Caution flags about Dreem company and my Dreem 2 use

[2SleepBetta]

This is not a call for therapy help, but is just a device, software and data dense matter that is off the main trail (Main Forum) of sleep apnea therapy. It begins with a couple of things prospective Dreem buyers may want to review. An example of Dreem 2 use among several OSCAR graphs is shown.

First, a cautionary tale:

WatchPAT price far out of reach, I had been visiting the Dreem site, read its backup research, etc. I didn't bite to buy until on Thursday, October 1, 2351 hrs. Bingo, the price offered for the Dreem 2 (D2) was $299, down from usual $499. Not finding an input box for credit card and address, etc., I accepted the offer immediately in the accompanying CHAT box. Auto CHAT responded I'd get response in 30 minutes, but not so. 

The next day, I received (from an eponymous principal or executive of Dreem) a breezy message saying the offer ended after September. I emailed back to that individual and Dreem's CEO asking for reconsideration, . No reply. (See redacted copy of purchase offer and acceptance below, if interested. My request for reconsideration is only referred to in the header of the copy of the offer and denial I sent back).

What Dreem did, of course, is legal so far as I know and within their rights. But it is the first time a firm I consider and expect to be consumer and service oriented has reneged on a written offer to me, my being mistaken or not. Costco particularly, for example, and many other firms over a lifetime just have not done that. A sheltered life, huh? 

Upshot: 
A fellow who has tried all the sleep and other actigraphy devices had a slightly used one for sale and I bought and have used it the past three nights. Last night's example is below.

My beginning to use the Dreem: night or sleep session 3
I knew going in that Dreem would not detect arousals shorter than 2 minutes (Dreem calls those "micro arousals"), but I had hope there would be discernible and useful signals, all taken together, from our OSCAR graphs (including, possibly, oximetry and accelerations), to show why there were sleep stage changes, in one to three steps, from DEEP, REM or LIGHT to WAKE. I do not presently see such indicators but will be looking to see what I can learn.

          Another yellow flag for potential user/buyers:
It may be that my rate adjusting pacemaker (set to a minimum  of 60 bpm) prevents detection of DEEP SLEEP, heart rate variation being needed, possibly, by the algorithm for pattern recognition. I have to hope I do get some DEEP sleep, but my dubious 128 actual-minutes-slept test (12/2015) showed none as DREEM has done so far. But I have lots of long REM periods and LIGHT sleep according to the D2.

          Comments about the example below:
     1. Red and green vertical bars mark times of what I understand to be arousals and a couple of sneezes, respectively, as expanded views of the flow rate (FR) curve show.
     2. I think it is accidental but at about 0700 there could be one indication for the night that an arousal triggered a stage change out of REM.
     3. I don't show stage 4 (DEEP SLEEP) because there was none.
     4. There may be a sleep cycle based regularity in 3 out of the 5 long sleep segments : a higher density of the blue trace during the first half of a the segment than during the last.


   

What motivates a post like this?
I delayed posting to another thread of mine on UARS related topics as I awaited getting such a device to gain more insight into identifying arousals. To what extent does a given pattern cause awakening? The goal of the thread is to gather member help and present information to develop a practical method and answers that would be useful to those with fragmented sleep, UARS, RERA and such. 

It may be a fool's errand, but it seems to me and others that there is a lot of unused, little mentioned but good information to be seen now in OSCAR-expanded views of our FR curves now. A lot of researchers and the most recent of ResMed's patents (2018) on flow limit (FL) detection matters confirm that. Those sleep analysts whose sleep is still troubled after AHI<1 could use a simple shape and frequency scoring system to objectively assess more quickly whether one or more life changes, of one kind or another or two, has helped their sleep--whether they think so or not, at a moment,  after a few day, week, or months of trial. The standard admonition "adopt good Sleep Hygiene" is apt and great, but for those who have lots of time, few changes to make. It would be good to zero in on a trouble factor and decide whether the urge to get better sleep outweighs foregoing something.

Redacted copy of the Dreem offer, my acceptance and Dreem's withdrawal of the offer (denial):

   


2SB

[slowriter]

FWIW, I don't see Dreem having done anything wrong here.

It's an online vendor, where offer "acceptance" consists of clicking "buy" on the product page and paying.

I will also mention on measurement quality, one key variable is to ensure you get a good fit, and in turn "signal quality."

