Hacker News Reader

Comment by DenisM | Hacker News Reader

Another view on the Oura thing: https://www.medpagetoday.com/practicemanagement/informationt...

It's rather skeptical. Choice quote:

[...] The SRI researchers studied 41 healthy adolescents and young adults (average age 17). Sleep data were recorded using the Oura ring and standard PSG on a single laboratory overnight. Metrics were compared using Bland-Altman plots and epoch-by-epoch analysis.

The ring accurately detected "light" and "deep" sleep in 65% and 51% of the sleep epochs, respectively. It also accurately detected 61% of REM sleep epochs, with an overall overestimation of PSG REM sleep (by about 17 min). When the ring misclassified PSG REM sleep, the algorithm usually classified the epoch as "light sleep" (76%).

These data suggest that the Oura Ring is virtually useless in telling you if you are in REM sleep versus deep or light sleep. [...]

View on HN

Matthew Walker, probably the best known sleep scientist, has been critical of Oura and pretty much all the sleep trackers for sleep phase data. Unsurprisingly, it's difficult to guess at brain EM wave patterns from a band on the finger, wrist, etc.

I'm 6 months on Oura, and I don't think it's actually that consistent, i.e. I don't think you can say "Oura shows 50% more deep sleep today, so even if the base mumber is wrong, I must have gotten 50% more deep sleep". I wear a fitbit to cross validate each night, and the readings on both often contradict each other seemingly arbitrarily, night to night.

Overall I've found they're great for tracking overall sleep time (HRV, heart rate, temp too) but I don't read too much into the actual sleep phase data. (Day-to-day self asessment of cognitive performance is probably the best proxy.)

> Day-to-day self asessment of cognitive performance is probably the best proxy.

How do you do that?

Measure it using a series of cognitive performance tests. You could even just grade your own experience day to day - assuming you feel everything is a little worse when you sleep poorly (a sensible assumption), you will probably get representative data.

It doesn't matter so much how accurate it is, as long as it's consistently inaccurate.

I don't care if my scale at home is perfectly accurate. I care about knowing that my weight is going up/down.

At 50-65% accuracy rates it’s pretty useless even for binary decision making (better/worse).

I care if my scale at home is accurate. I don't want to know if my weight is going up or down, I want to know what I currently weigh so I can decide what I need to do about it.

Cant you just use the mirror for that?

Hard to use a mirror in your sleep.

Uhh what? They were talking about weight

It's also difficult to weigh yourself with a mirror.

Your mirror’s hard? Let me take a rain check on that, mine’s sleeping.

Your example seems to disagree with your premise. If someone tells me I weigh "9239" this means nothing to me, unless in context of whether it's going up or down, and how it is related to others of my height (and age and gender, perhaps).

I don't care that I weigh some multiple of an SI unit, or some fraction of a car. I care about my weight relative to others and my history.

If you care about your weight relative to others, then it has to be accurate, no? Because how would you compare your weight unless you had a bunch of people using your scale.

True, but given that it's more difficult to change your weight than it is to find someone of a different weight, I'd rather live my whole life with a scale that might be reading (say) 5lbs high or low, rather than a scale that fluctuates between 5lbs high or low from measurement to measurement.

Sure. Weight relative to 1 lb or 1kg is essentially the definition of “absolute weight.”

No, it does not. Say I'm 6' tall. "Normal" BMI is between 18.5 and 25, which translates to between 136.4 lbs - 184.3 lbs. Telling me my weight is going up or down tells me nothing about whether I fall into this healthy range. And, depending where I fall (below, within, or above the range), the actions I would wish to take differ. It's that simple.

What instrument do you have that measures your "weight relative to others and your history"? Don't you only want to measure it relative to other healthy individuals? And, if you are particularly underweight or overweight, wouldn't you want to know when you're in the normal range? An instrument that only tells you what direction the change is doesn't tell you any of that.

But weight by itself (and BMI) are fairly crude measures. You could be 6' tall, be a muscular 210 lbs with a body fat of 10% and be classed as obese. The next level of the "onion" is body composition (fancy scales do this, but not that well). Then you're getting dexa scans....and there are probably other levels after that. I think I'd rather know if my weight and body fat % are going up or down, rather than their absolute value.

Le sigh. I should have known someone would throw that particular nitpick.

That’s an edge case. People who fall in that category already know that. And they’re very interested in their accurate, absolute weight.

Besides, BMI is just a convenient way of quantifying what a “healthy weight range” is. I could have used any range and said I was interested in weighing between X and Y.

I am kind of in that category...and I'm interested in relative fluctuations in body fat % even though I don't give much credence to the absolute accuracy of the body fat measure.

How would you even know if it is consistently inaccurate? The variability in biological signals might be so high day to day that it could be highly correlated but with frequent pathological cases that make the data useless.

I have a cheap Mi Band 2 and I'm impressed with how well it measures when I'm sleeping and when I'm awake. I don't know how good it is to measure deep sleep. I does not pretend to measure REM.

Are there any weareable device that measures it well?

I use an Apple Watch and an app called Pillow and it does a plausible job of guessing when I went to sleep and when I woke up. It also identifies light/deep sleep and REM, but I have no way of telling how accurate. It seems to detect awake periods at night as well as I can remember them.

But ... I kinda assume it's a crapshoot and I use it more as a novelty and an estimate of total sleep than any kind of accurate representation of sleep cycles.

I was using a Fitbit Charge (the one before the just released one). One night I was awake looking at my phone from 1-3am and it said I was asleep. Took it off and put it into a drawer right then.

I just got the Charge 3 and I must toss and turn like crazy because it sometimes doesn't even register my sleep. I didn't expect it to be the most accurate sleep tracker, so I am not super upset. It still gives me some level of insight into my sleep, which is better than nothing.

>I didn't expect it to be the most accurate sleep tracker

I work for Fitbit. I don't speak for Fitbit, but in my opinion you should expect it to be the most accurate sleep tracker, and if it isn't you should contact customer service.

Maybe I will then. I just assume that the technology to accurately determine REM vs Deep or Light sleep is not really available, so I can't fault Fitbit or any manufacturer for not having perfect sleep trackers. But, the marketing material does claim it's very accurate, so I should probably get more out of it since that's what I was sold.

Thanks for the advice.

On the contrary, my Charge 2 is generally pretty accurate in terms of when I'm asleep vs awake.

I've got a Charge 3, the one with the oxygen sensor, and I've also seen it think I was asleep when I wasn't. On the other hand, it seems weirdly reliable at telling if I'm in REM sleep or not. Whenever I wake up directly from what it says was REM sleep I remember my dreams but not otherwise.

Reply on news.ycombinator.com