Using Behavioural Measures to Evaluate Route Safety Schemes

Using behavioural

measures to evaluate

route safety schemes

Guidance for practitioners

By H A Ward, S Helman, N Christie, F P McKenna

CPR102 9

1. Context: The need for evaluation

Route safety schemes are complex interventions which may involve more than one agency. They use multiple methods to reduce the

number of casualties along sections of road by changing driver behaviour through elements of engineering, education and enforcement. Evaluation is the mechanism by which the effectiveness of the intervention is assessed in achieving its objectives. Route safety schemes bring new challenges in the form of developing the best way

to evaluate their effectiveness.

1 This note is a short guide for practitioners on using behavioural measures to evaluate route safety schemes.

http://www.adeptnet.org.uk/assets/userfiles/documents/000282.pdf

Above is a link to the location of the main report (which should be read in conjunction with this guide) to direct readers to a fuller description

of:

the top candidate behaviours which should be considered as relevant

to route safety schemes;

how to measure the impact of a scheme through changes in these behaviours;

the key issues related to the design of interventions including how to define what „success‟ looks like, how to develop a cost effective monitoring programme and how to analyse and interpret data so that a robust conclusion can be drawn from the evaluation; and

case studies of good practice from local authorities.

1 The first author drafted this guidance document based on the accompanying main report. The remaining authors are listed in the same order in which they appear on the main report. This order was determined randomly. All four authors worked together and contributed in equal measure to the content of the main report (on which this guidance document is based) and its recommendations.

In addition to road safety, schemes can have wider objectives, so a set of metrics is essential to a rounded and robust evaluation of the impacts across a number of objectives such as: a reduction in: o the number of casualties o the number of people exceeding the speed limit

o the number of people close following

o the number of people using mobile phones while driving

o the number of people overtaking o the number of people driving whilst impaired through alcohol, drugs or fatigue

o the number of people driving without licences, tax or insurance o social exclusion

an increase in: o seat belt wearing or child restraint use

o knowledge of road safety messages

o community involvement

Why monitor behaviour?

Evaluation using behavioural outcomes provides evidence to the road safety team regarding whether to:

continue the intervention or stop it;

expand or reduce it; or

modify it or leave alone.

Assessing the impact of a route scheme on the occurrence and severity of casualties is an important element in evaluating its effectiveness but numbers are often small and collision sites scattered along a route, rarely clustering at locations in numbers large enough to apply specific engineering measures. The traditional approach of assessing three years before data with three years after data may mean that, given the high variability inherent in small numbers, emerging problems are not spotted quickly enough to intervene to prevent further collisions and injury.

Measuring changes in road user behaviour gives the evidence needed to enable the detection of impacts of interventions over short timescales. This can include evidence that behaviours are changing as intended, or whether there are unintended consequences that lead to increases in casualties or unwanted road user behaviours.

While it is generally acknowledged that practitioners act with the best of intentions there has been a tendency for some in road safety to design interventions on the basis of „enthusiastic intuition‟ or the overall plausibility of the intervention, neither of which provide a guarantee of success. Evaluation is not an optional activity, separate from the key task of producing a successful intervention; it is not an alternative or competitor to more implementation. It is the core provider of evidence to aid learning from experience of what works and what doesn‟t for designing and implementing future schemes and for assessing cost effectiveness of current schemes. Evidence obtained through reliable evaluation can underpin the justification for route safety programmes and support bids for future funding.

Measurement of behavioural variables can help to establish the effects of route safety treatments over relatively short timescales

2. Which behaviours and attitudes should be measured and by what means?

It has been clear for some time that human factors and road user behaviour play an important role in accident involvement. The

key questions that need to be answered are:

1. Which human factors / behaviours play a role? 2. What is the evidence relating them to collision risk? 3. How can they be

measured? Human factors are known to play a role in collisions

In this guide, a concise set of answers is provided to these questions. Importantly, the aim is to include only human factors that are known (or very strongly suspected) to be related to accident risk. The intention is to focus practitioners‟ attentions (when designing interventions and when evaluating them) on behaviours that are most likely to lead to safety outcomes.

