This study explores the determinants of crash injury severity of police-recorded crashes involving e-scooters in the UK, with crash severity outcomes referring to the most severely injured individual involved in each crash. To that end, latent classes binary logit models were estimated using crash-level data from the STATS19 database. Due to their differences in the injury generation mechanism, separate statistical models were estimated for single- and multiple-vehicle crashes. The analysis accounted for unobserved heterogeneity by allowing the identification of unobserved (latent) classes within the crash population where determinants exert homogeneous effects within each class but varying effects across different classes. Key socio-demographic characteristics, including rider’s age and gender were critical in defining class membership in both crash types, with crash characteristics being also pivotal for defining latent classes in multiple-vehicle crashes. The results revealed that lighting conditions have pronounced effects in both crash types, but with divergent effects across classes. For single-vehicle crashes, interactions of pedestrians with e-scooters, are found to significantly influence crash severity outcomes, even though such events are often associated with slight injuries. In addition, older male riders tend to be associated with factors linked to greater crash severity. In multiple-vehicle crashes, the involvement of bicycles and e-scooters significantly increases the likelihood of severe injuries, especially if the crash occurred at a roundabout or on a non-dry road surface. The role of socio-demographic characteristics in defining the latent classes highlights the need for policy interventions tailored to the characteristics and behavioral nuances of different user groups. Infrastructure improvements, targeted education and clear legislative communication can enhance e-scooter safety, ensuring a safer integration of e-scooters into urban transportation systems.
Fountas et al. (Wed,) studied this question.