|

Predictors of bicycling injury in a mid-sized North American city: an assessment using crowdsourced data

Author(s): Fischer, Nelson, Winters

Slidedeck Presentation Only (no paper submitted):

7B - Fischer

Abstract:

Many cities are working to increase the number of people who bicycle, however, safety concerns are the primary deterrent to bicycling uptake. Sparse data on bicycling risk are limiting safety research and bicycle friendly urban design. Traditional reporting systems such as police and insurance reports miss 60-70% of bicycling incidents and few record injury circumstances.

BikeMaps.org is an opensource tool for citizen reporting of bicycling incidents that is generating worldwide data and providing new opportunities to study bicycling safety and injury. Citizen mappers report bicycling collisions and near misses and provide information on bicycling event circumstances. Our aim is to model predictors of bicycling injury in Victoria, BC, Canada to better understand conditions that lead to injury in bicyclists. To meet this goal we leverage novel citizen reports of bicycling incidents available from BikeMaps.org. We used a geospatial dataset of bicycling incidents reported to BikeMaps.org (111 collisions and 234 near misses) for the City of Victoria. The data included 24 variables representing infrastructure, environmental, and trip characteristics for each event. To evaluate how these variables (predictors) relate to injury we created a model using random forest classification. Random forests are an extension of classification and regression tree methods and are useful in ranking predictor variables for a given outcome. Overall, the top predictor of injury was bicyclist-only crashes. These crashes involved falls to avoid collisions (11%), collisions with a surface feature such as train tracks or a pothole (53%) and road conditions such as ice (26%); the remaining 10% were the result of bicyclist error. A bicyclist colliding with an open door of a parked vehicle (dooring) was also a top predictor. Arterial roads, painted on-street bicycle lanes and multi-use paths were the strongest predictors of injury compared to other infrastructure types. Environmental conditions that strongly predicted injury were sightlines obscured by glare or reflection. Bicyclist-only crashes often result in injury, but are rarely captured in traditional reports. Our findings parallel past research using surveys, interviews, or hospital administrative data, which identify road surface features and conditions along the route as risk factors in bicyclist-only crashes. Similarly, in crashes involving other road users, our results reflect research that identifies higher risk for infrastructure such as multi-use paths, painted bike lanes and major roads with parked cars compared to bicycle-specific infrastructure. That our findings align with past research suggests that citizen science data can be complementary to bicycling safety research: specifically, that BikeMaps.org reports have the potential to address gaps in safety data and provide more nuanced information on the circumstances of bicyclist crashes than available in traditional reports. We highlight that in Victoria, bicyclist-only crashes and doorings are strong predictors of injury, and find that arterial roads, painted on-street bicycle lanes, multi-use paths, icy road conditions and obscured sightlines are also associated with elevated risk. Our approach provides a foundation for understanding predictors of bicycling injury using a novel and growing dataset available from BikeMaps.org, and demonstrates how BikeMaps.org data can help to address data gaps in traditional reporting.