It knows when you are sleeping, but will it tell you?

[Rubicon]

Originally from way back from David Rapoport and Bob Norman.

https://portal.unifiedpatents.com/paten ... nt/6988994

That patent in turn is listed on some F&P devices.

[MJS_]

When he got around to explaining the Auto-Ramp feature, I said "This thing can't really tell when you are asleep or awake, can it?" He assured me that it does a good job detecting whether your asleep or awake based on breathing patterns.

Sounds like you're not aware that the following two statements:

• Sleep/awake cannot be accurately determined without EEG, and
• ResMed's Auto-Ramp feature can do a good job detecting whether you're asleep or awake based on breathing pattern

Can both be true.

It is debatable whether these machines can be said to do a "good job" detecting whether you're asleep if they cannot be accurate. Apparently, ResMed and F&P do not trust their sleep detection enough to provide an index of sleep efficiency or to use this information when calculating the AHI. However, I was only challenging the assertion that it is "common knowledge" that these machines are incapable of accurately detecting whether or not the user is sleeping.

[chunkyfrog]

It doesn't even "know" you are alive.
It functions as programmed.
Hopefully, the programmer anticipated you.

[MJS_]

[snip]

Information about Resmed's AutoRamp was pulled from https://www.resmed.com/en-us/sleep-apne ... -pressure/

In the case of Resmed's AutoRamp, the machine is looking for one of the following three things before it starts to ramp the pressure up:

• 30 breaths of stable breathing (roughly 3 minutes)
• 5 consecutive snore breaths
• 3 obstructive apneas or hypopneas within 2 minutes
• 30 minutes have passed since the machine was turned on

[snip]

Thank you for this useful information! Luckily, it appears I can control my apnea with a minimum air pressure setting that I find tolerable. My CPAP struggles have been mostly caused by an ill-fitting noisy and leaky face mask and F&P's overly heated air tube with no temperature setting control.

Those four criteria used by ResMed to detect sleep have convinced me to avoid their AutoRamp feature. I do not want to experience breathing problems before the machine delivers a therapeutic level of air pressure. Also, a technique I've used to overcome insomnia (pre-CPAP) is to enter a meditative state using controlled stable breathing, which ResMed may interpret as sleeping. [I've found it challenging to mediate while feeling like I'm suffocating due to hot humid air being blown in my face and while my uncomfortable face mask is making a variety of intermittent noises. ]

[billbolton]

He assured me that it does a good job detecting whether your asleep or awake based on breathing patterns.

You seem to be reading what you want to hear into simple statements.

The Resmed alogrithm is only useful for the gross purpose of commencing or ceasing treatment. It will not reliably "score" anything but autonomous sleep breathing. The "auto" ramp sensing is solely to determine whether autonomous sleep breathing has commenced, and has no other useful purpose.

[MJS_]

He assured me that it does a good job detecting whether your asleep or awake based on breathing patterns.

You seem to be reading what you want to hear into simple statements.

The Resmed alogrithm is only useful for the gross purpose of commencing or ceasing treatment. It will not reliably "score" anything but autonomous sleep breathing. The "auto" ramp sensing is solely to determine whether autonomous sleep breathing has commenced, and has no other useful purpose.

Could it be that I am interpreting simple statements in a straight forward manner the way a typical lay person would, and you are reading in nuances and caveats based on your knowledge of this field and what you want to hear?

[robysue1]

When he got around to explaining the Auto-Ramp feature, I said "This thing can't really tell when you are asleep or awake, can it?" He assured me that it does a good job detecting whether your asleep or awake based on breathing patterns.

Sounds like you're not aware that the following two statements:

• Sleep/awake cannot be accurately determined without EEG, and
• ResMed's Auto-Ramp feature can do a good job detecting whether you're asleep or awake based on breathing pattern

Can both be true.

It is debatable whether these machines can be said to do a "good job" detecting whether you're asleep if they cannot be accurate. Apparently, ResMed and F&P do not trust their sleep detection enough to provide an index of sleep efficiency or to use this information when calculating the AHI. However, I was only challenging the assertion that it is "common knowledge" that these machines are incapable of accurately detecting whether or not the user is sleeping.

and

He assured me that it does a good job detecting whether your asleep or awake based on breathing patterns.

