of the 890 Census Super
Output Areas (SOAs) and road collision data for the period 1999-2008 supplied by the PSNI.
The road collision data comprised records for 64,651 collisions involving 109,385 casualties of which
8,829 were pedestrians. The number of child pedestrian casualties was 3,235.
The ten year data have been comprehensively tabulated according to a number of collision factors
recorded by the police.
In addition to the ten year dataset, a two year dataset was provided by the PSNI for the period April
2007 – March 2009 (inclusive). Postcode information relating to casualties has been recorded by the
PSNI since April 2007. This allowed an analysis to take place of how the location of collision and
location of residence of the casualty are distributed and whether or not – for the ten year data – the
location of collision could be used as a proxy for location of residence of the casualty. The conclusion
was that the use of location of collision as a proxy for location of residence of casualty was appropriate
in the context of this study. This view was supported by other research considered as part of the
literature review.
Tabulations from the ten year data provided an important quantitative insight into the relationship
between pedestrian casualty rates and area deprivation. The overall average child pedestrian
casualty rate is 0.81 casualties per year per 1,000 population with males in the in the 5 – 7 year age
group showing the highest rate of 3.34 casualties per year per 1,000 population. Across all child age
groups (ages 0 -15) it is shown that children resident in the most deprived areas are 4.8 times more
likely to be injured as a pedestrian in a road collision than those resident in the most affluent areas.
This figure varies by age and sex with females resident in the most deprived areas being 14.7 times
more likely to be injured as a pedestrian. These results were highly statistically significant.
The analysis was carried out in stages involving three levels of statistical modelling. In the Level 1
modelling it was firmly established that there were clear and unequivocal links between pedestrian
casualty rates and deprivation for children and adults and that the rates differed by age, sex of
casualty and an Urban-Rural Definition Group (URDG) classification. This finding was very much in
keeping with outcomes of studies elsewhere in the UK.
The Level 2 modelling looked in more detail at factors other than those used to establish the link in the
Level 1 modelling. These included roads infrastructure data (road lengths by road class and number
of junctions), Local Government District, workplace employment, population density and sunshine
hours. The Level 2 modelling established the key variables which are statistically significant in
explaining pedestrian casualty rates.
Finally, the Level 3 model included only those variables which had statistically “lasted the course” for
retention in the model. These variables related to:
The Multiple Deprivation Measure (MDM) score
MDM Crime domain score (a component of the MDM score)
MDM Proximity to Services domain score (a component of the MDM score)
‘Traffic proxy’ – as an alternative measure of relative traffic movements in SOAs
The number of school places
Sunshine hours
Annual Rainfall
1
Final Report
URDG classification
The coefficients for the three MDM related variables, traffic proxy and school places are highly
statistically significant across most age groups and sexes (p generally <<0.001). The URDG
classification is highly significant for adult pedestrian casualties.
On the basis of the Level 3 model outputs some implications for road safety planning are also
ventured. The key implication relates to fact that deprivation remains a highly significant factor when a
number of other explanatory factors have been included in the model. This suggests that road safety
strategies should recognise the effect of area deprivation on child pedestrian casualty rates and that
interventions should be distributed with an appropriate focus on deprived areas. Specific interventions
might include:
a national “traffic club” scheme
road crossing training for children (Kerbcraft)
complementary speed enforcement devices and speed management programmes
improvements to infrastructure (including traffic calming and 20mph zones)
road safety inputs to the master planning of regeneration projects
local transport plans should include pedestrian reduction targets
a distribution of road safety measures which targets socio-economic environments
2
Final Report
1 Introduction
1.1 Objectives of the study

1.1.1 Colin Buchanan was commissioned by the Department of the Environment’s Road Safety
Division to:
carry out a systematic literature review of recent (last 10 years) published and
unpublished key research relevant to the project (UK or Irish only).
carry out a 10 year analysis of available PSNI casualty data for child pedestrian
casualties aged 0-15 years.
determine whether or not there is evidence suggesting a link between child

pedestrian casualties in NI and the deprivation of the area (as measured by the
Northern Ireland Multiple Deprivation Measure (NIMDM) in which the collisions
occurred, and:
- if there is evidence suggesting a link then a statistical model should be

developed to measure the effect of deprivation, taking account of
environmental factors. The reliability of this model should be assessed.
- if there is no evidence of a link then outline the reasons why this might be
the case (i.e. any data issues or NI-specific factors which mean that the
deprivation effect is not as prevalent here as in the rest of the UK).
1.2 Report structure

1.2.1 In Chapter 2 we set out the literature review. Chapter 3 summarises the data sources for
the data used in the study. A 10 year analysis of the data is set out in Chapter 4 together
with a series of tabulations and figures which provide background and context to the
study. Statistical modelling is the subject of Chapter 5. Chapter 6 presents the key
conclusions and implications for road safety planning.
3
Final Report
2 Literature Review
2.1 Purpose of the review

2.1.1 The purpose of the wider commission is to:
determine if the Northern Ireland (NI) child pedestrian casualty rates are – as
suggested in previous research – linked to social deprivation; and
- if a link is found, identify those factors which contribute to the relationship
and the range of remedial measures that may be used to reduce the
casualty rate and the disparities between the rates for children living in more
affluent and more deprived areas
- if no link is found, outline reasons why this might be so.
2.1.2 It is not the purpose of this commission to conduct original research to test new
hypotheses into the links between deprivation and the child pedestrian casualty rates.
2.1.3 This literature review, therefore, has the purpose of identifying the body of existing
relevant research which will inform various aspects of the current commission, viz:
assess the likelihood that a link will be found
identify the key factors which will help to explain the existence of a link
inform the choice of statistical analyses
identify counter-measures which could help reduce the child pedestrian casualty
rate
2.2 Context
2.2.1 Up until the 1990s there was largely anecdotal evidence of the link between social
deprivation and road accident casualty rates. However, during the 1990s with a greater
emphasis being placed on social inclusion, this issue came to the fore in road safety. At
the time there was no specific formal data source for road accident casualty addresses
which could be used in research.
2.2.2 In Great Britain (GB), the Department for Transport undertakes a five year (quinquennial)
review of the STATS19 system for collecting and collating road collision data. In this
review the values of existing data items are reviewed together with assessments of new
data items which might add value to the data collected.
2.2.3 In 1997 Quinquennial Review it was recommended, with effect from January 1999, that
the postcode of the person's home address should be included as part of the statistical
records for each of the drivers and the casualties who were involved in an injury road
accident. It was recognised that the driver and casualty postcodes (and the use of the
special code for non-UK residents) are potentially of great value in the analysis of the
road accident statistics data including the analyses of "socio-economic" background
information about the kinds of areas in which people involved in road accidents live
(SGLGRAS (2000)).
2.2.4 The new PSNI road traffic collision data collection system implemented in NI on 1 April
2007 included for the first time in Northern Ireland details for drivers and casualties.
2.2.5 In GB there has now been just over 10 years worth of postcode data and this has allowed
detailed research to be undertaken for the first time on a national data set incorporating
links to place of residence and social deprivation indices. It is, therefore, this more recent
research which will have the most relevance to the present study. Our search has
4
Final Report
focussed on literature since 1999 but includes relevant earlier work which is of value to
the current study.
2.3 Literature search

2.3.1 The literature search was originally targeted at a number of e-resources but this was later
1
limited to the Web of Science (a subset of the Web of Knowledge ) resource available to
readers at the National Library of Scotland as this was found to produce at least the same
plus additional results as compared with the other resources.
2.3.2 In addition, a search was also conducted using Google.
2.3.3 The keywords adopted were:

2
depriv? - for deprivation and deprived
child? - for child and children
pedestrian
casualt? - for casualty and casualties
injur? - for injury and injuries
accident? - for accident or accidents
collision? – for collision of collisions
2.3.4 As expected, a large number of references (>10,000) were found using all or groups of
these keywords combined with “OR” statements. The search was subsequently fine
tuned to identify relevant literature on the basis of using “AND” combinations on three and
two keywords as set out in Table 2.1:
Table 2.1: Combination of keyword searches
Keyword Searches undertaken using a logical "AND"

depriv?
child?
pedestrian
casualt?
injur?
accident?
collision?
2.3.5 This search revealed several hundred different papers and these were examined to
identify relevant material. The majority of the texts identified were either not based on
British or Irish research or on subjects not relevant to the present study (eg child
casualties in deprived areas caused by fire).
2.3.6 The citations in the relevant literature reviewed were also used to identify further
references. Finally, a Google search was also carried out to find items not found on the
databases searched. The Google search used the same keywords as used to search the
Web of Science.
2.3.7 All relevant items were read and where the content was considered to be of particular
interest to the present study it has been distilled and set out in the review.
1
See http://isiwebofknowledge.com/
2
a question mark (?) is used in a search term to indicate a wildcard, matching on zero or more characters in
place of the ?.
5
Final Report
2.3.8 The review is based on a cross-section of literature with a variety of approaches and
conclusions using a number of datasets from across the UK.
2.3.9 Items included in the review are set out in the references section. Items read but not
used are listed in a short bibliography.
2.4 Literature review

2.4.1 The review is presented in four sections.
1. A review of research into the links between deprivation and child pedestrian
casualty rates.
2. A review of the link between casualty residence and collision location
3. A review of potential measures that can be taken to reduce child pedestrian
casualty rates.
4. A summary of the key points of relevance to the present study
2.4.2 All references and bibliographical items are listed in Appendix 1. References are
documents which have been reviewed and cited. The bibliography lists documents read
but which were not reviewed.
The link between deprivation and child pedestrian casualty rates

2.4.3 Much of the early work undertaken into the relationship between deprivation and road
collision involvement sought to establish and measure the link. It did not seek to explain
or refine it through more detailed analysis involving explanatory variables in addition to
those relating to deprivation itself.
2.4.4 Kendrick (1993) noted that existing work on road traffic accidents was based on
analysing accident clusters despite evidence that accidents tended to be more dispersed.
Her study analysed 573 child pedestrian casualties (ages 0 -11) in Nottingham by
deprivation score for the years 1988-90. Casualty rates were mapped against deprivation
score and tabulations were presented. The analysis showed a highly significant
relationship between casualty rates and deprivation score (Chi square tests with p<0.001
and correlation coefficients (rs). Figure 2.1 shows Figure 2 from the report which shows
casualty rates mapped against deprivation score for children aged 5 – 11 years.
6
Final Report
Figure 2.1: Pedestrian accidents by deprivation: age 5-ll years, 1988-90

(Kendrick (1993))
2.4.5 McGuigan, Bull and Gorman (1996) in a paper reviewing links between casualty rates
and deprivation based on over 16,000 casualty records in Lothian Region, Scotland
3
tabulated (their Figure 4) pedestrian casualty rates by social deprivation category which
showed a clear relationship between casualty rates and deprivation (see Figure 2.2).
3
In Figure 2.2 the x axis relates to the Carstairs and Morris Index of Deprivation (called” depcat”) which was
widely used in Scotland as a measure of quantifying relative socioeconomic deprivation or affluence in different
localities across the country.
7
Final Report
Figure 2.2: Figure 4 reproduced from McGuigan, Bull and Gorman (1996)
2.4.6 Abdalla et al (1997) using a three year data set (1990 – 1992) comprising 8,560 casualty
records for Lothian Region, Scotland conducted a wide ranging study into the
relationships between casualty rates and social deprivation indicators for the casualties’
zone of residence. The study identified a statistically significant link between child
pedestrian casualty rates and deprivation.
2.4.7 Table 7 from their paper is reproduced below (Figure 2.3) where pedestrian casualty
rates are stratified by age. It is clear that the disparity between the most deprived and
most affluent groups differs by age. Of relevance to the present study, for children aged
0-4, 5-11 and 12-16 the ratios of the rates (15% most deprived / 15% most affluent) are
4
5.2, 7.6 and 2.4 respectively. Also the SCR values are over 1 for the most deprived
areas and under 1 for the most affluent areas. The difference in rates is highly significant
(p<0.01) based on a statistical test of the difference between the number of casualties per
resident population in the 15% most deprived and 15% most affluent zones.
4
SCR or standardised casualty ratio was a concept adopted by Abdalla et al in their paper. It is calculated as
the observed casualties ÷ expected casualties.
8
Final Report
Figure 2.3: Table7 reproduced from Abdalla et al (1997)

Table 7. Average number and rates (1990 - 1992) of road accident pedestrian casualties
by area type and by age structure
15% most deprived
15% most affluent areas Lothian
areas
Age Cas. Rate* SCR† Cas. Rate* SCR† Cas. Rate*
0-4 16 18.8 2.0 3 3.6 0.4 40 8.9
5-11 48 50.4 2.1 8 6.6 0.3 141 23.8
12-16 24 50.7 1.8 16 21.0 0.8 99 27.6
17-24 23 17.6 1.4 8 7.3 0.6 102 12.6
25-29 55 12.8 2.2 14 2.2 0.4 203 5.9
≥60 30 16.1 1.4 13 7.3 0.7 160 11.1
*Rate per 10,000 per year.
†SCR (standardised casualty ratio) = observed casualties (Cas.)/expected casualties.
2.4.8 In a similar study, Chichester et al (1998), analysed 1,300 attendance records at the
Accident and Emergency unit at Monklands/Bellshill Hospital in North Lanarkshire,
Scotland, to seek to understand more about associations between road traffic accidents
and deprivation. Of the 1,300 records, 1,020 were linked to socio-economic data. The
study found significant associations indicating increasing casualty rates for all causes of
injury as deprivation increased. Of relevance to the present study, it was also noted that
analyses on child casualties showed particularly strong significance (Chi square test,
p<0.005).
2.4.9 Building on earlier work in Lothian Region, Scotland, White and Raeside (2001) sought to
understand the underlying reasons why children from socially excluded backgrounds are
more at risk. The authors undertook pupil interviews in schools in the both the most
deprived and most affluent areas in total 128 interviews were conducted at four schools
(two in an affluent area and two in a deprived area). The interviews focused on how
children travelled to/from school, how they played and participation in after school
activities. The conclusions of this study were:
the risk of child involvement in road accidents is highly class related
children from disadvantaged areas tend to suffer more serious injury
children from disadvantaged areas have greater exposure to risk relating to where
they play and lack of parental supervision when travelling to/from school
that there may also be cognitive factors at work because children up to about nine
years of age do not recognise dangerous situations
children in poor areas primarily walk to school, however, children from affluent
background cross more busy roads than their poorer counterparts – probably
reflecting greater travel distances for the affluent group.
there is less accompaniment by adults and more accompaniment by other children
for those in the poorest area compared to those in the most affluent areas
it is speculated that children from poorer areas will be more exposed to risk as they
are less likely to be driven than their affluent counterparts and are more likely to
play in the street
2.4.10 These findings suggest different behaviour patterns by children related to deprivation and
suggest that there may be more complex factors at work than deprivation itself.
2.4.11 Grayling et al (2002) point out that Britain’s good record in reducing road accident
casualties is marred by a relatively poor record on child pedestrian safety. This study for
9
Final Report
the Institute of Public Policy Research mapped pedestrian casualties for the first time in
England and tested the links with deprivation and other area characteristics including:
number of children
traffic generating potential
residential and employment density
length and type of roads
weather and other specific local factors
2.4.12 In total there were more than 16,000 child pedestrian casualties in England in the
dataset. For some of the analysis the data were stratified by deprivation decile. The key
findings and aspects of the study were:
the use of deciles as the basis for analysis
children in the most deprived decile were more than three times as likely to be
pedestrian casualties than children in the least deprived decile – this result was
highly statistically significant
there was a considerable amount of unexplained variation between areas with
similar deprivation scores
the central conclusion of the study was that having controlled for other (non-
deprivation) factors there was still a distinct deprivation effect
population density had little impact on child pedestrian casualty rates
higher numbers of road junctions and A road length are linked to higher child
pedestrian casualty rates
higher annual rainfall and sunshine hours are linked to higher child pedestrian
casualty rates
the components of the deprivation measure were tested separately and it was
found that different components had different effects: in summary the components
relating to income, health, education and housing deprivation were associated with
higher pedestrian casualty rates whilst the components of employment and
accessibility were associated with lower rates
the effect of deprivation varied by region across England; for example, in London
the deprivation effect is particularly strong for children but less so for adults (with
respect to the average for the study)
the report ends with a discussion of possible remedial measures to reduce the
effects of deprivation on the rate of child pedestrian casualties
2.4.13 Figure 2.4 reproduces Figure 1.9 of the report which shows casualty rates by ward decile
and demonstrates how the casualty rate rises with increasing deprivation.
2.4.14 The key recommendations for the study were:

traffic calming combined with a maximum 20mph speed limit should become the
norm in residential and other built up areas
priority should be given to traffic calmed 20mph zones in deprived areas with high
pedestrian casualty rates
all local transport plans should include pedestrian casualty reduction targets
all local transport plans should include a speed management strategy
home zones in Britain should rapidly move beyond demonstration projects to
widespread implementation
the government should develop a national strategy for pedestrians to promote an
increase in walking at the same time as a reduction in pedestrian casualties
10
Final Report
Figure 2.4: Figure 9 reproduced from Grayling et al (2002)
2.4.15 Graham et al (2004) report on a study into child and adult pedestrian casualties in
England focusing on the influence of socio-economic deprivation on casualty rates. The
study was based on STATS19 data for the years 1999 and 2000 provided by the
Department for Transport (DfT). The dataset comprised 28,228 child pedestrian casualty
records in 8,414 wards in England.
2.4.16 The report authors begin by stating that it is easy to see in the “raw” data that there is an
apparent strong relationship between casualty numbers and measures of deprivation.
They warn that one should not jump to the conclusion that there is a simple relationship
between deprivation and casualties which is unrelated to other factors. The authors
identified the possibility that deprivation is more commonly found in dense urban areas
and that collisions occur more frequently in dense urban areas and that this could have
more to do the relationship than deprivation. The work described in the paper sought to
disentangle the effects of area type from the influence of deprivation.
2.4.17 The independent variables used to define area type included:

population (stratified by age)
employment statistics
number of road junctions (nodes)
number of road junctions/km of road
road lengths by class of road
weather: rainfall and sunshine
2.4.18 Regression modelling was used to link the independent variables to the casualty rates.
The key finding was that in addition to the area type variables used in the statistical
models, deprivation provides an additional explanation of the pedestrian casualty rates.
This outcome was rigorously tested and finding was confirmed as highly statistically
significant (t-statistic = 14.7 (p<< 0.001)) . The central conclusion was that “not only is
deprivation associated with an increase in the probability of a pedestrian casualty taking
place, but also with an increased probability of incidents involving a serious injury or a
fatality”. This deprivation effect was present for both adults and children but to a much
larger extent for children.
11
Final Report
2.4.19 In Wales, a wide ranging report on deprivation and health by the National Public Health
Service (NPHS (2004)) sets out child pedestrian casualty rates by quintiles for child
pedestrian casualties reported to the Police and also for hospital inpatient episodes.
These are reproduced in Table 2.2. The data relate to slightly different time periods, age
grouping and casualty severity. Also, for the STATS19 data it is the location of accident
that provides the deprivation measure in contrast to the hospital inpatient episode
database which assigns the deprivation measure on the basis of place of residence.
Nevertheless, both datasets show a clear link between deprivation and child pedestrian
casualty rates. On the basis of the STATS19 data, children in the most deprived areas
were found to be two and a half times more likely to be injured in a road collision as a
pedestrian.
Table 2.2: Child pedestrian casualty rates by deprivation quintile reported to

police and for hospital inpatient episodes in Wales
1 2
Reported To Police Hospital Inpatient Episodes
3 3
Deprivation fifth Children aged 4−16 years Children aged 5−14 years
1 (least deprived) 82.6 18.3
2 119.3 24.8
3 161.6 26.3
4 178.3 31.1
5 (most deprived) 209.3 39.3
Overall 159.8 29.6
Rate Ratio 5:1 2.52 2.15
1 STATS19 (1995-2000)
2 Patient Episode Database foe Wales (PEDW) (1997-2002)
3 Rate per 100,000 persons per year
2.4.20 The value of this report to the present study is that it complements the findings of the
Scottish and English research reviewed above on the basis of an independent dataset for
Wales.
2.4.21 Noland and Quddus (2004) pointed out that land use characteristics of an area had not
featured in earlier analyses and that deprived neighbourhoods may be subject to higher
traffic volumes than more affluent areas. In a detailed and wide ranging study based on
1999 casualty data for England casualty rates were modelled against a number of land
use, road network, traffic and demographic characteristics. The data comprised:
an undisclosed number of casualty records for 1999 across 8,414 wards
land use data
road characteristic data
demographic characteristic data
proxy for traffic flow data
2.4.22 The study did not look specifically at child pedestrian casualty rates. The main
conclusions were:
lower population densities were linked to higher casualty rates
higher employment densities were linked to higher casualty rates
larger levels of urbanisation were associated with fewer casualties
road lengths are associated with increases in serious injuries
deprivation is linked to higher casualty rates
higher exposure to traffic was linked to higher casualty rates
12
Final Report
2.4.23 Silversides et al (2005) recognised that injuries in childhood represent a major public
health concern particularly in North and West Belfast where there is a high rate of
emergency department attendance due to injury and in where social deprivation is
widespread. In a study of nearly 479 children (aged 0 – 12) presenting to emergency
departments in North and West Belfast, Silversides et al (2005) identified inter alia a
highly significant correlation (p < 0.001) between economic deprivation (measured by the
Noble Index) and injuries due to collisions on the road. This was a small and prospective
study but it does add a Northern Ireland dimension to the link between deprivation and
road collisions.
2.4.24 In 2006 a wide-ranging report to the London Road Safety Unit based on STATS19 data
looked at the links between road collision casualty rates and deprivation (Edwards et al
(2006)). The purpose of the study was to examine the strength of the relationships
between deprivation and child pedestrian rates to help develop road safety policies and
interventions. The casualty data comprised 5,823 child pedestrian casualty records
(ages 0 – 15) for the six year period 1999 to 2004. The study used an adjusted Index of
Multiple Deprivation (IMD) to measure levels of deprivation in each of 4,765 Super Output
Areas (SOAs) in London. The adjustment was to remove an element of the
environmental component of the IMD which itself includes an indicator that relates to road
traffic injuries to pedestrians and cyclists. This adjusted IMD was called IMD2. This
problem will not arise with the NI data as the multiple deprivation measure (MDM)
adopted for Northern Ireland does not include an element for pedestrian and cyclist
casualties (NISRA (2005)).
2.4.25 The research methodology involved ranking the SOAs from 1 (least deprived) to 10 (most
deprived) and then basing the statistical analysis on this stratification. These were
referred to as deprivation deciles. The study also considered the effect that road network
variables had on the relationship.
2.4.26 The key aspects and findings of the London study of relevance to the NI study are:
the use of deciles as the basis for analysis
the pedestrian injury rate for children living in the most deprived decile were found
to 2.93 times as high as that for children resident in the least deprived decile
road length per hectare and number of junctions/km of road were included in the
multivariable regression analysis carried out
the adjustment to create IMD2 did not offer any material benefits in the analysis
over using the unadjusted IMD
for certain casualty types – especially child pedestrians and cyclist casualties – it is
acceptable to conduct an analysis linking casualties to the resident population of
the area in which the collision occurred (but not other casualty types)
2.4.27 The London study concluded that it is not necessarily the deprivation of an area per se
that causes the increased risk. For example, the risk of being involved in a road collision
increases with the time spent travelling and people resident in deprived areas may be
more likely to be injured as pedestrians because they don’t have a car and are more
likely to be walking.
2.4.28 London data were also used to monitor the relationship between injury and deprivation
over time. The outcome of this was an observed downward shift in overall child casualty
rates over time but that there had been little change in the relative distribution of
casualties by IMD. This result is broadly confirmed by a similar analysis of data for the
City of Edinburgh where child pedestrian casualty rates for 1991-94 were compared with
equivalent data for 2000-03 (McGuigan et al (2006)). Figure 2.5 clearly demonstrates the
overall fall in casualty rates between the two periods but note that the slope of the trend
line for 2000-03 is not as steep as that for 1991-94. This suggests that the overall
casualty rates are falling and that the disparity between rates for the most affluent and
13
Final Report
most deprived are also falling. In 1991-94 the child pedestrian casualty rate for children
living in the most deprived areas (depcat = 7) was 4.58 times higher than children living in
the least deprived areas (depcat = 1). The equivalent figure for the 2000-03 data was
4.05).
5
Figure 2.5: Casualty rates by deprivation category (depcat ) for pedestrians in
age group 5-14 for 1991-94 and 2000-03 (City of Edinburgh)
2.4.29 The London report (Edwards et al (2006)) went on to look at:

remedial measures to address deprivation and road accident injuries
policy and practice issues
2.4.30 These parts of the report may be of value to Road Safety Division in any policy
development work that may be required following consideration of the final report for this
study.
2.4.31 Later work by Graham and Stephens (2008) looked specifically at the impact of
deprivation on child pedestrian casualties in England. The STATS19 data for an
undisclosed number of casualties for the five year period 1998 to 2002 across 7,925
small areas were used in the study. An objective of the study was to decomposed the
deprivation effect by considering six separate dimensions related to:
income
employment
health
education
housing and services
crime
2.4.32 Also included in the dataset were five local factors:

the number of children resident in the area
the volume of traffic - based not on measured data but on gravity model estimates
5
See footnote on page 6.
14
Final Report
nature of the local environment – e.g. population per road junction

local road infrastructure – road lengths by road class
other factors – weather (based on 31 weather stations)
2.4.33 Area subsets were also coded to provide a relative measure of urbanisation according to
10 factors (ranging from Central London to rural) based on Department for Transport
(DfT) classifications. Separately generalised linear regression models were used. .
2.4.34 The findings were complex but consistent with previous research.
2.4.35 The results indicated:

a substantial effect on the incidence of child pedestrian casualties for the
characteristics associated with area deprivation.
the deprivation components have different effects across environmental factors
interestingly, it is the crime component which has the most compelling and
consistent positive association the incidence of child pedestrian casualties
the variables which were included to control for activity levels and traffic flows were
found to be correlated to child pedestrian casualties
the Greater London conurbation was an area of elevated risk
2.4.36 A study of fatal casualties in England by Clarke et al (2008) looked only at vehicle
occupant fatalities. This study, however, does have some relevance to the present study
because for every pedestrian road casualty there will be a vehicle driver involved: both
parties are likely to have contributed to the cause of the collision. [Earlier unpublished
work by McGuigan did show that there was a relationship between casualty and driver
indexes of deprivation. As most pedestrian casualties tend to be injured close to home it
is likely that the distribution of driver home addresses will reflect to an extent the area in
which the collision occurred – hence the relationship]. In the study by Clarke et al it was
found that there was a relationship between the IMD scores for drivers killed in road
collisions and deliberate risk taking. This may suggest that drivers resident in deprived
areas are more likely than drivers in less deprived areas to take more risks and drive less
responsibly and – as a consequence – are more likely to be over-represented in
pedestrian/vehicle collisions.
2.4.37 Edwards et al (2008) in a study to examine the variation in serious injuries in children by
the level of deprivation in an area and by type of settlement in England. Hospital
admission rates for serious injury to children aged 0-15 years (7,840 No) in census super
output areas in England during the 5-year period 1 April 1999 to 31 March 2004 were
analysed
2.4.38 They found:

rates of serious injury as a pedestrian in a road traffic accident in children was 4.1
times higher in the most deprived areas than in the least deprived (statistically
significant: p <0.05)
no evidence that there were differences between settlement type (urban/rural) in
the effect on pedestrian casualty rates.
2.4.39 The first finding confirms the findings of other studies in relation to link between
deprivation and child pedestrian casualty rates. The second finding is of relevance to the
present study as the effect of an urban rural classification for NI will be used as an
independent variable.
The link between casualty residence and collision location

