Hacker News new | comments | show | ask | jobs | submit login
The Apple Watch can detect hypertension and sleep apnea, a new study suggests (techcrunch.com)
227 points by helloworld 4 days ago | hide | past | web | 177 comments | favorite





The only downside to the Apple Watch in my book is the short battery life which should theoretically make it impractical for sleep tracking.

I was recently stupefied to discover Nokia Steel [1] and how virtually unknown it is to the general public, my former self included. It comes with great design, no bullshit digital screen, the usual tracking features, and most importantly seems like it's a return to form for Nokia, considering their country's outstanding design heritage.

edit: lol I forgot to mention the best part. THE BATTERY LASTS EIGHT MONTHS

1: https://health.nokia.com/es/en/steel


This perception of low battery life has really lasted from the first apple watch. It was the reason, along with that it wasn't water proof enough for me, that I didn't get the first one.

I wear my apple watch 2 all day, sleep in it, and charge it in the time I'm getting ready in the morning. that's all it needs and it never runs out of battery.


I completely agree; I wore my watch series 3 (no cell) all day yesterday, tracked sleep last night, and this morning I still have 65% battery left. I quick charge while I shower in the morning is all it's needed and I'm blown away by how long it lasts. (I had a Pebble Time previously.)

I had low expectations when I bought the Series 3 because of what has been said in the past about the battery life but my findings are similar: I put it on in the morning at 7:30am and it's still got ~65% when I'm getting ready for bed. I took it off last night around 11:30 and it was charged back to 100% by 12:15.

I too had a Pebble Time and while that lasted seemingly forever, I'm very very happy with the Series 3's battery life and charge time.


My Series 1 used to last 2 full days. Now when I wake up it's at 40% and will die around 5pm (on the 2nd day). Pretty sure the watchOS 4 update caused the battery life to drop significantly.

I also have a Series 3. What app do you use to track sleep?

I'm using AutoSleep, but it seems to really overestimate the time I'm actually asleep, and as far as I can tell it doesn't take into account tossing/turning during the night (either that, or I'm getting way better sleep than I think I'm getting).


You can tweak its sensitivity +/- dramatically, even for the past night, in the Tweak tab in the app. That did it for me.

I'll try that. Thanks for the tip!

With my Pebble Time I can travel for a week and not even bring the charging cable. Still disappointed that they went under.

with my Apple Watch I am more physically fit.

And I can leave my phone in my pocket and still not miss anything important. Improved focus.

And I don't have to put my phone under my freaking pillow to get my alarm to go off at the right time.

And I also have a night-light on my wrist, which is super useful when I gotta get up and take a pee or something.

Only need to take it off when I wake up for about half an hour. While I shower and what not. Works for me.


   >>with my Apple Watch I am more physically fit.
Can you expand on that? I would be careful before making such a claim because of a watch; unless of course you're referring to the extra weight of the device... /s

   >>And I can leave my phone in my pocket and still not miss anything important. Improved focus.
Have you actually used a Pebble Time? You also get notifications, so I don't see why you would miss stuff

   >>And I don't have to put my phone under my freaking pillow to get my alarm to go off at the right time.
You don't need to do that either with any smartwatch. Why do you compare it to a phone?

   >>And I also have a night-light on my wrist, which is super useful when I gotta get up and take a pee or something.
Other smartwatches (and traditional watches) including the Pebble Time have backlight.

   >>Only need to take it off when I wake up for about half an hour. While I shower and what not. Works for me.
With the Pebble Time (or others) you don't even need to do that; for an entire week.

I'm sure there are good reasons to purchase an Apple Watch specifically, but I believe the ones listed above are true for most smartwatches


I feel like there is a huge difference in the capabilities of the Apple Watch and the Nokia Steel. The Steel has a motion sensor, accelerometer, bluetooth and maybe wifi? The Apple Watch, needless to say, has a lot more than that: GPS, GSM, wifi, bluetooth, heart rate sensor, accelerometer, light sensor, etc. I'm sure the Nokia Steel fills a niche but I feel like the Apple Watch is much more interesting from a capabilities standpoint.

And if the Apple watch also came with a fold-out bottle opener and a tool for getting stones out of horse's hooves, would that make it a better watch? It's not about the number of "features", but whether it's a productive, convenient item to have in your life. A 1-day battery life on a watch is a serious flaw.

The only smartwatch that was ever a good watch, at least that I've seen, was the Pebble. The one-week battery and always-on sunlight-readable screen are the two giant features that make it a comfortable object. It's very disappointing that nobody's making anything like that anymore.

The Nokia Steel looks like a fantastic watch, although not much of a smartwatch. I'd still rather have one than an Apple watch.


When you’re critiquing a hardware device, it might help to be factually accurate instead of basing criticisms on your own misunderstandings.

First, the Apple Watch easily lasts at least two days with headroom. Second, most of those extra hardware features — which you essentially dismiss outright — directly support what is likely the largest selling point and driver of demand: fitness tracking.

To be honest, at this point, the Apple Watch is to a wristwatch what the iPhone is to a traditional cellphone. Arguing that it’s not the best (or even a particularly good) timekeeping device isn’t particularly compelling, because such functions are secondary or even tertiary uses of the device. It’s about as silly as reviewing smartphones solely on their telecommunications stack, which is likely amongst the least-used features of a modern smartphone.

It’s fine if you really just want a timekeeping device, but your criticisms are going to go over about as well as someone wishing for a return of the Nokia candybar days.


> It's not about the number of "features", but whether it's a productive, convenient item to have in your life. A 1-day battery life on a watch is a serious flaw.

And the Apple watch is a productive and convenient item to have in many people's lives. I have a series 2 and easily get 2 days of battery life. Apparently the series 3s can get 3 days.

For me at least, 1 week battery life is in some ways worse. It means that I will not charge it regularly, and it will likely die when I want to use it. Right now I never think about the battery on my Apple Watch, and just charge it when taking a shower in the morning.


There is Amazfit Pace on the low spectrum of price and Garmin on the higher end. All with transflective screens and battery that varies from 4 to 10 days, depending if you want to receive notifications on your watch (and how many, Slack for example is pretty heavy due to the number of notifications).

I own a Pace and I am pretty happy with it. Not a complete smartwatch but has all the important features for me: always on screen, stylish, acceptable battery, activity tracking and notifications.


Never heard of the Pace, that's pretty cool. And GPS too! And round! And fits standard bands!

Maybe I'll get one when my Pebble dies.


I believe these watches used to be known under the Withings brand which got bought by Nokia. Very nice design (Max Bill inspired I guess).

Yes, I should've mentioned Withings. I think I've heard good things about them in the past as well, they used to do a scale that now, wow, looks even better [1].

