What constitutes traffic event severity in terms of human danger perception?

This study focuses on the severity gradation of non-collision events. Earlier theoretical work has suggested that a proper severity measure for an event should reflect the risk of personal injury, which is often split into two components, including the risk of a collision and the potential consequences had the collision taken place. While a great number of severity measures have been suggested, most of them fail to address both components, thus resulting in counter-intuitive event gradations and inconclusive outcomes in validation studies. Conversely, it has been shown that human observers often show very good agreement when given a task to rank traffic situations by their severity or level of danger. The aim of this study is to investigate in depth how human judgements of the severity of traffic situations can be expressed by means of objective safety indicators. In this study, a set of video-recorded traffic situations, in which a cyclist passes straight through an intersection while a left or right-turning motor vehicle crosses the cyclist's path, were analysed. Binary logistic regression was used to develop models assessing the most important traffic severity indicators associated with human feelings of danger. The results showed that the initial conditions of a traffic event, defined as a start of an evasive action, contain the most important information for explaining its severity. Moreover, variables related to both proximity and collision consequences are important and should be integrated into severity measures.