Hacker News Reader

Comment by DuskStar | Hacker News Reader

"What we list for legal purposes" and "what the device actually does" might not be the same thing. I imagine saying "my device detects heart attacks" gets the FDA involved and makes everything ten times more expensive...

View on HN

The watch doesn't have enough data to detect a heart attack. It's not possible for the device to do it (with current, or even near-future technology it's not possible for any wrist worn device to detect a heart attack as it's happening).

Source: he's an EMT.

Technically I'm a paramedic...

I can never recall which is which. I assume paramedic is the better-credentialed?

Yeah, technically they're all levels of EMT, but "EMT" generally refers to "EMT-Basic" (a few months of training, a couple nights a week). EMT-Paramedic is ~18 months of a couple nights a week, plus 1,000+ hours of clinical time.

Thank you for your service.

It is theoretically possible now, with clunky technology that is not wrist-worn.

Development of a 12-Lead ECG Signal Processing Algorithm Using NI LabVIEW® and NI ELVIS: https://ieeexplore.ieee.org/document/8574627

You would most likely need a 12 lead ECG/EKG too, to detect it accurately.

Well the watch isn't measuring troponin levels so it's hard to imagine how useful it could ever be at MI detection, especially given the limitations of a poor contact single lead ECG.

Even if it could check a trop, that's a pretty late indicator. If you're having the big one, you're going to be dead before your troponin spikes.

Is that right? When I worked in medical instrumentation, I was always told that TnI was a leading marker.

It will take up to 8 hours for troponin levels to rise after the onset of symptoms (typically 3-4).

ER nurse here - some STEMIs I've taken care of had negative initial trop's (not all though). The 4 hr delta should of course be positive though - I don't typically check the 4 hr repeat for STEMIs though because they're long gone out of the ED to the cath lab then to CVICU.

I could imagine the thing detecting Vfib but it would need to use some good AI signal processing to exclude artifacts. Elevated ST segments, probably not, even with good AI.

OTOH doesn't the EKG feature only work when you touch the watch with your other hand? In which case it would be useless for heart attacks that caused you to lose consciousness.

The watch did an emergency sos. AFAIK no heart related notification triggers that - only fall detection can.

Outside possibility is he felt bad, saw a heart rhythm notification, and manually triggered emergency sos by holding the power button. But a fall alert seems much more likely.

The specific language you are looking for here is "indications for use". The way the FDA works, you cannot advertise (even directly) a medical device for a usage you have not indicated in your filing and backed up with the required evidence. The level and type of evidence will depend on the type of filing and claims. A medical professional can advise for a different use (called "off label") but you, as the device developer, cannot even hint at it.

Aside: as others note, though, you can't really detect heart attacks well from something this simple.

Whether you officially list it or not, if your device is diagnosing and reporting heart attacks to people, the FDA will be involved sooner or later.

If they only diagnosed but didn't report to the user but instead anonymized the data and used it for marketing would the fda care?

The FDA was involved in the Apple Watch.

That being said of course they will state that the watch can’t detect accurately hearth attack to avoid any liability.

And because it can’t...

Differentiating between cardiac arrest and taking the watch off your wrist may be interesting. And I kinda wonder how good any single-lead EKG would be at detecting a heart attack when your heart is still beating. Even the expensive multi-lead machines usually auto-diagnose me with an old infarction, which is false.

I'd guess taking the watch off would produce a classic flat line. A heart in arrest which is in the process of dying will produce irregular electrical output.

Sometimes the rhythm may degrade into a ventricular rhythm, like v-tach then into v-fib if not treated. A heart can still pump in v-tach, albeit very inefficiently, but in v-fib, no blood is pumped. On the other hand, taking the watch off results in no signal input which is different than what you would see on the monitor for asystole.

Apart from the immediate “flatline” the watch also detects that it is not proximate to the skin. As soon as you take it off the watch disables things like ApplePay. It knows when it being/has been taken off.

Reply on news.ycombinator.com