I note that the study was of the correlation of population death rates with pollution from coal plants. Attributing the entire improvement to the coal pollution reduction misses things like coal-polluted areas tending to have lower average income than cleaner ones, with all the other health hazards and service shortages that brings. It's across time, and as the areas clean up they also tend to gentrify, changing that factor. (Though some of the resulting mortality improvement might fairly be credited to the pollution reduction - even though not due to direct chemical / micromechanical effects on bodies.)
Since it's across time, EVERYTHING ELSE changed as well, and some of that (such as improved medical treatment) might also have caused much of the improvement. To sort that out you need to do a multivariable analysis against pretty much everything (and then you STILL have a correlation-as-causation issue.)
(I'd also like to see how much percentage-wise the coal pollution dropped across the study groups, which would also be indicative.)
Nevertheless, though the number might not be right, pollution from coal combustion IS some very nasty stuff, and a lot of it has been eliminated. So perhaps most of this stat IS from that cleanup. (Or even "more than all of it", if the total of conflating effects was net negative.)