Sleep metric for sleep quality?

Is there any particular measurement that would tell someone if they are getting a good quality sleep?
HRV? Estimated time in REM (like some sleep trackers like Basis Peak)? Or some other formula?
What is the best sleep tracker available now that Zeo is not readily available?
Alan

If you feel refreshed in the morning and don’t get sleepy in the early afternoon, you probably slept well :slight_smile:

Sleep tracker data can be used to rule out certain problems such as frequent waking due to sleep apnea, and are a convenient way of keeping track of your sleeping habits, but I don’t expect much more from the current generation of consumer-grade sleep tracking devices.

1 Like

One well-known test for sleepiness is the Epworth Sleepiness Scale:

How likely are you to doze off or fall asleep in the following situations, in comparison to just feeling tired?

Personally, I also keep an eye on the number, frequency and intensity of yawns throughout the day, though that definitely depends on the activity I’m engaged in.

2 Likes

Has anyone tried a cognitive test as a way to measure tiredness?
If so what would you recommend?

Perhaps you could try a reaction time test? First take some consistent measurements at different points in the day to establish baselines (wake up, afternoon, evening, before bed) along with a qualitative score (1-10) of how tired you feel, then over time see how your readings compare/trend?

1 Like

Seth Roberts used a simple “rested rating” that he found to be quite sensitive to small changes. Some background is in this paper: http://media.sethroberts.net/blog/pdf/2012-09-24-The-Growth-of-Personal-Science-Implications-For-Statistics.pdf

1 Like

Thanks! Those are great ideas.

Interesting paper, but the heterogeneity in different people is mostly greater than the measurement variation. Perhaps effort should be put into classifying subgroups of people based upon biometric responses. For example some people react to stress with high blood pressure. In others the BP lowers.

If you could access all the paper by scientists and PHD thesis papers you might find a lot has been done. But the cost and time of research…

For me there seems to be a strong correlation with Quantified Mind tests first thing in the morning vs Basis Band measurements. The deep sleep score in particular:

http://myquantifiedbrain.com/post/128728873365/mental-focus-impacted-by-sleep

1 Like

Interesting! I hardly find anything in my Basis sleep data, what is your opinion about your Basis sleep data?

It seems like coding has the strongest correlation with coding, how do you do your (coding) tests?

Hey Justin :slight_smile:

Coding test is the test from http://www.quantified-mind.com/. Yeah, blog not too clear on that. In terms of cognitive function, its "Visual Perception & Cognition Skills

Great info on the “Quantified Mind” site on the science of each test here:

I find the basis sleep data not so accurate. I use a Beddit also which I find gives much better indicator. The basis ‘deep sleep’ I think is most useful out of all the metrics. Overnight heart rate is a pretty good metric to use also. I wrote some scripts to calculate.

What were you looking for in your basis sleep data? Did you do much analysis yourself? I really want to dig deeper into the data here with multivariate correlations & a few other analysis tools. Love to get your feedback on that :slight_smile:

Hi Justin,

Thanks for the data share, very interesting stuff and you got me curious.
What unit are those data in? Amount of sleep in each sleep type or % of the night in that sleep type?

I have been tracking my sleep quality through a 4 level score system and went back to look if I see any correlation with the amount of light, deep, REM sleep or the sleep quality calculated by BASIS.
I have 631 nights with B1 and 284 nights with Peak and I cannot see any correlation.

This was looking quickly at the data using that sleep quality metric. If I have some time I could look at those data against other sleep trackers I am using concomitantly and/or my mood that day (after the sleep period).

I concur with you that despite the nice amount of data provided by BASIS not much I really trust. Even the heart rate data is questionable in my opinion.

Thanks for sharing your approach.

That’s an important kind of calibration, pretty interesting that for all the work they put into this (and many of us know some of the principals) the sleep tracking data fails to correlate with subjective report.

Apologies Luke & Gary - don’t know how I missed this post.

The metrics I used in my calculations on the blog or correlations between minutes in each phase of sleep (as per Basis band) and the results from Quantified Mind.

I also self-tracked subjectively (1-10 score) in the morning how well I slept, and correlated with the Basis Band metrics.

Strongest correlations are with the REM & deep sleep - but they’re still not great (0.328 & 0.314 coefficients).

I’ve got 2 months worth of data from the Beddit also, which is giving slightly better results at 0.568 with the sleep score, but data set isn’t as big as with Basis.

If I have the time, I might post something on analysing just the sleep tracking.

Thanks,
Justin

Nice that you were able to find some correlations between subjective and Basis scores. However, I have about two years of sleepdata both Basis and subjective(1-10) and couldn’t find any significant correlations between them.

My subjective sleep scores do have correlations with my psychological variables (happiness/health etc.) while my Basis scores does not have any correlations. Therefore, I value my subjective scores more than the Basis scores.

Also, I went to a sleeplab (as a research participant) and got my validated sleepdata from the study from the researchers, I want to compare this data with my Basis data but didn’t have the time to do this yet. I’ll keep you updated when I post a blog about this.

That’s interesting that you didn’t find any significant correlations Justin – I wonder if that points to the validity of the Basis at measuring sleep. Sleep quality and quantity tend to correlate highly, and consumer grade devices are generally pretty good at measuring sleep quantity (within a margin of error), although they are known to overestimate a bit.

For example, in this study, the researchers found that the SenseWear (validation study wtih polysomnography: r = .84) correlated strongly with the Misfit shine (r = 0.82, bias = 44 mins), Jawbone UP (r = 0.89, bias = 23.5mins) Withings pulse (r = 0.92, bias = 24.4), and the Fitbit one (r = 0.92, bias = 15.9) in every day settings. As you can see, the correlations ® are high, but most devices tend to overestimate sleep. The shine overestimates sleep by as much as 44 minutes, and the Fitbit One by just 15 minutes. Conversely, in an older study (n = 23) with the Fitbit One, the fitbit one was only accurate for about half of participants. I’m not sure if the algorithm was updated between these studies leading to the varying results – it’s a bit perplexing.

Alternatively, simple worn Actigraph devices have shown high correlations with subjective sleep quality (which is what Alan seemed to be particularly interested in). In a study using a GT3M actigraphs for example, my colleague reported a correlation of about .88 (p < .01) with subjective measures of sleep quality (n = 113) – where sleep quality was measured via the Pittsburgh Sleep Diary. This suggest that subjective measures of sleep which merely ascertain perception of sleep quality are pretty accurate. So more complex testing might not be necessary.

So to answer @Dr_Mush’s question, I’m not sure what specialised devices are particularly most accurate right now, but in our research we simply use the Pittsburg Sleep diary for subject insight on sleep quality – which is surprisingly accurate. In an upcoming study we will use consumer grade devices (i.e., Fitbit One) to get an estimation of within-person variability in sleep quantity, which according to the aforementioned studies are reasonably accurate at estimating duration of sleep in healthy adults.

3 Likes