Capturing Spatial Heterogeneity in Cycling Accidents using a Latent Class Discrete Outcome Model

Abstract

Cities are striving for sustainable transportation, with cycling playing a pivotal role. Despite its health benefits, cyclists face various hazards, leading to accidents and injuries. To enhance cyclist safety, understanding the factors contributing to accidents is crucial. This study introduces a novel modeling framework employing a latent class discrete outcome model, integrating machine learning and econometric approaches. Results reveal the effectiveness of our approach in identifying risk factors based on accident location and their distinctive contributions. Complex relationships between built environments and accident characteristics are unveiled, illustrating variations in the impact of specific factors across environment typologies. These findings emphasize the capability to capture accident heterogeneity and its correlation with the built environment, enabling the targeted design of effective countermeasures and policies for specific risk scenarios.