There are a number of behaviours that are either known or strongly suspected to be linked to collision risk. In this guide and the main report, such behaviours are listed as candidate behavioural metrics for

use in the evaluation of route safety schemes. The behaviours are split into a taxonomy based on whether they need to be measured by observing (on the route in question) vehicles or drivers/road users, or by talking directly to road users who use the route. The main report reviews the evidence linking these behaviours to collision risk and the methods

by which they can be measured are outlined.

Table 1 summarises these Table 1: Measurable behaviours involved factors associated with in collision risk or influencing the collision risk and the severity of the outcome relative ease with which Observable behaviour that can be the authors believe they measured from vehicles can be measured within a route safety context. Easier to measure Speed One important point to make is that the authors Speed profile are advocating the measurement of Following distance behavioural and attitudinal variables directly; this Overtaking allows practitioners to get a direct handle on the Gap acceptance causal mechanisms by which drivers and road More Lateral position users on a given route difficult to measure may be increasing (or decreasing) the risk of Observable behaviour that can be accidents. measured from drivers

Obviously not all these behaviours need be Easier to Unlicensed and uninsured driving measured for each scheme measure but the choice of Mobile phone use and other distractions appropriate ones will give an indication of how Seat belt use progress towards achieving a variety of More Fatigue and impairment objectives can be assessed difficult to (see Section 3 in main measure report and following in this Non-observable attitudes and behaviour guide). that can measured from drivers

The measurement of Attitudes regarding specific behaviours behaviours in the table are especially violations briefly summarised below with more information in Attitudes regarding interventions the main report. especially perceived likelihood of detection and collision risk

The evidence for the relationship between speed and collision involvement is well known and its measurement is routine for Local Authorities (the Road Safety Good Practice Guide gives details – see Section 5 in main report).

Other behaviours which can be measured from observing vehicles include speed and journey time over the route which are relevant to assessing measures to reduce speed along the route. By the use of instrumented cars and a range of drivers, profiles of deceleration and acceleration can be obtained which are useful in assessing effectiveness of programmes to reduce „harsh‟ driving styles that have been linked to collision involvement.

Close following is an aggressive driving style with links to collision involvement. Following distance is known as headway and is measured in distance or time from the front of (leading) Vehicle A to

the front of (following) Vehicle B. It is usually measured using loop detectors (some speed measuring devices have this facility) but can also be undertaken with a stop watch and a reference point at the Close following is known to be related side of the road to collision involvement

People who commit driving violations are more likely to be involved in collisions. Tendency to commit violations and self reported rates can be measured by using a questionnaire which can be found in the appendices of the main report. Red light running can be measured directly.

Injudicious overtaking is associated with collision risk as it exposes drivers to head-on collisions. It can be measured by observation and speed measurements.

It is commonly accepted that if drivers were to increase the size of gaps they accept when turning across or joining traffic, safety would improve. It is usually measured by filming vehicles and measuring the size of gaps accepted and rejected.

Lateral position on the carriageway may be relevant to determining risk of head on and run- off collisions, especially at bends. It can be measured on-site using filmed footage though this is labour intensive.

Hazard perception is related to collision risk and distraction – including mobile phone use reduces the ability of drivers to detect hazards. Video photography and still photography are potential techniques though the most straightforward might be road side observation of the numbers of drivers using phones, eating, and interacting with in-car technology. In-car distractions are an important risk factor for road accidents

Fatigue leads to sleep related collisions which often have serious consequences but this is the behaviour perhaps most difficult to identify and measure. Local road users could be contacted to complete a specially designed survey.

Impairment due to alcohol or drugs; there is an established relationship between alcohol and collisions but that for drugs is more complex. The only way to measure impairment is through police stopping and testing drivers.

