A Review of the Delivery of the Road Safety Strategy - UCL Report

Appendix 3: Target setting and accident data

A3.1 Review of the national target setting process

The numerical context for the setting of the current casualty reduction targets for 2010 is described by Broughton et al (2000). The detailed statistical work described there was carried out at the TRL by Broughton himself, and the interpretation and judgements based on it were made by a subgroup of the STAR (Strategy and Targets for Accident Reduction) group set up by the Road Safety Division of the then Department of the Environment, Transport and the Regions (DETR) to oversee the development of the road safety strategy. At that time it was already envisaged on the basis of experience with the target for 2000 set in 1987 (Department of Transport 1987) that there would be separate targets for killed or seriously injured (KSI) and for slight casualties, but it was not yet envisaged that there would be a separate target for children.

A3.1.1 Basis for numerical advice on target setting

The exact form of the target was a matter for decision by Ministers, and although their eventual decisions about targets for numbers KSI and the slight casualty rate implied roughly an extrapolation to 2010 of the rates of reduction and increase respectively in annual numbers of KSI and slightly injured over the period of the previous target, the professional advice provided to them to help them in their decision was based on a wider range of considerations than simply experience with the previous target.

The method adopted for developing advice on the numerical levels at which the targets should be set was influenced by two considerations:

1 The Government's integrated transport policy (DETR 1998a) seeks deliberately to alter the trends of recent decades in road use by encouraging walking, cycling and the use of public transport whilst moderating growth in the use of cars.

2 The Government had committed itself to 3-yearly reviews of progress towards the target and of priorities within the strategy, so that the numerical basis of advice on target-setting had to be transparent and the calculations repeatable in the context of the reviews.

These two factors together meant that analysis had to distinguish explicitly between the effects upon future casualty numbers of safety policies on the one hand and changes in use of the roads on the other. Forecasts of future numbers of casualties were therefore made by

• forecasting casualty rates per unit of road use in the absence of new safety policies;
• reducing the resulting forecast casualty rates to reflect the likely effects of new safety policies; and
• applying these reduced rates to a range of possible future scenarios for road use.

A3.1.2 Forecasting of casualty rates

The method of forecasting casualty rates for each main road user group was to extrapolate the longstanding tendency for the logarithms of such rates to change roughly linearly over time in the short to medium term. Extrapolation to 2010 was based upon examination of the changes in the various rates over the period 1983-1998, after adjustment for the estimated effects of three particular kinds of road safety measure whose effects over that period had been estimated in separate exercises. The graphs used for this purpose are presented by Broughton et al (2000). The main judgement required was whether the rate of change to be extrapolated should be the average for the whole period 1983-1998 or for a shorter period ending in 1998.

The amounts by which the extrapolated rates were reduced to reflect the likely effects of new safety policies are discussed in Section A3.1.4.

A3.1.3 Scenarios for road use

Scenarios for road use in 2010, 36 in number to cover a wide range of possibilities, were constructed from the National Road Traffic Forecasts of 1997 (DETR 1997) and the relevant aspects of the integrated transport policy (DETR 1998).

A3.1.4 Sources of reduction in casualties

The reductions required to meet the targets are expected to stem from a combination of two processes. The first is the continuation of the general downward tendency in numbers KSI and in the slight casualty rate that is the product of increased motorisation and associated adaptation by society, including those road safety policies and their implementation which have become accepted policy and practice. If past trends continue, this should provide a reduction of 20-25 per cent in the numbers KSI between 1996 and 2010. The second is the implementation of new policies and measures of twelve kinds which, pursued as far as seems foreseeably practicable up to 2010 should reduce the number of KSI by 35 per cent by then. Combining these two estimated reductions multiplicatively gives a reduction of about 50 per cent, and the lower target of 40 per cent recognises that the past tendency may not continue in every respect, and that not all of the envisaged policies may be able to be pursued to the full.

Judgements about the likely effects of new safety policies on casualty rates were made separately for each road user group, leading to the following aggregate effects on the numbers KSI.

1. New road safety engineering programme - 7.7 per cent
2. Improved secondary safety in cars - 8.6 per cent
3. Other vehicle safety improvements - 4.6 per cent
4. Motorcycle and pedal cycle helmets - 1.4 per cent
5. Safety on rural single-carriageway roads - 3.4 per cent
6. Reducing accident involvement of novice drivers - 1.9 per cent
7. Additional measures to protect pedestrians and cyclists - 1.2 per cent
8. Additional measures to reduce speeds - 5.0 per cent
9. Additional measures to protect children - 1.7 per cent
10. Reducing casualties in drink-driving accidents - 1.2 per cent
11. Reducing accidents during long-distance work driving - 1.9 per cent
12. Additional measures for improved driver behaviour - 1.0 per cent

Effects of measures of all these kinds combined multiplicatively - 35 per cent.