You can confirm this with the app realtime, and you can also email Dreem support to ask them to check it for you.

[davidcmoors]

I had an email recently from Dreem, indicating that they will no longer sell direct to consumers, only to health or research organisations.

I mention this here only because their website has not yet been updated to make this clear. It still shows the previous "currently unavailable in your country" message.

When I first tried to buy, it was unavailable. I registered my interest and was informed when it became available. Fortunately I bought immediately at the lower price, as it increased shortly afterwards.

If anyone has an individual Dreem 2 research subscription, I'd be interested to know if the device is the same as the consumer version, and what additional data and/or software is provided.

[2SleepBetta]

I'd be interested to know if the device is the same as the consumer version, and what additional data and/or software is provided.

Please consider sharing your thoughts about your Dreem 2. I bought one knowing it had limited usefulness for me, just thought it would offer some EEG-like help. I hoped to catalogue typical flow rate bursts, those marked by brief but suddenly higher amplitudes as well as some missing inspiration half waves. The plan was and is to note those that occur most often at or just before sleep stage "drops" (to lower quality sleep or wakefulness). If some regularity in the bursts appeared at those drops a self-scoring system might then be developed for PAP users with RERA to adopt. It would provide a  quantitative way to make and assess changes made to reduce arousals.

I'd be interested to learn whether the research device has better reslolution and wakefulness detection than my Dreem 2. I'd like to compare my Dreem's detection with that of the $2800 WatchPat: see how many arousals and mini-arousals are being missed by the Dreem. My sense is that the Dreem is useless for detection of RERA and UARS and that deep and REM, if not Light sleep, are all over reported for persons with a lot of  inspiratory flow limits (IFL). Similarly, I expect sleep disturbance is under/not reported by Dreem for IFL, for below scoring-threshold "rera" and for all UARS. In short, the Dreem could be a helpful device for people to work on sleep hygiene of their basically-normal sleep.

As I recall, the Itamar WatchPAT measures peripheral arterial tone (PAT) and may be, or is, superior to sleep lab techs' subjective readings of EEGs for detection and scoring of arousals.

I'll ask my MD to order a lower cost at-home WatchPat test . I'd run its test concurrent with the Dreem 2 for comparison. I'll try, but doubt insurance will accept WatchPat results and relax requirement of an in-lab sleep test to enable a "covered" prescription for replacement of my out of pocket and aging AirCurve VAuto with a new one. I bought it because of high density inspiratory flow limits (which it cuts dramatically for those affected). 

My initial in-home screening test and then follow-up lab test supported a prescription (in 2015) for the AirSense AutoSet I got as a PAP newbie. My DME (Medicare) supplier was quick to try pushing another AutoSet onto me at the 5-year mark last September. The AutoSet did a great job eliminating OSA, my scoreable and "treatable" problem.

[davidcmoors]

My experience of Dreem 2 after 3 months use is that it has provided the most credible measure of sleep stages of the various consumer devices I have tried, the only one that uses EEG data, and the only one that is a medically approved device.

I use a Resmed Airsense 10 Autoset and Rescan software, so I can see the graphs of respiratory events, and using the Dreem hypnogram in the phone app, can scroll across the timeline to see if any arousals or sleep stage changes are related to respiratory events seen in the Rescan graphs.

The Dreem sleep stage data seems subjectively pretty credible, having compared it with the hypnograms from 4 PSG studies spaced several years apart, with periods of deep sleep early in the night and more and longer periods of REM later in the night. So basically what you would expect to see.

The only area of doubt is sleep onset time, which on several early use occasions had been up to 2 hours, when subjectively I usually fall asleep within 20 minutes. On one occasion it reported an early awakening 2 hours earlier than Rescan data would suggest. I got a friend to cut the back and sides of my hair much shorter, which seems to have improved the sleep onset reporting problem, though it's still longer than Rescan data would suggest. The early awakening issue has not recurred, and wake time now seems always accurate.

