"Premature deaths" is not a useful measurement because it does not take into account how premature the death was. You could say everybody who dies of a respiratory problem dies prematurely, because pollution made them die a few hours/minutes/seconds faster than they would have breathing perfectly clean air. The correct unit for measuring the harm caused is the disability-adjusted life year:

https://en.wikipedia.org/wiki/Disability-adjusted_life_year

Good points. I sort of touched on this in another comment, but perhaps the emission standard on these vehicles is what's unreasonable.

We have a more-or-less arbitrarily set maximum allowed tailpipe emissions for a given class of vehicles. Is it fair to apply this arbitrary standard uniformly across all vehicles in the same class?

If vehicle A and vehicle B both meet the same emissions standards, but vehicle B is capable of better fuel economy, isn't it reasonable to suspect vehicle B's net pollution is less over the vehicle's lifetime?

In that light, perhaps these standards are unfair to the companies who are trying to make this class of vehicle. In the case of VW, the vehicles will have to lose fuel economy in order to adhere to this standard, which means more fuel is being consumed over the vehicle's lifetime, which may very well equate to a net gain of pollution (not just for the vehicle's tailpipe, but also the entire system from the oil well through the pump).

None of this excuses blatant falsification of tests - however, perhaps it's time we took a better look at what exactly pollution is, and how tweaking our vehicles may impact the output.