I would now agree with this FWIW - but note that the NYT article is also making a point about age. And on that topic they are probably correct - albeit using statistics that, as Dew correctly pointed out (and I would agree with more than I initially did), are exaggerated.
An interesting question is how representative of the atrial fib patient population was the Pradaxa registrational trial. The registrational trial intentionally targeted new patients as 50% of the enrolled population - who would obviously be younger than the average atrial fib patient. So given the bleeding problems that show up as the patients get older the registrational trial would make the risk reward for pradaxa vs warfarin look a lot better than it actually is.