Indeed, while there are problems with comparing accident rates on a per km basis, especially between different modes where exposure is not comparable (e.g. cycling short trips within cities vs car acccident rates that include motorways) it is a reasonable basis for comparing accident rates for cycling between different countries. We are talking about similar types of journeys in comparable locations, we would expect risk to be approximately in proportion to total distance travelled. Nonetheless, risk is still not homogeously distributed throughout any cycle trip: more than two thirds of accidents occur at junctions. So, while Amsterdam and Copenhagen have hundreds of route km of segregated cycle route, at junctions physical segregation ends; instead different signal phases are provided to avoid conflict between cyclists and other traffic. But here's the interesting thought: the physical segregation is only able to tackle around 30% of accidents, the non-physical measures at junctions have to deal with the rest. So, where should we focus our resources? Expensively segregating between junctions to address a minority of accidents, or at the junctions where the vast majority occur?