Citation

Wang S, Ma J, Ding H, Lu Y. J. Transp. Saf. Secur. 2023; 15(9): 918-942.

Copyright

(Copyright © 2023, Southeastern Transportation Center, and Beijing Jiaotong University, Publisher Informa - Taylor and Francis Group)

DOI

10.1080/19439962.2022.2137869

PMID

unavailable

Abstract

Despite the benefits of cycling being widely accepted, bicycle safety--especially severe injury--has received increasing attention due to the vulnerability of bicyclists on the road. Factors contributing to varying bicycle injury severity have been identified in the literature. For the zonal factors, variables related to sociodemographic and household characteristics, built environments, land use, and traffic conditions are considered. However, it is rare that the heterogeneity and hierarchal features of bicycle injury severity are simultaneously considered. This study contributes to the literature by investigating the heterogeneous effects of zonal factors on varying bicycle injury severity, using a 3-year crash data set from the Lower Layer Super Output Areas of London. A combination of latent class clustering and partial proportional odds methods was developed. First, five subgroups of bicycle crashes were identified based on the latent class clustering method. Afterward, partial proportional models were developed separately for different clusters.

RESULTS indicate that a series of factors is found to be associated with the occurrence of severe bicycle injuries. However, effects of these factors could be distinctive among different clusters. For example, some factors only have significant impacts in the specific crash clusters. Furthermore, heterogeneous effects of the same factors in one or different clusters are discovered. The findings of this study can be helpful for the development of cycle infrastructures, traffic management, and safety education that can enhance the risk perception of bicyclists and reduce the occurrence of severe bicycle injuries.

Language: en

Keywords

bicycle safety; injury severity; latent class clustering; partial proportional odds models; zonal factors