1: https://health.nokia.com/uk/en/body


The steel has no heart rate sensor. Nokias Steel HR has one, however after the aquisition the app quality tanked [1], and my SO had to return both Steel HR and scale for this reason. Connection issues, synch problems and the needed weekly factory reset for both were too much of a hassle.

[1]: https://play.google.com/store/apps/details?id=com.withings.w...


Yeah I was going to say, that design looked really familiar.

I have a lot of friends who love the Steel!

Cardiogram is compatible with Apple Watch, and launched for Android Wear in October:

https://itunes.apple.com/us/app/cardiogram/id1000017994?ls=1...

https://play.google.com/store/apps/details?id=com.cardiogram...

Our goal is to support any device with a heart rate sensor over time.


Maybe the app I'm using for sleep tracking does not track enough things (sound is tracked with the phone, for instance, but heartbeat is tracked with the watch), but it only seems to consume 3-5% of battery of my watch (Series 2, small). When I started tracking my sleep I changed my charging pattern from night to "while preparing breakfast" (after waking up) and "while at shower" (evening). Of course, 8 months is a different game :)

I use my series 2 for sleep tracking and as an alarm clock every night. The battery lasts for ~2 days, and since the battery is so small it charges really quickly, so topping it off before/after bed for a few minutes is enough so it's always charged.

What app do you use for sleep tracking?

Sleep Watch, it runs automatically so I don't have to bother with it. I'm not sure how accurate it is, though.

I use AutoSleep, which has been very accurate after tuning for me. (not OP)

Battery life has come a long way since the original Apple Watch (which I had to charge every night). With the Series 2 I only use about 15-20% of the battery during the day and can charge it back up to 100% in the time between getting into bed and going to sleep.

I recently got an Apple Watch series 3 (non-cellular) and I can use it for three days before needing to charge it. It’s my understanding that this wasn’t the case with the previous versions, the series 3 battey life has been drastically improved.

My series 2 lasts two days. It'd be nice if it lasted three so I wouldn't have to take my charger with me on weekends away, but I haven't found it to be much of a pain either.

The other impediment to sleep tracking with Apple Watch is its heart rate measurements. One can get acceptable sleep tracking just from accelerometer data, but to get something like Fitbit's sleep tracking you need to measure heart-rate variability. Apple Watch's intermittent heart rate measurements (outside of workout mode) might make certain health metrics difficult to gather without killing the battery.

Theoretically Apple could catch up with Fitbit in this area at some point and achieve continuous 24-hour heart rate measurements.


The apple watch measures heart rate every 5 minutes, and it does do HRV as of watch OS 4.

Fitbit Ionic "Stores heart rate data at 1 second intervals during exercise tracking and at 5 second intervals all other times" - https://www.fitbit.com/shop/ionic#specs

I don't know how much useful information you can get from the variability of a metric you only sample every 5 minutes.


The parent's description isn't 100% accurate but it's accurate enough. The Apple Watch tracks every 5 minutes unless it detects an elevation in heart rate. If it detects something higher than 10 or 20 bpm, it polls more frequently. The standard, resting heart rate polling is at 5 minutes which, according to everything I've read, is more than accurate enough for most medical uses.

Looks like Nokia bought Withings (or maybe it was always a subbrand)?

Anyways, I wish somebody would built a watch that split the difference between the Steel and the Apple Watch. So far, the Garmin Fenix 5S is the closest I've found, though I prefer an analog face (or good digital rendering) and a slightly less casual case.


I charge my gen 1 Apple Watch for ~45 minutes before bed every night and in a year and a half have only run out of battery during the day twice. I use it for sleep tracking every night.

The watch charges so quickly that dropping on the charger at 9pm for a bit is all I ever need.


Damn, their app ia simple and beautiful - I wish fitbit or garmin connect would work like this!

Also, Garmin, you do 24/7 fitness tracking but still no smart alarm after years?


This is interesting, I was searching for replacements for Pebble since the acquisition and this is a promising option.

Same here! I love my pebble classic but it's starting to wear out. This looks like a decent, minimal alternative.

I only lose 10-15% of a charge at night.

I'm skeptical about the accuracy of applications like this running on Apple Watch. I bought the Apple Watch Series 3 precisely because of articles like this. I now routinely get messages about how I've achieved my standing goal - while lying in bed watching TV or working on my laptop. If it can't accurately tell whether you're lying down or standing, it is unimaginable to me that it could be remotely accurate in detecting far more complex conditions.

You get the message that you achieved your standing goal right after the hour in which you achieved it. So it's entirely possible that you stood at 22:10, went to bed at 22:30, and get the achievement message at 23:00.

That's certainly possible, though I have noticed more quirky behavior than just this - the stand goal was just the most amusing one to me and that is why I posted that particular one here. I sometimes meet the exercise goal on a very sedentary day, while I don't meet it when I vigorously exercise, etc. I have a two story house and I climb and descend the steps probably 15+ times per day, yet for the current week, it shows that I have climbed 4 flights of steps. Then other days it seems like it's right on the mark. It just seems to be all over the place.

As the old joke goes, How do you measure altitude with a barometer?

A: Drop it and measure how long it takes to hit the ground.


When I work from home I have a standing desk, and it looks like the watch does not pick me as standing, so I almost never reach the standing goal there

I have had some entertainment downloading an app that just feeds the accelerometer data to the screen. I did it first as a diagnostic to see if the accelerometer was broken (it was on this machine), but then I have some fun just watching the data on the screen and trying to imagine how on earth you're supposed to extract whether the user is standing or not, or going up stairs, etc.

And I say that from a position of being educated in signal processing, not just staring slackjawed at all the numbers. It's a very noisy signal, even after processing, and while it's correlated with the values you're trying to get out of the data it's not necessarily a very strong correlation. There's a lot of headwinds against getting this data out of the accelerometer. It's a minor miracle that they are as accurate as they are.


I would expect it to be accurate enough to detect the changes: if it is accurate enough to detect stand up and move, it also should pick stood up, moved a bit, didn't stand down. I guess the profile of accelerometer/altimeter data of standing up and moving from sitting is different from stand up and moving without having been seated before.

This happened to me an hour ago. I stood up at my desk and continued working but it didn't count according to the watch. I'd speculate that it wants some lateral movement as well - stand up and get a drink of water, maybe?

It wants you to repeat the process of standing. So, if you stand for an hour, that is only 1 stand procedure. You have to sit for 5 minutes, then stand again.

No, you need to stand and move around for at least 1 minute. Just standing won't do it. The point is to not be sedentary, not just to stand upright.

The stand goal is measured by standing up and moving around for at least one minute https://support.apple.com/en-us/HT204517.

> Even if you stand all day, you still need to move around.

