Interesting validation of the DAPT Score in a large real world population. I'm most interested in the authors discussion, because the data are largely consistent with everything else that has been published. It's the spin that is different here. 1/
Let's start with the validation. The DAPT score is built on 2 separate Cox models. The authors validated each one in their study. Here is how they performed: C-statistic 0.67 for ischemia and 0.67 for bleeding. That's basically how well they performed in the original cohort. 
Somehow, that didn't make it to the abstract, but instead, they focused on this.
The problem w this analysis is that the DAPT Score was not made to predict a bleed or predict an MI. If you wanted to do that, you would just use the individual models upon which the score is built. The c-stat isn't the right measure for its utility. cc: @JeremySussman
The score IS meant for is to identify individual pts at high risk for ischemic events but NOT at high risk for bleeding (and vice versa). Until now, every score that identifies high ischemic risk also finds those same pts to be at high bleeding risk. Here are relevant results:
To my eyes, it sure looks like the DAPT score once again finds that high score patients are at higher ischemic risk (1.5 times) without any elevated bleeding risk (HR 0.87, non sig). It uncouples ischemic and bleeding risk, a "conundrum" not previously dealt with by other scores.
The other critique the study tries to make is that the DAPT Study was not a real world population. True, but interestingly, this registry had a LOWER ischemic and bleeding event rates than the trial. Usually, we find real world events to be higher, not lower than trials.
How did they ascertain bleeding? Administrative codes. How good are administrative codes for ascertaining bleeding? Well, it turns out that the NHLBI has funded us to look at exactly this. And without spoiling too much, let me say, in the most scientific way possible, they suck.
Maybe claims are better in Sweden. Apparently they are, per the cited study. Their codes have good specificity. But sensitivity is in the 85% range.... in other words, they undercount true bleeding events which might explain part of their findings about miscalibration.
I'm not going tell you that the score is perfect. But if you're looking for a tool to identify pts at high ischemic risk but not high bleed risk to help augment clinical decisions for DAPT duration, the DAPT score does that. It has done it now in at least 5 external populations.
If you'd like instead to keep guessing based on intuition alone, or treating all patients the same regardless of their risk factors, I'd suggest that there is significantly less data supporting your choice.
