If you're an engineer who wants to make the vision Steve paints into a reality--but for everybody, not just for Apple Watch owners--I'd love to talk with you.
I'm a co-founder at Cardiogram. In multiple N=14,011 studies with UCSF Cardiology, we've shown that heart rate sensors on consumer wearables can detect multiple major health conditions like sleep apnea, hypertension, diabetes, and atrial fibrillation (https://cardiogr.am/research). These conditions affect more than one billion people worldwide, and are frequently undiagnosed -- for example, more than 80% of people with sleep apnea don't know they have it.
We're live on iOS and Android, and we could use help in several areas: mobile and frontend engineering, senior machine learning engineers, and payer relations (the reimbursement from health insurers that Steve talks about). If that sounds like you, please email me at brandon@cardiogr.am!
> heart rate sensors on consumer wearables can detect multiple major health conditions
It sounds so logical and inevitable where this technology is heading for health care. Over 10 years ago, I paid $1200 to wear a Holter monitor[1] for 24 hours.
In contrast, a consumer wearable like Apple Watch is $399. Arguably, the Holter monitor fee also included the doctor's time to read the ECG recording. But it also seems like a lot of the rudimentary analysis of heart data from wearables will be done in the cloud with machine learning. Hence, the price of basic diagnosis goes way down.
With the increasing aging population in USA, Europe, and Japan that will want to stay on top of medical issues, it seems like a big business opportunity. A bunch of players are going to compete in this space. Good luck to you!
You're never going to diagnose something based on a watch (or other consumer focused wearable). The watch is just telling you that you should go talk to your doctor (so they can get you on a real cardiac monitor and get the level of information needed to make a diagnosis). Consumer wearables aren't going to be replacing Holter monitors.
EDIT: A lot of folks seem to be interpreting my comment as implying there is no value in this technology. That is absolutely not the case. There is huge value in widely deployed screening like this (especially since it looks like the specificity is quite high). My point is that these devices will not be replacing diagnostic equipment (in fact, it may very well increase its use as more conditions that would otherwise have gone undetected will now be flagged for further diagnostic testing).
There is value attached to having a device that monitors over the long haul.
A tech journalist who was diagnosed with afib after taking part in the Apple Watch heart study reported that he had undergone traditional EKGs more than once without a problem being detected.
>It's not the kind of thing that comes up in an EKG unless it is actually happening when the test is occurring. I've had EKGs a number of times, and there was never any indication anything was wrong.
I've even had heart tests as a precondition for having bariatric surgery four years ago, and been under general anesthesia, with no adverse results of any kind.
There's definitely value. I'm not disagreeing with that.
The tech journalist in question still underwent more testing after the watch alerted them to make the actual diagnosis. The watch is a great screening tool, but it's not a replacement for more complete diagnostic tests.
Another point most folks seem to be missing: even if you are screened early by one of these devices, it's not at all obvious that subsequent early treatment will deliver extra benefit. It's even less obvious that the extra benefit of early screening and treatment will outweigh the high false positive (and therefore possible overtreatment) costs.
One of the major worries with untreated Afib is that it will trigger blood clots that cause strokes. (Not to mention other possible side effects of not treating the condition)
Modern cardiac event monitors that the patient tapes onto their own chest, wears for a period of time, and then mails back in for analysis are so much cheaper than treating a stroke that it tilts the playing field.
Many people have mentioned a Holter monitor, but those are not only more expensive, but would be overkill as a first step in seeing if there is a real issue.
Here's a comparison between a Holter monitor and one brand of a more modern cardiac event monitor.
The Apple Watch can perform an FDA-cleared diagnosis of certain forms of Atrial Fibrillation, and if what they showed on stage is to be believed, it can produce a really clear ECG that's also FDA-cleared. Just because you're redirected to a doctor doesn't mean the watch did nothing, it just means the watch can't provide a complete patient care cycle. The condition is identified, now what? False-positives can happen even with current medical-grade technology. It's just a question of how the false-positive rate compares.