The purpose of the standing goal isn't just to be stood up, it's to stretch your legs and get you away from a desk (it doubles as a good reminder to take a break from your computer screen).


I don't stand still, I move around, probably more often than when I'm working on a chair

It's not about whether you've been on your feet. It's about whether you've been moving that hour. The health benefit comes from moving on your feet every hour. When you're standing still for an hour you might as well be sitting.

The thing is, I'm not standing for hours, I move around, specially when thinking through a problem. I'd expect the system to realise there has been a change there.

How would it actually know if you're standing, especially if you're hands are on a desk? Heart rate, blood pressure?

It's a somewhat confusingly named metric. It actually measures "have you been moving around for at least one minute this hour?", where I think "moving around" is basically defined as generating steps.

My reason for this theory is that I sometimes bounce my leg when I'm sitting, and if I have my arm resting on said leg this sometimes triggers the stand goal for the hour.


Most likely that the absence of movement data is perceived as sitting/standing. Since the stand goal requires standing and moving for at least one minute (the purpose of the stand goal is more to get you moving for at least a minute rather than just stand up - just standing up doesn't count), if you haven't moved for a while (50 minutes in an hour) then it gives you a nudge to do so.

Yeah I have the same issue. It seems the primary key in triggering a stand hour is having your arms down at your side, or moving around a bit - so keyboard posture doesn't work out.

> I now routinely get messages about how I've achieved my standing goal - while lying in bed watching TV or working on my laptop.

I routinely get messages about having to stand up while standing up on my standing desk at work. (Series 2 here)


That’s supposedly expected behaviour. When I’m standing at my rig in the lab it pings me to stand, just like when I sit at my desk. This actually means get up and/or walk about — which usually works to fill that segment of the ring. But I also get false stands when sitting too.

I’ve a series 0 however with the lesser sensor heart rate sensor so I don’t get any of these new cardio features.


Apple calling it a "standing goal" is a misnomer. Its more of a walking goal. I'm pretty confident that they're actually looking for 1 minute of consistent steps every hour.

It doesn't have to be walking. It needs to be one minute of movement in a standing position. You can do jumping jacks or walk in place and it counts.

You are supposed to go for a walk, which is a good thing.

That's right - the notification itself tells you to get up and walk around for a minute.

After 1 year with the Series 2, I never had an issue with he watch not understanding properly if I was active or not, standing or not. I did calibrate it on the first day though: https://support.apple.com/en-us/HT204516 and http://www.idownloadblog.com/2015/05/12/how-to-calibrate-app....

I have completed the calorie/move daily goal message while eating a donut. That made me chuckle.

This has never happened to me with the Stand goal, but it sometimes does with the Move goal / calorie count.

It is more like a copy-text / UI design issue. The point is to move around regularly to improve your health.

Danger, danger: heavily regulated waters ahead! Remember these words well: "This statement has not been evaluated by the Food and Drug Administration. This product is not intended to diagnose, treat, cure, or prevent any disease."

I came to say "danger, danger" but for another reason.

I think a whole generation of fitbit users are going to be horrified when their insurance premiums go up based on data that was purchased by insurers. If Apple watch starts making a health play, I don't see why it would be any different.


(I'm a study author and worked on HealthCare.gov rescue team)

The good news is you're protected by the law. Under the Affordable Care Act, insurers can only vary your rate based on your age, location, and whether you smoke. They can't take into account prior health conditions or biomarkers, including novel biomarkers measured on wearables.

Every major Republican "repeal & replace" bill has preserved this part of the law, called "rating factor limitations", although sometimes they vary the constants. For example, the Paul Ryan plan let insurers charge the elderly 5x more than the young, compared to the current law of 3x, but otherwise preserved the rule that insurers can only vary rate based on age, location, and whether you smoke.

What insurers can do is offer you a 30% (maximum) premium rebate for taking healthy actions, like going for a run, enrolling in a diabetes prevention program, or getting your blood pressure checked. But they can't tie the rebate to the outcome of the test (health status), only your participation.

This is a good point, though, and shows a case where regulation helps technology have a positive impact. If people fear their insurance premiums will go up, they'll avoid all forms of diagnosis, not just Apple Watch.


They can't take into account prior health conditions or biomarkers, including novel biomarkers measured on wearables.

Yet.


Apple has been aggressively branding as the company that does not need to sell your data because their margins on hardware allow them to.

It might not always be the case but they can future-proof that claim by keeping that data on your device or protect it using differential privacy.

What might happen if health plans become habitually based on tracking (which I expect will be the case) is that having an Apple watch (or nothing) will be considered suspicious and charged a premium.


Typical PR spin-doctored hogwash. They don't need it, arguably, but they sure as hell want it, as evidenced by:

1. The extremely lucrative market of personal data

2. Their continued collection of such personal data


The lucrative market is for advertising placement which demands that you don’t sell the data that you have (otherwise the targetting that you provide is no longer unique).

Some company do sell the data they have behind the scene (Experian is one of them) but there are few buyers, those buyers can tell their data is of bad quality and they have competitors with similar data, so it is hardly lucrative.

As far as I can tell, Apple’s efforts to make articles readable by removing JS-embeds a while ago, encrypt iMessage, fight the FBI about weakening their security or to fight against cross-site cookies more recently, all that seems both real and representative of how they value their product. I have not seen crowds ask for differential privacy and yet they spend precious time in their recent keynotes about it.


That's nonsense. The market being lucrative doesn't provide any evidence for your claim and your second point is completely false. Apple doesn't collect any of this data. It's stored completely on the device itself and all processing of health data is done on the device.

If you don't see any reason why it would be different, you're not looking.

Apple is aggressively privacy focused. Apple is the most valuable company in the world, and none of that money has come from selling users' data despite the fact that they're sitting on probably the second most valuable corporate dataset in the world (behind Google).

And while Fitbit might also be very clear about their users' privacy, they don't share one quality Apple has: a virtual guarantee that they won't go under new ownership in the near future.


Vitality Health has already started down this path [0], and are doing a fantastic job habituating people to accepting sharing this data.

They have borrowed an identical copy of the auto insurance industry playbook [1]. Start out by offering "discounts" in exchange for tracking, and then slowly the cost of the insurance ramps up, unless you have the discount.

[0] https://www.macrumors.com/2017/10/24/vitality-uk-free-apple-...

[1] https://www.progressive.com/auto/discounts/snapshot/


Current law in the US for individual coverage is that they can vary pricing based on age and tobacco use and can't deny coverage to anyone.

Short term regressions to that seem likely enough. Long term I think you have to consider whether we even stick to anything resembling the model we have now (with mandatory private insurance being something like the baseline expectation).


I work for Fitbit, but I don't speak for Fitbit.

