This article describes a study in which a mathematical model wasused to predict the risk factors contributing to death or injury ofold people in traffic accidents in Croatia. The variables chosen for study initially were as follows: a) weather; b) season, month, andtype of day (weekend; weekday); c) fuel consumption; d) road length; e) number of registered traffic vehicles; f) number of crossings per kilometre of road; and g) traffic density. The most significant variables were found to be: fuel consumption; length of road; number of registered vehicles and number of inhabitants. In order to increase the sensitivity of the model, the number of variables used was decreased by using two compound variables: a) the specific traffic density G (the ratio of annual vehicle fuel consumption to length of road) and b) the specific motorization level F (the ratio of the number of registered motor vehicles to the number of inhabitants). The accident rates predicted using these variables were compared with actual accident rates occurring in the 1971-1985 period. The variables showed a very small correlation. This feature was used to improve theaccuracy of the model in a stepwise manner using a rank correlationmethod. The model indicates that the specific traffic density, G, is the main cause of accidents amongst the elderly
Abstract