2.4.40 In the Abdalla et al (1997)research reviewed above, the pedestrian distance from
residence zone to accident location is tabulated by age group. Table 2 from the paper is
15
Final Report
reproduced in Figure 2.6. In this table it can be seen that a quarter of children in the 5-11
age group are injured as pedestrians at a distance of over 4.5km from their zone of
residence.
Figure 2.6: Table 2 reproduced from Abdalla et al (1997)

Table 2. Quantiles: pedestrian distances (km) from residence
zone to accident location by age group
Pedestrian
Lowest Second Middle Fourth Highest
age
0-4 0.009 0.055 0.102 0.436 31.844
5-11 0.004 0.097 0.211 4.505 35.810
12-16 0.011 0.280 0.674 2.742 39.810
17-24 0.021 0.599 2.231 4.852 40.500
25-59 0.023 0.360 1.378 3.980 42.679
Elderly (≥ 60) 0.009 0.150 0.457 2.304 32.286
2.4.41 Petch and Henson (2000) in a study of factors relating to child accidents in Salford,
England looked at a number of factors to explain the distribution of casualties and
casualty rates. Of interest to the present study is a tabulation of the distance between
accident location and place of residence. Figures for child pedestrian casualties were set
out in the report and are reproduced here in Table 2.3. It can be seen that 78% of child
pedestrian casualties in Salford lived less than 1km away from the location of the
accident with 22% who lived more than 1km away. The proportion of casualties occurring
in collisions in excess of 2km from home is 11.4%
Table 2.3: Distance between accident location and place of residence (Petch
and Henson (2000))
Distance between location of

accident and location of residence Cumulative Child Pedestrian
(km) Casualties (%)
0.00 - 0.25 41.5
0.25 - 0.50 62.3
0.50 - 0.75 71.2
0.75 - 1.00 78.0
1.00 - 1.25 80.5
1.25 - 1.50 85.2
1.50 - 1.75 88.1
1.75 - 2.00 88.6
2.4.42 These previous studies do not present direct comparisons of casualty residence and
collision location but they do suggest that up to one in four children may be injured in a
zone outside their zone of residence. It should noted, however, that this does not
necessarily mean that the deprivation score assigned on the basis of the accident
location will be different from the deprivation score of their home zone.
2.4.43 Hewson (2004) studied child pedestrian casualties in Devon. This study confirmed that
there were links between deprivation and increased child pedestrian casualty risk.
Hewson specifically looked at the question of whether it is better to aggregate the
casualty data on the basis of the ward in which the collision occurred or on the basis of
the ward in which the casualty was resident. It was found that where the data had been
16
Final Report
aggregated by the ward in which the collision occurred there is evidence that that urban
wards have a higher casualty rate than rural wards. The same cannot be said where the
data were aggregated by ward of residence. This suggests that those collisions which
involve child pedestrians in areas in which they are not resident tend to be urban in
nature. Indeed, the author raises the question of whether or not the casualty rate is more
likely to be associated with the deprivation of the zone in which the collision occurred
than with that of the home address of the child (the author warns - given the limitations of
the underlying data - that conclusions have to be drawn with great caution).
2.4.44 A key finding of the London Study (Edwards et al (2006)) was that for certain casualty
types – especially child pedestrians and cyclist casualties – it is acceptable to conduct an
analysis linking casualties to the resident population of the area in which the collision
occurred. This needs to be compared with the finding by Abdalla et al (1997) that a
quarter of children in the 5-11 age group are injured as pedestrians at a distance of over
4.5km from their zone of residence (Figure 2.6 refers). Indeed, McGuigan et al (2006)
identified the proportion of “home” casualties by deprivation category. In this case, the
area level adopted was postcode sector and a “home” casualty is defined as a casualty
who was both injured and resident in the same postcode sector. Figure 2.7 shows that on
average about 60% of child casualties in Edinburgh were “home” casualties. The
average number of “home” casualties fell by a small amount between 1991-94 and 2000-
03. It should be noted the Edinburgh data are for all child casualties.
Figure 2.7: Percentage of “home” child casualties (aged 5-14) by deprivation

6
category (depcat ) 1991-91 and 2000-03 (City of Edinburgh)
Potential measures that can be taken to reduce child pedestrian casualty

rates
2.4.45 In true fashion many researchers suggest in their conclusions that more research is
needed to understand fully the relationship between deprivation and accident risk. As
there is some caution expressed by researchers into the nature of the observed link
between deprivation and child pedestrian casualty rates, few put forward suggestions as
6
See footnote relating to Figure 2.2.
17
Final Report
to what measures can be put in place to reduce the rate deficit between children living
affluent and deprived neighbourhoods.
2.4.46 Perhaps the first intervention was undertaken in Lothian Region, Scotland when on 1
April 1994 Lothian Regional Council and Lothian Health Board initiated the first region-
wide universally free Children's Traffic Club (CTC) in the UK. This intervention was
introduced directly as a result of the casualty rate deficit between children living in more
affluent and less affluent areas. Take-up rates for the first six month period showed a
clear gradation of take-up rates ranging from 62% in the most affluent areas to 39% in the
most deprived areas. Parallel measures were put in place to encourage membership of
the club and these included:
free distribution of CTC materials to nurseries and pre-school establishments
health visits to children included encouragement to join
mail drops
2.4.47 On the back of Lothian’s initiative, the then Scottish Office introduced a national CTCS
(Children’s Traffic Club for Scotland) in November 1995. This was supported by a
television advertising campaign. This initiative has recently been reviewed and a report
of consultation has been published (Road Safety Scotland (2009)). This report concludes
that there were many positive aspects of the resource but that it was some concern about
its general appropriateness. It was felt that the resource was too “busy” and “wordy” for
the age group it is aimed at. There was a strong sense that the CTCS was a worthwhile
but – after 13 years - a dated resource which needed refreshing. This report is currently
with Road Safety Scotland for consideration and future action.
2.4.48 The link between problem behaviour and social values is confirmed by the Department
for Transport DfT (1999) in a road safety research report which looked at problem
behaviour in children and accident risk. The report included a focused literature review of
previous research on accident risk and behavioural issues together with some original
research based on child and parent interviews. The conclusion is that social values
explained the link between problem behaviour and traffic collisions but not other types of
accidents. The study looked at the possibility that problem behaviour led to greater time
spent on the street exposed to risk and this was controlled for in the analysis. Taking this
into consideration, the report concluded with a high level of confidence that there is a link
between problem behaviour and traffic collision rates in children and that this “was not
simply a result of delinquent children spending more time in the traffic environment”. The
report also concluded that tackling this problem will require quite intensive and prolonged
intervention to establish significant improvements. The report, unfortunately, does not
indicate what these interventions might be.
2.4.49 Edwards et al (2008) in a wider study of serious injuries to children and deprivation
concluded that the major policy need related to safer road environments for pedestrians
and cyclists involving speed reduction. It is also suggested that, in the interests of
developing a “critical mass” effect, more walking and cycling should be encouraged as
this may well reduce the overall risks for these modes.
2.4.50 In an earlier report for the London Road Safety Unit, Edwards et al (2006) looked at a
range of potential remedial measures. They spelt out a general strategy in which they
recommended:
promotion of walking and cycling and discouragement of motor vehicle use
car free planning and designing particular urban areas for minimal motor vehicle
use
pedestrian-orientated commercial streets
public parks that encourage or require non-automotive access
18
Final Report
2.4.51 They also highlighted the need for well targeted traffic calming and speed enforcement
action (eg radar, speed cameras etc). Education, training and publicity also have a place
in a balanced and holistic response to high casualty rates.
2.4.52 Grayling et al (2002) indicate that there are some highly cost-effective ways of reducing
pedestrian casualties including:
‘spot’ measures such as installing pedestrian crossings
more 20mph zones
all local transport plans should include a speed management strategy
2.4.53 Since this report there is evidence that more local authorities are looking closely at risks
in areas of deprivation and are taking wider action to reduce the impact of deprivation
socially and also specifically in terms of road safety. More 20mph zones (either
mandatory or advisory) have also been introduced.
2.4.54 Graham and Stephens (2008) suggest that:

local authorities’ road safety strategies can be compatible with other local
regeneration or redistributive policy measures aimed at alleviating deprivation
local authorities should examine if road safety measures are unevenly distributed
with respect to their socio-economic environments.
2.4.55 The research reviewed has shown that there is a specific deprivation effect over and
above environmental factors relating to traffic, weather, population density etc. It also
appears that there are links between deprivation, behaviour and risk taking. Most of the
remedial measures discussed above do not relate specifically to this increased risk-taking
element of behaviour.
2.4.56 Whelan et al (2008) in a comprehensive report look at the Kerbcraft initiative which has
been introduced in a number of deprived areas in Great Britain. The initiative is shown to
improve children’s crossing skills and behaviour and the report looks at ways in which
Kerbcraft can be improved and rolled out to other areas. This initiative may well be
suitable for areas of deprivation in Northern Ireland. It relies heavily on volunteers and
requires a strong commitment from officials to make it work successfully and sustainably.
Key points of relevance to the present study
- the link between deprivation and child pedestrian casualty rates

2.4.57 The key points of relevance to the current study are:
significant links between deprivation and child pedestrian casualty rates are
evident [all studies reviewed]
in London, the pedestrian injury rate for children living in the most deprived areas
were found to 2.93 times higher than equivalent rates for children resident in the
least deprived areas (based on deciles). For the Welsh data the figure was 2.5
(based on quintiles). For data relating to Edinburgh for 2000-03 the equivalent
figure was 4.05 (based on seven quantiles)
the risk of serious injury or death in a road collision increases with increasing
deprivation [Graham et al (2004)]
these links have been found to exist for different areas of the UK in England,
Wales, Scotland and Northern Ireland for datasets based on STATS19, accident
and emergency and hospital episodes [England - Kendrick (1993), Grayling et al
(2002), Graham et al (2004), Noland and Quddus (2004), Graham and Stephens
(2008) and Edwards et al (2008): Scotland - McGuigan, Bull and Gorman (1996),
19
Final Report
Abdalla et al (1997) and Chichester et al (1998),: Wales - NPHS (2004): Northern

Ireland - Silversides et al (2005)]
the individual factors which make up an index of multiple deprivation are also
linked to child pedestrian casualty rates but with different statistical relationships
[Edwards et al (2006) and Graham and Stephens (2008)]
other variables, such as population, road infrastructure, employment density,
weather (rainfall and sunshine hours) are also found to be linked to child
pedestrian casualty rates [Petch and Henson (2000), Graham et al (2004),
Graham et al (2006) and Noland and Quddus (2004)]
the effect of deprivation varied by region across England; for example, in London
the deprivation effect is particularly strong for children but less so for adults (wrt to
the average for the study) [Grayling et al (2002) and Graham and Stephens (2008)]
the use of quantiles (particularly, quintiles or deciles) by a number of the research
projects to stratify deprivation measure values [Abdalla et al (1997), Grayling et al
(2002) and Edwards et al (2006)]
some research has indicated that it is acceptable to use location of accident as a
proxy for place of residence for child pedestrian casualties [Hewson (2004) and
Edwards et al (2008)]. Others have indicated that significant proportions of children
are injured as pedestrians in excess of 1km from home [Abdalla et al (1997), and
Petch and Henson (2000)]
2.4.58 It is perhaps worth noting that much of the research reviewed which relates to England
although independently carried out will have been based, in part, on the same data. The
sources for the casualty data would have been the STATS19 and the other data sources
for deprivation measures, population, road infrastructure, weather etc will have been
based on General Register Office, Met Office, Department for Communities and Local
Government, DfT etc. Given this, the findings are not based on entirely different datasets
and it would be surprising if interpretations and conclusions differed significantly. The
complementary findings by other researchers using different data sources are, therefore,
of particular relevance in supporting the overall conclusions [Scotland - McGuigan, Bull
and Gorman (1996), Abdalla et al (1997) and Chichester et al (1998),: Wales - NPHS
(2004): Northern Ireland - Silversides et al (2005)].
- the link between casualty residence and collision location

2.4.59 There is no hard and fast evidence in the literature to be categorical about the value of
using location of residence or of collision as the most appropriate way of assigning
deprivation to an individual casualty. There is clearly a very strong argument (not
explicitly supported in the literature reviewed) to suggest at least theoretically that the
location of residence should be used because:
the behaviour of a child is more likely to be linked to the characteristics of the zone
of residence than the zone of collision which may be some distance from home
assignment by zone of collision will mean that in some cases the numerator for the
casualty rate calculation will include casualties of a different deprivation score for
the denominator
2.4.60 This does not necessarily mean, however, from a practical point of view – particularly
where the zonal units are small – that the location of collision would not provide an
appropriate proxy for location of residence [e.g. Edwards et al (2006)].
- potential remedial measures

2.4.61 It is clear that there is no single solution to reducing the effect deprivation has on child
pedestrian casualty rates. The research points to a balanced approach which involves a
number of different but complementary interventions relating to the wider local
20
Final Report
environment, road infrastructure, traffic calming, education, training and enforcement. It

is likely any approach will require intensive effort and investment over a long period of
time. The mix of these will depend on the observed casualty rate, the area, budgets
available and opportunities for change (e.g. though regeneration). A strategy might be
based on:
a national “traffic club” scheme supported, where appropriate, by Kerbcraft training
[Road Safety Scotland (2009) and Whelan et al (2008)
[Grayling et al (2002), Edwards et al (2006) and Edwards et al (2008)]
[Grayling et al (2002)]
road safety inputs to the master planning of regeneration projects [Graham and
Stephens (2008)]
local transport plans should include pedestrian reduction targets [Grayling et al
(2002)]
local authorities should examine if road safety measures are unevenly distributed
with respect to their socio-economic environments and rectify if found to be so
[Graham and Stephens (2008)]
21
Final Report
3 Data sources
3.1 Road collision casualty data

3.1.1 Road collision casualty data were provided by the PSNI. The data covered the period
January 1999 to March 2009 (inclusive). These data provide:
a full calendar ten year dataset for January 1999 to December 2008 (inclusive)
a two year dataset for April 2007 to March 2009 (inclusive) providing location of
residence information for drivers and casualties based on the home postcode
(these data became part of the standard collision reporting in April 2007)
3.1.2 The first dataset will be used as the main source of data to be used to investigate the link
between area deprivation and casualties. The second dataset will be used as the source
for looking at the accuracy of using location of collision as a proxy for location of
residence.
3.1.3 The ten year dataset January 1999 to December 2008 (inclusive) comprises:
64,651 collision records
121,260 vehicle records
109,385 casualty records
3.1.4 The distribution of casualties by casualty class and age group is presented in Table 3.1.
Table 3.1: Casualties by casualty class and age group (January 1999 to
December 2008)
Age group
16
and All
1 2
Road user group 0 1-4 5-7 8-11 12-15 over ages
1. Driver 20 10 60 67 145 55,652 55,954
2. Pillion passenger 1 10 38 210 259
3. Vehicle passenger - Front 25 263 257 625 1,128 19,449 21,747
4. Vehicle passenger - Rear 35 1,506 1,198 1641 1,894 10,699 16973
5. Pedestrian 11 499 672 960 1,104 5,583 8,829
6. Motorcyclist 1 9 44 3,871 3,925
7. Pedal cyclist 25 150 264 262 997 1,698
All road users 92 2,304 2,337 3,576 4,615 96,461 109,385
1
age 0 is used to record an age which is unknown
2
age 1 is used to record casualties of age 0 (ie before first birthday) and age 1
See also paragraph 4.1.8
3.1.5 The collision data include all collisions in NI in which there was an injury to at least one
road user. The severity of injury is recorded as killed, serious or slight where
Killed is assigned if the casualty dies at the scene of the collision or within 30 days
as a result of the injuries sustained in the collision
Serious is assigned if the casualty is detained in hospital as an ‘in-patient’, or if any
of the following injuries whether or not the casualty is detained in hospital:
- fractures
- concussion
- internal injuries
22
Final Report
- crushings
- burns
- severe cuts and lacerations
- severe general shock requiring medical treatment
Slight is assigned if the injury is of a minor character such as a sprain, bruise or cut
not judged to be severe, or slight shock requiring roadside attention.
3.2 Population data

3.2.1 The population data are those relating the 2001 Northern Ireland Census. The data were
downloaded from the Northern Ireland Statistics and Research Agency (NISRA) website
from Table EXT20060206. The web link is:
(http://www.nisranew.nisra.gov.uk/census/Excel/commissioned_output/EXT20060206.xls)
3.2.2 The table presents these data by single year of age and by sex.
3.3 Deprivation data

3.3.1 A table containing the deprivation data for NI was downloaded from the NISRA website
(http://www.nisra.gov.uk/aboutus/default.asp88.htm) from Table
NIMDM2005_SOA_level1.
3.3.2 The Multiple Deprivation Measure for 2005 (MDM 2005) was constructed by combining
the seven transformed domain scores, using the following weights:
Income (25%)
Employment (25%)
Health Deprivation and Disability (15%)
Education, Skills and Training (15%)
Proximity to Services (10%)
Crime and Disorder (5%)
Living Environment (5%)
3.3.3 In addition the table provides two further deprivation measures:
the Income Deprivation Affecting Children (IDAC)
the Income Deprivation Affecting Older People (IDAOP)
3.3.4 The distribution of the MDM (organised by decile) is presented in Figure 4.1 in Appendix
4.
3.4 Urban Rural Classification

3.4.1 The urban rural classification was obtained through a web link from a NISRA file which
contained a two level factor (U = urban and R = rural). The file contains a list an urban
rural classification for each Super Output Area (SOA) assigned from the Statistical
Classification of Settlements Report February 2005. This is known as the Urban-Rural
Definition Group (URDG) Classification. The web link is:
http://www.ninis.nisra.gov.uk/mapxtreme_deprivation2005/Viewdata/UrbanRuralClassific
ation2005.xls
3.5 Road infrastructure data

3.5.1 From the Department for Regional Development’s (DRD) Roads Services digitised road
network, road infrastructure data were obtained at the SOA level which comprise:
road length by road type (Motorway, A Class, B Class, C Class and Unclassified)
23
Final Report
number of junctions (i.e. nodes on the digitised network with three or more
attached road links)
3.5.2 From these a junction density can be calculated as:
number of junctions / road length (junctions / km)
3.5.3 The junction density is not disaggregated by road class.
3.6 People in employment

3.6.1 People in employment figures for the 2001 census are taken from Table KS12a available
on the NISRA website. The data are presented at the SOA level with numbers and
percentages of the employment status of people resident in the SOA. The table was
downloaded from the following webpage:
http://www.nisranew.nisra.gov.uk/census/Census2001Output/KeyStatistics/ks_soa.html
3.6.2 The data are disaggregated according to the following occupation groups:
Managers and senior officials
Professional occupations
Associate professional and technical occupations
Administrative and secretarial occupations
Skilled trades occupations
Personal service occupations
Sales and customer service occupations
Process, plant and machine operatives
Elementary occupations
3.7 Workplace population

3.7.1 Workplace population data were provided by Customer Services for NISRA at the SOA
level. The data provided were in an Excel format entitled EXT20090820.xls. The data
provide a single figure for each SOA representing the number of persons aged 16-74 in
employment in the area.
3.8 School pupil data

3.8.1 School pupil data for 2008/2009 were provided in Excel format by the Statistics and
Research Branch of the Department of Education. All data were sourced from the Annual
School Census for that year. The data list the number of school pupils attending schools
by the SOA of the school (not the SOA of the pupil). The pupil numbers were
disaggregated as follows:
Age 5 - 7 Primary
Age 5 - 7 Special
Total age 5 - 7
Age 8 - 11 Primary
Age 8 - 11 Special
Age 11 Post primary
Total age 8 - 11
Age 12 Primary
Age 12 - 15 Post primary
Age 12 - 15 Special
Total age 12 - 15
24
Final Report
3.9 Car ownership data

3.9.1 Car ownership figures for the 2001 census are taken from Table KS17 available on the
NISRA website. The data are presented at the SOA level with numbers and percentages
of households with none, 1, 2, 3 or 4 or more cars or vans. The table was downloaded
from the following webpage:
3.10 Household tenure data

3.10.1 Household tenure data for the 2001 census are taken from Table KS18 available on the
NISRA website. The data are presented at the SOA level with numbers and percentages
of households by three different ownership and four different rented categories. The
table was downloaded from the following webpage:
3.11 Weather data

3.11.1 Weather data were provided by the Met Office following a specific request for the
information. The data are:
annual rainfall measurements (mm) for thirty weather stations in Northern Ireland (1999 -
2008)
annual sunshine hours (hrs) for eleven weather stations in Northern Ireland (1999 -2008)
3.12 Creation of composite variables

3.12.1 A number of composite variables were created as functions of the data compiled for the
study. These composite variables match in a number of instances variables used in
research outlined in the literature review. These composite variables are:
3.12.2 Traffic proxy: this variable is designed to provide a proxy for travel activity levels in each
SOA. This variable was described by Graham et al (2005). It is a function of the two
variables:
people in employment (i.e. the number of people in employment resident in the
SOA)
workplace population (i.e. the number of people with jobs within the SOA)
taking the form:
traffic proxy = √(people in employment * workplace population)
3.12.3 Road length per hectare: this variable is designed to provide a measure for road density
in each SOA. This variable was described by Noland and Quddus (2004). It is a function
of the two variables:
total road length
SOA area (hectares)
taking the form:
road length per hectare = total road length / area
3.12.4 Junctions per road km: this variable is designed to provide a measure for junction
density in terms of road length in each SOA. It is a function of the two variables:
number of junctions
25
Final Report
total road length

taking the form:
junctions per road km = number of junctions / total road length
3.12.5 Junctions per hectare: this variable is designed to provide a measure for junction
density in each SOA. This variable was described by Graham et al (2005) and Noland
and Quddus (2004). It is a function of the two variables:
number of junctions
SOA area (hectares)
taking the form:
junctions per hectare = number of junctions / area
3.12.6 Traffic density per road km: this variable is designed to provide a relative measure for
traffic density levels in each SOA in terms of road length. It is a function of the two
variables:
traffic proxy
total road length
taking the form:
traffic density per road km = traffic proxy / total road length
3.12.7 Traffic density per junction: this variable is designed to provide a relative measure for
traffic density levels in each SOA in terms of the number of road junctions. It is a function
of the two variables:
traffic proxy
number of junctions
taking the form:
traffic density per road km = traffic proxy / number of junctions
3.12.8 Resident population per road junction: this variable is designed to provide a relative
measure for population exposure at junctions in each SOA. This variable was described
by Graham et al (2005). It is a function of the two variables:
total resident population
number of junctions
taking the form:
resident population per road junction = total resident population / number of

junctions
3.12.9 Employed residents per road junction: this variable is designed to provide a relative
measure for employed resident exposure at junctions in each SOA in terms of the
number of road junctions. This variable was described by Graham et al (2005). It is a
function of the two variables:
employed residents
number of junctions
taking the form:
employed residents per road junction = employed residents / number of junctions
3.12.10 Workplace population per road junction: this variable is designed to provide a relative
measure for workplace population at junctions in each SOA in terms of the number of
26
Final Report
road junctions. This variable was described by Graham et al (2005). It is a function of

the two variables:
workplace population
number of junctions
taking the form:
workplace population per road junction = employed residents / number of junctions
27
Final Report
4 Ten Year Analysis
4.1 Datasets
Data Sources
4.1.2 The data used in the study have been either downloaded from the Northern Ireland
Statistics and Research Agency (NISRA) website or have been provided directly by the
appropriate agency. For a detailed description see Chapter 3.
Two road casualty datasets

4.1.3 Two overlapping datasets have been defined for use in the study. These have been
provided by the PSNI.
4.1.4 The main dataset comprises data for the ten year period covering the calendar years
1999 to 2008. This dataset will be used for the main purpose of the study to examine if,
in Northern Ireland, a relationship exists between deprivation and the child pedestrian
casualty rate.
4.1.5 The other dataset comprises data for the two year period from April 2007 to March 2009
(inclusive). This data set will be used to test the assumption that the location of collision
may be used as a proxy for location of casualty residence. It is important to investigate
this because information regarding people’s deprivation status is usually related to their
home address. In Northern Ireland (NI), casualty postcodes have only recently been
assigned to the collision data (from April 2007 onwards) and it is not, therefore, possible
to use the socio-economic status of the home address for the ten year dataset. Using the
two year dataset, however, it is possible to compare the assignment of deprivation to
individual casualties based on location of collision and location of residence to determine
the differences that exist and draw conclusions about the use of location of collision as a
proxy for location of residence.
The 10 year road casualty dataset

4.1.6 The ten year dataset covers the calendar years 1999-2008 and comprises
4.1.7 The distribution of casualties by casualty class and age group is presented in Table 1 of
7
Appendix 2 .
4.1.8 The PSNI reserve the use of an age of 0 for casualties of an unknown age and the
dataset includes 92 such casualties. Accordingly, these 92 casualties over the 10 year
period have been excluded from the analysis. Casualties who are aged under one year
are coded as age “1”.
4.1.9 As can be seen in Table 2 there are four age groups for children accounting for 3,235
child pedestrian casualties.
7
All Tables and Figures referred to in this Chapter are presented in Appendix 2.
28
Final Report
The 2 year road casualty dataset

4.1.10 The two year dataset covers the period from April 2007 to March 2009 (inclusive) and
comprises
4.1.11 The distribution of casualties by casualty class and age group is presented in Table 3.
4.1.12 As can be seen in Table 4 there are four age groups for children totalling 476 child
pedestrian casualties.
4.2 Deprivation dataset

4.2.1 The deprivation dataset comprises individual multiple deprivation measures (MDMs) for
each of Northern Ireland’s 890 Census super output areas (SOAs). The MDM was
constructed from seven independent factors including income, employment and proximity
to services. The analyses which follow use the MDM values.
4.2.2 The SOAs have been sorted by MDM values and split into ten groups ranging from 1
(most affluent) to 10 (most deprived). These groupings are referred to as deciles.
4.3 Population dataset

4.3.1 Population data from the NI Census of 2001 has been downloaded to provide population
by year of age and sex for each SOA.
4.4 Casualty rate analysis