However, this page does speak for Fitbit: https://www.fitbit.com/legal/privacy

For insurers to purchase data on Fitbit users, they'll have to purchase it from the users themselves.


Is there a legal structure in place that would prevent a new owner from rewriting the privacy policy and doing whatever the hell they wanted?

I'm not a lawyer. In high school I took a class that covered some aspects of contract law. Want my speculation?

I do know something about the financial incentives, though. Fitbit's users are its customers. Insurers and employers are a small part of revenue. The people Fitbit needs to keep happy are users.


No need to speculate. Of course it would be reassuring if there was something in the privacy policy that directly addressed the question.

There is stuff like If we are involved in a merger, acquisition, or sale of assets, we will continue to take measures to protect the confidentiality of personal information and give affected users notice before transferring any personal information to a new entity.

But "measures" and "notice" are pretty weak sauce if they can result in the new entity still doing whatever it wants with the data. How about an explicit opt in policy that would actually impact the sale value of the data?


IMO something that specifically addressed an acquisition wouldn't add anything.

Either Fitbit's acquirer is contractually bound by the privacy policy or it isn't. If it isn't, then the specifics about acquisition don't add anything. If it is, then "We pledge...to never sell your personal data" is enough.

FWIW, Fitbit's publicly traded stock is non-voting, so hostile takeover is not possible.


Something about offering to delete a users data before material changes to the privacy policy took effect would be meaningful in more circumstances than acquisition.

It's kinda maybe in there already, but not directly (there is an offer to delete data and the language about continuing to use the service given a change).

I consider the pledge to be nearly meaningless. It's certainly a statement of good faith from the people working at and controlling the company today, but without any sort of legal teeth, it's also little more than a wet napkin.


In some circumstances privacy policies have had legal teeth, "particularly when parties claiming a breach have alleged that they read and subsequently relied on the policy prior to transacting business with the site operator" according to this page:

http://lawprofessors.typepad.com/contractsprof_blog/2010/08/...


Ehhh... one of the big fights a decade ago was over DRM being built into CPUs and displays to the behest of entertainment companies. (They seem to have won that fight, as I now cannot find a lightning to hdmi cable that is “compatible” with Netflix.) During the years over which that controversy raged, I was always struck by the fact that companies like Microsoft and Intel were bending over backward to cater to Hollywood, not their customers.

Fitbit seems like a cool company and I don’t personally distrust them, but I don’t think that particular argument holds water.


The DRM situation was such that Hollywood had the power to deny access to something Microsoft and Intel's customers wanted. Those companies might fear that if they didn't implement DRM, customers would buy a competitor's display or CPU that did implement DRM and let them watch movies.

I don't really see what the analogous situation would be with insurers and Fitbit. Fitbit won't lose customers to a competitor that violates their privacy.


And those that can't provide that data will see even greater hikes. Insurers hate uncertainty.

  “And then you’d guide them through the appropriate final diagnosis, which would be through a blood pressure cuff and then treatment.”
so the intention is that this would generate a warning to go to your doctor?

It's a little more nuanced—the exact next step depends on the health condition, co-morbidities, previous tests you've taken, which interventions are covered by your insurance, etc.

Think this diagram of how Slack decides to send a notification, but for health: https://twitter.com/codevisuals/status/838881724016787457?la...

:)


I'm one of the co-founders of Cardiogram and involved in this study -- happy to answer questions if folks have them!

(EDIT: heading to bed, but will answer questions in the morning!)


What input signals do you have access to through the Apple Watch? Is it only Heart Rate output or do you get the actual interbeat intervals? Any chance of getting to the raw ADC output of the photodiode?

I've written a blog about this a while ago [0]. You're limited to what HealthKit allows you to query and the RR intervals are not available. With the release of WatchOS 4 however the SDNN heart rate variability can be queried.

[0] https://www.vanmil.org/heart-rate-variability-and-brain-wave...


Yes. Apple Watch takes measurements every five minutes (no RR intervals yet). There are a few popular watches, like Garmin, that do support RR intervals.

Interesting, thanks!

I didn't see a reference in the TechCrunch article -- is there a technical paper out?

Edit: A presentation is linked, as "study":

Link to presentation: http://www.abstractsonline.com/pp8/#!/4412/presentation/4522...

A followup: Does the neural network provide some extra insight into what part of the waveform/time-series of pulses is able to provide predictive information?


Correct -- the abstract will be included in a special issue of Circulation: http://circ.ahajournals.org/content/136/Suppl_1/A21042

The study was presented at the American Heart Association's Scientific Sessions earlier today. We included some details beyond what appears in the TechCrunch article in a blog post here: https://blog.cardiogr.am/screening-for-hypertension-and-slee...

I'd expect a more mature version of this work to appear in a medical journal, but that often takes 6-12 extra months.


Couldn't find the measurement process.

What is the common root cause of error? Inaccuracies in measurement or the model classifying data incorrectly?


Sorry for the habitual rant, but I feel the need to point out those are not open articles. Your case is typically something of interest to people who have no reason to have access to medical journal stores and… :(

This thread is not the right place to grab pitchforks, just a reminder open source science was still an on-going effort.


What do you do with the data collected via your app?

Whats the end result of your research?

Did the 70% people you collected data are aware?


The 70% of people were aware—they enrolled in an IRB-approved study with UCSF Health eHeart Study, and went through an explicit informed consent process (as medical research generally requires). We wrote a bit more about the end goal in the last section of the blog post here: https://blog.cardiogr.am/screening-for-hypertension-and-slee...

What are the false positive and false negative rates for sleep apnea detection? Would it be able to detect sub-clinical symptoms?

Second the other comment on technical paper. Is it available?


The accuracies quoted here are c-statistics, or area under the ROC curve. An example operating point for sleep apnea is 97% specificity with a sensitivity of 52% (compared to a 20% diagnosis rate in today's healthcare system). If a specificity of 82% is acceptable, then we can detect about 75% of sleep apnea.

No technical paper yet (this was a cardiology conference), but we have some follow-up work in submission to an academic artificial intelligence conference, so hopefully we'll have a technical paper with more details out soon.


Is it this https://itunes.apple.com/de/app/cardiogram-for-apple-watch/i... ?

Could you please state, if the app can currently do that?


Do you have plans to commercialize this?

> Do you have plans to commercialize this?

Cardiogram is a venture-backed company.


That's correct. More from our investors:

  https://a16z.com/2016/10/20/cardiogram/

  https://homebrewvc.tumblr.com/post/152080260668/your-heart-beats-102000-times-a-day-cardiogram

As someone with diagnosed apnea I have so many doubts about this.

Machines to diagnose apnea have lots of sensors including electrodes, "chest-band" pulsometers, and "saturationmeters". I have asked doctors about capabilities of smart devices for apnea control or detection, all I got was laughs.

