Spatial variation in bicycling risk based on crowdsourced safety data

Abstract

Bicycling-related injury data are difficult to obtain from official reports, which capture only about 20% of crashes and often lack coordinates, injury outcomes, and narratives needed for understanding where and why incidents occurred. Crowdsourced data on bicycling safety provides new opportunities for the study of bicycling injury and risk. Our goal was to quantify factors that influence the spatial variation in unsafe bicycling across a city, based on self-reports of bicycling incidents. To meet this goal, we leveraged BikeMaps.org, a global tool for reporting bicycling safety incidents, drawing on data from Metro Vancouver. We summarized incident conditions that led to injury, developed a model to identify predictors of injury using random forest regression, and mapped bicycling incident hot spots. Our results demonstrate that injuries from bicycling incidents are associated with older and younger bicyclists, downhill slopes, parked cars, recreation and weekend rides, falls, and single bicycle incidents with infrastructure, roads, and railroads. The broad range of incidents reported to BikeMaps.org allows us to add evidence that falls and single bicycle collisions are major causes of injury. Also, we demonstrate the value of attributing safety hot spots with contextual details to identify infrastructure interventions that can reduce injury for bicyclists.