Scrutable

Mishra, T., Wang, M., Metwally, A.A. et al.
Pre-symptomatic detection of COVID-19 from smartwatch data
Nat Biomed Eng 4, 1208–1220 (2020).

https://doi.org/10.1038/s41551-020-00640-6

... we show that data from consumer smartwatches can be used for the pre-symptomatic detection of coronavirus disease 2019 (COVID-19). We analysed physiological and activity data from 32 individuals infected with COVID-19, identified from a cohort of nearly 5,300 participants, and found that 26 of them (81%) had alterations in their heart rate, number of daily steps or time asleep. Of the 25 cases of COVID-19 with detected physiological alterations for which we had symptom information, 22 were detected before (or at) symptom onset, with four cases detected at least nine days earlier. Using retrospective smartwatch data, we show that 63% of the COVID-19 cases could have been detected before symptom onset in real time via a two-tiered warning system based on the occurrence of extreme elevations in resting heart rate relative to the individual baseline.

Stanford Wearables Project

Based on the quoted text, it sounds like yet another gee-whiz test with low sensitivity and unknown specificity. How many places are using sewage samples, as extensively discussed a few months ago?

Nice trochaic tetrameter. Very unusual - but evocative - metre.

Hiawatha, Hiawatha
On the day of the explosion
Teenage Mutant Ninja Turtles
Mighty Morphing Power Rangers
Adolf Hitler; Margaret Thatcher
Albert Einstein; Scooby Dooby
Pirate Zombie Jungle Island
Symptom tracking via fitbit.


https://en.wikipedia.org/wiki/Trochaic_tetrameter

Bird on a Fire wrote: ↑

Thu Dec 10, 2020 12:58 am

Nice trochaic tetrameter. Very unusual - but evocative - metre.

Hiawatha, Hiawatha
On the day of the explosion
Teenage Mutant Ninja Turtles
Mighty Morphing Power Rangers
Adolf Hitler; Margaret Thatcher
Albert Einstein; Scooby Dooby
Pirate Zombie Jungle Island
Symptom tracking via fitbit.


https://en.wikipedia.org/wiki/Trochaic_tetrameter

xkcd has examples
meta, eh

sTeamTraen wrote: ↑

Thu Dec 10, 2020 12:10 am

Based on the quoted text, it sounds like yet another gee-whiz test with low sensitivity and unknown specificity. How many places are using sewage samples, as extensively discussed a few months ago?

I think you are correct about the sensitivity and specificity problems. It means you would have to identify a way to use this usefully - perhaps requiring people who's resting heart rate was >3 above baseline to get a COVID-19 test.

Having looked at the data shown in the paper, this is more a finding patterns after the effect exercise. It does not look very convincing to me. Of the individual patterns they show, the first individual shows an elevate resting heart rate for the whole infective period - never really going below the baseline. However, the other 3 are all over the place. I know from my own data that resting heart rate can increase by as much as 7 when I am under stress and always goes up by at least 3 if I have a cold (I am also likely to sleep more and do less exercise).

Even a composite of several data sets is unlikely to be able to distinguish COVID-19 from other infections. Looking at the data they present, more worryingly, there will be a lot of false negatives.

I recall hearing a Digital Planet Podcast (back in the days when it was the Click! Podcast) that included mention of a medical centre that was able to use Fitbit data (other trackers are available) to help diagnose a person who had been found collapsed.
The data allowed them to determine when when the heart attack had actually occurred (some treatments should only be attempted within a certain time of the attack).

Chris Preston wrote: ↑

Thu Dec 10, 2020 2:31 am

sTeamTraen wrote: ↑

Thu Dec 10, 2020 12:10 am

Based on the quoted text, it sounds like yet another gee-whiz test with low sensitivity and unknown specificity. How many places are using sewage samples, as extensively discussed a few months ago?

I think you are correct about the sensitivity and specificity problems. It means you would have to identify a way to use this usefully - perhaps requiring people who's resting heart rate was >3 above baseline to get a COVID-19 test.

Having looked at the data shown in the paper, this is more a finding patterns after the effect exercise. It does not look very convincing to me. Of the individual patterns they show, the first individual shows an elevate resting heart rate for the whole infective period - never really going below the baseline. However, the other 3 are all over the place. I know from my own data that resting heart rate can increase by as much as 7 when I am under stress and always goes up by at least 3 if I have a cold (I am also likely to sleep more and do less exercise).

Even a composite of several data sets is unlikely to be able to distinguish COVID-19 from other infections. Looking at the data they present, more worryingly, there will be a lot of false negatives.