Sorry but I think this is not true. Go and see your doctor.


They say "90% accuracy" for apnea - definitely not high enough for a diagnosis. The study's lead investigator is talking about this as a screening test, not as a firm diagnostic.

And given the involvement of UCSF in the study, I'd take this pretty seriously - just make sure to pay attention to the claims the study is actually making, and not the science-journalism hype.


It’s good enough to raise awareness and I believe sleep apnea could use that.

> Machines to diagnose apnea have lots of sensors including electrodes, "chest-band" pulsometers, and "saturationmeters".

It seems like a fallacy to claim that since diagnosis required complicated equipment in the past, that it will be necessary going forward. Machine learning is going to make tons of current diagnostic equipment look archaic in comparison.

> I have asked doctors about capabilities of smart devices for apnea control or detection, all I got was laughs.

Medical doctors are rarely up to date on the latest technology. I wouldn't be surprised if my doctor has never heard the term machine learning before in his life.

Also, their livelihood kind of depends on this technology not being available. Just food for thought.


It seems a fallacy to say that "machine learning" will replace the sensors needed to produce the data required to make a diagnosis. Sure the sensors can always be in a smaller, simpler form factor, but that's engineering the sensors specifically. You need data to make a diagnosis, and the magic of machine learning can't replace raw data coming from sensors.

I see this kind of confusion a lot, where "machine learning" is mistakenly used to mean obsoleting sensor data instead of obsoleting the hardcoded processing of said data.


> It seems a fallacy to say that "machine learning" will replace the sensors needed to produce the data required to make a diagnosis.

I agree. Good thing I didn't say that :) My comment implied exactly what you wrote, that the machinery will get smaller, and machine learning will be able to make more accurate predictions from a more limited data set. Although, if we consider that the Apple Watch will be strapped to your wrist for 16 hours a day, it might actually turn out to be an expanded data set, but a smaller number of sources.

> I see this kind of confusion a lot, where "machine learning" is mistakenly used to mean obsoleting sensor data instead of obsoleting the hardcoded processing of said data.

I believe that's a misinterpretation on your part. I haven't seen anyone claim that machine learning doesn't require data from sensors.


> It seems like a fallacy to claim that since diagnosis required complicated equipment in the past, that it will be necessary going forward. Machine learning is going to make tons of current diagnostic equipment look archaic in comparison.

As someone sais below "ML is not going to replace raw input from sensors."

> Medical doctors are rarely up to date on the latest technology. I wouldn't be surprised if my doctor has never heard the term machine learning before in his life.

Current dignose equipament is big, invasive and unconfortable; often they require you to stay one night at the hospital.

There are tests programs about using those devices to replace current tech (obvious reasons: less costs, less invasive tests, subject owns the hardware…)

This kind of procedures are not used because "doctors are old and non-tech people", it's because it's not working.

Sure, apps and smart devices could replace some day those devices but they need more and accurate sensors. Extrapolation of data from a good-enough pulse rate sensor it's not a replacement.


While I understand and mostly agree with your general sentiment, I think you're downplaying the extent of how valuable the machine learning is in this process. The reason why we need all these sensors and machines currently is because we need a reasonable picture of how all that data works together so that a human being can look at that picture as a whole and make some deductions.

This isn't the first time or the last time that computers have been able to find patterns in much simpler measurements simply because they're not human. A similar example, although not quite as advanced, is the ability for a computer to extract sound information from a black and white video. Computers have been able to recreate sounds from behind double-paned glass by analyzing vibrations captured through videos. Humans have had to rely on various arrays of lasers, sonar, and other directional instruments to get 1/10th of the accuracy that a computer algorithm has been able to achieve from a simple, low-quality video camera. The point is that more machines and sensors doesn't always equal a better diagnosis. Better analysis of existing sensors and tech, even if it seems to be lower fidelity, can actually yield more accurate results.


> As someone sais below "ML is not going to replace raw input from sensors."

You've misinterpreted my comment. I didn't claim sensors weren't required. I said that it's a fallacy to claim that the current diagnostic equipment is required. There are sensors on smart watches. Cardiogram is using data from those sensors.

> This kind of procedures are not used because "doctors are old and non-tech people", it's because it's not working.

It's not working yet (at least not enough to make a 100% accurate diagnosis). No one claimed it was. I hate to break it to you, but the equipment currently used in sleep studies isn't 100% accurate either. Neither is the human interpretation made by sleep doctors. However, to say that doctors are laughing it off makes those doctors look stupid, not Cardiogram. Based on the studies they've run so far, it's not hard to envision a future where smart watches and apps will replace a large percentage of doctors whose main source of income is sleep studies.

> Sure, apps and smart devices could replace some day those devices but they need more and accurate sensors. Extrapolation of data from a good-enough pulse rate sensor it's not a replacement.

Your claim seems to be that sleep doctors need input from a large number of sensors, so it can't be possible to make a diagnosis from a single heart rate sensor. It would appear that Cardiogram has presented evidence to the contrary:

https://blog.cardiogr.am/screening-for-hypertension-and-slee...

If you disagree, make a specific criticism of their evidence.


Nobody is saying to use this to replace going to the doctors.

It's for you to just download the app, try it for a night, and see the doctor if the result concern you. It doesn't have to be hundreds percent accurate.

Are you against teaching people to do self check-up of breast cancer, too? Because I'm sure medical process of checking for breast cancer requires more complicate tools than just feeling your breast with a hand.


Apnea has obvious signs (fatigue, day sleep, depression, …) that are the equivalent of the "self check-up" in your example of breast cancer.

If you suspect that you have apnea episodes you should go to a doctor.


80% of people with sleep apnea don't know they have it. Unfortunately, the signs are not obvious to most.

I'm against breast self exams, and so is the US Preventive Services Task Force. The evidence indicates there's no benefit but instead the practice calls harm (in anxiety, overdiagnosis, unneeded biopsies, etc.

https://www.uspreventiveservicestaskforce.org/Page/Document/...


> I have so many doubts about this. As you should. "New study suggests" is basically the same as saying "new study didn't find". It's clickbait or a PR piece.

I went to sleep wearing my Apple Watch one night and awoke to a few hundred irregular heartbeat notifications.

They weren’t legit. The watch just can’t get a good heart rate reading if it’s not situated perfectly on your wrist, and while sleeping it was pushed one way or another often enough to be useless.


That's probably the case with all HR monitors, no? If it's not securely in place, any HR monitor will have the same problem. I bought a knock-off velcro wrist strap for my watch and it keeps it in place. I have mixed success with traditional watch straps as my wrist size always seems to be in-between notches.

All the optical wrist-based ones, probably.