You seem to be reading what you want to hear into simple statements.

The Resmed alogrithm is only useful for the gross purpose of commencing or ceasing treatment. It will not reliably "score" anything but autonomous sleep breathing. The "auto" ramp sensing is solely to determine whether autonomous sleep breathing has commenced, and has no other useful purpose.

Could it be that I am interpreting simple statements in a straight forward manner the way a typical lay person would, and you are reading in nuances and caveats based on your knowledge of this field and what you want to hear?

Not contradictory at all: An EEG is needed to determine precisely when you first fall asleep, as well as precisely when you wake up in the middle of the night and precisely how long each wake lasts. And this stuff is important for accurately determining the AHI value for a diagnostic sleep test.

But once sound sleep is, in fact, established, it's pretty easy to just look at a person and say with some confidence, "that person is asleep" even though we can't tell precisely when that sleeping person's consciousness changed from Wake to Sleep just by looking at them.

Same thing with the flow rate from our CPAPs: Statistics combined with shape of inhalations in the flow rate can be used to determine that there is a high probability that the person using the machine is asleep.

In the best (and easiest) scenario, once sound sleep is, in fact, established, the person's breathing should even out and become very stable and very regular with inhalations that are both more regular and smaller than waking inhalations. With decent programming, a APAP machine can be programmed to start ramping up the pressure from the starting Ramp pressure to the Min pressure setting once it is clear that very stable and very regular inhalations have been going on for a long enough time for it to be reasonable to assume the person is, in fact, asleep. And that's exactly what Resmed's programmers use for the first criterium for ending the auto ramp: Three minutes (30 breaths) of stable breathing has a high probability of being associated with the onset of sustained sleep; hence it's reasonable to start increasing the pressure from Ramp pressure to Min pressure.

Two of the other criteria that Resmed's programmers use to decide there is a high probability the person is asleep are based on the fact that in developing the Auto algorithm, they've already discovered ways to use the flow rate data to flag certain anomalies in the flow rate data as "snoring" and "OAs" and "Hs". Because the device is designed to treat sleep disordered breathing, Resmed's programmers have made decision to assume that if enough of those particular breathing patterns are detected during the AutoRamp period, the person is likely asleep and the bad breathing should be treated by ending the ramp and increasing the pressure. Notably, for both snoring and OAs/Hs, the Resmed programmers require more than one such event---5 consecutive snores or 3 OAs/Hs within a two minute period---specifically because some people's wake breathing can contain flow patterns that resemble the flow pattern of sleep disrupted breathing events, and they don't want the machine to prematurely start raising the pressure before the person using the machine is asleep. (The whole idea behind the auto ramp is that increasing the pressure while a person is awake may make it (much) harder for the person to fall asleep.)

The final criteria that Resmed uses to end the auto ramp is based on statistics: Most people have a latency to sleep that is between 10 and 20 minutes long. So statistically speaking, most people should be asleep by the time 30 minutes have elapsed since the machine was turned on. And because problems with sleep disordered breathing can start any time after a person falls asleep, they don't want the machine to be sitting a Ramp pressure for extended periods of time when it is likely the person is asleep, but their breathing has not yet been regular enough to meet the first criterium or irregular enough to meet the second or third criteria for the machine to meet the second or third criteria to end auto ramp.

Now it's also important to understand something else about what we mean when we say these machines can't really tell if we are awake or asleep without an EEG: Part of sleep disordered breathing is the body's response to the OAs and Hs: For many people, events come with arousals and mini-wakes that last 3-30 seconds. To further complicate matters, some people have spontaneous arousals that are not related to sleep disordered breathing, and these also can be as short as 3-10 seconds. These arousals and mini-wakes can be picked up by an EEG, but any change in the breathing pattern is too short to pick up (on the fly) and label as "Wake" with a high probability of certainty. People who look at lots of flow rate data can make intelligent guesses about what kinds of patterns in flow rate might be related to an arousal of some sort, but the auto algorithms and event tracking algorithms in APAP/CPAP machines don't label these as "wakes" because they're not programmed to try to tease out every arousal or mini-wake from just the flow rate data: Teasing them out with a reasonable amount of accuracy on the fly requires some additional data.