There is a relationship between uninsured and unlicensed driving and collision involvement which is especially strong for young drivers from more deprived backgrounds. A first indication is to look at the „hit and run‟ record within STATS19 and then work with the police to identify drivers by using ANPR (automatic number plate recognition).

Seat belt wearing is well known to reduce injury and their use (and child restraints) can be measured by road side observation or by video. The measurement of attitudes and self reported behaviours requires the use of appropriate questionnaires or survey tools which often require considerable development work before practitioners can be sure that they are valid and reliable measures of the behaviours and attitudes they want to assess. Generally speaking, when designing a questionnaire

„from scratch‟, practitioners should seek help from those with expertise in the area (e.g. academics and behavioural scientists). However there are also existing questionnaire Seat belts have a proven tools that can be used by safety benefit, and their use can be observed from the practitioners who are focusing on roadside or from video some of the key behaviours listed in footage this guide.

Evidence suggests that road safety behaviour change campaigns need to focus on four key areas:

Drivers‟ attitudes to risky driving behaviour i.e. whether they view it as positive or negative.

Drivers‟ attitudes towards the penalties and enforcement of risky behaviour i.e. whether they view it as positive or negative.

Propensity to violate driving regulations.

Perception of the risk of being caught or having an accident.

Attitudes represent the way people „favour or disfavour‟ something and may predict the way people behave. For example if we strongly disfavour drinking and driving we are less likely to drink and drive.

There is some evidence of a relationship between attitudes, self reported driving behaviour, and collision involvement.

2 The Driver Attitude Questionnaire (DAQ) and the Driver Behaviour Questionnaire (DBQ)3 are two validated tools to measure attitudes and behaviours that are linked to collision involvement and both can be downloaded for use from the main report (scales available in the appendices, including scoring instructions).

The DAQ is especially relevant for driver attitudes to speeding, close following, dangerous overtaking, and drinking and driving. In addition in the main report the authors have suggested a set of questions on mobile phone use (although it should be noted that this scale is not yet validated).

The DBQ refers to three aspects of driving behaviour: errors, lapses, and violations. It is recommended that practitioners only use the violation questions as these are most likely to be related to collisions,

and are also the most plausibly targeted by a route safety intervention. It is recommended that these questions are related to the specific route in which you are interested.

Attitudes represent the

way people “favour or disfavour” something and may predict the way people behave.

There is some evidence of a relationship between attitudes, self reported driving behaviour, and

collision involvement

Surveys can be carried out with users of the route

2 Parker, D., Stradling, S. G., and Manstead, A. (1996). Modifying beliefs and attitudes to exceeding the speed limit: An intervention study based on the theory of planned behaviour. Journal of Applied Social Psychology, 26, 1–19. 3 Parker, D., Reason, J., Manstead, A. M., and Stradling, S. G. (1995) Driving errors, driving violations and accident involvement. Ergonomics, 38(5), 1036–1048.

Surveys for intervention exposure can collect specific information about what intervention measures the road users have experienced in terms of enforcement and education on the route. This can be gleaned by asking questions about whether they have been stopped by the police or detected by a camera on the route within a given time period, or whether they are aware of education measures such as signage or the occurrence of roadside education initiatives. These surveys give evaluators an idea of whether or not the intervention is achieving penetration of the target road user groups. If it can be shown that the people being exposed to the intervention are changing their behaviour more than those who have not been exposed, this makes it easier to ascribe these effects to the intervention.

Attitudes to enforcement. It can be argued that interventions, especially educational interventions, can have indirect effects on road safety even in the absence of direct effects. For example, an educational intervention might be designed not to change speeds directly, but instead to make drivers on a route more tolerant of the enforcement of speed limits, and thus make it easier to introduce such enforcement later in order to If there is evidence of increasing bring about the desired reductions acceptance of enforcement on a in speed. To measure attitudes to route, this can suggest an indirect benefit of the route safety different enforcement approaches intervention would again require well-designed questionnaires.

