Something doesn't seem right here. All of these are northern cities; maybe the snow has something to do with skewing the stats, but Wisconsin looks over-represented here. I wonder if the statistics used in the methodology are reported differently by states... for example, consider the following:
Scenario A: Car A runs a red light and both hits Car B and kills its driver. Driver A is sober, but the toxicology on Driver B indicates a BAC of 0.2
Scenario B: Car A is legally stopped at an intersection when Car B rear-ends Car A, killing the passenger in Car A. Both drivers are completely sober but the deceased passenger had a BAC >0
Scenario C: Car A and Car B are involved in a minor collision that results in the death of a pedestrian who was intoxicated. Both drivers were sober.
I don't know about other states, but in Wisconsin, all three of the above scenarios would be reported as an alcohol-related driving fatality. If Wisconsin is the only one that reports alcohol-related fatalities so liberally, then that would certainly explain 12 of 20. But having lived in several states outside Wisconsin, I can't say that the differences in drinking habits are on the magnitude that goes along with having every single one of your metro/metropolitan areas in the top 20.