I imagine if I wore my chest strap monitor I'd have better results as long as I put some kind of jelly on the electrodes to make up for the lack of sweat through the night.


Interesting, but are people really sleeping with their Apple Watch on? I don't you need to charge it every night?

You are overestimating how long it takes to charge an Apple Watch and underestimating the battery. 30-45 minutes while you're showering and having breakfast is enough.

That's why you buy two, one for charging one for wearing.

You could sleep with it and put it to charge when you wake up and it will probably be charged when you leave the house.

Hey, just let the bed charge the watch by induction. Health is important!!!

Does anyone know how we can use this today?

Good question—this clinical validation study is a first, necessary step to prove that this technology is accurate enough to be used in a real-world setting, but the research study alone is a milestone rather than the end goal.

We're now working on implementing care pathways within the Cardiogram app. If all goes well, you should hear something from us in the next few months.

(In general, a digital health startup requires more patience than most other types of startups: you need to run rigorous, IRB-approved clinical studies to prove the technology really works; work with health systems and insurers rigorously show health economics; and comply with applicable regulations like HIPAA and FDA. There are just a lot of details to think through.)


Okay. Any way for me to get updates on when it's available for use?

It'll appear directly in the Cardiogram app. (Would you personally find it helpful to have another mechanism, like a mailing list?)

Yeah. Don't have an Apple watch but if it can detect sleep apnea or other issues, I'd consider buying one for my partner. Mailing list would be ideal.

Yes I want to know this too. I feel like I may have mild sleep apnea and would love to see what my results look like!

It's quite interesting - the way we have evolved as a society has completely broken the link between "us and our body". We have stopped listening to it a long time ago. If we really listened to our bodies, we wouldn't eat/drink/consume stuff we don't need, we wouldn't be overweight, and we'd rapidly be aware of any health-related imbalance which needs restoring.

But no, we forget our original (and free-of-charge) way of functioning and desperately turn to commercial and unnatural solutions.

As a start, I recommend looking into the mindful meditation technique, championed by pioneers like John Kabat-Zinn.


If I really listen to my body it's telling me to eat and drink everything I can get my hands on. That's what it's programmed to do. What you're trying to say is that the more primal cravings should be ignored and we should be very conscious of what we eat / drink / do.

Body still tells me to sit in a couch and eat myself to an early death though.


If your body appears to be telling you that it is probably because there is something wrong with the way you’re living your life.

Now, you can wait until you’re on your death bed before you realise it, or you can pay attention to this signal your body is giving you right now that something needs to change in your life. The latter is too terrifying to consider for most people, and the former feels quite distant, so most people just carry on as “normal” with their mostly dysfunctional lives.

Listening do your body doesn’t mean acting on every impulse you feel. If I feel like smashing something, that doesn’t mean I need to immediately do it. But it is wise to ask myself, why do I feel this? Am I angry? Why am I angry? What just happened that cause this reaction in me? What clarity can this anger bring me? Should I do something about it?

The same for your body’s signals. Why do you feel like this? What has caused this unhealthy craving? Your hypothesis that this is an evolutionary pressure is interesting but easily disproved by speaking to a wide cross section of people with balanced lifestyles who don’t seem to be perpetually driven by cravings to eat everything in sight. Are they fundamentally different from you? Probably not. So what is different about the choices you make each day that lead you to live in a constant state of craving food and drink, rathet than a more balanced state of being?

Those are important but tough questions. Most people don’t bother.


Hypertension presents no symptoms and left undiagnosed can lead to kidney failure. It's often referred to by doctors as 'the silent killer', and you don't need to eat like shit and be fat to have it.

Going to your GP 2-3 times a year and getting your blood pressure checked (and your blood tested for the 'regular stuff') solves the problem. 'Listening to your body' pseudoscience doesn't really solve the problem.


Meditation isn't "listening to your body" BS, it's a discipline that needs to be taught and practiced and developed as a skill. You can't just sit, close your eyes and try it on your own. And there are actual studies [1][2] that show the benefits (particularly related to stress and blood pressure), so not exactly pseudoscience either. It isn't particularly difficult to learn to recognize the feeling of different blood pressures in the body during meditation and develop some measure of conscious control over it.

GP is absolutely right that we're a society that has taught ourselves to tune out important signals from the body. If we can unlearn that and better attune ourselves to the relevant sensations in the body, we'd be a lot healthier.

[1] https://www.ncbi.nlm.nih.gov/pubmed/19798037

[2] https://news.harvard.edu/gazette/story/2012/11/meditations-p...


When I had a medical test for my Australian permanent residency - full thorax x-ray, HIV test, the works - I asked the doctor doing the final once-over if "a routine checkup" is something I should be doing.

He asked me how old I was. 30-ish, at the time. He asked me if I felt healthy. Yep.

He said no, don't bother.


Pretty sure it's an American thing, the idea that you need several checkups a year as if you were a car that constantly needs maintenance, and we can debate forever why that is (my bet is money). It doesn't seem to happen in countries with socialized healthcare where care is distributed accordning to criticality and population risk rather than what generates the most revenue. If you're in your 20's or 30's the odds of you being affected by something are miniscule. It doesn't hurt to get a checkup every few years I guess, but several times a year is ridiculous.

Exactly; no need to pressure an already overworked health care system with a "what if". Especially when you can check a lot of basic things yourself with e.g. a smartwatch, as this article suggests. I'm fairly sure an at-home BP measuring thing isn't too expensive either.

It goes against typical advice, then.

Or don't do it, maybe you'll develop cancer and not find out until it's too late!

EDIT: It's also free to go to the doctors in Australia so at least go once a year and do your routine bloods. Abnormalities can catch things before they progress too far that you have worse outcomes. You're paying for Medicare, use it! :)


The "annual physical" is not evidence based. It might seem to "make sense," however, that's not how science works.

Not only that but there's no such thing as "routine bloods." The reason I say that is because 1) to my knowledge there's no evidence based group that recommends any sort of blood test annually and 2)I have been to many GPs over my lifetime (~10?) and I was well into my 30s until I saw one that ordered routine bloodwork as a matter or course. All the ones before that just listened to my heart, took vitals, felt organs, etc. Then that doctor retired and I her replacement also ordered routine bloodwork. The two doctors differed greatly on which test they considered routine.

https://sciencebasedmedicine.org/re-thinking-the-annual-phys...

