Risky riding in context: built environment moderation of injury severity in food delivery E-bike crashes

Traffic violations by food delivery e-bike riders may be associated with different injury consequences across roadway and land use contexts. Identifying spatial contexts associated with more severe injury outcomes may help inform safety interventions for food delivery riders. Using 882 delivery e-bike crashes in Hangzhou, China, this study classifies injury outcomes as no injury, light injury, or injury-fatality. In addition to rider violations, peak time, and delivery platform affiliation, the analysis incorporates 18 built environment indicators for each crash location. These indicators describe the surrounding road network and land use, such as food delivery restaurant density and residential street density. A two-stage modeling approach is proposed to investigate the impacts of context factors on injury severity of e-bike crashes. First, a spatial generalized ordered logit model estimates associations with injury severity while accounting for street-level spatial dependence. Second, a spatial generalized additive model examines how built environments modify selected violation-injury associations. The spatial model had better model fit and higher in-sample classification accuracy than the nonspatial model. Secondary road density, recreational land proportion, and food delivery restaurant density were associated with a higher probability of injury-fatality crashes. The association between speeding and injury-fatality probability was stronger at higher secondary road density and weaker at higher sidewalk density and residential land proportion. Educational land proportion and expressway density were also associated with weaker associations between red-light running and injury-fatality probability. The injury consequences of rider violations vary across built environment contexts. Safety governance should combine behavioral interventions with spatially targeted measures that address the road and land use conditions associated with more severe crash outcomes.