Demographic variables. Risk of injury is not evenly spread across the population with males of all ages, especially teenage and young adult males having a higher casualty rate than females of the same age. Older people are at risk of injury as pedestrians and car and bus passengers. Some of this is due to exposure where people of certain age and gender are more active in the road environment as pedestrians, cyclists, motor cyclists or car occupants. For these reasons it is important to obtain information about the demographic mix (especially age and gender and these are relatively easy to observe) of those people from whom other behavioural data are being collected.

One reason for this is that if changes are observed in some behavioural

measures such as seat belt wearing and child restraint use it may then be possible to assess if this change is partly due to more females changing their behaviour than males. Another reason for knowing the numbers of males and females and people in different age groups is that the distribution of males and females in the driving population is not equal (there are still more male drivers than female) and even if five or ten year age bands are used there are different numbers of people in each age band.

3. Designing the evaluation

When an intervention is introduced it may have a positive, neutral or negative effect on casualties, target behaviours or attitudes. The aim of undertaking an evaluation of effectiveness is to establish whether the intervention has worked in the way intended and produced a change which is over and above that which could have occurred by chance or random factors alone.

The figure below illustrates the steps in evaluating a route safety intervention using behavioural measures.

Step 1.

When designing a route safety scheme and its evaluation the first step is to establish who are the road user groups most at risk of, or involved in, collisions along the route? There are several fundamental questions that need to be asked in order to target interventions and to evaluate the impact of route safety interventions:

1. Who is using the route? 2. Who is having collisions on the route? 3. Who is violating the regulations on the route? 4. Where do they come from? 5. What is known about which „social groups‟ they belong to? 6. Why are they using the route i.e. business or other?

Mapping data from various databases such as STATS19, police records of offences and socio-demographics information, using GIS and linking it with MAST and DVLA4 data gives a feel for where people come from, their social profile and whether or not they are registered to a company. These data will allow practitioners to target the groups they most want to influence and will provide the basis of a sample from which to evaluate the impact of interventions.

Ideally a monitoring plan should be set out at the outset when the scheme is being designed and objectives set. Data should be collected before the intervention and once or twice after There are various sources of data the intervention, usually four to six to establish what types of people weeks after and 12 months after to are using the route, and how they allow for driving behaviour to stabilise. are using it

4 It should be noted that DVLA data may be more easily obtained by the Police, meaning that this part of the partnership should take action on using such data.

Step 2.

The second step is to be clear about what it is the scheme is trying to achieve. In other words the objectives of the route safety scheme, and individual elements within it, need to be set out clearly. In this guide and in the report, a number of behaviours have been identified which the evidence indicates are associated with collision risk. A route safety scheme will have a set of objectives and a corresponding set of interventions to address them. Therefore it is only necessary to measure the behaviours that are being targeted for change.

In order to know what needs to be measured, an evaluation should always begin with the question:

“What is this intervention trying to achieve?” For example, if the objective of the intervention is to increase the number of people wearing seatbelts on a given route (having identified that wearing rates on the route are low), then this is the behaviour to be monitored in order to establish whether the intervention had been effective. Asking the question “what The specificity might refer to different types is this intervention trying of road users, and multiple measures, for to achieve?” is crucial in example: any evaluation

“The objective of the intervention is to reduce motorcycle speeds on bends, and also to discourage overtaking by motorcycles near junctions.” Again the specificity in the answer to the original “what is the intervention trying to achieve?” question, in terms of specific, measurable behaviours, leads naturally to the detail on what is required to monitor (and in whom, and where).

Asking this simple question, and answering it as described here, is essential to the evaluation, and no intervention should proceed until the objectives have been specified precisely, what the target behaviour(s) is/are and whether an increase or a decrease in the target behaviour is expected.