I don't think you're wrong, but could it be useful in that you could use it to get people tested before they develop symptoms or get spready? It doesn't have to be good at finding The Covids or discriminating between illnesses, just find candidates for testing earlier than they would normally. That seems useful to me. Kind of like how it works already, having two symptoms means you should get a test, but does not mean you have The Covids. The two symptom method would have a high false negative rate if you were using that for diagnosis, but it's useful for determining who needs a test.

monkey wrote: ↑

Thu Dec 10, 2020 4:26 pm

I don't think you're wrong, but could it be useful in that you could use it to get people tested before they develop symptoms or get spready? It doesn't have to be good at finding The Covids or discriminating between illnesses, just find candidates for testing earlier than they would normally. That seems useful to me. Kind of like how it works already, having two symptoms means you should get a test, but does not mean you have The Covids. The two symptom method would have a high false negative rate if you were using that for diagnosis, but it's useful for determining who needs a test.

But the number of tests available is also a finite resource (and reaching saturation in some places right now), so if every FitBit owner got a signal saying "get a test" and turned up to demand one, you would have a new problem.

Also, people use their FitBit every day, so you would have some people (with borderline bodies or borderline FitBits) getting alerts every couple of weeks. I suppose FitBit could use this to leverage their brand ("Present this alert at a participating Walgreen's to get 15% off your PCR test"), though. It's a fairly "lifestyle" device. (Mrs sTeamTraen took delivery of her third today. They last about two years before wear and tear gets to them. Impressive bit of kit for the size, though.)

sTeamTraen wrote: ↑

Thu Dec 10, 2020 5:54 pm

monkey wrote: ↑

Thu Dec 10, 2020 4:26 pm

I don't think you're wrong, but could it be useful in that you could use it to get people tested before they develop symptoms or get spready? It doesn't have to be good at finding The Covids or discriminating between illnesses, just find candidates for testing earlier than they would normally. That seems useful to me. Kind of like how it works already, having two symptoms means you should get a test, but does not mean you have The Covids. The two symptom method would have a high false negative rate if you were using that for diagnosis, but it's useful for determining who needs a test.

But the number of tests available is also a finite resource (and reaching saturation in some places right now), so if every FitBit owner got a signal saying "get a test" and turned up to demand one, you would have a new problem.

If r is greater than 1, for every person that is positve, you have to do more than one test, because you have to test the people they passed it on to too. If you can get them tested before they do the spreading, you can get them to isolate before they do the spreading, lowering r, so fewer tests needed. If r is less than one, testing capacity shouldn't be a problem, because it should be improving relative to the disease prevalence. If the false positive rate is high, that may mean there are more tests needed, so there'll be a threshold for that, but the specificity may not need to be that good for this to be useful. That'll depend on how much earlier you can detect the disease.

Also, I didn't mean to suggest that it could be done based on this data, they'd have to do an actual test on how good it is at predicting and work out what the false positive rate would be. Just thought that this might have potential, and shouldn't just be dismissed before more work is done.

But yeah, it might be pointless too, especially as I don't think too many people wear these things, and they might not be a useful subset of the population to target with your resources.

According to this, 20% of Americans wear a Fitbit or similar.

Such devices collect a bunch of data that obviously has health significance. It's good that someone is exploring how that data might be useful, even if it doesn't do quite what you might like it to in this particular case.

My favorite Fitbit feature is that I can tell from my heart rate graph when I was playing bridge, which it usually records as cardio exercise. Or at least, it does when I'm playing bridge well. When I'm just pushing cards around, my heart rate hardly goes up at all.

bolo wrote: ↑

Thu Dec 10, 2020 9:42 pm

According to this, 20% of Americans wear a Fitbit or similar.

205 of Americans' favourite sex position is now "on top", to get the Fitbit points.

sTeamTraen wrote: ↑

Thu Dec 10, 2020 9:51 pm

bolo wrote: ↑

Thu Dec 10, 2020 9:42 pm

According to this, 20% of Americans wear a Fitbit or similar.

205 of Americans' favourite sex position is now "on top", to get the Fitbit points.

Or you can strap your Fitbit to your dog, and sit on the couch watching TV while Fido racks up points for you.

bolo wrote: ↑

Thu Dec 10, 2020 9:53 pm

sTeamTraen wrote: ↑

Thu Dec 10, 2020 9:51 pm

bolo wrote: ↑

Thu Dec 10, 2020 9:42 pm

According to this, 20% of Americans wear a Fitbit or similar.

205 of Americans' favourite sex position is now "on top", to get the Fitbit points.

Or you can strap your Fitbit to your dog, and sit on the couch watching TV while Fido racks up points for you.

I suggested that to my wife, but she didn't seem too keen.

I did spot the last time I was off work with a bad cold a couple of years ago. The day I was off, my resting pulse rate rose to 62, from mid to low 50s.