Characteristics of Cyclist Crashes Using Polytomous Latent Class Analysis and Bias-Reduced Logistic Regression

Although the number of cyclist crashes is decreasing in Japan, fatality rate not. Thus, reducing their severity a major challenge. We used polytomous latent class analysis to understand characteristics and bias-reduced logistic regression analyze severity. Specifically, 90,696 combinations 139,955 accidents were divided into 17 classes. The variable contributing most classification was crash location. Common risks included older age groups rural areas, whereas other factors differed among locations. Median strips, stop signs, boundaries between sidewalk roadway affected at intersections. Moreover, existence median strip, collision partner, time period On sidewalks, risk higher when front part bicycle subjected collision.