>Not too long ago, the “if it ain’t broke don’t fix it” mindset changed. It became customary for everyone to have a yearly checkup with a doctor even if they were feeling perfectly well. The doctor would look in your eyes, ears and mouth, listen to your heart and lungs with a stethoscope and poke and prod other parts of your anatomy. He would do several routine tests, perhaps a blood count, urinalysis, EKG, chest-x-ray and TB tine test. There was even an “executive physical” based on the concept that more is better if you can afford it. Perhaps the need for maintenance of cars had an influence: the annual physical was analogous to the 30,000 mile checkup on your vehicle. The assumption was that this process would find and fix any problems and insure that any disease process would be detected at an early stage where earlier treatment would improve final outcomes. It would keep your body running like a well-tuned engine and possibly save your life.

>We have gradually come to realize that the routine physical did little or nothing to improve health outcomes and was largely a waste of time and money. Today the emphasis is on identifying factors that can be altered to improve outcomes. We are even seeing articles in the popular press telling the public that no medical group advises annual checkups for healthy adults. If patients see their doctor only when they have symptoms, the doctor can take advantage of those visits to update vaccinations and any indicated screening tests.

>The physical exam of a healthy, asymptomatic adult is unlikely to reveal any significant abnormality (1) that would not have been detected eventually when symptoms developed and (2) whose earlier detection and treatment would reduce morbidity and mortality in the long run.

>A directed physical exam is sometimes indicated in patients with risk factors for specific conditions. A Pap smear is indicated in most women, but not every year, and the accompanying pelvic exam is likely a waste of time.

>For healthy adults between the ages of 18 and 65, The American Academy of Family Physicians (AAFP) recommends only these components of the traditional physical exam:

>For men, a blood pressure measurement. >For women, a blood pressure measurement and a periodic Pap smear.

>They have other recommendations including vaccinations, counseling, and screening tests; but none of those require a physical exam.

>There is a general perception, among the public and among doctors, that there’s no such thing as a bad screening test, that early detection is important, that knowing is always better than not knowing. If something is wrong with you, you need to know because, if you find a problem in time, it can be treated effectively to prevent morbidity and mortality. If you get a checkup and everything looks OK, you can breathe a sigh of relief and relax. Unfortunately this is all wrong.

>A recent book explains why: Overdiagnosed: Making People Sick in the Pursuit of Health, by Drs. H. Gilbert Welch, Lisa M. Schwartz, and Steven Woloshin. It’s a comprehensive explanation of how test results make people sick and why visiting a doctor can be hazardous to your health.

>For a healthy, asymptomatic patient, the physical exam with the laying on of hands and stethoscope and other rituals is pretty much meaningless. If nothing is found, it can produce false reassurance. If something is found, it is not likely to prolong the patient’s life and it has a significant likelihood of leading to harm from unnecessary treatment or from a diagnostic cascade of tests, unnecessary surgeries, unnecessary expense, and unnecessary worry.

https://sciencebasedmedicine.org/a-skeptical-look-at-screeni...

https://sciencebasedmedicine.org/overdiagnosis/


> 'Listening to your body' pseudoscience doesn't really solve the problem.

Well, all the big health issues in our societies are about giving a shit about your body.


"Listening to your body" is an un-falsifiable statement, though. What does it mean to listen to your body? If you don't notice anything, are you not listening hard enough?

Wouldn't you say that things like hypertension are only the result of something?

What I am thinking of is this:

Listen to what your body tells you when you do the things that lead to problems like hypertension.


Genetics can be a factor, so you have to get it checked.

A bloke my sister knows (she's a nurse), was diagnosed with 'hypertension not otherwise specified', at the ripe old age of 31. But they didn't catch it early because he 'felt fine'. He wasn't overweight unhealthy. But he needs new kidneys now!

It's misinformation to say otherwise. I mean, I live in Australia, where you can visit the doctor for free, so it's like, no skin off our back to go and see one once or twice a year.

EDIT: Also, I have hypertension AND sleep apnea. I am overweight. It is most likely the cause, though we do have a history of high blood pressure in the family so until I drop 20% of my weight I'm not going to know for sure. I don't disagree with that you need to look after yourself physically and be glad if you don't have any other illness that ain't your goddamn fault.


> A bloke my sister knows (she's a nurse), was diagnosed with 'hypertension not otherwise specified', at the ripe old age of 31. But they didn't catch it early because he 'felt fine'. He wasn't overweight unhealthy. But he needs new kidneys now!

It's possible that kidney disease caused hypertension in that case. It's pretty much impossible to prove which one came first in many cases. None of this argues in favor or against regular checkups, it was just something I thought was interesting to refer.


I recommend reading Wilber’s Grace and Grit for a more balanced view of “aren’t things like hypertension the result of something”?

Otherwise you might find yourself on a slippery slope where you start implying that cancer sufferers brought it on themselves, etc.

Wilber’s book, about his wife who died of cancer, offers a powerful perspective that’s also grounded in mindfulness and explicitly discusses this topic and how to approach it more sensitively.


Mmmh… The watch tells you to meditate and can even help you do it.

It’s less about “listening to your body” than remind you regularly about obvious things. Some people (like me) can focus quite easily (which is great to write code) but can get lost in their task. I know that I regularly realise that the sun is up when I keep on coding after dinner. That’s eight to ten hours that have gone by without me moving an inch. Having a reminder to go to bed has helped.


97% accuracy seems high, but with an approximate sales number of 10 million Apple watches, it will give 300 thousand users an incorrect diagnosis.

Which is not so bad. They can follow up and figure it out.

Some people will go for decades before getting a proper diagnostic - or even die before they get one.


Only 300k would have to, if this device would tag its diagnosis with “but it is incorrect”. Otherwise, up to the full 100% would have to “follow up and figure it out”.

Also, of those 10M, where would those told “your blood pressure is OK” or “you don’t have sleep apnea” get the hint to “follow up and figure it out”?


Unfortunately, most arrhythmias are benign and do not lead to further treatment from a physician. So you are taking 300,000 worried people wrongly to a general practitioner, who in fact can do nothing but write down the event and perhaps correlate it with other data.

It's worth always keeping this in mind, and not just taking extra worried people to the gp, but also giving unwarranted peace of mind to those who have a problem.

Even just having them get in a car and go to their GP, or increase their stress levels, have the potential to cause harm. It may be an exceedingly low risk, but applied to a large enough number of people and it starts adding up. When applied to conditions that not everyone has, this can easily tilt the balance.

Consider a hypothetical condition and a population of 1000000. If 1% has the condition at any one time, and the accuracy of the test is 97%, then 10,000 should have the condition. Of those we'd expect 9,700 to get the correct diagnosis; 300 would remain undiagnosed; But 3% of those who don't have the condition - an additional 29,700 - will be wrongly diagnosed.

Let's say this is a condition where 10% would die untreated without this new intervention, it'd be tempting to suggest we've saved 970 people. But early detection might not make that much of a difference - for many things the most severe cases will be noticed anyway and/or late detection is still sufficient to save the patient.

