The Bayesian intrinsic conditional autoregressive convolution model was used to study the spatial variations in road mortality in the regions of Belgium, France and Germany. In all three countries, the spatial structure behind relative risk is significant and spatial heterogeneity predominates over unstructured heterogeneity. The maps of spatially-structured component of random effect enable the spatial structures of risk to be identified and highlight the zones and areas where the mortality risk exhibits spatial dependency. Hence, a west-east gradient in risk level is found in Germany and a north-south gradient in Belgium. In areas with high road network density (and relatively high population density), there is generally less heterogeneity in road mortality across neighbouring regions, while in areas with high regional disparities, there are significant spatial variations in mortality risk. Furthermore, a model was produced for 272 regions in 13 European continental countries, making it possible to investigate whether national borders have any specific effect on the distribution of road mortality risk compared to what happens within countries' administrative borders. Cross-border regions were found not to be particularly likely to have similar road mortality risk levels where they shared a common national border. National borders have no specific effect on the distribution of road mortality across Europe.