Now I don't have any personal experience with the F&P SensAwake algorithm to understand what it actually does. But from what's been posted by Rubicon and from what I have read, SensAwake is looking for at least a minute of what Rubicon called "Troubled Wakefulness", which is just another way of saying the breathing is clearly not good quality sleep breathing AND the breathing is not typical sleep disordered breathing. So there's a high probability that the person is either awake or drifting between wake and sleep without being able to establish real, continuous sleep. Again, if you look at lots of flow rate patterns, it's easy to make intelligent guesses about when a person is not sleeping well and the problem is not sleep disordered breathing. Around here, such periods are often called "sleep-wake-junk" of SWJ for short. But without the EEG data, you can't tell from the flow rate whether the person is awake for the whole period of time or is bouncing back and forth between wake and stage 1 sleep without making it to stage 2 sleep.

So my best guess is that SensAwake does a decent, but not fully accurate, job of realizing that something other than restful sleep is going on and that the "something" is not likely to be sleep disordered breathing. But whether the "something" is a real, genuine prolonged wake or whether the "something" is a restless period where the person is bouncing back and forth between wake and sleep is beyond what SensAwake is designed to do. Regardless of what that "something" is, the programmers of the SensAwake algorithm have made a decision that lowering the pressure (slightly?) should allow the person using the machine to more easily get back to a real, continuous, sleep rather than being stuck in a period of SWJ if that's what is happening at the time the "troubled wake" breathing has been going on long enough for SensAwake to kick in.

Finally, I will add this: CPAP/APAP manufacturers repeatedly say that machine reported AHI numbers are most important as trending data. They don't claim the overnight AHI is totally accurate every single night specifically because they know that the CPAP/APAP cannot determine the total sleep time (TST) with sufficient accuracy (as compared to a PSG). And they also know that the machine can flag certain events when a person is, in fact, awake. But the overall statistics indicates that there is sufficient accuracy in the data that for a typical PAPer if the long term machine reported AHI is substantially below 5.0, then the actual, real AHI (if it could be accurately computed) would also be substantially below 5.0. Likewise, there is sufficient accuracy in the data for a typical PAPer if the long term machine reported AHI is above 5.0, there is a high probability that the sleep disordered breathing is not yet being well treated by PAP. The assumption is that a typical PAPer does not spend exceptionally long periods of being awake while using the machine. In other words, the assumption is that a typical PAPer's sleep with the machine looks reasonably close to decent sleep in a person without sleep disordered breathing.

[lazarus]

Nice post, RS1! Thank you for that carefully worded explanation. I learned from it.

Although I am still naturally drawn to the conspiracy theory that my machine is somehow keeping secrets from me.

[zonker]

Nice post, RS1! Thank you for that carefully worded explanation. I learned from it.

Although I am still naturally drawn to the conspiracy theory that my machine is somehow keeping secrets from me.

your machine isn't the only one.

[chunkyfrog]

Learn all the secret handshakes you can.

[MJS_]

When he got around to explaining the Auto-Ramp feature, I said "This thing can't really tell when you are asleep or awake, can it?" He assured me that it does a good job detecting whether your asleep or awake based on breathing patterns.

Sounds like you're not aware that the following two statements:

• Sleep/awake cannot be accurately determined without EEG, and
• ResMed's Auto-Ramp feature can do a good job detecting whether you're asleep or awake based on breathing pattern

Can both be true.

It is debatable whether these machines can be said to do a "good job" detecting whether you're asleep if they cannot be accurate. Apparently, ResMed and F&P do not trust their sleep detection enough to provide an index of sleep efficiency or to use this information when calculating the AHI. However, I was only challenging the assertion that it is "common knowledge" that these machines are incapable of accurately detecting whether or not the user is sleeping.

and

He assured me that it does a good job detecting whether your asleep or awake based on breathing patterns.

You seem to be reading what you want to hear into simple statements.

The Resmed alogrithm is only useful for the gross purpose of commencing or ceasing treatment. It will not reliably "score" anything but autonomous sleep breathing. The "auto" ramp sensing is solely to determine whether autonomous sleep breathing has commenced, and has no other useful purpose.

