The actual paper discussed in that article isn't just comparing frequencies of accidents, but rather using them to calculate the odds that a given vehicle type

The actual paper discussed in that article isn't just comparing frequencies of accidents, but rather using them to calculate the odds that a given vehicle type causes a pedestrian accident in a variety of scenarios. In doing so the author determined that these vehicles were overinvolved in accidents relative to their frequency on the road.