Can road traffic accidents be reliably and usefully classified.

This paper is concerned with the question of whether accidents can be sensibly classified according to the driving behaviour involved. The database used for this exercise was a postal survey of 714 drivers who had previously formed part of the TRRL's national driver survey. From the survey, it was discovered that 120 drivers had had an accident in the past year; and they were asked to describe their most recent accident using a standard report form. The first major subdivision of accidents was 'active' versus 'passive'. This distinction did not seek to establish blame, but rather whether from the information provided there was anything the driver could have done in the circumstances to d the accident. The 79 active accidents were then examined to see whether they could be classified further. In the event a large proportion of the accident descriptions appeared to indicate a primary role either for excess speed, or a failure to look at or notice some crucial element in the traffic environment. Drivers with speed accidents rated poor control on their part as significantly more important than did drivers with observation accidents. They also rated poor road surface as a more important contributory factor (see table 6). The potential utility of the speed versus observation distinction is shown in table 6 by the fact that the two categories of accidents are associated with different values on a number of key variables. Speed accidents were associated with greater damage and injury severity. Drivers with speed accidents were more likely to engage in deviant driving behaviour and to drive fast. They were considerably younger and had much less driving experience than did drivers with observation accidents. It is concluded that accident descriptions can be reliably and usefully classified with a simple methodology and that there is some reason to believe that this classification bears some resemblance to actual accidents. A broadly-based classification system, not based on blame, may be one solution. In addition, it may be possible to identify specific accident scenarios which occur with sufficient frequency to make it worth targeting specific countermeasures on them.