Re: I'd add to that line of thinking:
It's fair to compare accidents\fatalities involved per mile, but you should also consider the number of times that the safety features engaged successfully and prevented or reduced the severity of an accident.
No, you shouldn't. The reason being, those numbers are already included in the headline "accidents/fatalities per million miles", or whatever number you're looking at.
The trouble is that if you get a number for "times safety feature engaged", you have nothing to compare that number with. Human drivers don't, typically, make a systematic count of every time they have to brake to avoid crashing into the car in front - and if they did, the answer would be so subjective as to be meaningless anyway. So that number can only, at best, be a distraction.
We need numbers that can actually be measured with a reasonable degree of certainty and consistency. Number of accidents, and especially number of fatalities, are the only metrics that come close to meeting that requirement.