Could it be that I am interpreting simple statements in a straight forward manner the way a typical lay person would, and you are reading in nuances and caveats based on your knowledge of this field and what you want to hear?

Not contradictory at all: An EEG is needed to determine precisely when you first fall asleep, as well as precisely when you wake up in the middle of the night and precisely how long each wake lasts. And this stuff is important for accurately determining the AHI value for a diagnostic sleep test.

But once sound sleep is, in fact, established, it's pretty easy to just look at a person and say with some confidence, "that person is asleep" even though we can't tell precisely when that sleeping person's consciousness changed from Wake to Sleep just by looking at them.

Same thing with the flow rate from our CPAPs: Statistics combined with shape of inhalations in the flow rate can be used to determine that there is a high probability that the person using the machine is asleep.

In the best (and easiest) scenario, once sound sleep is, in fact, established, the person's breathing should even out and become very stable and very regular with inhalations that are both more regular and smaller than waking inhalations. With decent programming, a APAP machine can be programmed to start ramping up the pressure from the starting Ramp pressure to the Min pressure setting once it is clear that very stable and very regular inhalations have been going on for a long enough time for it to be reasonable to assume the person is, in fact, asleep. And that's exactly what Resmed's programmers use for the first criterium for ending the auto ramp: Three minutes (30 breaths) of stable breathing has a high probability of being associated with the onset of sustained sleep; hence it's reasonable to start increasing the pressure from Ramp pressure to Min pressure.

Two of the other criteria that Resmed's programmers use to decide there is a high probability the person is asleep are based on the fact that in developing the Auto algorithm, they've already discovered ways to use the flow rate data to flag certain anomalies in the flow rate data as "snoring" and "OAs" and "Hs". Because the device is designed to treat sleep disordered breathing, Resmed's programmers have made decision to assume that if enough of those particular breathing patterns are detected during the AutoRamp period, the person is likely asleep and the bad breathing should be treated by ending the ramp and increasing the pressure. Notably, for both snoring and OAs/Hs, the Resmed programmers require more than one such event---5 consecutive snores or 3 OAs/Hs within a two minute period---specifically because some people's wake breathing can contain flow patterns that resemble the flow pattern of sleep disrupted breathing events, and they don't want the machine to prematurely start raising the pressure before the person using the machine is asleep. (The whole idea behind the auto ramp is that increasing the pressure while a person is awake may make it (much) harder for the person to fall asleep.)

The final criteria that Resmed uses to end the auto ramp is based on statistics: Most people have a latency to sleep that is between 10 and 20 minutes long. So statistically speaking, most people should be asleep by the time 30 minutes have elapsed since the machine was turned on. And because problems with sleep disordered breathing can start any time after a person falls asleep, they don't want the machine to be sitting a Ramp pressure for extended periods of time when it is likely the person is asleep, but their breathing has not yet been regular enough to meet the first criterium or irregular enough to meet the second or third criteria for the machine to meet the second or third criteria to end auto ramp.

Now it's also important to understand something else about what we mean when we say these machines can't really tell if we are awake or asleep without an EEG: Part of sleep disordered breathing is the body's response to the OAs and Hs: For many people, events come with arousals and mini-wakes that last 3-30 seconds. To further complicate matters, some people have spontaneous arousals that are not related to sleep disordered breathing, and these also can be as short as 3-10 seconds. These arousals and mini-wakes can be picked up by an EEG, but any change in the breathing pattern is too short to pick up (on the fly) and label as "Wake" with a high probability of certainty. People who look at lots of flow rate data can make intelligent guesses about what kinds of patterns in flow rate might be related to an arousal of some sort, but the auto algorithms and event tracking algorithms in APAP/CPAP machines don't label these as "wakes" because they're not programmed to try to tease out every arousal or mini-wake from just the flow rate data: Teasing them out with a reasonable amount of accuracy on the fly requires some additional data.

