“All-cause mortality” simply means death from all (or any) causes.
So for smokers, you got a buttload of people with this thing in common, and rather than look specifically at something like deaths from lung cancer, you take a step back and look at deaths from anything. And then go in and try to find correlations and help to understand those correlations.
It’s kind of a chicken and egg scenario, because some of those causes might not be from smoking, but from a person’s proclivity to smoke.
For example, smokers might possibly be more impulsive than non-smokers (generally speaking) and there might be a higher risk of motor vehicle fatalities in the smoker group, but the cause wouldn’t be smoking, it’d be underlying behavioral differences that would make someone more likely to smoke.
It’s basically looking at mortality from a distance as opposed to looking at very specific things up close (but with the data it lets people zoom in on everything)