A Typological Analysis of Pedestrian Accidents

For more than 20 years, in France as in many other countries, the number of pedestrians killed in road accidents has tended to decrease. The stakes remain high, however : 1,126 pedestrians killed in France in 1994, and this user category is particularly vulnerable in accidents. To further our knowledge of pedestrian accidents, an analysis of data supplied in reports drawn up by the police forces for road fatal accidents was carried out. Its aim is to highlight the different types of pedestrian accident victims. A typology of pedestrian accidents is proposed based on a multi-factor analysis, followed by a classification. This classification clearly identifies four groups : pedestrians age 65 or over crossing the road in a built-up area, children in accidents in built-up areas while running or playing, drunken pedestrians, intermodal changes, secondary accidents. The resulting typological breakdown should then serve as a basis for in-depth analysis to improve our understanding of these accidents and propose suitable action. With a view to learning more about road accidents, the INRETS "Accidentology" programme provides for the cross-sectional treatment of themes based on data collected at different levels of information. Pedestrian accidents are one of the themes addressed in this cross-sectional approach. The stakes are high : 1,126 pedestrians were killed in road accidents in France in 1994, i.e. 13% of all deaths on the road, children and elderly people being overrepresented. Research on accidentology is often based on corporal accident statistics, i.e. accidents in which at least one person is killed or seriously or lightly injured. Three levels of accident information may be taken into account : national statistical files based on BAACs (Corporal Accident Analysis Forms). They have the advantage of being exhaustive, since a BAAC is drawn up by the police for every corporal accident. The national data base into which these reports are fed provides substantial general information on all corporal accidents. The national statistics published by the French Inter-Ministerial Road Safety Observatory (ONISR, 1995) are based on this data. corporal accident reports. These serve as a basis for subsequent legal action. These reports, also drawn up by the police for every corporal accident, give a detailed description of events and mention the statements made by the different parties involved. They show a plan of the accident scene and sometimes include photographs. They provide an overall view of the accident, but are not as detailed as the detailed accident analyses (see below). detailed accident analyses. Detailed accident analyses drawn up by multi-disciplinary teams sent to the scene of the accident as quickly as possible are often a source of details that might otherwise be missed, such as traces of braking or skid marks, or interviews with witnesses in the heat of the moment (Ferrandez et al, 1986). The aim is to analyse the accidents, to reconstitute what happened in the preand post-collision phases, to pinpoint the processes that generate danger and injuries, to identify specific counter-measures and to evaluate safety measures. Much can be learned by taking these different levels of information into account. Statistical analysis reveals correlations between previously coded criteria. Rather than explain the causes of the accidents, these correlations reflect a "statistical proximity". The detailed accident analyses make it possible to go further in identifying accident factors and mechanisms and the effects thereof on a limited number of cases. This identification then has to be validated by statistical analyses of more representative accident samples and using criteria evidenced by more clinical approaches. This "to-and-fro" process between different levels of information helps to take our knowledge of accident processes a stage further. 1 killed : victim who dies on the spot or within 6 days of the accident seriously injured : victim whose condition requires more than 6 days in hospital lightly injured : victim whose condition requires between 0 and 6 days in hospital This is the context in which this first analysis of pedestrian accidents came about (Fontaine and al, 1995). The purpose here is to present a statistical description of this category of users, based on police reports on fatal accidents that occurred between March 1990 and February 1991 and collected exhaustively by the Accidentology and Biomechanics Laboratory run jointly by car manufacturer Peugeot and Renault. The aim of this descriptive analysis is to identify some very distinct groups of pedestrian accident victims. The results of this typology may then serve as a basis for more in-depth analysis to learn more about how these accidents happen and what might be done to prevent them. Methodology We adopted a multi-dimensional approach to reveal the interactions between the different analysis criteria and thereby deduce homogeneous types of pedestrian accident. The data in the fatal pedestrian accident file is either qualitative (sex) or quantitative (age). Multiple correspondence analysis is a technique for describing qualitative data. It helps to transform similarities or likenesses between individuals and relations between variables into geometric distances easy to illustrate on simple graphs (factorial plans). The method is applied to qualitative variables using their modes or items. Thus the quantitative variable 'age' is transformed into a qualitative variable with four modes : under 15, 15 to 29, 30 to 64 and 65 and over. The sample analysed comprises 1,275 pedestrians killed in road accidents. We decided to select 17 active variables that serve to define the axes of the factorial analysis : 9 describe the pedestrian involved in the accident : age, sex, reason for outing, type of outing (alone, in a group), action of pedestrian, position, obstacles to progress, intermodal change, alcohol ; 8 are related to the environment or the type of accident : situation in or outside builtup area, straight line or curve, day of week, month, light, weather conditions, type of vehicle, secondary accident. Criteria correlated with the above variables, such as socio-professional category and age (children come under the "schoolchildren" category and elderly people under "retired") were added as illustrative variables. These do not count in the calculation of factors but are shown on the factorial plans. The factors thus obtained are then used in an increasing hierarchical classification that reveals very distinct groups of pedestrians involved in fatal accidents. The aggregation criterion is the Ward criterion (Saporta, 1990). The dendogram or classification tree reveals a significant number of classes. Calculations were made using the SPAD programme. Results of the multiple-correspondence analysis Axis 1 distinguishes pedestrians in accidents in built-up areas in the daytime from those in rural areas at night. Axis 2 is linked to the age, action and alcohol rate of the pedestrian. The factorial plan formed by axes 1 and 2 (figure 1) reveals certain likenesses : the group of elderly pedestrians in accidents when crossing the road in built-up areas in the daytime, the group of children running or playing, the group of pedestrians with high alcohol rate in accidents in rural areas at night. The hierarchical classification based on the results of this multiplecorrespondence analysis will highlight these groups of pedestrians. Figure 1: Analysis of multi-factor correspondences in pedestrian accident victims axes 1 and 2