Now I don't have any personal experience with the F&P SensAwake algorithm to understand what it actually does. But from what's been posted by Rubicon and from what I have read, SensAwake is looking for at least a minute of what Rubicon called "Troubled Wakefulness", which is just another way of saying the breathing is clearly not good quality sleep breathing AND the breathing is not typical sleep disordered breathing. So there's a high probability that the person is either awake or drifting between wake and sleep without being able to establish real, continuous sleep. Again, if you look at lots of flow rate patterns, it's easy to make intelligent guesses about when a person is not sleeping well and the problem is not sleep disordered breathing. Around here, such periods are often called "sleep-wake-junk" of SWJ for short. But without the EEG data, you can't tell from the flow rate whether the person is awake for the whole period of time or is bouncing back and forth between wake and stage 1 sleep without making it to stage 2 sleep.

So my best guess is that SensAwake does a decent, but not fully accurate, job of realizing that something other than restful sleep is going on and that the "something" is not likely to be sleep disordered breathing. But whether the "something" is a real, genuine prolonged wake or whether the "something" is a restless period where the person is bouncing back and forth between wake and sleep is beyond what SensAwake is designed to do. Regardless of what that "something" is, the programmers of the SensAwake algorithm have made a decision that lowering the pressure (slightly?) should allow the person using the machine to more easily get back to a real, continuous, sleep rather than being stuck in a period of SWJ if that's what is happening at the time the "troubled wake" breathing has been going on long enough for SensAwake to kick in.

Finally, I will add this: CPAP/APAP manufacturers repeatedly say that machine reported AHI numbers are most important as trending data. They don't claim the overnight AHI is totally accurate every single night specifically because they know that the CPAP/APAP cannot determine the total sleep time (TST) with sufficient accuracy (as compared to a PSG). And they also know that the machine can flag certain events when a person is, in fact, awake. But the overall statistics indicates that there is sufficient accuracy in the data that for a typical PAPer if the long term machine reported AHI is substantially below 5.0, then the actual, real AHI (if it could be accurately computed) would also be substantially below 5.0. Likewise, there is sufficient accuracy in the data for a typical PAPer if the long term machine reported AHI is above 5.0, there is a high probability that the sleep disordered breathing is not yet being well treated by PAP. The assumption is that a typical PAPer does not spend exceptionally long periods of being awake while using the machine. In other words, the assumption is that a typical PAPer's sleep with the machine looks reasonably close to decent sleep in a person without sleep disordered breathing.

Thank you for your detailed explanation. I'm sticking with my assertions that this knowledge is not "common", and it is not reasonable to expect a lay person to deduce much of this from simple statements like: "When set to Auto, your device will ramp up to your prescribed pressure only when it detects that you've fallen asleep" or "the machine does a good job detecting whether you are asleep or awake."

[ChicagoGranny]

I'm sticking with my assertions

Stick with 'em. Too much emotional investment to back down now.

[lazarus]

Purdy much anything about treatment-machine data is like that.

"It knows whether the apnea was obstructive or central."

Eh, well, it guesses, based on airway status, not effort, since it has no effort belts like an NPSG has.

"It can detect apneas and hypopneas."

Eh, well, it uses proprietary definitions, ones that differ significantly from NPSG, to label the breathing events.

There is a lot of gray area and inexactitude in most all of what it does and reports, but it does it and reports it very well nevertheless. Brilliantly, in fact.

What is understood and reasonable to lay people depends a lot on how much time and effort lay people choose to invest in personal research and education beyond reading surface-level ad copy.

Just like everything else in life.

[palerider]

I'm sticking with my assertions

Stick with 'em. Too much emotional investment to back down now.

You've got to admire people who double down after being proven to be full of crap...

No, wait, you don't. Sorry, got that one backwards,

[MJS_]

What is understood and reasonable to lay people depends a lot on how much time and effort lay people choose to invest in personal research and education beyond reading surface-level ad copy.

Very true. Unfortunately, medical and public health agencies and professionals often discourage patients from researching beyond surface level ad copy. [And quite a few doctors are content to rely on surface-level ad copy themselves.] In fact, during a recent pandemic we witnessed medical and public health agencies cooperating with federal espionage agencies, law enforcement agencies, big tech companies, big media companies, and even medical journal publishers in an effort to deny the public access to scientific research and debate that could have helped them gain perspective and make informed decisions.