Deaths per capita might not be as useful a metric as deaths per 100,000 km driven could arguably be better. But then you’re perhaps not taking into account deaths of pedestrians & cyclists.

I would argue that deaths per distance traveled (even if it included modes other than driving) could be worse because it might skew the results in favor of sprawl-y countries with a lot of freeway driving.

The pursuit of "safety" (measured in deaths per distance traveled) has been the excuse for a lot of terrible design decisions in traffic engineering, because keeping the number of deaths the same while increasing the speed and distance traveled looks like a win.

I would usually agree with this kind of normalization, but in this case I actually think it would actually obfuscate the picture. Safer roads are a good thing, but if traffic deaths are reduced because more people bike or take the train, that's still a win. Roads and cars are inherently dangerous, and that danger needs to be minimized using multiple strategies. We need to focus on holistic changes that consider people's behaviour and their interactions with the built environment.

When people feel they absolutely need to drive, that's a failure of infrastructure.

Valid, but only if you also include the distance travelled by other means than by car.

Being compelled to drive more is kind of the problem.