Child pedestrian casualty rates
4.4.2 The total number of child pedestrian casualties aged 0-15 over the ten years 1999 to
2008 are set out in Table 5 stratified by MDM decile, age group and sex.
4.4.3 The corresponding populations stratified by MDM deciles, age group and sex are set out
in Table 6.
4.4.4 The casualty rates for these stratifications are set out in Table 7. The casualty rate is
expressed as the annual average number of casualties per 1,000 population, viz:
(10 year casualty total/10) / (population/1,000)
4.4.5 It can be seen in Table 7 that for each age group for both males and females, the
casualty rates show an upward trend as the MDM decile rises. In some cases the trend
is clear (e.g. for males aged 8 to 11) in others it is less clear (e.g. for females aged 12-
15). In all cases, however, the relationships are highly statistically significant at least at
the 2.5% level, with most of the relationships being statistically significant at the 0.5%
level (Table 8 refers).
4.4.6 Table 9 shows deprivation disparity ratios (DDRs) which are the ratios of rates for decile
10 (most deprived) to rates for decile 1 (most affluent). The DDR is a measure of the
additional risk of being involved as a pedestrian road collision casualty between children
resident in the most affluent and most deprived areas. It is noted that the overall DDR for
females is higher than the value for males. It would, however, be unwise to draw specific
conclusions from this result because in terms of casualty numbers the data for decile 1
29
Final Report
(most affluent) are the most sparsely populated area of the data. There are 46 female
casualties and 87 male casualties recorded for decile 1 (Table 5 refers).
4.4.7 The rates set out in Table 7 are shown in graphical form in Figures 1 – 8.
Child pedestrian casualty rates by decile (all ages both sexes)

4.4.8 Figure 1 shows the casualty rates for all casualties (i.e. all age groups for males and
females combined).
4.4.9 The key points to note are:

there is a clear trend of increasing rates with increasing area deprivation
2
the R statistic shows that just over 90% of the variation in casualty rates is
explained by an exponential equation
the DDR for the most deprived decile is 4.85 times the rate for the most affluent
8
decile
Child pedestrian casualty rates by decile and sex (all age groups)
4.4.10 Figure 2 shows the casualty rates for all child age groups grouped by sex.

there is a clear trend of increasing rates with increasing deprivation for males
there is a clear trend of increasing rates with increasing deprivation for females
although it is not as strong as it is for males
the rates for males are higher: overall the rate for males is 1.49 times that for
females
the DDR for males is 4.54
the DDR for females is 5.29
Child pedestrian casualty rates by decile and sex (ages 0 - 4)

4.4.12 Figure 3 shows the casualty rates for children aged 0 - 4 grouped by sex.

there are clear trends of increasing rates with increasing deprivation for males and
females
the rates for males are higher: overall the rate for males is more than double that
for females (2.1 times greater)


there are clear trends of increasing rates with increasing deprivation for males and
females
the rates for males are generally higher (with one exception in decile 6): overall the
rate for males is 1.45 times that for females
8
This disparity is rather greater than those identified in the literature where an equivalent figure of 2.93 was
observed for London data and a figure of 2.5 for Welsh data. The London data were based on deciles but the
Welsh data were based on quintiles which would be very likely to identify a lower disparity ratio.
30
Final Report



there is a clear trend of increasing rates with increasing deprivation for males
there is a trend of increasing rates with increasing deprivation for females although
it is not as clear as it is for males
the rates for males are higher: overall the rate for males is 1.52 times that for
females


there is a trend of increasing rates with increasing deprivation for males
the rates for females are generally lower (with one exception in decile 6): overall
the rate for males is 1.30 times that for females
Summary of child pedestrian casualty rates

4.4.20 Figures 7 and 8 combine the information provided in Figures 3 to 6 for males and
females respectively.
Adult pedestrian casualty rates

4.4.21 The total number of adult pedestrian casualties (aged 16 and over) for the ten years 1999
to 2008 are set out in Table 10 stratified by MDM decile, age group (16-24 and 25 and
over) and sex. The figures for all children (ages 0-15) are also included for ease of
reference.
4.4.22 The corresponding populations stratified by MDM deciles, age group and sex are set out
in Table 11.
4.4.23 The casualty rates for these stratifications are set out in Table 12.
4.4.24 It can be seen in Table 13 that for each age group for both males and females, the
casualty rates show an upward trend. In all cases the relationships are highly statistically
significant at the 0.5% level.
4.4.25 Table 14 shows deprivation disparity ratios (DDRs) which are the ratios of rates for decile
10 (most deprived) to rates for decile 1 (most affluent). Here the DDR is a measure of
the additional risk of being involved as a pedestrian road collision casualty between
adults resident in the most affluent and most deprived areas.
4.4.26 The rates set out in Table 12 are shown in graphical form in Figures 9 – 11.
31
Final Report
Young adult pedestrian casualty rates by decile and sex (ages 16 – 24)
4.4.27 Figure 9 shows the casualty rates for young adults aged 16-24 grouped by sex.

there is a trend of increasing casualty rates with increasing deprivation for males
and females
the casualty rates for males are higher: overall the rate for males is 1.55 times that
for females
Adult pedestrian casualty rates by decile and sex (ages 25 and over)
4.4.29 Figure 10 shows the casualty rates for young adults aged 25 and over grouped by sex.

there is a clear trend of increasing casualty rates with increasing deprivation for
males and females
for females
Pedestrian casualty rates for all ages by decile and sex (ages 0 and over)
4.4.31 Figure 11 shows the casualty rates for all ages (0 and over) grouped by sex.

there is a clear trend of increasing casualty rates with increasing deprivation for
males and females
for females
Summary of adult pedestrian casualty rates

4.4.33 Figures 12 and 13 combine the information shown in Figures 9 – 11 for males and
females respectively. The rates for all child pedestrian casualties (for ages 0 – 15) are
also shown for comparison.
4.4.34 For male casualties, the key points to note are:

generally children experience the highest casualty rates: this is particularly the
case in the more deprived areas (deciles 7 –10)
the rates for older adults (ages 25 and over) are the lowest.
the foregoing suggests a reducing male casualty rate with increasing age.
4.4.35 For female casualties, the key points to note are:
generally children experience the highest casualty rates: this is particularly the
case in the more deprived areas (deciles 6 –10)
the rates for older adults (ages 25 and over) are the lowest.
the foregoing suggests a reducing female casualty rate with increasing age.
32
Final Report
4.5 Location of collision as a proxy for location of residence
Overview
4.5.2 In this section the two year – April 2007 to March 2009 (inclusive) - dataset is used for
the analysis.
4.5.3 During this period there were 476 child pedestrian casualties. All but 28 (6%) of these
casualties were reported with a postcode which has been assigned to an SOA. The
location of each collision has also been assigned to an individual SOA and we describe a
“home” casualty as a person who was injured in a collision in the same SOA in which
he/she is resident. An “away” casualty is a person injured in an SOA outside the SOA of
his/her residence. Table 15 summarises these.
Analysis
4.5.4 Information regarding the SOA of residence has been recorded in Northern Ireland for
each collision since 1 April 2007. This means that the analysis of the ten year data set
cannot use the casualty “home” SOA information as the link does not exist for all but 21
months. There is a concern that using the location of collision as a proxy may impair the
analysis. This concern is not, however, generally, borne out by the literature review. The
purpose of this analysis is to examine the issue using data local to Northern Ireland.
4.5.5 Of the casualties for which a postcode was provided, 54% were injured in collisions
outside their home SOA with the remaining 46% injured within their home SOA. Although
the “away” collisions may be in any other SOA in the province, it is likely that the majority
will occur more close to home (perhaps in an adjacent SOA) and possibly in an SOA with
the same or similar MDM and deprivation decile.
4.5.6 The literature review includes a study by Petch and Henson (2000) which examined
factors relating to child accidents in Salford, England. In this study the authors reported
on a tabulation of the distance between accident location and place of residence. Figures
for child pedestrian casualties were set out in the report and are summarised in the first
two columns of Table 16. The third column of the table shows the equivalent data for NI.
The NI distances were provided by the PSNI for 447 of the 448 child pedestrian
casualties in the dataset. The data for NI show that whist 68.5% of casualties are injured
within 1km of their home address, the corollary is that 31.5% are injured at distances of
more than 1km with 21.4% at distances in excess of 2km.
4.5.7 The NI child pedestrian casualties tend to be involved in collisions further from their home
than is reported for Salford. Over one in five (21.4%) of NI casualties occur more than
2km from the home address compared with an equivalent figure for Salford of 11.4%.
The Salford data would be expected to show a different distribution for longer distances
because the dataset was limited to casualties resident in Salford and who were injured in
2
Salford. Salford has an area of 96.4 km which is very much smaller than the area of NI
2 9
which is 13,576.5 km (excluding major inland waterways) . It is interesting to note that
the proportions of casualties injured within .25km of home are almost identical for both
Salford and NI.
4.5.8 For two reasons this need not suggest that the location of the collision is necessarily a
poor proxy for the location of residence of casualties. These reasons are:
some longer distances – particularly in rural areas – may occur within an SOA
which covers a wide area
9
http://www.nisranew.nisra.gov.uk/census/area_measurement.html
33
Final Report
where the SOA for the collision is different from the SOA of residence of the
casualty, the SOAs may be adjacent and may share the same (or similar)
deprivation score.
4.5.9 Table 17 summarises the analysis by decile and crosstabulates the decile for the location
of the collision for “away” casualties. Columns 2 and 3 identify the number of “home” and
“away” casualties. The next ten columns tabulate the deprivation deciles for the location
of collision for “away” casualties. Although some of the numbers are small, it is evident
that there is some clustering of “away” collision deciles around the MDM decile value of
the casualty’s home SOA.
4.5.10 The shaded cells in Table 17 highlight the “away” deciles that are within two deciles of
the “home” decile. Two deciles have been selected as an indicative “buffer” to identify
broadly similar levels of deprivation. So, for example, where the decile of residence is 7,
the “buffer” area includes away deciles ranging from 5 to 9. The second last column
sums the number of casualties in the “buffer” and the final column adds these to the
decile of residence to create a “buffer total” value. This, therefore, creates a value for:
(the number of casualties occurring in the home SOA) +

(those in an away SOA with a decile within 2 deciles of the home SOA
decile)
4.5.11 The buffer total for all the data sums to 362 which comprises 81% of the total number of
child pedestrian casualties in the dataset. This means that for 81% of all child pedestrian
casualties the use of location of collision provides the same or similar (within 2 deciles)
deprivation score that would be assigned from location of residence. Or, to put it another
way, fewer than 20% of the casualties are injured in away SOAs with widely differing
MDM scores.
4.5.12 Figure 14 shows pedestrian casualty rates where the deprivation decile is based on
location of residence. Figure 15 shows pedestrian casualty rates where the deprivation
decile is based on location of collision. Figure 16 shows the difference in rates by
deprivation decile: here the difference is:
Collision based rate – residence based rate
4.5.13 Where the difference is positive it means that the collision based rate is higher and when
it is negative the opposite is true: the residence based rate is higher.
4.5.14 Figure 16 has deliberately been kept to the same scale as Figures 14 and 15 for
comparison. The key points to note are:
the value of the differences are generally low
there is no particular pattern or trend to the differences in relation to the decile
values
Conclusion
4.5.15 The analysis shows that:
for 81% of all child pedestrian casualties the use of location of collision provides
the same or similar (within 2 deciles) deprivation score that would be assigned
from location of residence
the difference between casualty rates based on location of collision and location of
residence are:
- generally low
- not biased in relation to deprivation decile
34
Final Report
4.5.16 The conclusion is that the use of location of collision as a proxy for location of residence
is very unlikely to skew results to an extent that would substantially impact on the
robustness of the findings of the study.
4.6 Analysis of collision factors

4.6.1 In this section the distribution of child pedestrian casualties is examined across a range of
factors reported in the ten year collision dataset. They are:
Severity of injury by sex and year
Age
Time of day / month
Speed limit
Weather
Light condition – daylight/dark and streetlights lit/unlit
Responsibility - causation factors
Junction type (if at junction) and junction control
Presence of Pedestrian Crossing Facilities
Carriageway type
Road surface conditions
Special conditions/hazards
Vehicle location/manoeuvre
On a journey to or from school
Pedestrian location/movement
Severity of injury by sex and year

4.6.2 The child pedestrian casualty data are presented by year, severity of casualty and sex for
10
killed and seriously injured (KSI) casualties and for all casualties in Figures 17 and 18
respectively. The key points to note are:
there is a clear downward trend by year for KSI and for all child pedestrian
casualties for both males and females. The trends are highly statistically
significant (p <0.005)
although there is a clear downward trend by year for girl casualties the proportional
decrease is less than it is for boy casualties
Casualties by sex, age group and year

11
4.6.3 The child pedestrian casualty data are presented by year of age , age group for males
and females in Figures 19 and 20 respectively. The key points to note are:
there is a clear downward trend by year for child pedestrian casualties for both
males and females in all age groups. The trends are statistically significant (p
<0.05)
Casualties by time of day and month

4.6.4 Tables 18 and 19 show the distribution of child pedestrian casualties by time of day and
month of year for weekdays and weekend respectively.
10
The figure for all casualties includes casualties whose severity is defined as slight. The difference between
KSI casualties and all casualties represents slights casualties.
11
The age groups 0-4, 8-11 and 12-15 cover periods of four years. The age group 5-8 covers three years. To
control for this difference, casualty totals in the former periods have been divided by 4 and the casualty totals for
the 5-7 age group have been divided by 3 to provide average values for each age group by year of age.
35
Final Report
4.6.5 From Table 18 it can be seen that for weekdays casualty numbers peak for the three
hour travel home from school period 15.00 (3pm) to 18.00 (6pm). There is a smaller
peak in the morning during the travel from home to school hour beginning 08.00 (8am).
These four hours account for 50% of all child pedestrian casualties during weekdays. In
terms of monthly distribution it can be seen that child pedestrian casualties tend to peak
during spring (April – June) and the autumn (September – October) although this is not
very pronounced.
4.6.6 For weekends (Table 19 refers) the picture is rather different with casualty numbers
peaking throughout the afternoon and into the evening over the nine hour period 12.00
(noon) until 21.00 (9pm). These nine hours account for 87% of all child pedestrian
casualties at weekends. As is the case for weekday casualties, in terms of monthly
distribution it can be seen that child pedestrian casualties tend to peak during spring
(April – June) and the autumn (September – October) although this is not very
pronounced.
Speed limit
4.6.7 The distribution of child pedestrian casualties by speed limit is shown in Table 20. The
data have been stratified by urban rural classification, sex and age group.
4.6.8 It is evident that in rural areas that the proportion of casualties injured on roads subject to
a 30mph speed limit falls with increasing age with a corresponding rise in the casualties
on unrestricted roads where the speed limit is 60mph. This is observed for both males
and females.
4.6.9 In urban areas there is similar but much less marked situation where the proportion of
casualties injured on roads subject to a 30mph speed limit falls with increasing age with a
counterbalancing rise in the casualties on roads where the speed limit is 40 to 60mph.
This is observed for both males and females.
Weather
4.6.10 The distribution of child pedestrian casualties by speed limit is shown in Table 21. The
data have been stratified by urban rural classification, sex and age group.
4.6.11 The significant majority of casualties are injured in collisions when fine weather prevails.
The urban rural classification has no clear effect. For females and males it is evident that
in urban areas the proportion of casualties injured during fine weather falls with
increasing age with a corresponding rise in the casualties injured when it is raining. The
same pattern is not evident for rural areas.
Light condition – lit/unlit

4.6.12 The distribution of light condition stratified by month of year is presented in Table 22. As
expected the time of year has a significant impact on the distribution of this variable.
4.6.13 In total, 80% of casualties are involved in collisions during daylight hours; this figure rises
to over 95% in the summer months (May – July) and falls to under 65% in the winter
months (November – January).
4.6.14 Conversely, in total, 20% of casualties are involved in collisions during darkness hours;
this figure rises to 50% in December and falls to under 5% in the summer months (May –
July).
36
Final Report
Principal causation factor

4.6.15 The principal causation factor stratified by age group for child pedestrian casualties is
summarised in Table 23. The key points to note are:
pedestrian only factors account for 84% of all factors
driver only factors account for 15% of all factors
the other factor groups (vehicle, weather, passenger and miscellaneous) account
for just 1% of all principal causations.
for pedestrians the most frequently occurring principal factors are:
- heedless of traffic crossing carriageway (35% of all reported factors)
- walk/run movement masked (25%)
- lack of supervision/escaping from custody (10%)
- walking or running onto carriageway (5%)
for children in the 0 – 4 year age group over half (54%) the principal causation
factor is lack of supervision/escaping from custody
for drivers the most frequently occurring principal factors are:
- inattention or attention diverted (5% of all reported factors)
- reversing without care (2%)
- excessive speed having regard to conditions (2%)
- disobeying pedestrian crossing (2%)
Junction detail and control

4.6.16 A cross tabulation of junction detail by junction control for child pedestrian casualties is
presented in Table 24. This table shows that 50% of all casualties do not occur at
junctions. Of the junction accidents, the key points to note are:
75% of casualties occur at T or staggered junctions
46% and 29% occur where the form of junction control is give way and
uncontrolled respectively (another 10% occur when the junction control has been
recorded as give way or uncontrolled)
11% occur where there are traffic signals
Pedestrian crossing facility

4.6.17 Table 25 presents the data for pedestrian crossing facility stratified by age group. Overall,
80% of child pedestrian casualties are injured in collisions with no pedestrian crossing
facility within 50m. This figure decreases from 88% for age group 0 – 4 years to 74% for
age group 12 – 15 years with a corresponding increase in the proportion of Pelican and
other light controlled crossings being present.
Carriageway type
4.6.18 The carriageway type associated with child pedestrian casualties is set out in Table 26
stratified by urban rural classification. Collisions on single carriageway roads account for
over 89% of all child pedestrian casualties. The urban rural classification does not
significantly influence the overall distribution of carriageway types associated with child
pedestrian casualties.
Road surface condition

4.6.19 The road surface condition associated with child pedestrian casualties is set out in Table
27 stratified by urban rural classification. Collisions on dry roads account for just under
three quarters (73%) of all child pedestrian casualties. Collisions on wet/damp roads
account for just under a quarter (24%) of all child pedestrian casualties. Other road
surface conditions are present in just 3% of collisions involving child pedestrian
37
Final Report
casualties. The urban rural classification does not significantly influence the overall
distribution of road surface conditions associated with child pedestrian casualties.
Special conditions at site

4.6.20 Table 28 sets out the special conditions at site for collisions involving child pedestrian
casualties. 99% of the collisions are reported to have no special conditions with 1%
relating to road works.
Carriageway hazards
4.6.21 Table 29 sets out the carriageway hazards present for collisions involving child
pedestrian casualties. 99% of the collisions are reported to have no carriageway hazards
with 1% relating to other object in carriageway.
Vehicle manoeuvre and vehicle location

4.6.22 Child pedestrian casualties by vehicle manoeuvre and vehicle location at time of impact
are cross-tabulated in Table 30. 60% and 32% of vehicle locations were on the major
and minor road respectively.
4.6.23 Two thirds (67%) of vehicles were recorded as “going ahead other” with 9% passing a
stationary vehicle on the off side.
Journey to/from school

4.6.24 Table 31 presents child pedestrian casualties by school pupil casualty by age group for
week and weekend days. Part A of the table shows all casualties and Part B shows the
average number of casualties per day (i.e. the weekday totals divided by 5 and the
weekend totals divided by 2) to allow for comparison between week days and weekend
days. The key points to note are:
there are no journey to/from school casualties at weekends (the road casualty data
provided by the PSNI do not record any collisions involving children as pedestrians
on journeys to or from school on a Saturday or a Sunday)
the number of school pupil casualties on a journey to/from school increases with
age from an average of 18.8 casualties per weekday for 5 – 7 year olds to 78.0 for
12 – 15 year olds
for all age groups the average daily casualty totals for weekdays are higher than
for weekend days
Pedestrian location and pedestrian movement

4.6.25 Child pedestrian casualties by pedestrian location and pedestrian movement are cross-
tabulated in Table 32. The key points to note are:
over half (53%) of casualties are reported to have been in carriageway crossing
elsewhere at the time of the collision
almost half of all casualties (47%) were crossing from driver’s nearside, split 30%
unmasked and 17% masked by a static vehicle
just over a third (35%) were crossing from driver’s offside, split 21% unmasked and
14% masked by a static vehicle
38
Final Report
5 Statistical Modelling
5.1 Overview
5.1.1 Statistical modelling is a process which requires a great deal of care and consideration to
avoid misuse and consequent misinterpretations of the model outputs. In the context of
this study there are three considerations which need to be taken into account.
1. In general terms it is widely accepted as good practice that there is a virtue in the
parsimonious use of independent variables to keep models practical and simple.
2. Where there is strong intercorrelation between independent variables a statistical
model does not need both independent variables in order to explain the dependent
variable. It can happen that one of the independent variables will be eliminated
from the regression and it can often be that it is the wrong variable. This is usually
called "nonsense-elimination". Another problem with intercorrelation is that the
confidence intervals on the regression coefficients will be very wide and may even
include zero, which means you can’t even be confident whether an increase in the
X value is associated with an increase, or a decrease, in Y. Where the confidence
intervals are so wide, excluding a variable (or adding a new one) can change the
coefficients dramatically – and may even change their signs.
3. Identify and manage any interactions there may be between variables. In simple
situations the effect of each independent variable is completely separate from the
other independent variables. Whatever the effects of the different independent
variables are, they just add up or accumulate in a simple way. These are called
additive effects (i.e. there is no interaction). In more complicated situations the
effect of one independent variable depends on another independent variable(s). In
terms of road safety males may take more risks than females but the level of
additional risk may be linked to age or possibly deprivation. This effect is
interaction.
5.1.2 These considerations will be germane as the analysis progresses.
5.2 Statistical package

12
5.2.1 The R statistical package was used to undertake the modelling work required. R is an
integrated software environment for statistical computing, modelling and graphics.
5.3 Four stages

5.3.1 This work was undertaken was undertaken in four stages:
Stage 1 relates to looking at possible intercorrelations to eliminate at an early
stage a number of variables which are clearly intercorrelated. This is not intended
to remove all intercorrelations because the modelling process may assist in
understanding the linkages between the intercorrelated variables and casualty
rates before deciding which (if any) should be eliminated from the model. The
output from Stage 1 is a list of variables to take forward into Stage 2.
The purpose of Stage 2 is to undertake a high level analysis (Level 1 Modelling) of
the relationship between casualty numbers and the key variables of deprivation,
age, sex and URDG classification. These data were collated in the form of a
contingency table. This analysis has been found to provide pointers to the
importance of these variables in explaining casualty numbers and to the shape and
12
R Development Core Team (2005). R: A language and environment for statistical computing. R Foundation
for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, URL http://www.R-project.org.
39
Final Report
form of the detailed modelling (Level 2 Modelling) in Stage 3. This analysis is

based on grouping the data according to factor values representing the data.
The purpose of Stage 3 is to look in detail - through a Level 2 modelling process -
at the variables which influence casualty numbers. This analysis is informed by
Level 1 modelling outputs and is based on ungrouped data representing each of
the 890 SOAs. This Level 2 modelling process will identify which variables should
be retained in and which should be eliminated from a final model.
In Stage 4 a final model (Level 3 Model) is developed which contains the key
variables which explain the casualty rates. This model is effectively a distilled
version of the Stage 3 model. The implications for road safety planning are also
examined.
5.4 Stage 1: intercorrelation of dependent variables

5.4.1 Before starting the modelling process in the light of the introductory remarks to this
chapter it, therefore, makes sense to look at the correlations there are between the
independent variables.
Multiple deprivation measure

5.4.2 The multiple deprivation measure (MDM) adopted for the study (source: NISRA) is itself a
weighted function of seven separate (domain) scores relating to:
Income
Employment
Health
Education
Proximity to Services
Crime
Living environment
5.4.3 A separate Income Deprivation Affecting Children (IDAC) measure has also been
calculated by NISRA.
5.4.4 For each of the 890 SOAs values for the seven domains have been determined and
these have been combined to generate an MDM for each SOA.
5.4.5 The intercorrelations between these domains (across all 890 SOAs) are shown in Table
5.1.
40
Final Report
2
Table 5.1: Intercorrelations (R values) between multiple deprivation measure,
IDAC and constituent domains
Environment
Proximity To
Employment
Education
Services
Income
Health
Living
Crime
IDAC
MDM
MDM domain
MDM -
Income .932 -
Employment .910 .878 -
Health .669 .551 .567 -
Education .796 .693 .666 .468 -
Proximity to Services .105 .100 .109 .271 .086 -
Crime .217 .182 .159 .342 .117 .440 -
Living environment .604 .521 .475 .436 .459 .066 .152 -
IDAC .874 .934 .789 .556 .689 .149 .228 .517 -
2
5.4.6 The R values effectively measure the proportion of variation in one variable explained by
the other. So, for example, the variation in the Income factor explains some 93% of the
variation in the MDM. At the other end of the scale, the variation in the Proximity to
Services factor explains just over 10% of the variation in the MDM. Although all the
values for R2 are highly statistically significant at least at the 0.1% level, many of them do
not explain very large proportions of variation in other variables.
5.4.7 At this point it was decided to drop the Income, Employment, Health, Education, Living
environment domains and IDAC from further analysis and take forward the MDM, with the
Crime and Proximity to Services domains to the detailed modelling stage. It was also
observed that there was a strong relationship between car (and van) ownership levels
and the MDM: accordingly it was decided to eliminate the level of car ownership from the
detailed modelling stage (see Figure 5.1). This was also the case with housing tenure
variables and they too were eliminated from the detailed modelling stage (see Figure 5.2)
41
Final Report
Figure 5.1: Plot of MDM Score with proportion of households with no car
Plot of MDM Score v proportion of households with no car
80
Proportion of households with no car(%)
60
40
20
0
0 20 40 60 80
MDM Score
Figure 5.2: Plot of MDM Score against proportion of rented households
Plot of MDM Score v proportion of rented households

100
Proportion of rented households(%)
80
60
40
20
0
0 20 40 60 80
MDM Score
5.4.8 At this stage it was decided to keep the composite variables generated above in the
dataset and deal with any intercorrelation effects as the modelling progressed. The
42
Final Report
reason for this was to keep in the dataset a number of variables that reported in the
literature review.
Independent variables taken forward to the Level 1 modelling

5.4.9 A dataset was assembled which stratified the population and casualty data by 7 age
group for females, males and both sexes combined to provide 21 stratifications. The age
groups are:
Preschool children (ages 0 – 4)
Infant/early primary (ages 5 – 7)
Primary (ages 8-11)
Secondary/post primary (ages 12 – 15)
All children (ages 0 – 15)
Young adult (ages 16 – 24)
Adults (ages 25 and over)
5.4.10 Each of the stratifications was linked to a set of other independent variables as set out in
Table 5.2.
Table 5.2: List of independent variables
Variable Variable type

Population (by age group and sex) number
Pedestrian casualty total (by age group and sex) number
Local Government District factor
URDG classification factor
Population density per number
School places (by age groups: 5-7, 8–11 and 12-15) number
MDM Score number
MDM Crime domain score number
MDM Proximity to Services domain score number
Workplace employment number
Employed residents number
Traffic proxy number
Road length (by M, A, B, C and unclassified roads) number
Total road length (all road classes combined) number
Road length density (km of road per hectare) number
Number of road junctions number
Junctions per km (junctions per km of road) number
Junctions per hectare number
Traffic proxy per road km number
Traffic proxy per road junction number
Average annual rainfall (mm) number
Average annual sunshine hours number
Free school meals – primary number
Free school meals – post primary number
Free school meals – primary and post-primary number
Resident population per road junction number
Employed residents per road junction number
Workplace population per road junction number
43
Final Report
5.5 Stage 2: Level 1 Modelling
Model form
5.5.2 If data conform to a Poisson distribution, the key assumption is that mean is equal to the
variance. In the case of the NI data, the distribution of the casualty totals by SOA show
variances in excess of the mean.
5.5.3 The average number of pedestrian casualties for all children (ages 0-15) is 3.6 and the
variance is 15.9. This indicates that the data are overdispersed and that it would be
inappropriate to use a Poisson model to analyse the data. The appropriate model would
be one with a negative binomial error structure.
5.5.4 Negative binomial regression models can be used to model both counts and rates.
Examples of each would be:
The year by year number of collisions at a particular junction
The number of casualties in collisions in a given time period occurring in SOAs
5.5.5 The number of collisions at a junction is purely a count outcome and can be successfully
modelled using standard regression techniques.
5.5.6 But how do the models handle rates? In the current study, the rate is just a count per
head of population. If there is a variation in the population this will affect the number of
casualties observed. Population, here, can be considered as a measure of exposure.
5.5.7 Statistical models can manage exposure variables by allowing users to define such
variables as offset variables. The final model structure adopted for the study was a
negative binomial model with log(population) defined as an offset variable. The R
package provides fully for this model structure.
Contingency table
5.5.8 In this stage a multi-way classification in the form of a contingency table was prepared
where the pedestrian casualty and population data were classified according to the key
characteristics of deprivation index, sex of casualty, age of casualty and URDG
Classification each of which is expressed as a non-numeric factor. The columns of the
contingency table were:
MDM decile (10 classifications/factors)
Sex of casualty (2)
Age of casualty (6)
URDG Classification (2)
The average number of casualties by year of year of age
The average population by year of age
5.5.9 The casualty numbers and population data have been standardised by dividing by the
number of years of age covered by the age group. The means that the modelled
coefficient values for age are relative. If this standardisation had not been undertaken the
coefficient values for age would have biased by the number of years in the age group
particularly in the case of the adult age group.
5.5.10 The table comprised 240 rows (ie 10 x 2 x 6 x 2). Table 5.3 shows the structure or the
data for the first 6 and last 6 rows in the table.
44
Final Report
Table 5.3: Contingency table format
Number by year of age

MDM URDG Age
decile Classification group Sex Casualties Population
M1 R A1 M 0 89
M1 R A1 F 0 88
M1 R A2 M 0 94.67
M1 R A2 F 0 112.67
M1 R A3 M 0.5 102.5
M1 R A3 F 0.25 91.25
.. .. .. .. .. ..
.. .. .. .. .. ..
M10 U A4 M 31.25 1615
M10 U A4 F 25.25 1,545.25
M10 U A5 M 24.67 1.281
M10 U A5 F 19.44 1,347.89
M10 U A6 M 7.29 574.42
M10 U A6 F 5.79 703.84
5.5.11 To demonstrate how the inclusion of variables affects the model and goodness of fit, the
following analysis outlines the effect of adding variables one by one and then looking at
interaction affects.
5.5.12 Tables A 1 – A 6 set out in Appendix 3 show the coefficient estimates for the intercept
value and each factor used in the model. Note that the coefficients for the first level of
each factor are aliased. Each table has six columns:
Column 1 Factor: identifies the factor and factor level (e.g. mdmdec3, indicates
the third decile for MDM decile)
Column 2 Estimate: provides the value for the coefficient relating to the factor
Column 3 Std, Error: is the standard error for the coefficient
Column 4 Z-value: is the estimate divided by the standard error
Column 5 Pr((>|Z|): is the probability of the value of Z occurring by chance - the
lower the number the more significant is the value for the estimate
Column 6 Significance code: is purely a visual presentation of the significance
(the more stars the more significant the estimate value is)
5.5.13 Figures A 1 – A 6 in Appendix 3 show graphically the goodness of fit associated with
these models.
MDM decile as independent variable

5.5.14 Table A 1 shows the output from the regression analysis where MDM decile has been
used as the only independent variable.
5.5.15 It is evident that deprivation is a highly significant factor with increasing deprivation and
higher pedestrian casualty rates.
5.5.16 Figure A 1 shows the relationship in graphical form where it can be seen that although
there is a strong relationship between MDM decile and pedestrian casualty rates, the
overall fit is poor. This would not be a good model to use for predictive purposes.
45
Final Report
MDM decile and age as independent variables

5.5.17 Table A 2 shows the output from the regression analysis where MDM decile and age
have been used as the independent variables.
5.5.18 It is evident that both deprivation and age are highly significant factors with:
increasing deprivation predicting higher casualty rates
casualty rates varying by age group
5.5.19 Figure A 2 shows the relationship in graphical form where it can be seen that the fit is
better than it is for MDM decile, there are a number of outliers particularly where the
observed number of casualties is low.
MDM decile, age and sex as independent variables

5.5.20 Table A 3 shows the output from the regression analysis where MDM decile, age and sex
have been used as the independent variables.
5.5.21 It is evident that the independent variables are highly significant factors with:
males giving rise to higher casualty rates than females
5.5.22 Figure A 3 shows the relationship in graphical form where it can be seen that although
sex is a significant factor, the fit is not observably better than it is for mdm decile and age.
There remain a number of outliers particularly where the observed number of casualties
is low.
MDM decile, age, sex and URDG classification as independent variables

5.5.23 Table A 4 shows the output from the regression analysis where MDM decile, age, sex
and URDG classification have been used as the independent variables.
5.5.24 It is evident that the independent variables are highly significant factors with:
the urban classification giving rise to higher casualty rates than the rural
classification
5.5.25 Figure A 4 shows the relationship in graphical form where it can be seen that the
inclusion of the URDG classification factor improves the fit measurably.

(with MDM decile and URDG classification modelled with interaction)
and URDG classification have been used as the independent variables (with MDM decile
and URDG classification modelled with interaction).
5.5.27 With interaction effects the output it is more difficult to analyse the model output. It is
evident that the independent variables and the interaction between MDM decile and
URDG classification are highly significant factors with:
46
Final Report
classification
the effect of the URDG classification is uneven with rural casualty rates for MDM
deciles 9 and 10 being less than for deciles 2 to 8. The high estimate value for the
MDM decile 10/urban interaction suggests that casualty rates for this category are
significantly higher than for other DMD decile values.
5.5.28 This model needs to be treated with a degree of caution because the populations in rural
areas with MDM deciles 9 and 10 particularly in the lower age groups are low. The
modelled outputs may reflect this with some deceptive or unexpected estimates.
5.5.29 Table 5.4 shows the combined estimates for male child pedestrian casualties (by age
group) in urban areas where it can be seen that the combination of the estimates relating
to the MDM decile*URDG interaction begin to make sense with the combined estimates
generally rising with increasing deprivation.
5.5.30 Figure A 5 shows the relationship in graphical form where it can be seen that the level of
fit is not observably improved by the introduction of the interaction term.
Table 5.4: Combined estimates for child pedestrian casualties (age 0-15)
Age group
MDM decile 0-4 5-7 8 - 11 12 - 15
1 0.40 1.21 1.27 1.40
2 0.45 1.26 1.32 1.45
3 0.68 1.49 1.54 1.68
4 0.66 1.47 1.53 1.66
5 0.49 1.30 1.36 1.49
6 1.00 1.81 1.87 2.00
7 1.01 1.83 1.88 2.02
8 1.27 2.08 2.14 2.27
9 1.68 2.49 2.55 2.68
10 2.33 3.14 3.20 3.33

(with MDM decile and URDG classification and age and sex modelled with
interaction)
and URDG classification have been used as the independent variables (with MDM decile
and URDG classification and age and sex modelled with interaction).
5.5.32 With interaction effects the output it is more difficult to analyse the model output. It is
evident that the independent variables and the interaction between MDM decile and
URDG classification are highly significant factors with:
classification
the effect of the URDG classification is uneven is discussed above
the effect of the age/sex interaction shows (because of the negative and significant
values for the interaction terms ageA2-6sexM) that the male casualty rates for
47
Final Report
ageA1 (ages 0-4) is higher than for the other age groups. The standard errors or
the interaction terms show that there are no significant differences between the
estimates calculated.
5.5.33 Figure A 6 shows the relationship in graphical form where it can be seen that the level of
fit is not observably improved by the introduction of the new interaction term.
5.5.34 This Level 1 model analysis demonstrates that there are clear and highly significant
relationships between casualty rate and deprivation, sex of casualty, age of casualty and
the URDG classification of the collision location. This is a key finding in the context of
this study as it leads on to the more detailed Stage 2 modelling.
5.6 Stage 3: Level 2 modelling
Overview
5.6.2 In the high level (Level 1) model it was established that there are clear and highly
significant relationships between casualty rate and deprivation, sex of casualty, age of
casualty and the URDG classification of the collision location. In this more detailed part
of the analysis datasets for each casualty age group, casualty sex and
urban/classification were created with each set being the subject of a similarly designed
model. In this way it is possible to see how the independent variables representing
environmental and other factors behave in the models across, age, sex and URDG
classification stratifications.
Initial analysis
5.6.3 In Stage 1 it was established that there are clear and highly significant relationships
between casualty rate and deprivation, sex of casualty, age of casualty and the URDG
classification of the collision location. In this more detailed part of the analysis datasets
for each casualty age group, casualty sex and urban/classification were created with
each set being the subject of a similarly designed model. In this way it is possible to see
how the independent variables behave in the models across, age, sex and URDG
classification stratifications.
5.6.4 The variables relating to school places and free school meals were made relevant to the
age group being modelled. When younger children walk to school they will generally be
accompanied by a responsible adult and may also be in the company of younger siblings;
accordingly, these adults and the younger children will be exposed to risk as pedestrians.
Following an initial analysis it was decided to use values for all school places and
proportion of all pupils entitled to free school meals as variables for the pre-school age
group (ages 0-4) and for the young adult and adult age groups. The variables for school
places and free school meals entitlement included in the modelling are set out in Table
5.5.
48
Final Report
Table 5.5: Descriptions of School Places and Free School Meals variables
Age Group School Places Free School Meals

0 - 14 All school places % of all pupils with free school meals
0-4 All school places % of all pupils with free school meals
5-7 School places for % of primary school pupils with free
5 – 7 year olds school meals
8 – 11 School places for % of primary school pupils with free
8 – 11 year olds school meals
12 - 15 School places for % of post primary school pupils with
12 – 15 year olds free school meals
16 – 24 All school places % of all pupils with free school meals
25 and over All school places % of all pupils with free school meals
All ages All school places % of all pupils with free school meals
5.6.5 In the initial analysis, all variables set out in Table 5.2 were tested in the model. Many of
these variables were found either not to be significant in explaining casualty rates or were
found to not as effective in explaining casualty rates as other similar variables. Table 5.6
summarises the outcome of this initial analysis
49
Final Report
Table 5.6: Outcome of initial analysis
Independent Variable Outcome

Variables taken forward
Local Government District A number of districts provided significant coefficients
Population density per hectare Highly correlated with population density per hectare
(by age group and sex) (entire population)
URDG classification Significant variable for some age groups
Population density per hectare Significant variable for some age groups
(entire population)
School places (by age group) Highly significant factor for school age groups
MDM Score Highly significant variable
MDM Crime domain score Highly significant variable
MDM Proximity to Services Highly significant variable
score
Traffic proxy Highly significant variable
Total road length (all road Significant variable for some age groups
classes combined)
Average annual rainfall (mm) Significant variable for some age groups
Average annual sunshine hours Significant variable for some age groups
Free school meals – primary Significant variable for some relevant age groups
Free school meals – post Significant variable for some relevant age groups
primary
Variables not taken forward
Population density per hectare Highly correlated with population density per hectare
(by age group and sex) (entire population)
Workplace employment Not a significant variable
Employed residents Highly correlated with traffic proxy
Road length (by M, A, B, C and Individual road lengths were not significant variables
unclassified roads)
Traffic proxy per road junction A significant variable for a minority of age groups
Road length density (km of road Not a significant variable
per hectare)
Number of road junctions Not a significant variable
Junctions per km (junctions per Not a significant variable
km of road)
Junctions per hectare Not a significant variable
Traffic proxy per road km Not a significant variable
Resident population per road Not a significant variable
junction
Employed residents per road Not a significant variable
junction
Workplace population per road Not a significant variable
junction
5.6.6 The significance of the coefficients are set out in Table A 7 and Table A 8 for urban and
rural areas respectively. Also shown is the sign associated with the coefficient. A plus
sign (“+”) indicates a positive correlation and a minus sing (“-“) indicates an inverse
correlation for the variable with casualty rates. In a positive correlation, as one variable
gets larger the other gets larger. In an inverse correlation, as one gets larger, the other
gets smaller.
50
Final Report
5.6.7 It should be noted that because of the smaller datasets for rural casualties a number of
the models did properly converge and the software indicated a warning message and
these are identified in the right hand column of Table A 8.
5.6.8 Looking at the urban models it is observed that:

the MDM score is a consistent and highly significant positively correlated variable
across all age groups and sexes – the more deprived the SOA, the greater the
casualty rate
the MDM Crime domain score is a consistent and highly significant positively
correlated variable across all age groups and sexes – the more crime related to the
SOA the greater the casualty rate
the MDM Proximity to Services domain score is a consistent and highly significant
inversely correlated variable across all age groups and sexes – the less deprived
the SOA is in terms of proximity to services related to the SOA the greater the
casualty rate.
population density is found to be a significant positively correlated variable for half
the models examined – the greater the density, the greater the casualty rate
road length is a significant positively correlated variable for most models – the
longer the road length, the greater the casualty rate
the traffic proxy variable which provides a relative measure of travel activity in the
SOA is a consistent and highly significant positively correlated variable across all
age groups and sexes – the more activity there is in the SOA, the greater the
casualty rate
the number of school places is a consistent and highly significant positively
correlated variable across all models – the more school places, the greater the
casualty rate
the proportion of free school meals is not seen to be a significant factor for most of
the models. Significant relationships are observed to be positive and inverse.
sunshine hours is seen to be a highly significant generally inversely correlated
variable in the models for most child age groups. It is, however, seen to be a
positively correlated variable for the young adult (16-24 age group) – the greater
the number of sunshine hours, the lower the casualty rate (for children) and the
higher the casualty rate (for young adults)
like sunshine hours, annual rainfall is seen to be a highly significant generally
inversely correlated variable in the models for most child age groups. It is,
however, seen to be a positively correlated variable for the young adult (16-24 age
group) – the greater the rainfall, the lower the casualty rate (for children) and the
5.6.9 Looking at the rural models it is observed that:
the MDM score is a consistent and highly significant positively correlated variable
casualty rate
the MDM Crime domain score is a consistent and highly significant positively
the MDM Proximity to Services domain score is a consistent and highly significant
inversely correlated variable across all age groups and sexes – the more deprived
casualty rate. This inverse correlation suggests the closer the population is to
services the more likely there will pedestrian trips which will increase exposure.
Where service are located further away fewer walking trips may be made.
population density is found to be a significant correlated variable for half the
models examined – for most of these the correlation is inverse- the greater the
density, the lower the casualty rate (in most instances)
51
Final Report
road length is found to be a significant correlated variable for half the models
examined – for most of these the correlation is inverse- the greater the road length,
the lower the casualty rate (in most instances)
the traffic proxy variable which provides a relative measure of travel activity in the
SOA is generally a highly significant positively correlated variable across all age
groups and sexes – the more activity there is in the SOA, the greater the casualty
rate
the number of school places is a highly significant positively correlated variable for
about half of the models – the more school places, the greater the casualty rate
the proportion of free school meals is seen to be a significant factor for about half
the models. Where is it significant the relationship is positive - the higher the
proportion of pupils with free school meals, the greater the casualty rate
sunshine hours is seen to be a highly significant generally positively correlated
variable in the models for most child age groups – the greater the number of
sunshine hours, the higher the casualty rate
like sunshine hours, annual rainfall is seen to be a highly significant generally
positively correlated variable in the models for most child age groups – the greater
rainfall, the higher the casualty rate
5.6.10 There are some interesting differences between the rural and urban models.
the population density is generally positively correlated with casualty rates for
urban areas and inversely for rural areas.
the length of road is generally positively correlated with casualty rates for urban
areas and inversely for rural areas.
sunshine hours is generally inversely correlated with casualty rates for urban areas
and positively for rural areas.
annual rainfall is generally inversely correlated with casualty rates for urban areas
Discussion
5.6.11 For the MDM score and the MDM Crime domain the relationships show positive
correlations and this is as intuition would suggest that the more deprived an area is the
higher the casualty rate.
5.6.12 The MDM Proximity to Services domain shows inverse correlations which may seem
counter-intuitive at first. The domain, however, is scored to give higher values to areas
which are further away from services and lower values to areas closer to services.
Higher values are, therefore, more deprived for this domain. This may suggest that the
closer the population is to services the more likely there will be pedestrian trips and
consequent increased exposure. Where services are located further away, fewer walking
trips may be made and there is less exposure with a consequent lower casualty rate.
5.6.13 Population density is generally positively correlated with casualty rates in urban areas
and mostly inversely correlated in rural areas. The reason for this may rely on the strong
relationship between population density and road length which is set out clearly in Figure
5.3 where it can be seen that higher population densities are associated with lower road
length and vice versa. This strong intercorrelation indicates why there may be some
confusing model outputs regarding these two variables and suggests that only one of
them need be in the model. Given that population density includes population as its
numerator and that road length is a wholly independent variable, it is road length that
should survive with population density being eliminated.
52
Final Report
Figure 5.3: Plot of population density with road length
Plot of population density v road length

150
Road length (km)
100
50
0
0 50 100 150
Population density (people/ha)
5.6.14 Total road length is generally positively correlated with casualty rates in urban areas and
generally inversely correlated in rural areas. There may be a parallel here with the MDM
Proximity to Services domain where in rural areas road length and distance to services
may be correlated. Indeed, this is borne out by a plot of road length against MDM
Proximity to Services which is shown in Figure 5.4. Road length itself not a measure of
distance required to be travelled whereas the MDM Proximity to Services domain is
certainly a social measure relating to distances that may need to be travelled as part of
day-to-day living. Accordingly road length was eliminated from the model in preference to
MDM Proximity to Services.
53
Final Report
Figure 5.4: Plot of MDM Proximity to Services with road length
Plot of MDM Proximity to Services v road length

2
MDM Proximity to Services
1
0
-1
-2
0 50 100 150
Road length (km)
5.6.15 Traffic proxy is highly significant and positively correlated with casualty rates. This result
follows intuition. Traffic proxy is a relative measure of the amount of activity in an area
both in terms of vehicular and pedestrian movements. High values for traffic proxy
suggest greater exposure to risk for pedestrians and consequent higher casualty rates:
this is borne out by the model outputs.
5.6.16 The number of school places in an area is highly significant and positively correlated with
casualty rates in urban areas. The models for rural areas show less conclusive outputs
but they do show some highly significant and positively correlated coefficients. This
result is in keeping with intuition. For the avoidance of misinterpretation of this result, it is
to be noted that the values for the variables used to measure school places relate to the
number of school places at schools in an SOA (not to the number of school children
resident in an SOA. Therefore, the relationship is most likely to be the result of collisions
around schools occurring on journeys to/from school. The relationships hold good for
models including children and adults. The fact that the relationships are much stronger
and more consistent in urban areas may be explained by the higher proportion of children
in rural areas who will be entitled to (and will use) free school transport. This would
reduce exposure as pedestrians on journeys to/from school for the pupils, parents and
younger siblings alike.
5.6.17 The proportion of children with free school meals does not demonstrate a consistent
pattern and it would be unwise to place emphasis on the model outputs for this variable.
Some coefficients are significant but show no definable pattern. Some of the correlations
are positive and others inverse. The reason for this may lie in the fact that there is
correlation between the proportion of free school meals and the MDM score and this is
likely to give rise to unreliable coefficient estimates. Figure 5.5 shows a plot of the
proportion of all pupils entitled to free school meals against MDM Score which shows a
clear correlation between the two variables. Note that most SOAs do not have schools
54
Final Report
within their boundaries and that it these SOAs which form the vertical column of data at
point “0” on the x-axis of the chart.
Figure 5.5: Plot of proportion of all pupils entitled to free school meals against
MDM Score
Proportion of all pupils entitled to free school meals v MDM Score

80
60
MDM Score
40
20
0
0 20 40 60
Proportion of all children entitled to free school meals(%)
5.6.18 Both sunshine hours and annual rainfall are generally positively correlated with casualty
rates in rural areas and inversely correlated in urban areas. The exception to this relates
to the young adult age group (16-24) where in urban areas both sunshine hours and
annual rainfall are positively correlated with casualty rates. A possible reason for these
differences is discussed below.
Regional differences in the Level 2 model

5.6.19 The literature review highlighted two pieces of research which identified regional
differences in casualty rates [Grayling et al (2002) and Graham and Stephens (2008)].
For Northern Ireland, the local government district has been used to stratify the data in
this study providing 26 groupings for the SOAs. Models combining urban and rural areas
for children (0-15) young adults (16-24) and adults (25 and over) have tested the impact
of adding local government district as a factor in the models. Table 5.7 shows the
districts which were found to be significant factors relating to pedestrian casualty rates
together with the nature of the relationship and the level of significance.
55
Final Report
Table 5.7: Local Government Districts which impact on casualty rates (Level 2
model)
Local Government Impact on Level of

Age group District casualty rates significance
Children Ards Inverse 0.05
(0 – 15) Carrickfergus Positive 0.01
Larne Positive 0.001
Lisburn Positive 0.05
Magherafelt Inverse 0.05
North Down Inverse 0.001
Newry and Mourne Inverse 0.05
Young Ards Inverse 0.05
Adults Craigavon Positive 0.001
(16 – 24) Dungannon Positive 0.001
Magherafelt Positive 0.01
Adults Coleraine Inverse 0.01
(25 and Derry Inverse 0.001
over) Moyle Positive 0.001
Omagh Inverse 0.05
Strabane Inverse 0.01
5.6.20 It is interesting to note that sunshine hours and rainfall are both correlated with Local
Government District. This being the case makes it difficult to judge which variable is
dominant. If Local Government District is the dominant factor it may explain why the
model outputs relating to sunshine hours and annual rainfall show differences between
rural and urban areas. It is difficult within the scope of this study to measure objectively
the differences between Local Government Districts which may impact on pedestrian
casualty rates; examples of such differences may relate to:
the level of pedestrianisation in the district
the proportion of children in the district entitled to and using free school transport
local policies and design standards relating to facilities for pedestrians
implementation of urban design standards at a district level which minimise
pedestrian/vehicle conflict
5.6.21 In the absence of a clear reason to believe that Local Government District is a dominant
factor it was decided to retain sunshine hours and annual rainfall in the model.
5.7 Stage 4: Level 3 modelling
Urban and rural models

5.7.2 Following on from the Level 2 modelling, the final urban and rural models include seven
variables, viz:
MDM Score
MDM Crime domain
MDM Proximity to Services domain
Traffic proxy
School places
Sunshine hours
Annual rainfall
56
Final Report
5.7.3 Table A 9 and Table A 10 set out the significance and signs for the coefficients for urban
and rural areas respectively.
5.7.4 Looking at the final urban models it is observed that:

The MDM score is a consistent and highly significant positively correlated variable
casualty rate
The MDM Crime domain score is a consistent and highly significant positively
The MDM Proximity to Services domain score is a consistent and highly significant
inversely correlated variable across all age groups and sexes – the less deprived
casualty rate.
The traffic proxy variable which provides a relative measure of travel activity in the
casualty rate
The number of school places is a consistent and highly significant positively
correlated variable across all models – the more school places, the greater the
casualty rate
Sunshine hours is seen to be a highly significant generally inversely correlated
variable in the models for most child age groups. It is, however, seen to be a
positively correlated variable for the young adult (16-24 age group) – the greater
the number of sunshine hours, the lower the casualty rate (for children) and the
Like sunshine hours, annual rainfall is seen to be a highly significant generally
inversely correlated variable in the models for most child age groups. It is,
however, seen to be a positively correlated variable for the young adult (16-24 age
group) – the greater the rainfall, the lower the casualty rate (for children) and the
5.7.5 Looking at the rural models it is observed that:

casualty rate
casualty rate. This inverse correlation suggests the closer the population is to
services the more likely there will pedestrian trips which will increase exposure.
Where service are located further away fewer walking trips may be made.
SOA is generally a highly significant positively correlated variable across all age
groups and sexes – the more activity there is in the SOA, the greater the casualty
rate
The number of school places is a highly significant positively correlated variable for
about half of the models – the more school places, the greater the casualty rate
Sunshine hours is seen to be a highly significant generally positively correlated
57
Final Report

5.7.6 There remain differences between the rural and urban models.
Sunshine hours is generally inversely correlated with casualty rates for urban areas
Annual rainfall is generally inversely correlated with casualty rates for urban areas
For young adults in urban areas are an exception where sunshine hours and
annual rainfall are positively correlated with casualty rates
5.7.7 The reasons for these differences relating to sunshine hours and annual rainfall are not
clear and were discussed earlier in paragraph 5.6.20.
Combined urban and rural model

5.7.8 Table A 11 shows the significance and signs for models where urban and rural areas are
combined. The variable for URDG classification has been added to the list of independent
variables for this set of models, viz:
MDM Score
MDM Crime domain
MDM Proximity to Services
Traffic proxy
School places
Sunshine hours
Annual Rainfall
URDG classification
5.7.9 For the combined urban and rural models it is observed that:
casualty rate
casualty rate.
casualty rate
The number of school places is a consistent and highly significant positively
correlated variable for about half of the models – the more school places, the
greater the casualty rate
Sunshine hours is seen to be a highly significant generally positively correlated
URDG classification is generally positively correlated with child casualty rates and
generally inversely correlated with adult casualty rates
58
Final Report
Regional differences in the Level 3 combined model

5.7.10 Regional differences for the Level 2 model are summarised in Table 5.7. For the Level 3
model the regional differences are similar but with a number of additional local
government districts highlighted ( refers).
Table 5.8: Local Government Districts which impact on casualty rates (Level 3
model)
Impact on Level of
Age group Local Government District casualty rates significance
Children Ballymena Positive 0.05
(0 – 15) Belfast Positive 0.05
Carrickfergus Positive 0.001
Castlereagh Positive 0.05
Derry Positive 0.05
Larne Positive 0.001
Limavady Positive 0.01
Lisburn Positive 0.01
Magherafelt Inverse 0.05
Newry and Mourne Inverse 0.05
Newtownabbey Positive 0.05
Young Adults Antrim Inverse 0.001
(16 – 24) Ards Inverse 0.01
Craigavon Positive 0.001
Dungannon Positive 0.001
Magherafelt Positive 0.05
Adults Derry Inverse 0.001
(25 and over) Dungannon Positive 0.01
Moyle Positive 0.001
Strabane Inverse 0.05
5.7.11 Figure 5.6 shows a Level 3 model plot for all child pedestrian casualties (age 0 -15).
Whilst there is a clear correlation between the modelled and observed values there is a
quite a lot of statistical “white noise”. The “white noise” was not so marked with the Level
1 model which is an entirely expected outcome. The Level 1 model is an aggregation of
data and the individual cell values in the contingency table represent average values
which will to a large extent have removed a significant proportion of variation which will
give rise to the “white noise”.
59
Final Report
Figure 5.6: Observed and modelled casualty totals for all child pedestrian
casualties (ages 0 – 15) (Level 3 model)
Observed and modelled casualty totals

35
30
Observed casualty totals
25
20
15
10
5
0
0 5 10 15 20 25 30 35
Modelled value
(Level 3 model)
Poorest and best performing SOAs

5.7.12 For each SOA the models provide a predicted value for the number casualties in the area
– this is called the fitted value. The fitted values can be subtracted from the observed
number of casualties to provide for each SOA a residual value which can be thought of as
a measure of road safety performance for the area. Where the residual is a positive
value it implies that the area has more casualties than might be expected and the higher
the value the poorer the performance. Conversely, where the residual is a negative value
it implies that the area has fewer casualties than might be expected and the higher the
value the better the performance.
5.7.13 Table A 12 in Appendix 3 tabulates the 50 poorest performing SOAs and Table A 13
tabulates the 50 best performing SOAs. A review of the poorest performing SOAs may
shed some light on the reasons why particular they have been listed and may assist in
developing remedial action. On the other hand, a review of the best performing SOAs
may highlight some examples of good practice which could be transferred to other areas
of Northern Ireland. Care, however, should be exercised in undertaking this work as the
data relate to casualty records over a ten year period and there may have been
significant demographic and infrastructure changes over this period.
5.7.14 Figure 4.1 in Appendix 4 maps the distribution of performance (as deciles) across
Northern Ireland. This map can be compared with the equivalent map for MDM decile.
The comparison shows that whilst deprivation tends to be clustered, the performance
shows less clustering. This could be investigated further using the processes outlined in
the previous two paragraphs.
60
Final Report
Increasing and decreasing trends at the SOA level

5.7.15 The ten year data were split to provide five year child pedestrian casualty totals for the
periods 1999 – 2003 and 2004 – 2008. The overall number of casualties in the first five
year period was 2,034 and the equivalent figure for the second period was 1,201.
5.7.16 The ratio of the two values (second five year period : first five year period) is 0.590.
Therefore, an estimate of the number of casualties expected in an SOA for the second
five year period (2004 - 2008) would be the number of casualties for the preceding five
years (1999 – 2003) times 0.590.
5.7.17 These estimated casualty totals can be compared with the observed casualty totals to
determine the change in the number of casualties. Here the value is:
observed casualty total – estimated casualty total
5.7.18 Where the number is positive this represents an increase in casualty totals. Conversely,
where the number is negative this represents a decrease in casualty totals.
5.7.19 Table A 14 in Appendix 3 tabulates the 50 SOAs with the greatest increases in casualty
totals and Table A 14 tabulates the 50 SOAs with the largest decreases. A review of
these tabulations may be helpful in identifying trends over the past ten years which may
have been occasioned by changes in demography, traffic patterns, school locations or
specific road safety interventions.
61
Final Report
6 Conclusions
6.1 Link between deprivation and child pedestrian casualties

6.1.1 The modelling process has established a clear link between deprivation and child
pedestrian road casualty rates. The effect is found to be relative to the degree of
deprivation: child pedestrian casualty rates rise as deprivation rises.
6.1.2 Overall children resident in SOAs comprising the most deprived 10% of SOAs are 4.78
times more likely to be injured as a pedestrian in a road collision as children resident in
the least deprived 10% of SOAs. This value is defined as the deprivation disparity ratio
(DDR). This overall figure of 4.78 masks different risks for children of different ages sexes
which are shown in Table 9 and reproduced below in Table 6.1
Table 6.1: Deprivation disparity ratio of child pedestrian casualty rates1 by sex,
Multiple Deprivation Measure (MDM) decile and age group (1999 –
13
2008)
Age Group Male Female Both

0-4 9.63 10.10 13.49
5-7 11.03 14.73 12.25
8-11 5.08 3.57 4.35
12-15 1.93 3.05 2.30
All ages 4.54 5.29 4.78
6.1.3 This finding fulfils one of the key objectives of the study. It is clear from this table that
there a clear differences in the DDR values by age and sex. For example:
Males in the 5-7 age group are 5.7 more likely than males in the 12-15 age group
to be injured as a pedestrian
Males in the 8-11 age group are 1.4 times more likely than females in the same
group to be injured as a pedestrian
Females in the 12-15 age group are 1.6 times more likely than males in the same
age group to be injured as a pedestrian
6.2 Statistical modelling

6.2.1 In the event of evidence supporting a link between deprivation and child pedestrian
casualty rates, the remit of the study included a requirement to develop a statistical model
to measure the effect of deprivation, taking account of environmental factors. The
reliability of this model should be assessed.
6.2.2 The statistical modelling carried out has shown that key variables relating to how busy an
area is, proximity to services, crime and number of school places are linked to the child
pedestrian casualty rates. Other variables - weather conditions, URDG classification and
local government district are also linked but less robustly to casualty rates.
6.2.3 The reliability of the models appears to be good with very high levels of consistency
across the age groups and sexes for the key variables with moderate consistency for the
other variables.
13
For notes on the data see Table 9.
62
Final Report
6.3 Implications for road safety planning

6.3.1 The modelling process has highlighted a number of key implications relating to road
safety planning.
6.3.2 The key finding is that there is a clear link between deprivation and child pedestrian
casualty rates in Northern Ireland. From a road safety planning point of view this would
strongly suggest that road safety initiatives should target or at least be prioritised in part
on areas of deprivation
6.3.3 The number of school places in an SOA (as a proxy for the number of children travelling
to and from a school in that area as opposed to the number of children living in the area)
is a consistent and highly significant variable. This suggests that the presence of schools
adds significant additional risk for children. This perhaps not surprising as most children
will attend school just under 200 days a year and many will travel as pedestrians either
alone or accompanied by an adult. The model, however, shows quantitatively that there
is a highly significant additional risk. This finding would support interventions which make
travel to school safer including through improved infrastructure and possible pedestrian
skills training. This effect is additional to impact deprivation has on casualty rates and
suggests that priorities for interventions of this type be given to deprived areas. Part time
20 mph zones may also be an effective counter measure.
6.3.4 The traffic proxy variable has also been found to be a consistent and highly significant
variable. This, too, is perhaps not a surprising conclusion but the modelling process has
quantitatively shown a strong and robust link between traffic and child pedestrian casualty
rates. This suggests that busier areas should be a focus for an intervention strategy
perhaps involving speed reduction strategies.
6.3.5 The model can be used to identify areas of concern (child pedestrian casualty blackspots)
which might respond to site specific intervention. The production of a list of areas of
concern for investigation could prove helpful in implementing low cost remedial
measures. Table A 12 in Appendix 3 sets out an indicative list based on areas with more
casualties than expected based on the model outputs. This list is indicative and care
needs to exercised to avoid an “over-interpretation” of the data (see paragraph 5.7.13).
6.3.6 The models could also be used to identify good practice by preparing a list of areas
where there are fewer casualties than predicted by the model outputs. These areas
could be the subject of a review to determine if there are any underlying reasons why
casualty rates are lower than expected. Table A 13 in Appendix 3 sets out an indicative
list of such areas.
6.3.7 Child pedestrian casualty rates are modelled to be higher in urban areas than in rural
areas. Whilst the traffic proxy variable is a measure of traffic in an area it is not a
measure of the traffic density on roads which will be higher in urban than in rural areas.
Also in urban areas it is likely that individual walking trips will involve more street
crossings than in rural areas. Consequently, some priority should be given to
interventions relating to urban areas.
6.3.8 Finally, proximity to services is seen to be linked to casualty rates where the closer
children live to services the higher the casualty rates. This is perhaps more widely
related to planning issues and suggests that road safety ought to be a key consideration
in the design of regeneration and major development schemes.
63
Final Report
6.4 Specific interventions

6.4.1 The literature review identified a number of potential specific interventions which would
be appropriate in combating high child pedestrian casualty rates (paragraph 2.4.61
refers). These are:
a national “traffic club” scheme
road crossing training for children (Kerbcraft)
road safety inputs to the master planning of regeneration projects
a distribution of road safety measures which targets socio-economic environments
6.4.2 It is stressed from the findings of the literature review and from the author’s experience
that interventions should not be ad hoc activities, rather they form part of a
complementary and integrated delivery strategy
64
Final Report
Appendix 1
Literature Review References and Bibliography
1
Final Report
References
Abdalla IM, Raeside R, Barker D and McGuigan DRD. An investigation into the relationships between
area social characteristics and road accident casualties. Accident Analysis and Prevention Vol 29,
pp583-593, 1997.
http://www.sciencedirect.com/science?_ob=ArticleURL&_udi=B6V5S-3SX1G85-
4&_user=10&_rdoc=1&_fmt=&_orig=search&_sort=d&_docanchor=&view=c&_acct=C000050221&_ve
rsion=1&_urlVersion=0&_userid=10&md5=15f57e30af5f3942421da81c871a28d2
Chichester BM, Gregan JA, Anderson DP and Kerr JM. Associations between road traffic accidents
and socio-economic deprivation in Scotland’s west coast. Scottish Medical Journal Vol 43 pp135-
1398, 1998.
Not available on the web.
Childhood accidents update: Promoting young people’s health using the Health Behaviour in School-
aged Children (HBSC) study and the Health Development Agency’s Evidence Base. NHS Health
Development Agency, undated.
http://www.wiredforhealth.gov.uk/Word/HBSCBriefingAccidentalInjury.doc.
Clarke DD, Ward P, Truman W and Bartle C (School of Psychology University of Nottingham). A poor
way to die: social deprivation and road traffic fatalities. Department for Transport 2008.
http://www.dft.gov.uk/pgr/roadsafety/research/rsrr/theme5/poorwaytodie.pdf.
Cran GW. Social deprivation and childhood injuries in North and West Belfast. The Ulster Medical
Journal, Volume 74 No. 1, pp. 22-28, May 2005.
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2475484/pdf/ulstermedj00001-0025.pdf
Department for Transport (DfT). Childhood accidents and their relationship with problem behaviour
(No.07). DfT, July 1999.
http://www.dft.gov.uk/pgr/roadsafety/research/rsrr/theme1/childhoodaccidentsandtheirre4729?page=5
Edwards P, Green J, Lachowycz K, Grundy C and Roberts I. Serious Injuries in children: variation by
area deprivation and settlement type. Archive of Disease in Childhood Vol 93 pp485-489 2008.
http://www.ncbi.nlm.nih.gov/pubmed/18381345?dopt=AbstractPlus
Edwards P, Green J, Roberts I, Grundy C and Lachowycz K. Deprivation and Road Safety in London:
A report to the London Road Safety Unit. London: London School of Hygiene and Tropical Medicine,
2006.
http://www.tfl.gov.uk/assets/downloads/deprivation-and-road-safety.pdf
Graham D, Glaister S and Anderson R. The effects of deprivation on the incidence of child and adult
pedestrian casualties in England. Accident Analysis and Prevention Vol 37, pp125-135, 2005.
http://www.sciencedirect.com/science?_ob=ArticleURL&_udi=B6V5S-4DKKDWP-
1&_user=10&_coverDate=01%2F31%2F2005&_alid=1018631002&_rdoc=1&_fmt=high&_orig=search
&_cdi=5794&_sort=r&_docanchor=&view=c&_ct=1&_acct=C000050221&_version=1&_urlVersion=0&
_userid=10&md5=e5afa54340109afb8f405df78f60134f
2
Final Report
Graham DJ and Stephens DA. Decomposing the impact of deprivation on child pedestrian casualties
in England. Accident Analysis & Prevention Vol 40, Issue 4, pp1351-1364, 2008.
http://www.sciencedirect.com/science?_ob=ArticleURL&_udi=B6V5S-4S26HCC-
&_cdi=5794&_sort=r&_docanchor=&view=c&_ct=59&_acct=C000050221&_version=1&_urlVersion=0
&_userid=10&md5=3ff2102fff5d9c73a950258f644af710
Grayling T, Hallam K, Graham D, Anderson R and Glaister S. Streets ahead safe and liveable streets
for children. Institute for Public Policy Research, 2002.
http://www.ippr.org.uk/members/download.asp?f=%2Fecomm%2Ffiles%2Fstreets%5Fahead%2Epdf
Hewson P. Deprived children or deprived neighbourhoods? A public health approach to the

investigation of links between deprivation and injury risk with specific reference to child road safety in
Devon County, UK. BMC Public Health 2004, 4:15doi:10.1186/1471-2458-4-15.
http://www.biomedcentral.com/1471-2458/4/15
Kendrick D. Prevention of pedestrian accidents. Archives of Disease in Childhood 1993; 68: 669-672
http://adc.bmj.com/cgi/reprint/68/5/669
McGuigan D, Bull M and Gorman D. Connecting social deprivation and involvement in road accidents:
results and implications. 24th PTRC European Transport Forum. Brunel University, England.
Proceedings of Seminar H. Traffic management and road safety, 2-6 Sept 1996. PTRC Education and
Research Services Ltd.
http://www.etcproceedings.org/paper/connecting-social-deprivation-and-involvement-in-road-
accidents-results-and-im
McGuigan D, Fraser D and McBride A. Road accidents and social deprivation. Scottish Transport
Applications and Research (STAR) Conference, Glasgow, 2006.
Not available on the web
National Public Health Service for Wales (NPHS). Deprivation and Health - A report by the National
Public Health Service for Wales. NPHS, 2004.
http://www2.nphs.wales.nhs.uk:8080/hiatdocs.nsf/1f8687d8da97650980256fa30051b0be/83086791e1
fb86538025726f0043322c/$FILE/Deprivationreport10Dec04.pdf
Noland RB and Quddus MA. A spatially disaggregate analysis of road casualties in England. Accident
Analysis and Prevention Vol 36 pp973-984, 2004.
http://www.sciencedirect.com/science?_ob=ArticleURL&_udi=B6V5S-4BM6B16-
_userid=10&md5=f342bc500a40767afb4f8ff554f9224e
Northern Ireland Statistics & Research Agency (NISRA). Northern Ireland Multiple Deprivation
Measure 2005. NISRA, May 2005.
http://www.nisra.gov.uk/deprivation/archive/NIMDM2005FullReport.pdf
3
Final Report
Petch RO and Henson RR. Child road safety in the urban environment. Journal of Transport
Geography Vol 8 pp197-211, 2000.
http://www.sciencedirect.com/science?_ob=ArticleURL&_udi=B6VG8-40JFYM1-
_userid=10&md5=37240e1564772964d94ff6a8a574a96b
Road Safety Scotland (RSS). Road Safety Early Years Resource Consultation: Analysis of
Responses. RSS, April 2009.
http://www.road-safety.org.uk/Images/EarlyYearsAnalysisFinalReport,ODS,April2009_tcm4-
549089.pdf
SGLGRAS. Paper LGRAS(00)6: Driver and casualty postcodes, and their use in a research project on
tourist road accidents in rural Scotland. Scottish Government Liaison Group on Road Accident
Statistics (SGLGRAS) 27th June 2000.
http://cci.scot.nhs.uk/Topics/Statistics/scotstat/transport/roadaccidentstats/paper-lgras-00-6
Silversides JA, Gibson A, Glasgow JFT, Mercer R and Cran GW. Social deprivation and childhood
injuries in North and West Belfast. The Ulster Medical Journal, Volume 74 No. 1, pp. 22-28, May 2005.
http://www.pubmedcentral.nih.gov/picrender.fcgi?artid=2475484&blobtype=pdf
Whelan K, Towner E, Errington G and Powell J (Centre for Child and Adolescent Health, University of
the West of England, Bristol). Evaluation of the National Child Pedestrian Training Pilot Projects. Road
Safety Research Report No. 82. Department for Transport., 2008.
http://www.dft.gov.uk/pgr/roadsafety/research/rsrr/theme1/childpedestrianprojects/kerbcraftevaluation
main.pdf
4
Final Report
Bibliography
Abdalla IM, Barker D and Raeside R. Road accident characteristics and socio-economic deprivation.
Traffic Engineering and Control Vol 38 pp672-676, December 1997.
http://www.tecmagazine.com/index.cfm?fuseaction=magazine.Article&ArticleID=327
Cattan M, Green J, Newell C, Ayrton R and Walker J. Child–Parent Interaction in Relation to Road
Safety Education: Part 1 – A Critical Literature Review. Road Safety Research Report 101,
Department for Transport, December 2008.
http://www.dft.gov.uk/pgr/roadsafety/research/rsrr/theme1/no101roadsafetyreport.pdf
Dowswell T and Towner E. Social deprivation and the prevention of unintentional injury in childhood: a
systematic review. Health Education Research, Vol. 17, No. 2, 221-237, April 2002.
http://her.oxfordjournals.org/cgi/content/full/17/2/221?maxtoshow=&HITS=10&hits=10&RESULTFOR
MAT=1&author1=dowswell&author2=towner&andorexacttitle=and&andorexacttitleabs=and&andorexa
ctfulltext=and&searchid=1&FIRSTINDEX=0&sortspec=relevance&resourcetype=HWCIT
Haynes R, Jones A, Harvey I, Tony Jewell T and Lea D. Geographical distribution of road traffic
deaths in England and Wales: place of accident compared with place of residence. Journal of Public
Health VoI. 27, No. 1, pp. 107–111, February 2005.
http://jpubhealth.oxfordjournals.org/cgi/reprint/27/1/107.pdf
Health Development Agency. Childhood accidents update. NHS (2004).

http://www.wiredforhealth.gov.uk/Word/HBSCBriefingAccidentalInjury.doc
Hewson PJ. Epidemiology of child pedestrian casualty rates: Can we assume spatial independence?
Accident Analysis & Prevention, Vol 37, Issue 4, Pages 651-659, July 2005.
http://www.sciencedirect.com/science?_ob=ArticleURL&_udi=B6V5S-4FWKMDM-
_userid=10&md5=a8f759fe2931f2dc77b7a7a856497973
Sanjay Rana. Investigation of geographical dependence of road traffic casualties in highly deprived
areas of Greater Manchester. Neighbourhood Road Safety Initiative (NRSI) 2006.
http://eprints.ucl.ac.uk/1731/1/CTS_Talk_Presentation.pdf
ODPM. Making the connections: final report on transport and social exclusion. ODPM Social
Exclusion Unit, 2003.
http://www.cabinetoffice.gov.uk/media/cabinetoffice/social_exclusion_task_force/assets/publications_1
997_to_2006/making_transport_2003.pdf
White D and Raeside R. The socio-economic influences on the risk of child involvement in road traffic
accidents. The Royal Society for the Prevention of Accidents Road Safety Congress 12, 14 March
2001.
http://www.rospa.org/roadsafety/conferences/congress2001/proceedings/white.pdf
White D, Raeside R and Barker D (Napier University). Road accidents and children living in
disadvantaged areas: a literature review. The Scottish Executive Central Research Unit, 2000.
http://www.scotland.gov.uk/Resource/Doc/156570/0042052.pdf
5
Final Report
Appendix 2
Preliminary Analysis Tables and Figures
1
Final Report
Table 1: Casualties by casualty class and age group (January 1999 to December 2008)
Age group
14
Casualty Class 0 0-4 5-7 8-11 12-15 16 and over All ages
1. Driver 20 10 60 67 145 55,652 55,954
2. Pillion passenger 1 10 38 210 259
3. Vehicle passenger - Front 25 263 257 625 1,128 19,449 21,,747
4. Vehicle passenger - Rear 35 1,506 1,198 1641 1,894 10,699 16973
5. Pedestrian 11 499 672 960 1,104 5,583 8,829
6. Motorcyclist 1 9 44 3,871 3,925
7. Pedal cyclist 25 150 264 262 997 1,698
All casualty classes 92 2,304 2,337 3,576 4,615 96,461 109,385
Table 2: Child pedestrian casualties by age group (1999-2008)
Casualties
Age Group Male Female Both sexes
0-4 Pre-school 344 155 499
5-7 Infant (lower primary) 407 265 672
8 - 11 Upper primary 590 370 960
12 - 15 Secondary 638 466 1,104
All age groups 1979 1256 3,235
Table 3: Casualties by casualty class and age group (April 2007 to March 2009)
Age group
Casualty Class 0 0-4 5-7 8-11 12-15 16 and over All ages
1. Driver 15 1 16 9,920 9,952
2. Pillion passenger 7 38 45
3. Vehicle passenger - Front 20 24 36 98 175 3,155 3,508
4. Vehicle passenger - Rear 29 266 170 273 299 1,641 2,678
5. Pedestrian 11 87 71 138 180 1,187 1674
6. Motorcyclist 1 7 905 913
7. Pedal cyclist 1 5 20 57 58 308 449
All casualty classes 77 382 297 567 742 1,7154 19,219
Table 4: Child pedestrian casualties by age group (April 2007 to March 2009)
Casualties
Age Group Male Female Both sexes
0-4 Pre-school 54 33 87
5-7 Infant (lower primary 44 27 71
8 - 11 Upper primary 78 60 138
12 - 15 Secondary 98 82 180
All age groups 274 202 476
14
Age “0” is reserved by the PSNI for casualties of unknown age. Casualties of 0 (ie aged under 1 year) are
coded as “1”.
2
Final Report
Table 5: Child pedestrian casualty numbers by sex, Multiple Deprivation Measure (MDM)
decile and age group (1999 - 2008)
MDM Age group All age

Sex decile 0-4 5-7 8-11 12-15 groups
1 9 10 22 46 87
2 16 14 24 48 102
3 8 21 51 44 124
4 23 29 50 36 138
5 23 24 45 54 146
Male
6 23 24 54 53 154
7 24 45 61 71 201
8 44 44 61 85 234
9 63 59 81 75 278
10 111 137 141 126 515
All male casualties 344 407 590 638 1,979
1 5 18 23 46
2 8 10 20 36 74
3 3 13 26 40 82
4 9 10 30 19 68
5 7 6 16 36 65
Female
6 20 24 36 64 144
7 16 20 28 46 110
8 26 41 41 41 149
9 22 44 66 60 192
10 44 92 89 101 326
All female casualties 155 265 370 466 1,256
1 9 15 40 69 133
2 24 24 44 84 176
3 11 34 77 84 206
4 32 39 80 55 206
5 30 30 61 90 211
Both
6 43 48 90 117 298
7 40 65 89 117 311
8 70 85 102 126 383
9 85 103 147 135 470
10 155 229 230 227 841
All casualties 499 672 960 1,104 3,235
3
Final Report
1
Table 6: Child population by sex, Multiple Deprivation Measure (MDM) decile and age
group (1999 - 2008)

Male 1 5,058 3,304 4,773 4,701 17,836
2 5,640 3,592 4,820 4,740 18,792
3 5,675 3,570 4,960 4,980 19,185
4 6,019 3,717 5,236 5,386 20,358
5 5,743 3,718 5,173 5,523 20,157
6 6,121 3,842 5,420 5,570 20,953
7 6,112 3,667 5,429 5,800 21,008
8 6,121 3,765 5,331 5,750 20,967
9 6,238 3,896 5,514 6,076 21,724
10 6,479 4,105 6,024 6,669 23,277
All males 59,206 37,176 52,680 55,195 204,257
Female 1 4,834 3,220 4,381 4,435 16,870
2 5,346 3,343 4,631 4,494 17,814
3 5,339 3,345 4,720 4,808 18,212
4 5,563 3,602 4,999 5,029 19,193
5 5,454 3,420 4,818 5,088 18,780
6 5,794 3,607 5,103 5,499 20,003
7 5,711 3,451 4,999 5,459 19,620
8 5,858 3,596 5,170 5,479 20,103
9 5,966 3,566 5,280 5,758 20,570
10 6,147 4,023 6,064 6,378 22,612
All females 56,012 35,173 50,165 52,427 193,777
Both 1 9,892 6,524 9,154 9,136 34,706
2 10,986 6,935 9,451 9,234 36,606
3 11,014 6,915 9,680 9,788 37,397
4 11,582 7,319 10,235 10,415 39,551
5 11,197 7,138 9,991 10,611 38,937
6 11,915 7,449 10,523 11,069 40,956
7 11,823 7,118 10,428 11,259 40,628
8 11,979 7,361 10,501 11,229 41,070
9 12,204 7,462 10,794 11,834 42,294
10 12,626 8,128 12,088 13,047 45,889
All 115,218 72,349 102,845 107,622 398,034
1
Source: Northern Ireland Census 2001.
4
Final Report
1
Table 7: Child pedestrian casualty rates by sex, Multiple Deprivation Measure (MDM) decile
and age group (1999 - 2008)

Male 1 0.18 0.30 0.46 0.98 0.49
2 0.28 0.39 0.50 1.01 0.54
3 0.14 0.59 1.03 0.88 0.65
4 0.38 0.78 0.95 0.67 0.68
5 0.40 0.65 0.87 0.98 0.72
6 0.38 0.62 1.00 0.95 0.73
7 0.39 1.23 1.12 1.22 0.96
8 0.72 1.17 1.14 1.48 1.12
9 1.01 1.51 1.47 1.23 1.28
10 1.71 3.34 2.34 1.89 2.21
All males 0.58 1.09 1.12 1.16 0.97
Female 1 0.00 0.16 0.41 0.52 0.27
2 0.15 0.30 0.43 0.80 0.42
3 0.06 0.39 0.55 0.83 0.45
4 0.16 0.28 0.60 0.38 0.35
5 0.13 0.18 0.33 0.71 0.35
6 0.35 0.67 0.71 1.16 0.72
7 0.28 0.58 0.56 0.84 0.56
8 0.44 1.14 0.79 0.75 0.74
9 0.37 1.23 1.25 1.04 0.93
10 0.72 2.29 1.47 1.58 1.44
All females 0.28 0.75 0.74 0.89 0.65
Both 1 0.09 0.23 0.44 0.76 0.09
2 0.22 0.35 0.47 0.91 0.22
3 0.10 0.49 0.80 0.86 0.10
4 0.28 0.53 0.78 0.53 0.28
5 0.27 0.42 0.61 0.85 0.27
6 0.36 0.64 0.86 1.06 0.36
7 0.34 0.91 0.85 1.04 0.34
8 0.58 1.15 0.97 1.12 0.58
9 0.70 1.38 1.36 1.14 0.70
10 1.23 2.82 1.90 1.74 1.23
All 0.43 0.93 0.93 1.03 0.81
Ratio of rates for all males to
2.10 1.45 1.52 1.30 1.49
all females
1
Average annual casualties per 1,000 population (see paragraph 4.4.4)
5
Final Report
Table 8: Spearman’s ranked correlation coefficients and significance levels for the
relationship between deprivation decile and child pedestrian casualty rates by sex
Age group All age

Sex 0-4 5-7 8-11 12-15 groups
Spearman’s Ranked Correlation Coefficient (ρ)
Male 0.903 0.939 0.915 0.673 1.000
Female 0.915 0.879 0.830 0.648 0.879
Both 0.964 0.964 0.939 0.842 0.964
1
Level of significance for Spearman’s Ranked Correlation Coefficient
Male 0.5% 0.5% 0.5% 2.5% 0.5%
Female 0.5% 0.5% 0.5% 2.5% 0.5%
Both 0.5% 0.5% 0.5% 0.5% 0.5%
1
The ρ values corresponding to probability levels are:
Probability 0.05 (5%) 0.025 (2.5%) 0.01 (1%) 0.005 (0.5%)
ρ .564 .648 .745 .794
This means, example, where the significance level is 0.5% there is less than 1 in 200 probability that
the relationship could have occurred by chance.
1
Table 9: Deprivation disparity ratio of child pedestrian casualty rates by sex, Multiple
Deprivation Measure (MDM) decile and age group (1999 – 2008)
Age group All age

Sex 0-4 5-7 8-11 12-15 groups
Male 9.63 11.03 5.08 1.93 4.54
2
Female 10.10 14.73 3.57 3.05 5.29
Both 13.49 12.25 4.35 2.30 4.78
1
The deprivation disparity ratio is the casualty rate for decile 10 (most deprived) divided by the rate for
decile 1 (most affluent). It is a measure of the additional risk of being involved as a pedestrian road
collision casualty between children resident in the most affluent and most deprived areas.
2
There were no casualties in MDM decile 1 recorded for females in the 0-4 age group. The average
casualty rate for deciles 1-3 has been used in this instance as a proxy.
6
Final Report
Table 10: Child and adult pedestrian casualty numbers by sex, Multiple Deprivation Measure
(MDM) decile and age group (1999 - 2008)
MDM Age group

Sex decile 0-15 16-24 25 and over All age groups
1 87 33 104 224
2 102 46 124 272
3 124 63 164 351
4 138 93 136 367
5 146 71 164 381
Male
6 154 121 208 483
7 201 102 230 533
8 234 88 267 589
9 278 133 261 672
10 515 224 561 1,300
All male casualties 1,979 974 2,219 5,172
1 46 26 80 152
2 74 28 90 192
3 82 44 109 235
4 68 46 104 218
5 65 41 127 233
Female
6 144 73 154 371
7 110 50 207 367
8 149 71 229 449
9 192 64 230 486
10 326 175 442 943
All female casualties 1,256 618 1,772 3,646
1 133 59 184 376
2 176 74 214 464
3 206 107 273 586
4 206 139 240 585
5 211 112 291 614
Both
6 298 194 362 854
7 311 152 437 900
8 383 159 496 1,038
9 470 197 491 1,158
10 841 399 1,003 2,243
All casualties 3,235 1,592 3,991 8,818
7
Final Report
Table 11: Child and adult population by sex, Multiple Deprivation Measure (MDM) decile and
age group (1999 - 2008)
MDM Age group

1 17,836 9,196 51,559 78,591
2 18,792 9,309 51,805 79,906
3 19,185 9,426 52,123 80,734
4 20,358 10,160 52,683 83,201
5 20,157 11,639 52,159 83,955
Male
6 20,953 12,028 53,184 86,165
7 21,008 11,058 52,210 84,276
8 20,967 10,715 51,662 83,344
9 21,724 11,221 48,034 80,979
10 17,836 9,196 51,559 78,591
All male casualties 204,257 106,624 510,561 821,442
1 16,870 8,998 57,508 83,376
2 17,814 8,665 55,532 82,011
3 18,212 8,674 56,949 83,835
4 19,193 9,850 56,933 85,976
5 18,780 11,474 56,082 86,336
Female
6 20,003 11,667 57,126 88,796
7 19,620 10,539 56,682 86,841
8 20,103 10,990 57,664 88,757
9 20,570 11,567 55,850 87,987
10 22,612 12,439 54,984 90,035
All female casualties 193,777 104,863 565,310 863,950
1 34,706 18,194 109,067 161,967
2 36,606 17,974 107,337 161,917
3 37,397 18,100 109,072 164,569
4 39,551 20,010 109,616 169,177
5 38,937 23,113 108,241 170,291
Both
6 40,956 23,695 110,310 174,961
7 40,628 21,597 108,892 171,117
8 41,070 21,705 109,326 172,101
9 42,294 22,788 103,884 168,966
10 45,889 24,311 100,126 170,326
All casualties 398,034 211,487 1,075,871 1,685,392
8
Final Report
1
Table 12: Child and adult pedestrian casualty rates by sex, Multiple Deprivation Measure
(MDM) decile and age group (1999 - 2008)
MDM Age group

1 0.49 0.36 0.20 0.29
2 0.54 0.49 0.24 0.34
3 0.65 0.67 0.31 0.43
4 0.68 0.92 0.26 0.44
5 0.72 0.61 0.31 0.45
Male
6 0.73 1.01 0.39 0.56
7 0.96 0.92 0.44 0.63
8 1.12 0.82 0.52 0.71
9 1.28 1.19 0.54 0.83
10 2.21 1.89 1.24 1.62
All male casualties 0.97 0.91 0.43 0.63
1 0.27 0.29 0.14 0.18
2 0.42 0.32 0.16 0.23
3 0.45 0.51 0.19 0.28
4 0.35 0.47 0.18 0.25
5 0.35 0.36 0.23 0.27
Female
6 0.72 0.63 0.27 0.42
7 0.56 0.47 0.37 0.42
8 0.74 0.65 0.40 0.51
9 0.93 0.55 0.41 0.55
10 0.27 0.29 0.14 0.18
All female casualties 0.65 0.59 0.31 0.42
1 0.38 0.32 0.17 0.23
2 0.48 0.41 0.20 0.29
3 0.55 0.59 0.25 0.36
4 0.52 0.69 0.22 0.35
5 0.54 0.48 0.27 0.36
Both
6 0.73 0.82 0.33 0.49
7 0.77 0.70 0.40 0.53
8 0.93 0.73 0.45 0.60
9 1.11 0.86 0.47 0.69
10 1.83 1.64 1.00 1.32
All casualties 0.81 0.75 0.37 0.52
Ratio of rates for all males to
1.49 1.55 1.39 1.49
all females
1
Average annual casualties per 1,000 population (see paragraph 4.4.4)
9
Final Report
Table 13: Spearman’s ranked correlation coefficients and significance levels for the
relationship between deprivation decile and child and adult pedestrian casualty
rates by sex and age group (1999 - 2008)
Age group All age

Sex 0-15 16-24 25 and over groups
Spearman’s Ranked Correlation Coefficient (ρ)
Male 1.000 0.867 0.964 1.000
Female 0.879 0.842 0.988 0.964
Both 0.964 0.927 0.988 0.988
1
Level of significance for Spearman’s Ranked Correlation Coefficient
Male 0.5% 0.5% 0.5% 0.5%
Female 0.5% 0.5% 0.5% 0.5%
Both 0.5% 0.5% 0.5% 0.5%
1
The ρ values corresponding to probability levels are:
Probability 0.05 (5%) 0.025 (2.5%) 0.01 (1%) 0.005 (0.5%)
ρ .564 .648 .745 .794
This means, example, where the significance level is 0.5% there is less than 1 in 200 probability that
the relationship could have occurred by chance.
1
Table 14: Deprivation disparity ratio of child and adult pedestrian casualty rates by sex,
Multiple Deprivation Measure (MDM) decile and age group (1999 – 2008)
Age group All age

Sex 0-15 16-24 25 and over groups
Male 4.54 5.26 6.16 5.68
Female 5.29 4.87 5.78 5.75
Both 4.78 5.06 5.94 5.67
1
The deprivation disparity ratio is the casualty rate for decile 10 (most deprived) divided by the rate for
decile 1 (most affluent). It is a measure of the additional risk of being involved as a pedestrian road
collision casualty between children resident in the most affluent and most deprived areas.
2
There were no casualties in MDM decile 1 recorded for females in the 0-4 age group. The average
casualty rate for deciles 1-3 has been used in this instance as a proxy.
10
Final Report
Table 15: Number of child pedestrian casualties by location of collision (April 2007 – March
2009
Home/Away Number of casualties %age of postcode

Postcode recorded
- Away 242 54%
- Home 206 46%
No postcode recorded 28 -
Total 476 -
Table 16: Distance between collision location and home address for child pedestrian
casualties in Salford and Northern Ireland
Percentages of casualties with stated distance between

Distance between collision and collision and home address
1 2
home address (km) Salford casualties Northern Ireland casualties
≤ 0.25 41.50% 41.40%
≤ 0.50 62.30% 56.20%
≤ 0.75 71.20% 63.50%
≤ 1.00 78.00% 68.50%
≤ 1.25 80.50% 72.90%
≤ 1.50 85.20% 74.90%
≤ 1.75 88.10% 77.60%
≤ 2.00 88.60% 79.60%
> 2.00 11.40% 21.40%
1
Source: Petch and Henson (2000) = base not stated (May 1995 – April 1998)
2
Source: PSNI (2009) – base 477 casualties (April 2007 – March 2009)
11
Final Report
Table 17: Decile of residence by decile of location of collision for child pedestrian casualties (April 2007 to March 2009)
Number of
child Home +
pedestrian Number of Number of
Decile of casualties away deciles away deciles
residence with known Decile of location of away collision within two within two
(home home Home Away deciles of deciles of
decile) postcode Casualties Casualties 1 2 3 4 5 6 7 8 9 10 home decile home decile
1 23 4 19 7 2 2 2 1 1 1 2 1 11 15
2 23 12 11 1 2 1 1 2 1 2 1 4 16
3 37 14 23 3 1 3 1 1 5 4 2 3 8 22
4 30 9 21 1 1 3 1 6 1 2 2 3 1 12 21
5 39 21 18 2 1 5 4 1 1 1 3 11 32
6 42 20 22 2 1 2 4 3 2 1 4 3 12 32
7 40 19 21 1 3 1 6 2 1 2 4 1 15 34
8 46 29 17 2 1 3 4 1 3 1 2 11 40
9 59 27 32 2 2 1 2 2 3 3 7 3 7 20 47
10 109 51 58 3 3 3 12 37 52 103
All deciles 448 206 242 19 11 17 15 23 21 20 26 35 55 156 362
12
Final Report
Table 18: Child pedestrian casualties by month and time of day (weekdays)
Hour beginning All

Month 0-6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 hours %age
January 2 2 23 7 2 2 8 11 43 30 19 10 9 12 5 5 1 191 8%
February 1 1 18 6 2 3 3 6 11 38 25 21 22 16 5 5 1 1 185 7%
March 1 1 34 7 3 9 11 37 30 20 15 20 10 3 2 4 207 8%
April 14 7 1 3 8 8 17 38 36 30 21 15 14 10 3 225 9%
May 2 30 9 2 1 6 7 14 41 33 30 17 27 22 6 2 249 10%
June 1 1 17 7 4 1 8 6 14 30 27 25 18 27 13 11 4 6 220 9%
July 1 4 9 10 11 15 17 14 18 22 19 16 11 2 1 170 7%
August 1 1 7 14 16 17 15 17 38 22 17 24 12 8 2 211 8%
September 2 22 9 1 3 3 6 10 35 28 28 21 24 15 8 1 216 9%
October 2 28 7 2 4 6 8 18 46 35 30 31 19 14 11 3 1 265 11%
November 1 2 19 6 2 2 5 2 10 38 36 28 14 10 8 9 2 2 196 8%
December 8 8 3 2 8 6 10 37 28 19 16 16 10 8 5 2 186 7%
All months 8 12 214 73 24 35 76 93 158 415 339 306 229 219 163 99 38 20 2521 100%
%age - - 8% 3% 1% 1% 3% 4% 6% 16% 13% 12% 9% 9% 6% 4% 2% 1% 100%
13
Final Report
Table 19: Child pedestrian casualties by month and time of day (weekends)
Hour beginning All

Month 0-6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 hours %age
January 2 4 4 4 8 7 7 4 4 3 1 1 49 7%
February 3 3 5 4 4 5 5 6 8 3 7 1 1 55 8%
March 3 3 3 2 6 3 8 4 2 9 1 44 6%
April 1 2 5 5 10 8 12 6 3 3 8 6 1 3 73 10%
May 2 1 2 1 3 9 4 13 7 6 7 9 7 2 2 75 11%
June 1 6 7 8 6 7 3 6 12 5 3 7 1 72 10%
July 2 3 7 5 7 7 4 2 3 7 4 1 1 53 7%
August 3 6 5 5 1 2 3 9 5 4 5 1 1 50 7%
September 1 2 9 5 6 8 9 6 10 11 5 2 1 75 11%
October 2 1 3 2 2 6 12 4 7 4 9 6 2 2 1 1 64 9%
November 2 1 5 4 7 3 8 6 4 5 7 2 54 8%
December 1 2 2 10 7 6 3 4 3 4 3 1 4 50 7%
All months 16 1 0 2 11 24 54 75 70 80 67 63 75 65 63 29 15 4 714 100%
%age 2% 0% 0% 0% 2% 3% 8% 11% 10% 11% 9% 9% 11% 9% 9% 4% 2% 1% 100%
14
Final Report
Table 20: Child pedestrian casualties by speed limit and age group for males and females
(1999 - 2008)
Speed limit
All
Urban rural Age speed
1
classification group <30 30 40 50/55 60 70 limits
A: Males
number
0-4 0 34 1 0 8 0 43
5-7 0 41 3 0 8 0 52
Rural
8 - 11 0 53 4 2 28 0 87
12 - 15 0 63 5 1 55 0 124
0-4 0 294 5 1 1 0 301
5-7 0 348 5 1 1 0 355
Urban
8 - 11 0 466 25 5 7 0 503
12 - 15 1 454 34 9 15 1 514
row percentage
0-4 79% 2% 0% 19% 100%
5-7 79% 6% 0% 15% 100%
Rural
8 - 11 61% 5% 2% 32% 100%
12 - 15 51% 4% 1% 44% 100%
0-4 98% 2% 0% 0% 100%
5-7 98% 1% 0% 0% 100%
Urban
8 - 11 93% 5% 1% 1% 100%
12 - 15 0% 88% 7% 2% 3% 0% 100%
B: Females
number
0-4 1 22 1 0 2 0 26
5-7 0 22 1 0 8 0 31
Rural
8 - 11 0 39 2 0 15 0 56
12 - 15 1 38 2 0 32 1 74
0-4 1 125 2 0 1 0 129
5-7 0 224 9 0 1 0 234
Urban
8 - 11 1 298 9 2 4 0 314
12 - 15 0 357 19 10 6 0 392
row percentage
0-4 4% 85% 4% 8% 100%
5-7 71% 3% 26% 100%
Rural
8 - 11 70% 4% 27% 100%
12 - 15 1% 51% 3% 43% 1% 100%
0-4 1% 97% 2% 1% 100%
5-7 96% 4% 0% 100%
Urban
8 - 11 0% 95% 3% 1% 1% 100%
12 - 15 91% 5% 3% 2% 100%
1
One speed limit was recorded at 55mph.
15
Final Report
Table 21: Child pedestrian casualties by weather condition by age group for males and
females (1999 - 2008)
Urban rural Age Weather condition

classification group Fine Rain Snow Mist/sun Unknown Other All
A: Males
number
0-4 35 5 0 0 1 2 43
5-7 49 3 0 0 0 0 52
Rural
8 - 11 71 13 0 0 1 2 87
12 - 15 101 17 2 4 0 0 124
0-4 261 25 1 3 1 10 301
5-7 313 33 0 3 2 4 355
Urban
8 - 11 416 68 3 3 7 6 503
12 - 15 404 86 5 1 8 10 514
row percentage
0-4 81% 12% 2% 5% 100%
5-7 94% 6% 100%
Rural
8 - 11 82% 15% 1% 2% 100%
12 - 15 81% 14% 2% 3% 100%
0-4 87% 8% 0% 1% 0% 3% 100%
5-7 88% 9% 1% 1% 1% 100%
Urban
8 - 11 83% 14% 1% 1% 1% 1% 100%
12 - 15 79% 17% 1% 0% 2% 2% 100%
B: Females
number
0-4 21 4 0 0 0 1 26
5-7 24 7 0 0 0 0 31
Rural
8 - 11 43 7 1 2 3 0 56
12 - 15 60 9 1 2 0 2 74
0-4 115 13 0 0 0 1 129
5-7 205 20 0 4 1 4 234
Urban
8 - 11 257 50 1 0 3 3 314
12 - 15 309 66 2 2 7 6 392
row percentage
0-4 81% 15% 4% 100%
5-7 77% 23% 0% 100%
Rural
8 - 11 77% 13% 2% 4% 5% 0% 100%
12 - 15 81% 12% 1% 3% 3% 100%
0-4 89% 10% 1% 100%
5-7 88% 9% 2% 0% 2% 100%
Urban
8 - 11 82% 16% 0% 1% 1% 100%
12 - 15 79% 17% 1% 1% 2% 2% 100%
1
Weather conditions include fine, rain and snow both with and without high winds. The number of
weather conditions reported with high winds was 80 (fine (38), rain (46) and snow (2)). The table
combines with and without high winds for fine, rain and snow.
16
Final Report
Table 22: Child pedestrian casualties by light condition (1999 - 2008)
Light condition All

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec months
number
Daylight: street lights present 131 147 196 250 294 256 187 201 225 204 115 104 2310
Daylight: no street lighting 14 12 3 24 21 21 26 29 20 25 13 13 221
Daylight: street lighting unknown 2 4 2 2 1 4 2 3 6 11 4 2 43
Darkness: street lights present and lit 83 73 42 21 8 11 7 25 37 82 105 100 594
Darkness: street lights present but unlit 3 2 4 1 1 3 2 2 18
Darkness: no street lighting 7 2 3 1 3 1 2 7 13 39
Darkness: street lighting unknown 1 1 2 4 2 10
All 240 240 251 298 324 292 223 261 291 329 250 236 3235
column percentage
Daylight: street lights present 55% 61% 78% 84% 91% 88% 84% 77% 77% 62% 46% 44% 72%
Daylight: no street lighting 6% 5% 1% 8% 6% 7% 12% 11% 7% 8% 5% 6% 7%
Daylight: street lighting unknown 1% 2% 1% 1% 0% 1% 1% 1% 2% 3% 2% 1% 1%
Darkness: street lights present and lit 35% 30% 17% 7% 2% 4% 3% 10% 13% 25% 42% 42% 18%
Darkness: street lights present but unlit 1% 1% 2% 0% 0% 1% 1% 1% 1%
Darkness: no street lighting 3% 1% 1% 0% 1% 0% 1% 3% 6% 1%
Darkness: street lighting unknown 0% 0% 1% 2% 1% 0%
All 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%
17
Final Report
Table 23: Child pedestrian casualties by principle causation factor (1999 – 2008)
% age by age group Causation factor group

Causation Number of All age Number of
factor group Principal causation factor casualties 0-4 5-7 8-11 12-15 groups casualties %age
Pedestrian Heedless of traffic crossing carriageway 1,138 12% 31% 44% 41% 35%
only Walk/run movement masked 804 14% 36% 28% 21% 25%
Lack of supervision/escaping from custody 321 54% 7% 1% 0% 10%
Walking or running onto carriageway 165 3% 6% 6% 5% 5%
Using pedestrian crossing without care 71 0% 2% 3% 3% 2%
Playing on carriageway 53 1% 3% 2% 1% 2%
Slipping or falling pedestrian 43 1% 1% 2% 2% 1% 2,712 83.8%
Heedless of traffic walking or standing in
34 0% 1% 1% 2% 1%
carriageway - not crossing
Holding onto vehicle 33 0% 1% 2% 1% 1%
Other pedestrian factor 30 0% 1% 1% 1% 1%
Alcohol or drugs - pedestrian 12 0% 0% 0% 1% 0%
Pedestrian wearing dark clothing 5 0% 0% 0% 0% 0%
Physical/mental illness or injury - pedestrian 3 0% 0% 0% 0% 0%
Driver/Rider Inattention or attention diverted 149 4% 3% 4% 7% 5% 494 15.3%
only Reversing without care 62 3% 2% 1% 2% 2%
Excessive speed having regard to conditions 50 1% 1% 1% 2% 2%
Disobeying pedestrian crossing 49 2% 2% 2% 1% 2%
Other driver/rider factor 37 0% 0% 1% 2% 1%
Alcohol or drugs - driver/rider 27 0% 0% 1% 1% 1%
Disobeying traffic sign/signal 19 0% 1% 0% 1% 1%
Emerging from private road/entrance without care 15 1% 0% 1% 0% 0%
Turning right without care 14 1% 0% 0% 1% 0%
Starting without care 11 1% 0% 0% 0% 0%
Wrong course/position 10 0% 0% 0% 1% 0%
Overtaking on offside without care 9 0% 0% 0% 1% 0%
Emerging from minor road without care 8 0% 0% 0% 0% 0%
18
Final Report
Turning left without care 8 0% 0% 0% 0% 0%

Stopping without care 6 0% 0% 0% 0% 0%
Inexperience with type of vehicle 6 0% 0% 0% 0% 0%
Driving too close 4 0% 0% 0% 0% 0%
U turning without care 2 0% 0% 0% 0% 0%
Crossing or entering road junction without care 2 0% 0% 0% 0% 0%
Distraction by action inside vehicle 2 0% 0% 0% 0% 0%
Overtaking on nearside without care 1 0% 0% 0% 0% 0%
Distraction by action outside vehicle 1 0% 0% 0% 0% 0%
Failing to give or giving faulty signal 1 0% 0% 0% 0% 0%
Fatigue 1 0% 0% 0% 0% 0%
Vehicle Other vehicle factor 5 0% 0% 0% 0% 0%
factors Insecure load 4 0% 0% 0% 0% 0%
12 0.4%
Unattended vehicle running away 2 0% 0% 0% 0% 0%
Tyre blow-out before impact 1 0% 0% 0% 0% 0%
Miscellaneous No apparent cause/ cause not traced. 6 0% 0% 0% 0% 0%
8 0.2%
Other miscellaneous factor 2 0% 0% 0% 0% 0%
Weather Ice, frost or snow 2 0% 0% 0% 0% 0%
conditions 4 0.1%
Glaring sun 2 0% 0% 0% 0% 0%
Road factors Road works in progress 2 0% 0% 0% 0% 0%
3 0.1%
Flooding 1 0% 0% 0% 0% 0%
Passenger Opening door without care - passenger 1 0% 0% 0% 0% 0%
only 2 0.1%
Other passenger factor 1 0% 0% 0% 0% 0%
All All 3,235 100% 100% 100% 100% 100% 3,235 100%
19
Final Report
Table 24: Child pedestrian casualties by junction detail and junction control (1999 – 2008)
Junction Control
Give way Automatic Give way or %age

Not at sign or Un- traffic un- Stop Authorised junction
Junction detail junction markings controlled signal controlled Sign person All %age all collisions
Not at or within 20m of junction 1,603 1,603 50% -
T Junction or Staggered Junction 623 324 96 130 46 3 1,222 38% 75%
Crossroads 38 10 56 7 5 1 117 4% 7%
Private Drive / Entrance 101 14 1 116 4% 7%
Roundabout 39 1 2 9 1 52 2% 3%
Other Junction 9 23 4 1 37 1% 2%
Multiple Junction 10 2 18 3 33 1% 2%
Y junction 16 6 6 28 1% 2%
Mini Roundabout 14 2 16 0% 1%
Slip Road 4 1 4 2 11 0% 1%
All 1,603 753 468 186 168 51 6 3,235 100% 100%
%age all 50% 23% 14% 6% 5% 2% 0% 100%
%age junction collisions - 46% 29% 11% 10% 3% 0% 100%
20
Final Report
Table 25: Child pedestrian casualties by pedestrian crossing facility and age
group (1999 – 2008)
Age group
All age
Pedestrian crossing facility 0-4 5-7 8-11 12-15 groups
number
None within 50 metres 439 570 747 821 2,577
Pelican 13 30 77 103 223
Other light controlled crossing 12 15 45 44 116
Pedestrian Phase at traffic signal junction 13 16 32 48 109
Zebra crossing 13 18 25 37 93
Central refuge - no control 6 7 19 26 58
Site controlled by crossing patrol 2 10 9 13 34
Zebra controlled by School Crossing Patrol 5 3 6 14
Footbridge or subway 1 1 2 4 8
Site controlled by other authorised person 1 2 3
All 499 672 960 1,104 3,235
column percentage
None within 50 metres 88% 85% 78% 74% 80%
Pelican 3% 4% 8% 9% 7%
Other light controlled crossing 2% 2% 5% 4% 4%
Pedestrian Phase at traffic signal junction 3% 2% 3% 4% 3%
Zebra crossing 3% 3% 3% 3% 3%
Central refuge - no control 1% 1% 2% 2% 2%
Site controlled by crossing patrol 0% 1% 1% 1% 1%
Zebra controlled by School Crossing Patrol 0% 1% 0% 1% 0%
Footbridge or subway 0% 0% 0% 0%
Site controlled by other authorised person 0% 0% 0%
All 100% 100% 100% 100% 100%
21
Final Report
Table 26: Child pedestrian casualties by carriageway type and urban rural classification
(1999 – 2008)
Number %ages
Carriageway type Rural Urban All Rural Urban All

Single carriageway 1 lane in
374 1,860 2,234 76% 68% 69%
each direction
Single carriageway 75 310 385 15% 11% 12%
Single carriageway 4 lanes 191 191 7% 6%
Other / unknown 14 106 120 3% 4% 4%
One way street 2 91 93 0% 3% 3%
Single track road 22 42 64 4% 2% 2%
Dual carriageway 3 58 61 1% 2% 2%
Single carriageway 3 lanes 1 48 49 0% 2% 2%
Roundabout 2 31 33 0% 1% 1%
Motorway 3 3 0% 0%
Slip road 2 2 0% 0%
All 493 2,742 3,235 100% 100% 100%
Table 27: Child pedestrian casualties by road surface condition and urban rural
classification (1999 – 2008)
Number %ages
Road surface condition Rural Urban All Rural Urban All
Dry 354 2,016 2,370 72% 74% 73%
Wet / Damp 127 660 787 26% 24% 24%
Other 1 25 26 0% 1% 1%
Frost / Ice 2 18 20 0% 1% 1%
Slippery (after dry Spell) 4 8 12 1% 0% 0%
Snow 2 9 11 0% 0% 0%
Mud 1 3 4 0% 0% 0%
Flood 2 1 3 0% 0% 0%
Oil 1 1 0% 0%
Leaves 1 1 0% 0%
All 493 2,742 3,235 100% 100% 100%
22
Final Report
Table 28: Child pedestrian casualties by special condition at site and urban rural
Special condition at site Rural Urban All %age

None 488 2,711 3,199 99%
Road Works 3 23 26 1%
Automatic Traffic Signal Out 4 4 0%
Road Surface Defective 2 2 4 0%
Permanent Road Signing Defective or Obscured 1 1 0%
Railway Level Crossing present 1 1 0%
All 493 2,742 3,235 100%
Table 29: Child pedestrian casualties by carriageway hazard and urban rural
Carriageway hazard Rural Urban All %Age

None 489 2,702 3,191 99%
Other object in Carriageway 3 36 39 1%
Dislodged Vehicle Load in Carriageway 1 1 0%
Involvement with Previous Accident 1 1 0%
Dog in carriageway 1 1 0%
Smoke, dust etc 1 1 0%
Pedestrian in carriageway - not injured 1 1 0%
All 493 2,742 3,235 100%
23
Final Report
Table 30: Child pedestrian casualties by vehicle manoeuvre and vehicle location at time of impact (1999 – 2008)
Vehicle location at time of impact
Pedestrian
and On lay- Leaving Entering
On Other Entering Leaving vehicle by or lay-by / Bus lay-by /
On main minor public main Footpath main shared hard hard lane / hard
Vehicle manoeuvre road road place road (pavement) road precinct shoulder shoulder busway shoulder All %age
1
Going Ahead other 1,431 666 16 12 17 7 7 3 2 4 2,166 67%
Passing Stationary Vehicle on o/s 176 115 1 1 1 1 295 9%
Reversing 20 42 20 11 16 5 9 2 2 1 128 4%
Other / not known 47 41 7 2 11 1 3 112 3%
. About to Go Ahead / Held up 51 17 1 1 1 71 2%
Turning right 19 22 2 6 2 18 1 70 2%
Stopping 36 19 4 1 1 2 1 1 1 66 2%
Starting 27 22 4 5 1 1 4 64 2%
Going ahead left hand bend 39 20 3 1 63 2%
Going ahead left hand bend 26 23 2 5 2 58 2%
Passing on Nearside 30 26 56 2%
Turning left 13 11 18 10 1 53 2%
Parked 7 5 1 2 15 0%
Overtaking Moving Vehicle on o/s 6 3 1 10 0%
Waiting to turn left 1 2 1 4 0%
Changing Lane to Right 2 1 3 0%
U turn 1 1 0%
All 1,931 1036 58 56 54 46 22 13 9 6 3 3,235 100%
%age 60% 32% 2% 2% 2% 1% 1% 0% 0% 0% 0% 100%
1
Includes one miscoded value for vehicle location at time of impact.
24
Final Report
Table 31: Child pedestrian casualties by school pupil casualty by age group for week and
weekend days (1999 – 2008)
Not a School pupil School pupil