These percentages were based partly upon estimates derived from research assembled by other subgroups of the STAR group, and partly upon aspirational judgements made by the subgroup responsible for the numerical work.

A3.1.5 Implications of subsequent events

Implications of subsequent events for the various judgements mentioned in the previous three subsections are discussed in Section 3.1 of the report.

A3.2 Prospects for national data on contributory factors

In the 1950s the national road accident data recording system known as STATS19 included a list of 'causes' which the reporting officer could attribute to the accident being recorded. Although this ceased to be part of the STATS19 system from 1959, many police forces continued to record such information and supply it to their local authorities as part of the information that it is common for police forces and local authorities to record by local agreement in addition to the nationally agreed coverage of STATS19. More recently, this type of information has come to be called contributory factors in the light of the understanding that most accidents arise not from a single cause, but from a combination of a number of circumstances, altering any one or more of which would change the probability of the accident happening.

By the mid 1990s rather more than half of the police forces in England and Wales were still recording such data, using either a simple list or a coding scheme (Maycock 1995), and there was a widespread view among road safety professionals that national collection of such data on a common basis would enable enhanced understanding of road safety problems and help in targeting remedial measures. Research was therefore undertaken (Broughton, Markey and Rowe 1996) to develop a prototype national coding system for contributory factors. The objectives were that the system should encourage collection of data of high quality by covering the great majority of accidents within a standard set of codes that is suitable for use without extensive training or documentation by police officers who may attend only one or two accidents per year.

The outcome was a coding form which trial use by a number of police forces showed to be usable from the forces' point of view and to yield data that were both broadly consistent between different police forces and, in terms of the broad types of factor recorded, consistent with earlier detailed 'on-the-spot' studies of accident occurrence. The form required the identification of a single precipitating factor describing the main thing that went wrong in the immediate run-up to the accident, and then the identification of up to four contributory factors that had helped to give rise to that precipitating factor. Degree of confidence in identifying a contributory factor was indicated by coding it as definite, probable or possible. The form offered 14 specific precipitating and 51 specific contributory factors, together with possibilities for nominating 'others', but in practice only 5 precipitating factors were identified in more than 10 per cent of accidents, and these accounted for 80 per cent of accidents. Similarly, only 7 contributory factors were identified in more than 10 per cent of accidents. Trial analyses illustrated the potential value of analysing the contributory factors data jointly with STATS19 data, which would be possible routinely if contributory factors data were once again to be included in the nationally agreed range of STATS19 data.

The STATS19 system is reviewed every 5 years by a responsible body, the Standing Committee for Road Accident Statistics (SCRAS), in consultation with users of the data. In the 1997 review, the proposed national form resulting from the research just described was considered for inclusion. The outcome was that some police forces adopted the form voluntarily, others continued using their previous systems, and the remainder continued to record no contributory factors data. In particular, 12 forces have been using the proposed national form since 1999, so that 3 full years of data for those forces have been collected.

The 2002 STATS19 review, whose report is nearing completion, returned to the question of a national system for collection of contributory factors data within STATS19, and in the course of that review, research has been undertaken to review the proposed national system and recommend possible improvements to it and ways to secure nation-wide implementation. This research has taken into account experience not only with the proposed national system but also with alternative systems used by other police forces. The outcome is that the STATS19 review is likely to recommend adoption of a modified national system within STATS19 from January 2005 along with other smaller changes to the system - changes of the kind that more typically result from the 5-yearly reviews.

In the meantime, any use of the data obtained using the previously proposed national form will need to bear in mind that:

• they are for a particular set of police forces; and
• they are not comparable with the nationwide data that is likely to become available from 2006.

Nevertheless, work with them in conjunction with STATS19 data for the same accidents in these police force areas in the relevant years may well provide useful indicative information about the relative importance of different contributory factors in various kinds of accident. Such work will also provide research experience in the joint use of these two kinds of data which will be valuable in accelerating the process of learning to make good use of the extended STATS19 data, including nationwide contributory factors data when these become available from 2006.

For the purposes of this Review, one of the Review Team at UCL has been given access to some trial tabulations of data for the year 2001 demonstrating conclusions such as the following about accidents in these police force areas as illustrations of the scope offered by data of this type.