Step 3. The third step is to establish which measures are suitable to use for the target behaviour, and how many observations of vehicles, drivers or road users are needed (or how many questionnaires or interviews).

Advice in Section 2 of this guide covers the first of these aspects. Evaluations of road safety schemes are quite resource intensive. A rule of thumb is that a monitoring and evaluation programme should be about 10% of the total scheme costs. So as to use resources wisely and efficiently it is important to measure the correct number of vehicles or road users to have a relatively high probability that a change in behaviour There are multiple behaviours you can can be detected (if one monitor. All require different numbers of occurs). observations to provide robust findings

The main report gives guidance on this important issue. What needs to be established at the outset is how big a change (or effect) is being looked for. This will in part be determined by the practical importance of the scheme to have a large or small effect on a behaviour or attitude. For example the rule of thumb that a 1mph reduction in average speed should result in a 5% reduction in collisions is well known. So if a 1mph reduction is all that is considered feasible from the types of measures being implemented, then if the starting mean speed was 67.2 mph then the speeds of about 1050 vehicles would need to be measured to determine whether a true change has occurred. If however, the scheme is designed to bring down average speeds by 1mph on a road with a current mean speed of 31.6 mph then about 545 vehicles would need to be measured. The sample sizes required for different measures, and different effect sizes, are listed in a table in the main report.

Step 4.

The fourth step is to take into account background trends, or to choose an appropriate comparison group or site. Staying with the speed example, national or local data on trends in speeds on certain road types should be easily available. Free flow speeds on different roads have not been constant over the years and these trends need to be taken into account over the before and after periods so as not to over- or underestimate the effects of the scheme. Similarly, the numbers of casualties have been decreasing at a rate of 3–4% a year and this needs to be taken into account so as not to overestimate the effect of the scheme especially where a 3–5 year before period is used and a similar length of after period.

Allied to the issues around background trends is the use of a second unrelated group of people or another route for comparison purposes. The same behavioural measurements can be taken as from the participants or route being subjected to the intervention. In both cases, the logic is the same. What is measured is the change in the behaviours which are of interest, both in the comparison and treatment groups/routes.

If using a comparison group of people (for example if using an educational intervention to which some people are exposed and some people are not) try to gather the comparison and treatment participants from the same population. Practitioners could ask for volunteers to take part in the trial of the educational intervention, and then select on a completely random basis those participants who will be exposed to the intervention, and those who will not. By using random group assignment in this way, the risks are minimised that any effects seen from the intervention will just be because the people taking part in it are more enthusiastic than those who are not. If possible, have the comparison group take some other, unrelated intervention (perhaps a course on something nothing to do with road safety), to control for possible „placebo‟ effects.

If your intervention is having an effect, then a bigger change should be seen in behaviour in the treatment group/route than in the comparison. The key thing is that the comparison participants or route are not subject to the intervention. In other words the comparison group or route represents an estimate of what would have happened if the intervention had never been introduced. This general design is illustrated in more detail in the main report.

Step 5.

The fifth step is to assess whether the route safety scheme (or interventions within it) has been successful. As already described the most common design for an evaluation is to collect data through the monitoring process before and after implementation of the intervention. This same data need to be collected simultaneously (or as close in time as possible) for both the route (or road user group) being treated and the control or comparison route (or road user group). The Road Safety Good Practice Guide (see links in Section 5) gives more detailed information about data collection and monitoring.

The collection of data is only half the process. The assessment of effectiveness requires proper analysis of the data to establish how different it is and whether this difference could have occurred by chance (i.e. as a result of random fluctuations in the behaviour being observed).

The way to distinguish whether the effect we are measuring is greater than could have occurred by chance is to use a statistical test. Generally the outcomes of such tests are expressed as probabilities and conventionally the value taken is p=<0.05 which means that we are 95% confident that the change we have observed is not due to chance.