Age school not on journey on journey All
Day group pupil to/from school to/from school Other ages %Age
A: All casualties
0-4 265 36 15 60 376 15%
5-7 31 326 94 47 498 20%
Weekday
8 - 11 35 426 216 74 751 30%
12 - 15 37 394 390 75 896 36%
All weekday 368 1,182 715 256 2,521 100%
%age 15% 47% 28% 10% 100%
0-4 92 10 21 123 17%

5-7 21 142 11 174 24%
Weekend
8 - 11 26 159 24 209 29%
12 - 15 9 174 25 208 29%
All weekend 148 485 81 714 100%
%age 21% 68% 0% 11% 100%
0-4 357 46 15 81 499 15%

5-7 52 468 94 58 672 21%
All
8 - 11 61 585 216 98 960 30%
12 - 15 46 568 390 100 1,104 34%
All 516 1,667 715 337 3,235 100%
%age 16% 52% 22% 10% 100%
1
B: Average casualties per day
0-4 53.0 7.2 3.0 12.0 75.2 15%
5-7 6.2 65.2 18.8 9.4 99.6 20%
Weekday
8 - 11 7.0 85.2 43.2 14.8 150.2 30%
12 - 15 7.4 78.8 78.0 15.0 179.2 36%
All weekday 73.6 236.4 143.0 51.2 504.2 100%
%age 15% 47% 28% 10% 100%
0-4 46.0 5.0 10.5 61.5 17%

5-7 10.5 71.0 5.5 87.0 24%
Weekend
8 - 11 13.0 79.5 12.0 104.5 29%
12 - 15 4.5 87.0 12.5 104.0 29%
All weekend 74.0 242.5 40.5 357.0 100%
%age 21% 68% 0% 11% 100%
0-4 51.0 6.6 2.1 11.6 71.3 15%

5-7 7.4 66.9 13.4 8.3 96.0 21%
All
8 - 11 8.7 83.6 30.9 14.0 137.1 30%
12 - 15 6.6 81.1 55.7 14.3 157.7 34%
All 73.7 238.1 102.1 48.1 462.1 100%
%age 16% 52% 22% 10% 100%
1
To allow comparison of weekday and weekend casualty numbers, the weekday figures in Part A of the table
have been divided by 5 to show the number of casualties on an average weekday. Likewise the weekend figures
have been divided by 2 to show the number of casualties on an average weekend day.
25
Final Report
Table 32: Child pedestrian casualties by pedestrian location and pedestrian movement (1999 – 2008)
Pedestrian movement
Crossing Crossing
from from
drivers drivers In c'way
Crossing Crossing nearside offside Walking In c'way Walking stationary
from from masked masked in c'way stationary in c'way - not
drivers drivers by static by static Unknown/ back to - not facing crossing -
Pedestrian location nearside offside vehicle vehicle other traffic crossing traffic masked All %age
In carriageway crossing elsewhere 627 373 375 286 31 8 3 4 3 1,710 53%
In centre of carriageway 94 140 71 66 12 6 25 1 1 416 13%
In carriageway crossing elsewhere
94 64 44 54 3 1 2 1 263 8%
within 50 m of crossing
On footway or verge 6 1 1 1 125 49 1 63 247 8%
In carriageway, not crossing 23 11 22 6 65 21 48 13 9 218 7%
In carriageway on pedestrian crossing 88 73 23 17 2 1 1 205 6%
Unknown or other 28 15 6 8 85 4 4 1 3 154 5%
On refuge, central island or central
2 9 1 1 13 0%
reservation
In carriageway, within zigzag lines at
3 2 1 6 0%
crossing approach
In carriageway within zigzag lines at
1 2 3 0%
crossing exit
All 966 688 545 439 323 89 84 82 19 3,235 100%
%age 30% 21% 17% 14% 10% 3% 3% 3% 1% 100%
26
Final Report
Figure 1: Child pedestrian casualty rate by decile

(ages 0 - 15 inclusive) (1999 - 2008)
2.00
1.80
1.60
Child pedestrian casualty rate
1.40
1.20
1.00
0.80
0.60
0.40
0.20
0.00
1 2 3 4 5 6 7 8 9 10
Deprivation decile
Figure 2: Child pedestrian casualty rate by

deprivation decile and sex (ages 0 - 15 inclusive)
1999 - 2008)
2.50
2.00
1.50
Male
Female
1.00
0.50
0.00
1 2 3 4 5 6 7 8 9 10
Deprivation decile
27
Final Report

deprivation decile and sex (ages 0 to 4 inclusive)
(1999 - 2008)
1.80
1.60
1.40
1.20
1.00
0.80
0.60
0.40
0.20
0.00
1 2 3 4 5 6 7 8 9 10
Deprivation decile

(1999 - 2008)
4.00
3.50
3.00
2.50
2.00
Male
1.50 Female
1.00
0.50
0.00
1 2 3 4 5 6 7 8 9 10
Deprivation decile
28
Final Report

(1999 - 2008)
2.50
2.00
1.50
Male
1.00 Female
0.50
0.00
1 2 3 4 5 6 7 8 9 10
Deprivation decile

(1999 - 2008)
2.00
1.80
1.60
1.40
1.20
1.00 Male
0.80 Female
0.60
0.40
0.20
0.00
1 2 3 4 5 6 7 8 9 10
Deprivation decile
29
Final Report
Figure 7: Child pedestrian casualty rates by deprivation decile and age

group for boys (1999 - 2008)
4
3.5
3
Casualty rate
1 - 4 years
2.5
5 - 7 years
2
8 - 11 years
1.5
12 -15 years
1
0.5
0
1 2 3 4 5 6 7 8 9 10
Deprivation decile
30
Final Report
Figure 8: Child pedestrian casualty rates by deprivation decile and age

group for girls (1999 - 2008)
2.5
2
Casualty rate
1 - 4 years
1.5
5 - 7 years
8 - 11 years
1
12 -15 years
0.5
0
1 2 3 4 5 6 7 8 9 10
Deprivation decile
31
Final Report
Figure 9: Young adult pedestrian casualty rate by

(1999 - 2008)
2.00
1.80
1.60
1.40
1.20
1.00 Male
0.80 Female
0.60
0.40
0.20
0.00
1 2 3 4 5 6 7 8 9 10
Deprivation decile
Figure 10: Adult pedestrian casualty rate by

deprivation decile and sex (ages 25 and over)
(1999 - 2008)
1.40
1.20
1.00
0.80 Male
Female
0.60
0.40
0.20
0.00
1 2 3 4 5 6 7 8 9 10
Deprivation decile
32
Final Report
Figure 11: Pedestrian casualty rate by deprivation

decile and sex (all ages) (1999 - 2008)
1.80
1.60
1.40
1.20
1.00
Male
0.80
Female
0.60
0.40
0.20
0.00
1 2 3 4 5 6 7 8 9 10
Deprivation decile
Figure 12: Pedestrian casualty rates by deprivation decile and age group
for all children and adults (males) (1999 - 2008)
2.50
2.00
Casualty rate
1.50 0 - 15 years
16 - 24 years
1.00 25 and over
0.50
0.00
1 2 3 4 5 6 7 8 9 10
Deprivation decile
33
Final Report
Figure 13: Pedestrian casualty rates by deprivation decile and age group
for all children and adults (females) (1999 - 2008)
1.60
1.40
1.20
Casualty rate
1.00 0 - 15 years
0.80 16 - 24 years
0.60 25 and over
0.40
0.20
0.00
1 2 3 4 5 6 7 8 9 10
Deprivation decile
Figure 14: Child pedestrian casualty rate by decile based on

location of residence (April 2007 to March 2009)
1.40
1.20
1.00
0.80
0.60
0.40
0.20
0.00
1 2 3 4 5 6 7 8 9 10
Deprivation decile
34
Final Report
Figure 15: Child pedestrian casualty rate by decile based on

location of collision (April 2007 to March 2009)
1.40
1.20
1.00
0.80
0.60
0.40
0.20
0.00
1 2 3 4 5 6 7 8 9 10
Deprivation decile
Figure 16: Difference between child pedestrian casualty rates

by decile based on location of collision and residence (April
2007 to March 2009)
0.70
rate difference
Difference in child pedestrian casualty rate
0.50
0.30
0.10
-0.10 1 2 3 4 5 6 7 8 9 10
-0.30
-0.50
-0.70
Deprivation decile
35
Final Report
Figure 17: KSI child pedestrian casualties by year and sex (1999 -
2008)
100
90
Number of casualties
80
70
60
Male
50
Female
40
30
20
10
0
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
Year
Figure 18: All child pedestrian casualties by year and sex (1999 -
2008)
350
300
250
200 Male
150 Female
100
50
0
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
Year
36
Final Report
Figure 19: Average male child pedestrian casualties by year of age and
age group (1999 - 2008)
30
25
1-4
20
5-7
15
8-11
10
12-15
5
0
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
Year
Figure 20: Average female child pedestrian casualties by year of age

20
15 1-4
5-7
10
8-11
5 12-15
0
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
Year
37
Final Report
Appendix 3
Modelling Tables and Figures
1
Final Report
Table A 1: Level 1 model output with MDM decile as independent variable
Significance
Factor Estimate Std. Error Z Value Pr(>|Z|) code
(Intercept) 0.76548 0.09195 8.325 <2e-16 ***
mdmdecM2 0.11573 0.12564 0.921 0.35702
mdmdecM3 0.32195 0.12334 2.610 0.00905 **
mdmdecM4 0.34826 0.12281 2.836 0.00457 **
mdmdecM5 0.31371 0.12311 2.548 0.01083 *
mdmdecM6 0.69341 0.12002 5.777 7.59e-09 ***
mdmdecM7 0.71597 0.11995 5.969 2.39e-09 ***
mdmdecM8 0.90485 0.11893 7.608 2.78e-14 ***
mdmdecM9 1.21956 0.11884 10.262 <2e-16 ***
mdmdecM10 1.98809 0.11933 16.660 <2e-16 ***
Significance codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Table A 2: Level 1 model output with MDM decile and age as independent variables
Significance
Coefficient Estimate Std. Error Z value Pr(>|z|) code
(Intercept) 0.06881 0.10296 0.668 0.503948
mdmdecM2 0.13670 0.11585 1.180 0.238005
mdmdecM3 0.32541 0.11340 2.869 0.004111 **
mdmdecM4 0.39946 0.11235 3.556 0.000377 ***
mdmdecM5 0.34637 0.11288 3.069 0.002151 **
mdmdecM6 0.72863 0.10927 6.668 2.59e-11 ***
mdmdecM7 0.75236 0.10915 6.893 5.47e-12 ***
mdmdecM8 0.96688 0.10774 8.974 <2e-16 ***
mdmdecM9 1.27366 0.10714 11.887 <2e-16 ***
mdmdecM10 2.05901 0.10640 19.351 <2e-16 ***
ageA2 0.77367 0.08016 9.652 <2e-16 ***
ageA3 0.96016 0.07906 12.145 <2e-16 ***
ageA4 1.15707 0.07810 14.815 <2e-16 ***
ageA5 0.65797 0.08093 8.131 4.27e-16 ***
ageA6 -0.56574 0.09620 -5.881 4.08e-09 ***
2
Final Report
Table A 3: Level 1 model output with MDM decile, age and sex as independent variables
Significance
Coefficient Estimate Std. Error Z Value Pr(>|Z|) Code
(Intercept) -0.20944 0.10420 -2.010 0.044433 *
mdmdecM2 0.14618 0.11335 1.290 0.197177
mdmdecM3 0.33244 0.11088 2.998 0.002717 **
mdmdecM4 0.38775 0.11001 3.525 0.000424 ***
mdmdecM5 0.33174 0.11060 2.999 0.002705 **
mdmdecM6 0.75155 0.10654 7.054 1.74e-12 ***
mdmdecM7 0.74415 0.10666 6.977 3.01e-12 ***
mdmdecM8 0.97624 0.10507 9.292 <2e-16 ***
mdmdecM9 1.28022 0.10438 12.265 <2e-16 ***
mdmdecM10 2.06573 0.10341 19.976 <2e-16 ***
ageA2 0.79233 0.07824 10.127 <2e-16 ***
ageA3 0.97366 0.07714 12.622 <2e-16 ***
ageA4 1.18034 0.07612 15.507 <2e-16 ***
ageA5 0.67209 0.07905 8.502 <2e-16 ***
ageA6 -0.53863 0.09439 -5.706 1.16e-08 ***
sexM 0.46494 0.04368 10.645 <2e-16 ***
Table A 4: Level 1 model output with MDM decile, age, sex and URDG as independent
variables
Significance
Coefficient Estimate Std. Error Z value Pr(>|z|) code
Intercept) -1.20367 0.07714 -15.604 <2e-16 ***
mdmdecM2 0.21272 0.07939 2.679 0.00737 **
mdmdecM3 0.43138 0.07594 5.680 1.35e-08 ***
mdmdecM4 0.46758 0.07539 6.202 5.57e-10 ***
mdmdecM5 0.45862 0.07588 6.044 1.51e-09 ***
mdmdecM6 0.83590 0.07099 11.776 <2e-16 ***
mdmdecM7 0.83913 0.07079 11.855 <2e-16 ***
mdmdecM8 1.01032 0.06896 14.650 <2e-16 ***
mdmdecM9 1.24538 0.06691 18.614 <2e-16 ***
mdmdecM10 1.91408 0.06339 30.196 <2e-16 ***
ageA2 0.80782 0.04698 17.195 <2e-16 ***
ageA3 0.87007 0.04647 18.722 <2e-16 ***
ageA4 1.00702 0.04556 22.103 <2e-16 ***
ageA5 0.56376 0.04889 11.532 <2e-16 ***
ageA6 -0.64881 0.06832 -9.496 <2e-16 ***
sexM 0.43388 0.02454 17.681 <2e-16 ***
urdgU 1.50940 0.03349 45.064 <2e-16 ***
3
Final Report
variables (where MDM decile and URDG are modelled with interaction)
Std. Significance
Coefficient Estimate Error Z Value Pr(>|Z|) Code
(Intercept) -2.26149 0.28385 -7.967 1.62e-15 ***
mdmdecM2 1.57541 0.30226 5.212 1.87e-07 ***
mdmdecM3 1.75154 0.29841 5.870 4.37e-09 ***
mdmdecM4 1.93064 0.29525 6.539 6.20e-11 ***
mdmdecM5 2.22592 0.29130 7.641 2.15e-14 ***
mdmdecM6 2.35219 0.29008 8.109 5.12e-16 ***
mdmdecM7 2.34555 0.29040 8.077 6.63e-16 ***
mdmdecM8 2.26973 0.29142 7.788 6.79e-15 ***
mdmdecM9 0.60789 0.34164 1.779 0.0752 .
mdmdecM10 -1.57153 0.73763 -2.131 0.0331 *
urdgU 2.66163 0.28741 9.261 <2e-16 ***
ageA2 0.81135 0.04637 17.497 <2e-16 ***
ageA3 0.86772 0.04591 18.902 <2e-16 ***
ageA4 1.00162 0.04501 22.254 <2e-16 ***
ageA5 0.56139 0.04835 11.611 <2e-16 ***
ageA6 -0.65403 0.06792 -9.629 <2e-16 ***
sexM 0.43230 0.02416 17.897 <2e-16 ***
mdmdecM2:urdgU -1.52785 0.31394 -4.867 1.13e-06 ***
mdmdecM3:urdgU -1.47565 0.30911 -4.774 1.81e-06 ***
mdmdecM4:urdgU -1.66892 0.30614 -5.451 5.00e-08 ***
mdmdecM5:urdgU -2.13537 0.30347 -7.037 1.97e-12 ***
mdmdecM6:urdgU -1.75458 0.29985 -5.851 4.87e-09 ***
mdmdecM7:urdgU -1.73094 0.30001 -5.770 7.95e-09 ***
mdmdecM8:urdgU -1.39892 0.30016 -4.661 3.15e-06 ***
mdmdecM9:urdgU 0.67255 0.34828 1.931 0.0535 .
mdmdecM10:urdgU 3.50347 0.74040 4.732 2.22e-06 ***
4
Final Report
variables (where MDM decile*URDG and age*sex are modelled with interaction)
Std. Significance
Coefficient Estimate Error Z value Pr(>|z|) code
(Intercept) -2.50359 0.28934 -8.653 <2e-16 ***
mdmdecM2 1.57537 0.30226 5.212 1.87e-07 ***
mdmdecM3 1.75136 0.29841 5.869 4.38e-09 ***
mdmdecM4 1.93054 0.29525 6.539 6.21e-11 ***
mdmdecM5 2.22586 0.29130 7.641 2.15e-14 ***
mdmdecM6 2.35200 0.29008 8.108 5.14e-16 ***
mdmdecM7 2.34539 0.29040 8.076 6.67e-16 ***
mdmdecM8 2.26958 0.29142 7.788 6.81e-15 ***
mdmdecM9 0.60784 0.34164 1.779 0.075208 .
mdmdecM10 -1.57174 0.73763 -2.131 0.033106 *
urdgU 2.66145 0.28741 9.260 <2e-16 ***
ageA2 1.05568 0.08081 13.063 <2e-16 ***
ageA3 1.09170 0.08032 13.591 <2e-16 ***
ageA4 1.31480 0.07825 16.803 <2e-16 ***
ageA5 0.79623 0.08376 9.506 <2e-16 ***
ageA6 -0.29110 0.10857 -2.681 0.007337 **
sexM 0.80430 0.08371 9.609 <2e-16 ***
mdmdecM2:urdgU -1.52786 0.31394 -4.867 1.13e-06 ***
mdmdecM3:urdgU -1.47551 0.30911 -4.773 1.81e-06 ***
mdmdecM4:urdgU -1.66889 0.30614 -5.451 5.00e-08 ***
mdmdecM5:urdgU -2.13537 0.30347 -7.037 1.97e-12 ***
mdmdecM6:urdgU -1.75446 0.29985 -5.851 4.89e-09 ***
mdmdecM7:urdgU -1.73079 0.30001 -5.769 7.97e-09 ***
mdmdecM8:urdgU -1.39876 0.30016 -4.660 3.16e-06 ***
mdmdecM9:urdgU 0.67259 0.34828 1.931 0.053463 .
mdmdecM10:urdgU 3.50367 0.74040 4.732 2.22e-06 ***
ageA2:sexM -0.37536 0.09884 -3.798 0.000146 ***
ageA3:sexM -0.34204 0.09802 -3.490 0.000484 ***
ageA4:sexM -0.49142 0.09591 -5.124 2.99e-07 ***
ageA5:sexM -0.35973 0.10276 -3.501 0.000464 ***
ageA6:sexM -0.58162 0.14007 -4.152 3.29e-05 ***
5
Final Report
Table A 7: Significance of coefficients for pedestrian casualties in urban areas by age and sex for the Level 2 Model
Free
(Intercep MDM MMD MDM Population Road Traffic School school Sunshine Annual
Age Group Sex t) Score Crime Proximity density length proxy places meals hours Rainfall
Child Both 0.001(+) 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) 0.001(+) 0.001(+) 0.001(+) 0.001(-) 0.01(-)
casualties Female NS 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) 0.001(+) 0.001(+) 0.001(+) 0.001(-) 0.01(-)
(0-14) Male 0.05(+) 0.001(+) 0.001(+) 0.001(-) 0.001(+) NS 0.001(+) 0.001(+) NS 0.001(-) 0.05(-)
Child Both NS 0.001(+) 0.001(+) 0.01(-) 0.05(+) NS 0.001(+) 0.001(+) NS 0.001(-) 0.001(-)
casualties Female NS 0.001(+) NS 0.01(-) 0.01(+) 0.05(+) 0.05(+) 0.01(+) 0.001(+) 0.05(-) NS
(0-4) Male NS 0.001(+) 0.001(+) NS NS NS 0.05(+) 0.01(+) NS 0.01(-) 0.001(-)
Child Both NS 0.001(+) 0.001(+) NS 0.001(+) 0.05(-) 0.001(+) 0.001(+) NS 0.001(-) 0.01(-)
casualties Female NS 0.001(+) 0.001(+) 0.05(+) NS NS NS 0.001(+) NS 0.001(-) 0.01(-)
(5-7) Male NS 0.001(+) 0.05(+) NS 0.001(+) NS 0.001(+) 0.001(+) NS 0.001(-) NS
Child Both 0.05(-) 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) 0.001(+) 0.001(+) 0.01(+) NS NS
casualties Female 0.001(-) 0.001(+) 0.05(+) 0.001(-) NS 0.05(+) 0.01(+) 0.001(+) 0.01(+) NS NS
(8-11) Male NS 0.001(+) 0.001(+) 0.01(-) 0.001(+) 0.05(+) 0.001(+) 0.001(+) NS NS NS
Child Both NS 0.001(+) 0.001(+) 0.001(-) NS 0.001(+) 0.001(+) 0.001(+) NS 0.05(-) 0.01(-)
casualties Female NS 0.05(+) 0.001(+) 0.001(-) NS 0.001(+) 0.001(+) 0.001(+) NS 0.001(-) 0.001(-)
(12-15) Male NS 0.001(+) 0.001(+) 0.001(-) NS NS 0.001(+) 0.001(+) 0.001(-) NS NS
Young adult Both 0.001(-) 0.001(+) 0.001(+) 0.001(-) 0.01(+) 0.01(+) 0.001(+) 0.001(+) 0.001(-) 0.01(+) 0.001(+)
casualties Female 0.001(-) 0.001(+) 0.001(+) 0.001(-) 0.01(+) NS 0.001(+) 0.001(+) 0.01(-) 0.01(+) 0.001(+)
(16-24) Male 0.001(-) 0.001(+) 0.001(+) 0.001(-) NS 0.05(+) 0.001(+) 0.001(+) NS 0.01(+) 0.01(+)
Adult Both NS 0.001(+) 0.001(+) 0.001(-) NS 0.001(+) 0.001(+) 0.001(+) 0.001(-) NS NS
casualties Female 0.001(-) 0.001(+) 0.001(+) 0.001(-) NS 0.01(+) 0.001(+) 0.001(+) 0.001(-) NS NS
(25 and over) Male NS 0.001(+) 0.001(+) 0.001(-) NS 0.001(+) 0.001(+) 0.001(+) NS 0.05(-) 0.001(-)
Both 0.001(+) 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) 0.001(+) 0.001(+) NS NS NS
All Casualties Female NS 0.001(+) 0.001(+) 0.001(-) 0.01(+) 0.001(+) 0.001(+) 0.001(+) NS NS NS
Male 0.05(+) 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) 0.001(+) 0.001(+) NS NS NS
6
Final Report
Table A 8: Significance of coefficients for pedestrian casualties in rural areas by age and sex for the Level 2 Model
Free
MDM MMD MDM Population Road Traffic School school Sunshine Annual
Age Group Sex (Intercept) Score Crime Proximity density length proxy places meals hours Rainfall
Child Both 0.01(-) 0.001(+) 0.01(+) 0.001(-) NS 0.05(-) 0.001(+) NS 0.001(+) 0.01(+) 0.001(+)
casualties (0- Female 0.01(-) 0.01(+) 0.05(+) 0.001(-) NS 0.01(-) 0.001(+) 0.05(+) 0.05(+) NS NS
14) Male 0.001(-) 0.001(+) 0.05(+) 0.001(-) NS NS 0.01(+) NS 0.01(+) 0.01(+) 0.001(+)
Child Both NS 0.001(+) NS 0.001(-) NS NS NS NS NS 0.01(-) NS W
casualties (0- Female NS 0.001(+) 0.05(+) 0.01(-) NS 0.05(+) 0.05(-) NS 0.001(-) 0.01(-) NS W
4) Male NS 0.001(+) NS 0.001(-) 0.05(-) NS 0.01(+) NS NS NS NS W
Child Both 0.01(-) 0.001(+) NS NS 0.05(+) 0.001(-) NS NS NS NS NS W
casualties (5- Female NS 0.001(+) NS NS NS 0.001(-) NS 0.001(+) NS NS NS W
7) Male 0.001(-) NS NS NS 0.05(+) NS NS NS NS 0.01(+) NS W
Child Both 0.01(-) NS 0.001(+) 0.05(-) NS 0.01(-) NS 0.001(+) 0.001(+) NS NS W
casualties (8- Female 0.01(-) NS 0.01(+) NS 0.05(-) 0.001(-) NS 0.001(+) 0.05(+) 0.05(+) NS W
11) Male 0.05(-) NS 0.001(+) 0.01(-) NS NS NS 0.001(+) 0.01(+) NS 0.05(+) W
Child Both 0.001(-) 0.001(+) NS 0.001(-) 0.01(-) NS 0.001(+) NS NS 0.01(+) 0.001(+)
casualties Female 0.001(-) 0.01(+) NS 0.001(-) 0.05(-) NS 0.001(+) NS NS NS 0.05(+)
(12-15) Male 0.001(-) 0.05(+) NS 0.01(-) NS NS 0.001(+) NS NS 0.05(+) 0.001(+) W
Young adult Both 0.001(-) 0.001(+) NS 0.05(-) NS NS 0.001(+) NS NS 0.001(+) 0.001(+)
casualties Female 0.001(-) NS NS NS 0.05(-) 0.001(-) 0.001(+) 0.001(+) NS 0.05(+) 0.01(+)
(16-24) Male 0.001(-) 0.001(+) NS 0.001(-) NS NS 0.001(+) NS NS 0.001(+) 0.001(+)
Adult Both NS 0.001(+) 0.001(+) 0.001(-) 0.01(-) NS 0.001(+) 0.01(+) NS NS 0.05(+)
casualties (25 Female 0.01(-) 0.001(+) 0.001(+) NS NS 0.001(-) 0.001(+) NS 0.001(+) NS 0.01(+)
and over) Male NS 0.001(+) 0.001(+) 0.001(-) 0.001(-) 0.01(+) 0.001(+) 0.001(+) 0.001(+) NS NS
Both NS 0.001(+) 0.001(+) 0.001(-) 0.01(-) NS 0.001(+) 0.01(+) NS 0.01(+) 0.001(+)
All Casualties Female 0.05(-) 0.01(+) 0.001(+) 0.01(-) NS 0.001(-) 0.001(+) 0.01(+) 0.010(+) NS 0.01(+)
Male 0.01(-) 0.001(+) 0.001(+) 0.001(-) 0.01(-) 0.01(+) 0.001(+) NS NS 0.001(+) 0.001(+)
A “W” in the right hand column means that the model did not fully converge and the outputs may not be robust
7
Final Report
Table A 9: Significance of coefficients for pedestrian casualties in urban areas by age and sex for the Level 3 Model
MDM School Sunshine Annual
Age Group Sex (Intercept) MDM Score MMD Crime Proximity Traffic proxy places hours Rainfall
Child Both 0.001(+) 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) 0.001(-) 0.001(-)
casualties (0- Female 0.001(+) 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) 0.001(-) 0.001(-)
14) Male 0.001(+) 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) 0.001(-) 0.001(-)
Child Both NS 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) 0.001(-) 0.001(-)
casualties (0- Female NS 0.001(+) NS 0.001(-) 0.01(+) 0.001(+) 0.01(-) NS
4) Male NS 0.001(+) 0.001(+) 0.01(-) 0.01(+) 0.01(+) 0.01(-) 0.001(-)
Child Both 0.01(+) 0.001(+) 0.01(+) 0.001(-) NS 0.001(+) 0.001(-) 0.001(-)
casualties (5- Female NS 0.001(+) 0.01(+) NS NS 0.001(+) 0.001(-) 0.01(-)
7) Male NS 0.001(+) NS 0.001(-) 0.05(+) 0.001(+) 0.001(-) 0.05(-)
Child Both NS 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) NS NS
casualties (8- Female 0.001(-) 0.001(+) 0.01(+) 0.001(-) 0.001(+) 0.001(+) NS NS
11) Male NS 0.001(+) 0.01(+) 0.001(-) 0.001(+) 0.001(+) 0.01(-) 0.05(-)
Child Both NS 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) 0.05(-) 0.01(-)
casualties Female NS 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) 0.001(-) 0.001(-)
(12-15) Male 0.05(-) 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) NS NS
Young adult Both 0.001(-) 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) 0.01(+) 0.01(+)
casualties Female 0.001(-) 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.01(+) 0.01(+) 0.05(+)
(16-24) Male 0.001(-) 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) NS 0.01(+)
Adult Both NS 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) NS 0.01(-)
casualties (25 Female 0.001(-) 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) NS NS
and over) Male NS 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) NS 0.001(-)
Both 0.001(+) 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) 0.001(-) 0.01(-)
All Casualties Female NS 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) NS NS
Male 0.001(+) 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) 0.01(-) 0.01(-)
8
Final Report
Table A 10: Significance of coefficients for pedestrian casualties in rural areas by age and sex for the Level 3 Model
MDM Sunshine Annual
Age Group Sex (Intercept) MDM Score MMD Crime Proximity Traffic proxy School places hours Rainfall
Child Both 0.01(-) 0.001(+) 0.01(+) 0.001(-) 0.001(+) 0.05(+) 0.01(+) 0.001(+)
casualties (0- Female 0.01(-) 0.001(+) 0.001(+) 0.001(-) 0.05(+) 0.01(+) NS NS
14) Male 0.01(-) 0.001(+) 0.001(+) 0.001(-) 0.05(+) 0.01(+) NS NS
Child Both NS 0.001(+) NS 0.001(-) NS NS 0.01(-) NS W
casualties (0- Female NS 0.001(+) NS 0.001(-) 0.05(-) NS 0.01(-) NS W
4) Male NS 0.001(+) NS 0.001(-) 0.01(+) NS NS NS W
Child Both 0.01(-) 0.001(+) NS 0.001(-) NS NS NS NS W
casualties (5- Female NS 0.001(+) NS 0.001(-) NS 0.01(+) NS NS W
7) Male 0.001(-) 0.05(+) NS NS NS NS 0.01(+) NS W
Child Both 0.01(-) 0.001(+) 0.001(+) 0.05(-) NS 0.001(+) NS NS W
casualties (8- Female 0.01(-) NS 0.001(+) 0.001(-) NS 0.01(+) 0.05(+) NS W
11) Male 0.05(-) 0.001(+) 0.001(+) 0.001(-) NS 0.001(+) NS 0.05(+) W
Child Both 0.001(-) 0.01(+) NS 0.001(-) 0.001(+) NS 0.01(+) 0.001(+)
casualties Female 0.001(-) 0.05(+) NS 0.001(-) 0.001(+) NS NS 0.05(+)
(12-15) Male 0.001(-) 0.05(+) NS 0.01(-) 0.001(+) NS 0.05(+) 0.001(+) W
Young adult Both 0.001(-) 0.001(+) NS 0.01(-) 0.001(+) NS 0.001(+) 0.001(+)
casualties Female 0.001(-) NS NS NS 0.01(+) 0.001(+) 0.01(+) 0.001(+)
(16-24) Male 0.001(-) 0.001(+) NS 0.001(-) 0.001(+) 0.01(-) 0.001(+) 0.001(+)
Adult Both NS 0.001(+) 0.001(+) 0.001(-) 0.001(+) NS NS 0.05(+)
casualties (25 Female 0.01(-) 0.001(+) 0.001(+) 0.001(-) 0.001(+) NS NS 0.01(+)
and over) Male NS 0.001(+) 0.001(+) NS 0.001(+) NS NS NS
Both NS 0.001(+) 0.001(+) 0.001(-) 0.001(+) NS 0.01(+) 0.001(+)
All Casualties Female 0.05(-) 0.01(+) 0.001(+) 0.001(-) 0.001(+) 0.01(+) NS 0.001(+)
Male 0.01(-) 0.001(+) 0.001(+) 0.001(-) 0.001(+) NS 0.01(+) 0.001(+)
A “W” in the right hand column means that the model did not fully converge and the outputs may not be robust
9
Final Report
Table A 11: Significance of coefficients for pedestrian casualties in all areas (urban and rural combined) by age and sex for the Level 3 Model
MDM School Sunshine Annual URDG

Age Group Sex (Intercept) MDM Score MMD Crime Proximity Traffic proxy places hours Rainfall classification
Child Both 0.001(+) 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) 0.001(-) 0.05(-) 0.05(+)
casualties (0- Female NS 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) 0.001(-) 0.001(-) NS
14) Male NS 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) 0.05(-) NS 0.05(+)
Child Both NS 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) 0.001(-) 0.01(-) NS
casualties (0- Female NS 0.001(+) NS 0.001(-) 0.05(+) 0.001(+) 0.001(-) NS NS
4) Male NS 0.001(+) 0.001(+) 0.001(-) 0.01(+) 0.001(+) 0.01(-) 0.001(-) 0.05(+)
Child Both NS 0.001(+) 0.01(+) 0.001(-) NS 0.001(+) 0.001(-) 0.001(-) 0.001(+)
casualties (5- Female NS 0.001(+) 0.01(+) NS NS 0.001(+) 0.001(-) 0.01(-) 0.001(+)
7) Male NS 0.001(+) NS 0.001(-) 0.05(+) 0.001(+) 0.01(-) NS NS
Child Both 0.05(-) 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) NS NS NS
casualties (8- Female 0.001(-) 0.001(+) 0.01(+) 0.001(-) 0.001(+) 0.001(+) 0.05(+) NS NS
11) Male NS 0.001(+) 0.001(+) 0.001(-) 0.01(+) 0.001(+) 0.01(-) NS 0.01(+)
Child Both 0.05(-) 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) NS NS NS
casualties Female NS 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) 0.05(-) 0.05(-) NS
(12-15) Male 0.001(-) 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) NS NS NS
Young adult Both 0.001(-) 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) 0.001(+) 0.001(+) 0.001(-)
casualties Female 0.001(-) 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) 0.01(+) 0.01(+) NS
(16-24) Male 0.001(-) 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) 0.001(+) 0.001(+) 0.001(-)
Adult Both NS 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) NS NS 0.001(-)
casualties (25 Female 0.001(-) 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) NS NS 0.001(-)
and over) Male NS 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) NS 0.05(-) 0.001(-)
Both 0.001(+) 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) NS NS 0.001(-)
All
Female NS 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) NS NS 0.001(-)
Casualties
Male 0.001(+) 0.001(+) 0.001(+) 0.001(-) 0.001(+) 0.001(+) NS NS 0.001(-)
10
Final Report
Table A 12: 50 poorest performing SOAs in terms of child pedestrian casualties (Level 3 Model)
Local Government Observed Fitted Residual

SOA Code District SOA Name (O) (F) (O-F)
95MM12S3 Derry Crevagh_3 16 3.3 12.7
95GG01S2 Belfast Andersonstown_2 20 7.6 12.4
95GG44S2 Belfast The Mount_2 24 11.6 12.4
95GG18S2 Belfast Clonard_2 23 10.8 12.2
95BB06W1 Ards Central_Ards 24 12.8 11.2
95SS06S3 Lisburn Collin Glen_3 17 7.0 10.0
95GG17S2 Belfast Cliftonville_2 15 5.5 9.5
95GG22S1 Belfast Falls Park_1 21 11.5 9.5
95II05S2 Castlereagh Carryduff East_2 12 2.7 9.3
95NN07S2 Down Cathedral_2 15 6.3 8.7
95XX05W1 North Down Bangor Castle 17 8.5 8.5
95GG29S2 Belfast Island_2 16 7.8 8.2
95KK04W1 Cookstown Gortalowry 11 2.9 8.1
95SS06S1 Lisburn Collin Glen_1 12 4.0 8.0
95GG25S1 Belfast Glen Road_1 17 9.1 7.9
95DD08W1 Ballymena Castle Demesne 16 8.2 7.8
95WW18W1 Newtownabbey Hightown 11 3.8 7.2
95GG28S2 Belfast Highfield_2 12 4.8 7.2
95GG06S2 Belfast Ballysillan_2 11 3.9 7.1
95GG05S1 Belfast Ballynafeigh_1 13 6.1 6.9
95VV18S1 Newry and Mourne Kilkeel Central_1 9 2.2 6.8
95ZZ14S2 Strabane South_2 9 2.4 6.6
95GG31S3 Belfast Ladybrook_3 12 5.5 6.5
95GG27S2 Belfast Glencolin_2 12 5.6 6.4
95GG48S1 Belfast Whiterock_1 15 8.7 6.3
95GG25S3 Belfast Glen Road_3 11 4.7 6.3
95GG44S1 Belfast The Mount_1 22 16.2 5.8
95NN01S1 Down Ardglass_1 8 2.2 5.8
95BB19S1 Ards Movilla_1 8 2.4 5.6
95QQ07W1 Larne Central_Larne 12 6.4 5.6
95AA19W1 Antrim Toome 7 1.7 5.3
95GG16S2 Belfast Chichester Park_2 11 5.7 5.3
95II21W1 Castlereagh Tullycarnet 9 3.8 5.2
95BB04S1 Ards Bradshaw's Brae_1 9 3.8 5.2
95GG02S2 Belfast Ardoyne_2 14 8.9 5.1
95MM12S1 Derry Crevagh_1 8 2.9 5.1
95SS07S1 Lisburn Derryaghy_1 8 2.9 5.1
95DD21W1 Ballymena Park 10 5.0 5.0
95MM25S1 Derry Shantallow West_1 11 6.0 5.0
95II13W1 Castlereagh Graham's Bridge 10 5.1 4.9
95GG38S3 Belfast Rosetta_3 8 3.2 4.8
95GG02S3 Belfast Ardoyne_3 18 13.2 4.8
95XX06S1 North Down Bloomfield_1_NorthDown 8 3.2 4.8
95WW20S2 Newtownabbey Mallusk_2 8 3.2 4.8
95GG31S1 Belfast Ladybrook_1 8 3.2 4.8
95XX22W1 North Down Rathgael 7 2.3 4.7
95BB07S1 Ards Comber East_1 8 3.3 4.7
95GG47S3 Belfast Water Works_3 19 14.4 4.6
95BB13S2 Ards Gregstown_2 7 2.5 4.5
95HH09W1 Carrickfergus Killycrot 11 6.5 4.5
11
Final Report
Table A 13: 50 best performing SOAs in terms of child pedestrian casualties (Level 3 Model)
Local Government Observed Fitted Residual

SOA Code District SOA Name (O) (F) (O-F)
95PP09W1 Fermanagh Devenish 2 13.2 -11.2
95GG47S1 Belfast Water Works_1 20 32.0 -12.0
95GG40S2 Belfast Shankill_2 10 23.1 -13.1
95GG19S2 Belfast Crumlin_2_Belfast 6 19.6 -13.6
95ZZ02W1 Strabane Ballycolman 2 5.5 -3.5
95CC01W1 Armagh Abbey Park 1 4.5 -3.5
95GG40S1 Belfast Shankill_1 11 14.5 -3.5
95LL16S2 Craigavon Kernan_2 0 3.7 -3.7
95SS30S1 Lisburn Wallace Park_1 0 3.7 -3.7
95MM05W1 Derry Brandywell 10 13.7 -3.7
95VV15S1 Newry and Mourne Drumgullion_1 3 6.8 -3.8
95AA02W1 Antrim Balloo 0 3.8 -3.8
95GG28S3 Belfast Highfield_3 2 5.9 -3.9
95MM12S2 Derry Crevagh_2 3 6.9 -3.9
95JJ06W1 Coleraine Churchland 2 6.0 -4.0
95MM14S2 Derry Ebrington_2 8 12.0 -4.0
95GG12S1 Belfast Botanic_1 6 10.2 -4.2
95GG21S3 Belfast Falls_3 20 24.2 -4.2
95GG10S1 Belfast Blackstaff_1 6 10.2 -4.2
95MM14S1 Derry Ebrington_1 0 4.3 -4.3
95SS29S1 Lisburn Twinbrook_1 2 6.3 -4.3
95MM06S2 Derry Carn Hill_2 3 7.3 -4.3
95AA15S1 Antrim Springfarm_1 0 4.4 -4.4
95NN18S1 Down Quoile_1 2 6.4 -4.4
95MM23W1 Derry Rosemount 4 8.4 -4.4
95GG48S2 Belfast Whiterock_2 6 10.4 -4.4
95MM06S1 Derry Carn Hill_1 2 6.5 -4.5
95ZZ16S2 Strabane West_2 0 4.6 -4.6
95GG07S3 Belfast Beechmount_3 3 7.7 -4.7
95GG51S1 Belfast Woodvale_1 2 6.7 -4.7
95RR03W1 Limavady Coolessan 4 8.7 -4.7
95LL09S2 Craigavon Court_2 4 8.8 -4.8
95GG19S1 Belfast Crumlin_1_Belfast 9 13.8 -4.8
95GG39S2 Belfast Shaftesbury_2 15 19.8 -4.8
95GG04S1 Belfast Ballymacarrett_1 5 10.0 -5.0
95GG35S2 Belfast New Lodge_2 12 17.2 -5.2
95LL13S2 Craigavon Drumgor_2 5 10.3 -5.3
95GG35S1 Belfast New Lodge_1 16 21.3 -5.3
95MM08S1 Derry Clondermot_1 1 6.5 -5.5
95GG29S1 Belfast Island_1 5 10.9 -5.9
95MM27S2 Derry Strand_2_Derry 1 7.1 -6.1
95GG04S3 Belfast Ballymacarrett_3 8 14.8 -6.8
95VV28S2 Newry and Mourne St Patrick's_2 1 8.0 -7.0
95MM27S1 Derry Strand_1_Derry 11 18.7 -7.7
95GG20S1 Belfast Duncairn_1 13 21.3 -8.3
95GG39S3 Belfast Shaftesbury_3 4 13.1 -9.1
12
Final Report
Table A 14: 50 SOAs with highest casualty increases in most recent 5 year period (2004 – 08)
Local
Government Observed Expected Observed
SOA Code District SOA Name 1999 - 03 2004 - 08 2004 - 08 Increase
95XX05W1 North Down Bangor Castle 5 2.95 12 9.05
95PP20W1 Fermanagh Portora 4 2.36 8 5.64
95GG05S3 Belfast Ballynafeigh_3 1 0.59 6 5.41
95GG35S3 Belfast New Lodge_3 7 4.13 9 4.87
95PP17W1 Fermanagh Lisnaskea 2 1.18 6 4.82
95KK04W1 Cookstown Gortalowry 4 2.36 7 4.64
95GG21S2 Belfast Falls_2 6 3.54 8 4.46
95GG17S1 Belfast Cliftonville_1 3 1.77 6 4.23
95YY01W1 Omagh Beragh 0 0.00 4 4.00
95NN07S2 Down Cathedral_2 7 4.13 8 3.87
95GG37S2 Belfast Ravenhill_2 2 1.18 5 3.82
95JJ16W1 Coleraine Portstewart 2 1.18 5 3.82
95WW14W1 Newtownabbey Dunanney 2 1.18 5 3.82
95DD21W1 Ballymena Park 4 2.36 6 3.64
95WW12W1 Newtownabbey Coole 4 2.36 6 3.64
95MM09S1 Derry Claudy_1 1 0.59 4 3.41
95MM13S2 Derry Culmore_2 1 0.59 4 3.41
95DD08W1 Ballymena Castle Demesne 8 4.72 8 3.28
95BB19S1 Ards Movilla_1 3 1.77 5 3.23
95GG16S1 Belfast Chichester Park_1 5 2.95 6 3.05
95GG07S3 Belfast Beechmount_3 0 0.00 3 3.00
95II15S2 Castlereagh Knockbracken_2 0 0.00 3 3.00
95II20W1 Castlereagh Newtownbreda 0 0.00 3 3.00
95LL02S1 Craigavon Annagh_1 0 0.00 3 3.00
95MM12S2 Derry Crevagh_2 0 0.00 3 3.00
95MM19S2 Derry Kilfennan_2 0 0.00 3 3.00
95NN19S1 Down Saintfield_1 0 0.00 3 3.00
95WW02S1 Newtownabbey Ballyclare North_1 0 0.00 3 3.00
95YY15S1 Omagh Lisanelly_1 0 0.00 3 3.00
95JJ22W1 Coleraine Waterside 2 1.18 4 2.82
95KK16W1 Cookstown Tullagh 2 1.18 4 2.82
95NN17W1 Down Murlough 2 1.18 4 2.82
95SS21S1 Lisburn Maghaberry_1 2 1.18 4 2.82
95GG50S3 Belfast Woodstock_3 4 2.36 5 2.64
95LL08S2 Craigavon Corcrain_2 4 2.36 5 2.64
95SS18S2 Lisburn Lagan Valley_2 4 2.36 5 2.64
95SS28W1 Lisburn Tonagh 4 2.36 5 2.64
95GG31S3 Belfast Ladybrook_3 6 3.54 6 2.46
95II05S2 Castlereagh Carryduff East_2 6 3.54 6 2.46
95GG32S1 Belfast Legoniel_1 1 0.59 3 2.41
95LL07W1 Craigavon Church 1 0.59 3 2.41
95LL09S2 Craigavon Court_2 1 0.59 3 2.41
95MM01S2 Derry Altnagelvin_2 1 0.59 3 2.41
95MM20S1 Derry Lisnagelvin_1 1 0.59 3 2.41
95OO21W1 Dungannon Mullaghmore 1 0.59 3 2.41
Florence Court
95PP12W1 Fermanagh and Kinawley 1 0.59 3 2.41
95RR03W1 Limavady Coolessan 1 0.59 3 2.41
95WW08S1 Newtownabbey Burnthill_1 1 0.59 3 2.41
95WW22S1 Newtownabbey Mossley_1 1 0.59 3 2.41
95GG39S2 Belfast Shaftesbury_2 8 4.72 7 2.28
13
Final Report
Table A 15: 50 SOAs with highest casualty reductions in most recent 5 year period (2004 – 08)
Local
Government Observed Expected Observed
SOA Code District SOA Name 1999 - 03 2004 - 08 2004 - 08 Reduction
95SS06S1 Lisburn Collin Glen_1 11 6.50 1 -5.50
95GG40S1 Belfast Shankill_1 10 5.90 1 -4.90
95GG04S2 Belfast Ballymacarrett_2 8 4.72 0 -4.72
95GG21S1 Belfast Falls_1 13 7.68 3 -4.68
95BB06W1 Ards Central_Ards 18 10.63 6 -4.63
95ZZ06W1 Strabane East 11 6.50 2 -4.50
95GG07S1 Belfast Beechmount_1 9 5.31 1 -4.31
95QQ07W1 Larne Central_Larne 10 5.90 2 -3.90
95GG47S1 Belfast Water Works_1 15 8.86 5 -3.86
95GG22S3 Belfast Falls Park_3 8 4.72 1 -3.72
95MM10S1 Derry Creggan Central_1 8 4.72 1 -3.72
95HH11W1 Carrickfergus Love Lane 6 3.54 0 -3.54
95II16W1 Castlereagh Lisnasharragh 6 3.54 0 -3.54
95ZZ12W1 Strabane Sion Mills 6 3.54 0 -3.54
95GG16S2 Belfast Chichester Park_2 9 5.31 2 -3.31
95MM25S1 Derry Shantallow West_1 9 5.31 2 -3.31
95VV10S2 Newry and Mourne Daisy hill_2 9 5.31 2 -3.31
95GG31S1 Belfast Ladybrook_1 7 4.13 1 -3.13
95GG38S3 Belfast Rosetta_3 7 4.13 1 -3.13
95GG17S3 Belfast Cliftonville_3 5 2.95 0 -2.95
95GG27S3 Belfast Glencolin_3 5 2.95 0 -2.95
95GG43S3 Belfast Sydenham_3 5 2.95 0 -2.95
95HH02S1 Carrickfergus Bluefield_1 5 2.95 0 -2.95
95OO05W1 Dungannon Ballysaggart 5 2.95 0 -2.95
95SS07S2 Lisburn Derryaghy_2 5 2.95 0 -2.95
95SS26S1 Lisburn Poleglass_1 5 2.95 0 -2.95
95VV07S1 Newry and Mourne Clonallan_1 5 2.95 0 -2.95
95GG47S2 Belfast Water Works_2 10 5.90 3 -2.90
95GG40S2 Belfast Shankill_2 8 4.72 2 -2.72
95AA19W1 Antrim Toome 6 3.54 1 -2.54
95MM01S1 Derry Altnagelvin_1 6 3.54 1 -2.54
95MM20S2 Derry Lisnagelvin_2 6 3.54 1 -2.54
95SS13S1 Lisburn Hilden_1 6 3.54 1 -2.54
95GG48S1 Belfast Whiterock_1 11 6.50 4 -2.50
95GG18S2 Belfast Clonard_2 16 9.45 7 -2.45
95AA16W1 Antrim Steeple 4 2.36 0 -2.36
95CC07W1 Armagh Derrynoose 4 2.36 0 -2.36
95CC14W1 Armagh Laurelvale 4 2.36 0 -2.36
95EE07W1 Ballymoney Fairhill 4 2.36 0 -2.36
95FF06S1 Banbridge Dromore South_1 4 2.36 0 -2.36
95GG05S2 Belfast Ballynafeigh_2 4 2.36 0 -2.36
95GG23S3 Belfast Finaghy_3 4 2.36 0 -2.36
95GG32S3 Belfast Legoniel_3 4 2.36 0 -2.36
95HH07W1 Carrickfergus Gortalee 4 2.36 0 -2.36
95JJ07W1 Coleraine Cross Glebe 4 2.36 0 -2.36
95JJ20S2 Coleraine The Cuts_2 4 2.36 0 -2.36
95JJ21W1 Coleraine University 4 2.36 0 -2.36
95KK09W1 Cookstown Newbuildings 4 2.36 0 -2.36
95LL14S2 Craigavon Drumnamoe_2 4 2.36 0 -2.36
95MM21S1 Derry New Buildings_1 4 2.36 0 -2.36
14
Final Report
Figure A 1: Graph of observed and modelled pedestrian casualty totals with MDM decile as
independent variable for the Level 1 Model
Figure A 2: Graph of observed and modelled pedestrian casualty totals with MDM decile and
age as independent variables for the Level 1 Model
15
Final Report
Figure A 3: Graph of observed and modelled pedestrian casualty totals with MDM decile, age
and sex as independent variables for the Level 1 Model
Figure A 4: Graph of observed and modelled pedestrian casualty totals with MDM decile, age,
sex and URDG as independent variables for the Level 1 Model
16
Final Report
sex and URDG as independent variables (where MDM decile and URDG are modelled with
interaction) for the Level 1 Model
sex and URDG as independent variables (where MDM decile*URDG and age*sex are modelled
with interaction) for the Level 1 Model
17
Final Report
Appendix 4
Mapping
0
Final Report
Figure 4.1: Multiple Deprivation Measure by Decile
1
Final Report
Figure 4.2: Safety Performance by Decile
2
Final Report
Figure 4.3: 5 year change by decile

Deprivation Child Pedestrian Road Casualties

Încărcat de

Informații document

Descriere originală:

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Deprivation Child Pedestrian Road Casualties

Încărcat de

Drepturi de autor:

Formate disponibile

DOENI

Project No: 16216-01-1

Prepared by: Approved by:

Status: Final Issue no: 2 Date: 12 January 2010

1.1 Objectives of the study

determine whether or not there is evidence suggesting a link between child

- if there is evidence suggesting a link then a statistical model should be

1.2 Report structure

2.1 Purpose of the review

2.3 Literature search

2.3.2 In addition, a search was also conducted using Google.

2.3.3 The keywords adopted were:

Table 2.1: Combination of keyword searches

Keyword Searches undertaken using a logical "AND"

2.4 Literature review

The link between deprivation and child pedestrian casualty rates

Figure 2.1: Pedestrian accidents by deprivation: age 5-ll years, 1988-90

Figure 2.3: Table7 reproduced from Abdalla et al (1997)

2.4.14 The key recommendations for the study were:

Figure 2.4: Figure 9 reproduced from Grayling et al (2002)

2.4.17 The independent variables used to define area type included:

Table 2.2: Child pedestrian casualty rates by deprivation quintile reported to

2.4.29 The London report (Edwards et al (2006)) went on to look at:

2.4.32 Also included in the dataset were five local factors:

nature of the local environment – e.g. population per road junction

2.4.35 The results indicated:

2.4.38 They found:

The link between casualty residence and collision location

Figure 2.6: Table 2 reproduced from Abdalla et al (1997)

Distance between location of

Figure 2.7: Percentage of “home” child casualties (aged 5-14) by deprivation

Potential measures that can be taken to reduce child pedestrian casualty

2.4.54 Graham and Stephens (2008) suggest that:

Key points of relevance to the present study

- the link between deprivation and child pedestrian casualty rates

Abdalla et al (1997) and Chichester et al (1998),: Wales - NPHS (2004): Northern

- the link between casualty residence and collision location

- potential remedial measures

environment, road infrastructure, traffic calming, education, training and enforcement. It

3.1 Road collision casualty data

3.2 Population data

3.3 Deprivation data

3.4 Urban Rural Classification

3.5 Road infrastructure data

number of junctions / road length (junctions / km)

3.5.3 The junction density is not disaggregated by road class.

3.6 People in employment

3.7 Workplace population

3.8 School pupil data

3.9 Car ownership data

3.10 Household tenure data

3.11 Weather data

3.12 Creation of composite variables

traffic proxy = √(people in employment * workplace population)

road length per hectare = total road length / area

total road length

junctions per road km = number of junctions / total road length

junctions per hectare = number of junctions / area

traffic density per road km = traffic proxy / total road length

traffic density per road km = traffic proxy / number of junctions

resident population per road junction = total resident population / number of

employed residents per road junction = employed residents / number of junctions

road junctions. This variable was described by Graham et al (2005). It is a function of