"Never" is just such a strong word, given how quickly technology has advanced, and given how you don't seem to be aware of what exactly the new Apple Watch is supposed to be able to do.
Do you work for a Holter-producing company or something?
I am very aware of what the watch does. A doctor is not going to diagnose atrial fibrillation based on a single ECG lead. They are going to do a more complete ECG (either in-office, or with a Holter monitor style device at home).
The watch absolutely plays an important role (it's a widely distributed, high SNR screening tool; that's definitely valuable), but it's not replacing more complete diagnostic tests.
As a meta comment... I think it's good to avoid using "you" too much when replying to someone. Respond to what someone said, not assumptions about who they are, what they know, or what their motivations are. For whatever it's worth, I am a software developer and paramedic.
>Do you work for a Holter-producing company or something?
This is against the rules [1]:
[...] Please don't impute astroturfing or shillage. That degrades discussion and is usually mistaken. If you're worried about it, email us and we'll look at the data. [...]
Didn't realize, sorry. I was legitimately curious, because it was such an unusually hard-lined perspective to take in this discussion, one that they repeated in several very similar comments on this thread.
Even if they don't replace higher end gear, consumer devices will allow much better data to determine who needs to be hooked to a dedicated device. More people who need in depth can be brought in to do them, and less people who don't need it can avoid those tests.
Even if the amount of uses for high end devices (Holter monitor or whatever) stays the same, consumer wearables should lead to that usage being allocated in a more productive way, which will bring don't costs and improve outcomes.
I agree. I've seen some electrocardiogram peripherals for phones and they apparently only support lead II. Ability to instantly start recording a long II anywhere anytime is awesome but it is not a complete electrocardiogram.
I'd love to open a second office one day. I'm from Seattle and my co-founder is from Vancouver, so the Pacific Northwest is a natural choice. Right now we're 10 people, so a second office isn't imminent, but we're definitely aware that not everybody can live or wants to live in the bay area.
I need a list of seriously precise devices that are below 200$. By ~seriously I mean something I could rely medically and not a tiny wrist watch that will have too much noise / jitter to be useful. It's for family members (including me).
For sub-$200 price range, I'd suggest some of the less expensive Garmin models (which Cardiogram is currently compatible with) or the Fitbit Charge HR 3 (which we'll hopefully be compatible with in the future).
Is there any interest in developing software to help arrive at a diagnosis? As that would somewhat overlap with the ability to track users vitals.
I suffer from chronic pain, and doctors are seriously unreliable when it comes to diagnosing something relatively complex, even though the research is there and has been available for many years.
I would be interested in joining a project that would become a tool that helps diagnose complex conditions, or at least provide a path for diagnosis.
For the Cardiogram app, most of the UI is developed in React, Redux, and D3 (for interactive charts). For functionality that needs to be native--for example, integrating with the Health app or Google Fit--we use Swift (for both iOS and watchOS) or Java (for both Android and WearOS by Google, which is itself based on Android).
The machine learning code is written in Python, using Keras, TensorFlow, Pandas, scikit-learn, statsmodels. Our AAAI-2018 paper has a bit more info on how DeepHeart (our deep neural network) works. We also do feature engineering and traditional machine learning models when appropriate: https://www.aaai.org/ocs/index.php/AAAI/AAAI18/paper/view/16...
I'm a co-founder at Cardiogram. In multiple N=14,011 studies with UCSF Cardiology, we've shown that heart rate sensors on consumer wearables can detect multiple major health conditions like sleep apnea, hypertension, diabetes, and atrial fibrillation (https://cardiogr.am/research). These conditions affect more than one billion people worldwide, and are frequently undiagnosed -- for example, more than 80% of people with sleep apnea don't know they have it.
We're live on iOS and Android, and we could use help in several areas: mobile and frontend engineering, senior machine learning engineers, and payer relations (the reimbursement from health insurers that Steve talks about). If that sounds like you, please email me at brandon@cardiogr.am!