• Over 80 per cent of the accidents had one of the following five precipitating factors: failed to give way; failed to avoid vehicle or object in carriageway; loss of control of vehicle; pedestrian entered carriageway without due care; and poor turn/manoeuvre. No other precipitating factor was recorded for more than 5 per cent of accidents.
• The following contributory factors were recorded for more than 5 per cent of accidents: impairment - alcohol; behaviour - careless/thoughtless/reckless; behaviour - in a hurry; failed to judge other person's path or speed; failed to look; looked but did not see; inattention; excessive speed; following too close; lack of judgement of own path; and slippery road. An average of just over 2 contributory factors were recorded per accident.
• The factors of both kinds and the relative frequency with which they were identified differed greatly between different types of accident. For example, in collisions between a car and a motorcycle in which the precipitating factor was linked to the motorcycle, 'poor overtaking' replaced 'pedestrian entered carriageway without due care' among the only five precipitating factors recorded for more than 5 per cent of accidents. In such collisions, 'inexperience of driving' and 'aggressive driving' replaced 'impairment - alcohol' among the contributory factors recorded for more than 5 per cent of accidents.
• In contrast, for collisions between a car and a motorcycle in which the precipitating factor was linked to the car (which numbered about 60 per cent of collisions between a car and a motorcycle), the only three precipitating factors recorded in more than 5 per cent of accidents were: failed to give way; failed to avoid vehicle or object in carriageway; and poor turn/manoeuvre. In such collisions, neither 'impairment - alcohol', 'excessive speed' nor 'following too close' were recorded for more than 5 per cent of accidents.
• In single vehicle accidents between a car and a pedestrian, the only two precipitating factors recorded for more than 2 per cent of accidents were 'failed to avoid pedestrian (pedestrian not to blame)' and 'pedestrian entered carriageway without due care', and these were recorded in the ratio 2 to 7. The contributory factors recorded differed markedly, as would be expected, between accidents with these two precipitating factors.

In short, the availability of nationwide data concerning contributory factors will open up new possibilities for analysis and understanding of the ways in which various kinds of accidents occur, but it will be several years before such findings will be available to influence the rate of progress in casualty reduction.

A3.3 Data from on-the-spot accident investigations

A separate important source of information about factors contributing to road accidents and the mechanisms by which people are injured in such accidents is the detailed study of particular accidents by multi-disciplinary research teams who are called to the scene of the accident as soon as practical after the accident becomes known. These studies are necessarily resource intensive and confined to small numbers of accidents within a limited radius from the base where the research team concerned is located.

Two studies of this kind are currently in progress in Britain. One of these, the co-operative crash injury studies, has been ongoing for many years and is mainly concerned with the biomechanics of injury in relation to structural features of cars. The other, the on-the-spot studies, is taking place over a few years and is concerned with contributory factors of all kinds, those associated with the road environment, the vehicle or the road user. This will update the widely quoted picture provided by similar studies in the early and late 1970s and a related study of urban accidents in the late 1980s.

For practical reasons, the accidents covered by these studies cannot be a statistically representative sample of all accidents that occur. Instead, they form a wide-ranging set of examples from which extensive insights into contributory factors can be obtained. Any quantitative extrapolation of the results to the national patterns of accident occurrence depends upon careful judgements of associations between identified contributory factors and particular categories of accident that can be identified from STATS 19 data.

The national contributory factors data discussed in Section A3.2 will extend the range and reliability of estimation of the incidence of particular contributory factors and thus complement the findings of the on-the-spot studies. The new data will, however, in no way replace the data from on-the-spot studies, because the judgement of the reporting officer however valuable and experienced, cannot substitute for in-depth multi-disciplinary investigation.

A3.4 References

Broughton J, R E Allsop, D A Lynam and C M McMahon (2000) The numerical context for setting national casualty reduction targets. TRL Report 382. Crowthorne: Transport Research Laboratory.

Broughton J, K A Markey and D Rowe (1996) A new system for recording contributory factors in road accidents. Unpublished TRL Project Report PR/SE/229/96. Crowthorne: Transport Research Laboratory.

Department of the Environment, Transport and the Regions (1997) National road traffic forecasts (Great Britain) 1997. London.

Department of the Environment, Transport and the Regions (1998). A new deal for transport: better for everyone. London: TSO.

Department of Transport (1987) Road safety - the next steps. London.

Maycock G (1995) Contributory factors in accidents: police databases. Unpublished TRL Project Report PR/SE/229/96. Crowthorne: Transport Research Laboratory.

[ Previous ] [ Contents ] [ Next ]