Let's assume for this hypothetical condition, earlier detection will save 10% more of those. In the above that'd save 97 people. Awesome. But lets say it causes 10% more of the undiagnosed ones to die because they wait longer. That's 3 more dead. So net 94?

But if only 0.33% of those wrongfully diagnosed end up dying as the result of the misdiagnosis, be it car crashes on the way to the GP, or the visit leading to a follow up error leading to an intervention that is generally but not always safe, then the wrongful diagnosis would kill 98 people.

I'm not saying that would be the case here. But it's an illustration of why it is important to consider not how many problems are detected, but what benefits early detection gives vs. the cost of wrongly diagnosing people, even when it seems benign enough (just another test, and they'll confirm it was a mistake?)

E.g. there's been pushback against large scale breast cancer screening for this very reason: They're very successful at finding cancer early, but most gets found eventually anyway, and the improvement in long term survival rates from the earlier detection is reasonably small.

Couple this with low overall rates of occurrence across the total population at any given point in time (cancer rates look high when considering "will X get cancer at some point in their life?", but most people will be free of cancer at any given single point in their life; even most people who have had cancer), and the impact of false positives and potential health implications of a (very small but still existent) risk involved in even something "simple" like a biopsy, and it's unclear if large scale screening is ethical (over general populations; screening of groups with sufficiently high risk factors is another matter entirely).


Very rough back-of-the-envelope calculation:

we have 3000 GP's in the Netherlands. Lets say the proposal will cause 30.000 patients in the Netherlands to go to the GP incorrectly per year. That will cause 8 man-year of time (20 minutes per consult, including paperwork), 4 workable consulting-hours per GP, 300 days per year. That's 0.3% of all available GP time, wasted. That time needs to come from somewhere: patients having actionable issues. Healthcare is not an unlimited resource.


> Which is not so bad. They can follow up and figure it out.

Following up is not free. So that's cost incurred for the ones who have no problem in the first place.


Think about the cost of not following up. Personal health, productivity, life quality.

Also here in the Netherlands we have healthcare. I'd rather see these emerging devices offer good warnings to go see a doctor then people not going at all or going to often for every little thing that they feel and googled on the internet because they are to unsure.


See my long response elsewhere. Depending on incidence rates and whether or not the effect of following up earlier is sufficiently large, there can - and sometimes is - a net detrimental effect to wrongfully diagnosing someone with a condition. Even when following up is relatively simple and risk free, and lives are saved by finding conditions early.

E.g. cause an extra million doctors visits due to wrongly diagnosing people, and the incident rate of e.g. extra infections transmitted from another patient, or compounded misdiagnosis leading to extra interventions which have some risk of failure, or even just accidents on the way to the doctors office, very quickly leads to extra harm just from the sheer volume.

That extra harm might be worth it, if the early diagnosis of others save enough lives - the same way seat belts kills some people but save far more. But it also might not, e.g. if the condition is rare enough, or early detection have little enough effect in improving outcomes.


Give 1000 people apple watches with the app versus 1000 people other smart watches without the capability. At the end of the year, find out how well diagnosed the two groups are.

Yep—this is a good sketch of a randomized control trial to measure the primary outcome, change in diagnosis rate.

Secondary outcomes you might look at are overall change in medical spend, e.g., is this intervention cost-saving, or at least cost-effective? And change in treatment, which may be independent of diagnosis rate.

The tweak I'd make is that you don't need to give out Apple Watches, Garmins, Fitbits, or Android Wear watches. People are buying these devices on their own already. Given ubiquity of hardware, continuous screening may be achievable at close to zero marginal cost.


Before doing an expensive and inaccurate field experiment, we can calculate the expected cost per QALJ [1] by taking the following variables:

- Estimated accuracy and recall of the test

- Estimated probability of user doing a follow-up based on feedback from watch

- Estimated price of a doctors visit

- Estimated probability of a follow-up test using medical grade equipment (based on additional data)

- Estimated cost per QALJ of the extra test and follow-ups

- Estimated prevalence of arrhythmia among Apple Watch users based on age-group

I expect, but do not know, the Apple Watch tests to prove to be a net-negative compared to doing nothing.

[1] Quality Adjusted Life Year


Aetna, the US health insurance company, is planning to give away or offer a reduced price for Apple Watches to its 23 M customers. Apparently, following prescriptions is one of several planned apps to help increase health and reduce costs. https://www.engadget.com/2017/08/15/aetna-in-talks-for-disco...

“Versus Apple watches without the app” is the better control, I would think. It would make it easier to keep their GP’s in the dark as to who is in the control group and who isn’t.

It would help as its more of a “you should see a doctor” test rather than a official diagnosis.

I’m fine with it providing the marketing does not mislead people into thinking it’s foolproof. Basically, they need to say it can only detect obvious cases, and minor/rare cases are missed.


For this project like for self-driving cars, I believe that releasing machine-learning early in the wild “too early” has a dynamic benefit: we learn earlier and the bump in misdiagnostics pays for better diagnostic down the line.

To argue that, you need an estimate of the rate of progress with and without a deployment which I would assume is not a stretch.


I love my Apple Watch series 1.

The one thing that’s frustrating is installing any apps on it takes a really long time.


In the Future Everyone will Wear a Smartwatch

https://h4labs.wordpress.com/2015/07/28/in-the-future-everyo...


Unless you have a heart murmur like I do.

Then: it just lies.


I suspect that the technology (and data) from cases like yours will make building either dedicated software or hardware for your case cheaper.

My snarky reply: "It's too bad I have to pay for the priviledge of being included in a data set used for product refinement."

Also: no. When I was working as an intern at St. Jude Medical we were training neural networks to recognize heart conditions and adjust other measurements. It was 2002. That work was later published and widely adopted. Modern medical hardware can account for it (inexpensive or older hardware often asks the technician to diagnose and calibrate the machine).

It's very frustrating that Apple continues to pretend economy of scale is breaking into new territory.


Adapting your model from 2002 is not a trivial operation. Apple and others are building Machine learning platforms that make that work easier but saying that you worked on it, therefore it should be done by others comes off as a little naive about software integration. Going through other people’s code is not instant or trivial.

I'm saying Apple's system fails to catch edge cases that prior ML systems, system using a tiny fraction of the computing power and ultimately based off animal data, achieved.

I think the job they're doing is unimpressive. I think it's doubly impressive to imply I have to buy a watch to help them. I don't see why they wouldn't pay me for my unique data. I don't need or want to help them build a shoddy diagnosis product.


I think this is a great feature provided Apple won't collect the data.

[edit] Btw, I really don't mind getting down voted, but would appreciate to see your opinion on why you think Apple collecting that data is a good idea.




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | DMCA | Apply to YC | Contact

Search: