APPLE HEART Study, Big Data, May 23, 2019
<p>APPLE HEART Study, Big Data, May 23, 2019
This is Thursday afternoon at Stanford in Berg Hall at the
Big Data in Precision Medicine Conference.
This is the panel presentation on the Apple Heart Study (of <strong>AF = atrial fibrillation</strong>)
using Apple watches.</p>
<p> </p>
<p>moderator is Prof. Paul Yock...
presenters are</p>
<p> </p>
<p>Manisha Desai.. Stanford</p>
<p>Marco Perez Stanford and</p>
<p>and Mintu Turakhia, Stanford.</p>
<p> </p>
<p></p>
<p> </p>
<p><strong>APPLE HEART STUDY</strong>...</p>
<p> </p>
<p><strong>Prof Mintu Turakhia </strong>begins....</p>
<p> </p>
<p>A FIB....is irregular and fast (key outcome is a stroke)</p>
<p> </p>
<p>prob <strong>about 700,000 people out there that are undiagnosed who have</strong></p>
<p><strong>A fib...</strong></p>
<p> </p>
<p>Apple Watch... the algorithm looks at tachograms.... classify them...</p>
<p> </p>
<p>some early efforts of automated AF analysis had lots of false positives...</p>
<p> </p>
<p>Apple developed this algo...</p>
<p> </p>
<p>with the<strong> newest Apple Watch... you simply put finger on the crown and it does ECG</strong></p>
<p> </p>
<p>But... <strong>THIS study was with the old Apple watches </strong>(with just a light sensor)</p>
<p> </p>
<p>how well did it detect AF?...</p>
<p> </p>
<p>Manisha...<strong> we looked at concordance of chest ECG and watch....</strong></p>
<p> </p>
<p>first study... we only included participants. with no hx of AF.</p>
<p> </p>
<p>then <strong>we'd invite them to do a telehealth visit with an MD.</strong></p>
<p> </p>
<p>then mail them an ECG patch to wear (via telehealth visit) then they would wear it and mail it back.</p>
<p> </p>
<p><strong>within 8 months we had 400,000 people</strong>!!! (Mark says)</p>
<p> </p>
<p>overall cohort was 419,000 !!!
</p>
<p><strong>we notified 2161 this was 2161 divided by 419k.</strong></p>
<p><strong> </strong></p>
<p><strong>= 0.52% </strong> (that is, 1/2 of one percent had AF)</p>
<p> </p>
<p>much higher percent in the older people.</p>
<p> </p>
<p><strong>most were long episodes of AF.</strong></p>
<p> </p>
<p><strong>overall 1489 had AF on ECG patch .....</strong></p>
<p> </p>
<p>and of 2089 with positive tachograms</p>
<p>548 of ECG patch positive were over 65 years. 36% were over 65.
once you got notified you got a patch.</p>
<p> </p>
<p><strong>84% had AF by watch if you got a notification.</strong></p>
<p> </p>
<p>age > 22;</p>
<p>must have iPhone 5S or better and Apple watch series 1 thru 3.</p>
<p> </p>
<p>exclusion criteria: hx of AFib or Flut</p>
<p>or current use of anticoag.</p>
<p> </p>
<p>start wth overall cohort 419k then</p>
<p> </p>
<p>get notfied if watch detects an irreg rhythm</p>
<p> </p>
<p>then connect to telehealth MD then</p>
<p>get mailed an ECG patch</p>
<p> </p>
<p>then discuss results with MD</p>
<p> </p>
<p></p>
<p> </p>
<p><strong>only half of people who were notifed contacted the telehealth provider.</strong></p>
<p> </p>
<p> </p>
<p>final thing was EOS survey... <strong>end of study survey...</strong></p>
<p>90 day survey....</p>
<p> </p>
<p>Mintu... how much should you nudge people to participate</p>
<p>(only 50% particiapted; who were notifed that they had an irreg rhythm)</p>
<p>we didn't want to scare them....</p>
<p> </p>
<p>we didn't want to annoy them...</p>
<p> </p>
<p>baseline demographics</p>
<p> </p>
<p>overall cohort of 419k</p>
<p> </p>
<p>42% female...</p>
<p>mean age was 41...</p>
<p><strong> </strong></p>
<p><strong>notification... 2161....21</strong>% only were female 461 women...</p>
<p> </p>
<p>avg age was 57;</p>
<p> </p>
<p><strong>775 over age 65 (36%)</strong></p>
<p> </p>
<p>high concordance with ECG patch....</p>
<p> </p>
<p>450 wore the patch.... 181 had AF (40%)</p>
<p> </p>
<p>this was not just PACs...</p>
<p> </p>
<p>safety.... </p>
<p> </p>
<p>our initial concern was about false positives...
