The basic idea is that, of the 450 people who had a warning for AF in the study AND got this confirmed with an actual EKG, only 34% actually had AF THIS IS NOT GREATpic.twitter.com/1snetnaQkE
-
Show this thread
-
(Smart readers might note that this means that, in a sample of 420,000 PEOPLE, the Apple Watch correctly diagnosed just ~150 with AF which is PRETTY DAMN BAD)
1 reply 1 retweet 7 likesShow this thread -
Now, in an unspecified sample of other people who wore both the Apple Watch and a continuous EKG, the prediction was better - 71% We have no idea if this is meaningful, because there's no sample size, which brings us to another issue here
2 replies 1 retweet 5 likesShow this thread -
None of this research is published yet I honestly can't believe I have to say this, but if you're reporting on a study that you can't read you should be VERY SKEPTICALpic.twitter.com/pnvYns380n
2 replies 0 retweets 23 likesShow this thread -
Anyway, in ideal conditions 71% of people who the watch said had AF actually had the condition Interesting, but largely unimportant A much more interesting figure is that, even in the best of conditions, 29% of people the watch said had AF probably didn't
1 reply 1 retweet 8 likesShow this thread -
Without knowing the NPV, it's not possible to calculate the false positive rate, but given the information we have it's likely to be pretty high
1 reply 0 retweets 6 likesShow this thread -
Health Nerd Retweeted Health Nerd
And coming back to the original tweet, that's a really big problem because the population prevalence of AF in people who wear Apple Watches is very lowhttps://twitter.com/GidMK/status/1107871498667094016?s=19 …
Health Nerd added,
Health NerdVerified account @GidMKThe reporting of this new Apple Watch study is not great - 0.5% rate of atrial fibrillation in the study (HUGE PROBLEM) - 34% positive predictive value in real-world conditions (rarely correct) - 71% PPV in ideal clinical conditions - watch mostly useless clinically https://twitter.com/CNBC/status/1107856005097054211 …Show this thread1 reply 0 retweets 5 likesShow this thread -
Without delving too far into the specifics, if you're screening large populations for a disease and only a few of them have it, your test has to be NEAR PERFECT Being wrong 29% of the time does not qualifypic.twitter.com/eluDH0aOBZ
2 replies 2 retweets 11 likesShow this thread -
And on top of this all - something that NO ONE SEEMS TO MENTION ANYWHERE - is that AF screening >65 years is currently not recommended because we are unsure of the risks/benefits https://bjgp.org/content/67/660/296 …pic.twitter.com/q24z1MTL7o
1 reply 1 retweet 14 likesShow this thread -
What about those <65 years? We haven't even looked at the question properly, because the population prevalence is so low that most epidemiologists agree it'd be a pointless waste of timepic.twitter.com/m6cOP1I9yC
1 reply 0 retweets 8 likesShow this thread
So on the one hand, we've got the Apple Watch study saying that the watch is pretty useless clinically On the other, we've got a dubious rationale for doing this in the first place and no demonstrated benefit And yet, the headlinespic.twitter.com/20e8PbkiLY
-
-
(I'd like to point out that, in a wonderful twist, the Daily Mail has one of the best articles out on this all https://www.dailymail.co.uk/news/article-6816853/Apple-Watch-spot-heart-problem-research-needed.html …)
1 reply 0 retweets 8 likesShow this thread -
Anyway, Apple is apparently doing a proper RCT looking at true clinical risk/benefit, so until those results are formally published and I can pick them apart I'm staying skeptical that this is anything interesting at all
1 reply 2 retweets 12 likesShow this thread
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.