This guide is not the place for detailed descriptions on statistical analysis of data. The main report and Road Safety Good Practice Guide (Appendix B) gives information on which test to use under which circumstances and how to conduct the analysis. Many of the tests are straightforward and a helpful spreadsheet for practitioners to use is also given by the TMS Analysis using the correct statistical Consultancy.5 techniques will help you to assess if any changes are likely to be attributable to your intervention

5 http://www.tmsconsultancy.co.uk/resources/analysis-tools/statistics-tests-fyrr http://www.tmsconsultancy.co.uk/files/file3.xls

In general the t-distribution is used for data which are continuous such as speeds or flows. By this we mean the data can have any value in the range (including decimal places) and usually follow a normal (bell shaped) distribution. Tables for the distribution of t values are given in any statistical textbook and can be found on the internet.

The next most common test for differences between data is the chi- squared test. This is used to test for differences between before and after in the numbers of accidents or collisions and is used because accident numbers can only be whole numbers and cannot take a negative value. Some caution needs to be exercised when analysing casualty numbers because there can be varying numbers of casualties per accident. However, the chi-squared test is commonly used on casualty numbers to give an approximation.

The Road Safety Good Practice Guide gives information on how to conduct a test on proportions. This is really useful in the case of cyclists and pedestrians where it is difficult to match flows across routes (intervention and control for example) where the proportion on one route showing a particular behaviour vs. the proportion on another is a better indicator than raw numbers.

There is generally a lack of help for practitioners in the road safety literature on how to analyse the results of questionnaires which use scales; gathering advice from relevant experts is worthwhile on these issues.

A second important task for an evaluation is to help distinguish whether the effects observed are the result of the scheme and not the result of some extraneous factor for which there is an alternative explanation. A good example can be found in considering the design of a hypothetical road safety education programme where the findings showed that that those who have been on the course have safer crash records than those who have not been on the course. The difference between the two groups is statistically significant so there is a low probability that this is a chance result. This result, however, may have had nothing to do with the intervention but simply reflect an alternative scenario where, for example, those who were already safe were attracted to safety education courses and those who were not safe had no interest in attending safety education courses. Such pre-existing differences can in part be controlled for by the use of carefully selected study designs which include a robust comparison group.

4 Examples of good practice

Examples of evaluation of two high profile route safety schemes are included in the report and are briefly described below.

The first is Devon County Council‟s (DCC) Rural Road Demonstration „Country Mile‟ Project which is a combined area and route approach. DCC took an evidence led, strategic approach to develop a set of intervention measures and evaluate their effectiveness in what was known locally as the „Country Mile‟ project. The overall aims for the country mile project were to reduce the incidence and severity of road collisions that occur in the project area, to achieve improvements to public perceptions and awareness regarding road safety, and to identify and document the methodologies used to deliver the programme for wider demonstration purposes. Route treatments were adopted as the desired approach, including an aspiration to improve the „readability‟ and consistency of signing and marking of the rural road network, especially road curvature on „A‟ roads within the overall project. The project area was chosen on the basis of the distribution of injury collisions across the county, taking into account the shared road safety concerns of partnership agencies. Intervention design was based on extensive research to understand the pattern of road user casualties, offence levels and socio-demographic profiling to identify high risk groups, technical surveys were undertaken of lateral position at bends and journey speeds. Evaluation techniques included before and after monitoring of casualties, speed and flows, enforcement data, attitudinal surveys about local assessments of the route and perceptions of safety, surveys of awareness of changes to the road, speed limits and signage, and likelihood of being detected for speeding. In addition, an instrumented vehicle study was run to understand driver speed choice at bends.

The second is The Cat and Fiddle route safety project: Cheshire‟s Road

Safety Partnership (CRSP). The A537 Cat and Fiddle road between Macclesfield in Cheshire and Buxton in Derbyshire is a 14km stretch of road that runs through the Peak District National Park. Named as the highest risk road in Great Britain in the 2008 EuroRAP report, the Cat and Fiddle provides a consistent challenge to the Road Safety profession. Despite falling levels of casualties there are still too many people killed and injured on the road, with the main group being motorcyclists.