people calling their doctors and being v worried.</p>
<p>we wanted to make it v specific....</p>
<p> </p>
<p>we also have a tachy notif and if too slow they get notified...</p>
<p> </p>
<p>Ambient sensors....and d note: our next Aple watch has built in ECG....</p>
<p> </p>
<p>Q1 (from U Toronto hijab... how aobut the 50% who did not respond to telehealth</p>
<p>A we did look at it... no diff btwn people who followed up versus those who did not.</p>
<p> </p>
<p>Q2 who shoul;d we have on the team to do a similar study....</p>
<p> </p>
<p>(bob: ask Mintu about length Afib... how long does it usu last.)</p>
<p> </p>
<p>recomm having a pilot study first...</p>
<p> </p>
<p>Stfd Center for Clinical Research...</p>
<p> </p>
<p>Mintu Center for Digital Health...</p>
<p> </p>
<p>this was team science...</p>
<p> </p>
<p>Ingo Beinlich asks
did you report car accidents, suicides, un-related adverse events.</p>
<p> </p>
<p>unrelated events; </p>
<p> </p>
<p>a rash from the watch... anxiety of getting a notification</p>
<p> </p>
<p>Ann Hopfield OTL office of tech lic.</p>
<p> </p>
<p>about 20 people went to the ER...</p>
<p> </p>
<p>Q underserved populations....</p>
<p>cuz you're using Apple watches...</p>
<p> </p>
<p>get r leg and l leg ecg every time you brush your teeth.</p>
<p> </p>
<p><strong>Noel Ostravsky... </strong>conf organizer (in Lloyd Minor's office)</p>
<p> </p>
<p>humorous tweets;</p>
<p> </p>
<p><strong>Jeff Dean real world data is messy; </strong></p>
<p> </p>
<p>this is wrap up by Euan....</p>
<p> </p>
<p><strong>Don Rucker: our goal is to have an app economy....</strong></p>
<p> </p>
<p>muffins with blueberries mistaken for chihuahuas....</p>
<p> </p>
<p>Matt Lundgren...</p>
<p> </p>
<p>Jonathan Wilt</p>
<p> </p>
<p>Moonshot ID heart failure from your voice.</p>
<p> </p>
<p><strong>Serena Yeung... Ambient intelligence</strong>...</p>
<p> </p>
<p>Euan Ashley...</p>
<p> </p>
<p>52 posters....</p>
<p> </p>
<p>Euan is Scottish... his mother is watching...</p>
<p> </p>
<p>Mimi H... human labels are error prone.</p>
<p> </p>
<p>Will Salerno... Regeneron... special kind of snowflake.</p>
<p> </p>
<p>DJ Patil...tells us his fialures... failed at math and his failed startup...</p>
<p> </p>
<p>PEEPS are marshmallows at Easter...</p>
<p> </p>
<p>humans can sweat up to 14 liters per day...</p>
<p> </p>
<p>Where the patient spends his day... Dina Katabi....</p>
<p> </p>
<p>sensor can perceive even thru walls</p>
<p> </p>
<p>see patient in bed or on a chair.
</p>
<p>Maja Mataric baby with a robot....</p>
<p> </p>
<p> </p>
<p> </p>