Consistency
I don't mind what measure you use in particular, as long as it's consistent. For example, showing the number of deaths from nuclear against the number of deaths from coal is useful.
Then do the same with illnesses, impairment, quality of life or whatever.