Cycling is a healthy, green, and low-cost mode of transport and numerous policies and initiatives are employed for promoting cycling in the Netherlands. Despite various individual and societal advantages of prevalent bicycle use, the increasing number of fatal and severe injury crashes involving cyclists has raised concerns over safety of cyclists. In this regard, past studies have investigated effective environmental and traffic factors on cycling safety. In evaluating the relationships between cycling safety and probable effective factors, accounting for a reliable risk measure is critical.
Exposure and the number of incidences are two important factors in crash risk measurement. Unlike vehicular crash analysis, a common limitation in analyzing cycling crash risk is the availability of extensive bicycle counts as the exposure variable. In the past studies, bicycle counts, bicycle trips, and bicyclekilometer-traveled are some proxies used as exposure variable. In this regard, Ding, Sze revealed that application of trip-based exposure variable (expressed in bicycle-kilometer-traveled) resulted
in outperformance of cycling safety models. Moreover, new machine learning methods used for prediction of number of crashes have enhanced crash risk estimations. Despite the new ideas developed in cycling safety analysis, past studies have failed to address how cyclists as “travelers” are exposed to traffic crashes along their trips from an origin to a desired destination. Thus, this paper suggests a methodological approach to introduce a trip-based risk index to estimate crash risk associated with specific trip purposes at specific locations. We applied this approach to estimate a trip-based index for crash risk indicating the cost associated with work trips by bicycle in the municipality of Utrecht. This city is known as the best-bicycle-city in the world in 2022 with a 51% cyclists population. This study is focused on postcode-level 5 (PC5) zones in Utrecht and a 5 km buffer area. Various databases including road and cycling network, zonal socioeconomic and demographic chacrteristics, crash data, and Dutch national travel behavior data were used in the analysis. [From introduction]
