I am very sorry about your friend.
I think the real problem is that people are really poor at assessing risk. And I think we can make some headway there, educationally, and it might actually affect how people reason around drunk driving (or their friends, assuming they still have their faculties).
Let's take the example of driving home drunk without hurting anyone or having an accident. Suppose that (being optimistic) there's a 1% chance of an accident and a 0.1% chance of causing a fatality (including to self). Seems like an easy risk to take, right? But observe what happens if you drive home drunk 40 times:
99% chance of causing no accident each time, to the 40th power = 0.99^40 is roughly 67% chance that none of those 40 times results in an accident. 80 times? 45% chance of no accident. Now you're talking about flipping a coin to determine whether you cause an accident (potentially a fatal one, we'll get to that) at all over 80 attempts. (I feel like that is optimistic.)
If I have a 99.9% chance of not killing someone when drunk-driving one time, after 80 times I have a 92% chance of not killing someone (that is, an 8% chance of killing someone). Again, this seems optimistic.
Try tweaking the numbers to a 2% chance of an accident and a 1.2% chance of causing a fatality.
Anyway, my point is that people are really terrible at evaluating the whole "re-rolling the dice multiple times" angle, since a single hit is a HUGE, potentially life-changing loss.
(People are just as bad at evaluating success risk, as well, for similar reasons- a single large success is a potentially life-changing event)