My first Dreem 2 device had a wi-fi and bluetooth fault and was replaced under warranty, so I also reported the sleep onset issue which affected both devices. They kindly provided me with 3 nights raw data, one for a normal night, one for a long sleep onset night, and one for an early awakening night. They were not (unsurprisingly) able to offer any assistance with analysing the 3 nights raw data, provided in h5 format, so I'm working on doing that myself in Matlab. That's a steep and interesting learning curve but I'm making progress and hope to finish before my Matlab free trial expires. My last experience of using mathematical array analysis software was with IBM's APL for mainframes in the 1980s. At least Matlab doesn't need a special keyboard and monitor 

That's why I asked in my earlier post for any feedback from users who might have a Dreem research subscription, to discover if the device hardware is different, and what data access, in what format, and software is provided. I haven't had any responses yet.

The other devices I tried were a Fitbit Inspire HR and a Garmin Vivosmart 4, both of which attempt to derive sleep stage from heartrate variations. The Garmin also has SpO2 recording for part of each night. The Fitbit sleep stages were reasonably typical, but on many nights it failed to sync any sleep data at all, or would fail to sync in the morning but might work if re-tried later. It was impossible to tell if it had failed to record, or failed to sync, and it happened so often that I returned it to the retailer and swapped it for the Garmin.

The Garmin sleep stage data appears to be fairly random and unreliable, often with no deep sleep at all, or occasionally a very long deep sleep period late in the night. I can tell for certain it's pretty random and unreliable, as if I take it off my wrist when I wake up, it always continues to record ongoing sleep even in the absence of any heartrate data, until the next sync is performed, sometimes several hours later. The SpO2 data was so worryingly low, suggesting that I'd need oxygen with my CPAP, that I bought a standalone recording Oximeter and had my doctor check that against his device. The Garmin measurement is typically low by around 5%. Conclusion: It's fine as an activity tracker and pretty useless for sleep, in my view.

[2SleepBetta]

davidmoors,

Thank you for a most informative response. I hope to refer to it to direct others to it (if you don't) in another member's thread about activity trackers. 

Good to see the Dreem 2 is of more value to you than I can see for me. I agree it sometimes errs about sleep vs known wakefulness and am convinced I am often awake during reported REM late in a sleep session after a break. My DEEP sleep needs major improvement efforts if Dreem reports its infrequency and brevity accurately.

My preoccupation is with arousals and mini-arousals as have been reported in my earlier sleep studies. In more than 4 months of Dreem use I've seen only about 3 of them. Those were a couple of minutes. But, as I understand, frequent arousals of a few to, say, 30 seconds seriously impair sleep restfulness and are not sensed by my Dreem. 

In OSCAR I see many large amplitude FR bursts  together with large motions, but without rotation. The D2 SLEEP stage is undisturbed. Yet there is a bodily impact, sometime quite large. Close synchronization, fortunately, is possible (two imports, both a rough timed and a sigh-dot synched one must be done) so I think I can tell when a motion is comfort seeking, not a likely arousal. I believe air is suddenly drawn in before a comfort seeking motion (I hope I know what I'm writing about ).

I'm too many years beyond years of good sleep of most of the young to remember it but do think, despite all, that my sleep is reasonably restorative. Nevetheless, it's a quest to learn here, make my sleep better and then share about what works and doesn't for me.

[davidcmoors]

For anyone who might be interested, here's a link to the clinical study comparing the accuracy of Dreem sleep stage measurements to manually scored polysomnography.
I think this was the basis for approval by the FDA of Dreem as a medical device.

https://www.researchgate.net/publication...ep_Staging

The full pdf can be downloaded without payment using the button on the page.

Looks like the least accurate aspect of sleep stage measurement for both Dreem and expert PSG scorers was differentiating between light sleep and wakefulness. I should say that I'm no statistician though.

[selfcoacher]

I have a Dreem 2 as well. The version with the stimulations for deep sleep, which my data suggests it is useful for me and I respond well. My deep sleep numbers are pretty low as well. I recently got some raw data and can confirm that sensor contact is very important as most the channels in my data are noise and of no use. Something simple as position change can render sensor contact poor and data useless. With that said, no other consumer device can achieve such sleep stage accuracy assuming proper sensor contact. However, you cannot be sure of data accuracy unless perfect fit and sensor contact throughout the night.

I believe Dreem 2 hardware is the same for research, you just get access to raw data every night and their H5 viewer.

The consumer version of Dreem 2 is useless for Reras. You can see the micro arousals only in the raw data, but the report you get is auto scored and it does not report anything below 1minute arousal. I find heart rate changes of 6bpm or greater as a good proxy for arousals.