The aim of the intervention was to make the Cat and Fiddle and surrounding roads safe to use, visit and enjoy. The main intervention approach was to install average speed cameras along the full length of the

A537 between Macclesfield and Buxton with proactive marketing of the Cat and Fiddle as a safe place to ride with no fear of race track riders providing a threatening environment. Policing was proactive on the A537 and on associated leisure rider roads. Information was provided to riders of the specific risks on riding the roads and provided tactics to mitigate those risks. A wider communications plan was launched to reduce the number of bikers who feel it is inevitable that people are killed or seriously injured on the roads. The evaluation approach included measuring speed and flows along the road as well as casualty numbers. Local residents were included in surveys as well as the local motorcycling groups who use the road. The monitoring and evaluation plan was designed so as to quickly identify unintended consequences.

These two examples are described more fully in the report and demonstrate some of the general principles of good practice in terms of design of an evaluation. These included:

Injury monitoring 3 years before and planned for 3 years after the

various interventions; this ensures that a large enough dataset of accident data is gathered to permit meaningful statistical comparisons.

Monitoring of speed and traffic counter data planned before, during and after; this is useful as it means that traffic flow (i.e. the numbers of people exposed to risk potentially from different road

user groups) can be taken into account in analyses, and also any casualty savings attributable to speed changes can be accounted for.

Monitoring of enforcement data and enforcement schedules planned before, during and after; this is important as it makes it possible to attribute any short term or local changes to the enforcement-related activities separately from other parts of the intervention.

Attitudinal and self-reported behaviour surveys planned before, during and after the intervention which included a range of appropriate measures. The sample at each stage was large enough to detect relatively modest changes in attitudes (significant at the 5% level). Such surveys represent a high level of public engagement.

A clear analysis plan seeking to explore the impact of measures on different groups such as drivers of different ages, motorcyclists and

occupational drivers; this is important as road safety targets tend to be broken down into these different road user groups.

Exploration of both intended and unintended outcomes.

Clear project governance and accountability management structures for the project and its evaluation.

It should be noted that these schemes are high profile, and well funded, compared to everyday schemes run by local authorities. However they illustrate best-practice, and even in smaller schemes, a moderate amount of well-targeted evaluation following the principles in this guidance document should be seen as an essential part of route safety interventions.

5 Further resources

There a numerous resources that can be consulted regarding the evaluation of road safety schemes in general. These are listed here, and more detail can be found in the main report. DfT and RoSPA have created an evaluation toolkit („E-valu-it‟) to go live online from October 2010. The URL is www.roadsafetyevaluation.com The Department for Transport provides general guidance on evaluation here: http://www.dft.gov.uk/pgr/evaluation/evaluationlinks The Road Safety Good Practice Guide is also a useful resource. The section most relevant to evaluation is archived here: http://webarchive.nationalarchives.gov.uk/+/http://www.dft.gov.uk/pgr/r oadsafety/laguidance/roadsafetygoodpracticeguide?page=6 In addition, .pdf documents of the entire Guide (and the appendices, which are not currently available at the archived link above) have been made available on the ADEPT website, at these links: http://www.adeptnet.org.uk/assets/userfiles/documents/000283.pdf http://www.adeptnet.org.uk/assets/userfiles/documents/000284.pdf http://www.adeptnet.org.uk/assets/userfiles/documents/000285.pdf DfT has produced recent guidance for evaluators to help choose an evaluation approach to achieve better attribution of effectiveness to interventions rather than confounding factors. http://www.dft.gov.uk/pgr/evaluation/evaluationguidance/transportimpac t/ Another relevant resource is a review of health behaviour interventions from other related domains. http://www.dft.gov.uk/pgr/roadsafety/research/rsrr/theme3/inventionmo dalities.pdf