ResScan UnderReporting or SmartLink OverReporting or ?

jdm2857 wrote:Maybe you should try straight CPAP (at your typical 90- or 95-centile pressure) on both machines to factor out the effects of the auto algorithm. You'll get a much better picture of the machines' event detection abilities.

Good point, and important, I think it's worth trying, especially since you feel more or less the same, and are disturbed by the higher pressure.

DHC, some of the DeVilbiss events you listed are not what DeVilbiss calls "responders". And note that the DeVilBiss definitions for apnea and hypopnea are very different from the one's on Velbor's chart. Everybody has a low AI and a higher HI on a DeVilbiss, by definition (literally!), since for DeVilbiss, obstructive apneas only happen between a 90% and a 95% flow drop. Above 95% you're in "non-responsive" territory, less than 90% and you're in hypopnea territory.

A study comparing ResMed and Respironics previous model autos (Titration Efficacy of Two Auto-Adjustable Continuous Positive Airway Pressure Devices Using Different Flow Limitation-Based Algorithms) showed that a ResMed will go to higher pressure, and leave a higher residual AHI that will a Respironics. And that's when we're counting events in a PSG, not what the machines report.

[edited to correct the typo that was caught by DHC. My original sentence was absurd...]

This all makes me wonder if having a Respironics APAP is the best to have. Would an S9 perhaps be better? I need a backup machine and could get an S9 instead of a System one APAP.

At least I could do the experiment if I got one. I am beginning to wonder if the reports of these machines are valuable at all. Well, I see my sleep Dr. on Monday - and I'm sure she would write me an Rx for a Resmed S9 autopap if I asked. Then I could do a definitive study ... or could I? I love to design experiments - and this fascinates me. It actually perturbs me greatly as I see these machines as a wealth of potential misinformation.

I've got to admit, I feel great since APAP. But I like data. Real data! Data you can believe in! Where have I heard crap like that before?

ozij wrote: DHC, some of the DeVilbiss events you listed are not what DeVilbiss calls "responders". And note that the DeVilBiss definitions for apnea and hypopnea are very different from the one's on Velbor's chart. Everybody has a low AI and a higher AI on a DeVilbiss, by definition (literally!), since for DeVilbiss, obstructive apneas only happen between a 90% and a 95% flow drop. Above 95% you're in "non-responsive" territory, less than 90% and you're in hypopnea territory.

A study comparing ResMed and Respironics previous model autos (Titration Efficacy of Two Auto-Adjustable Continuous Positive Airway Pressure Devices Using Different Flow Limitation-Based Algorithms) showed that a ResMed will go to higher pressure, and leave a higher residual AHI that will a Respironics. And that's when we're counting events in a PSG, not what the machines report.

I need to apologize in advance for apparently being dense, but I really do not understand your point about everyone using a DeVilbiss machine will have a "higher AI."

>>DHC, some of the DeVilbiss events you listed are not what DeVilbiss calls "responders".<<

True enough. I was merely recounting the events that are detected by DeVilbiss. Which of those generate an active response is different. From what I can piece together, it looks like the different events that generate an active response are:

ResMed - snores, obstructive apneas and flow limitation.
DeVilbiss - snores, obstructive apneas and hypopneas.

Which leads us back to the definitions of the events for purposes of detection and response. When I looked at the chart shown upthread for several different machines (though notably absent the DeVilbiss), the ResMed S8 AutoSet definitions are shown for an Apnea and a Hypopnea. While there are some 'fine points' in the definition, it seems simple enough to resolve it down to:

ResMed S8 definition of Apnea = 75 % decrease in ventilation (flow rate) for at least 10 consecutive seconds.

DeVilbiss IntelliPAP definition of Apnea = 90 % decrease in flow for at least 10 consecutive seconds.

Given these definitions, the S9 (assuming the S9 uses the same algorithm as the S8) should indicate a larger number of apneas than the IP. What am I missing?

jonquiljo wrote:This all makes me wonder if having a Respironics APAP is the best to have. Would an S9 perhaps be better? I need a backup machine and could get an S9 instead of a System one APAP.

At least I could do the experiment if I got one. I am beginning to wonder if the reports of these machines are valuable at all. Well, I see my sleep Dr. on Monday - and I'm sure she would write me an Rx for a Resmed S9 autopap if I asked. Then I could do a definitive study ... or could I? I love to design experiments - and this fascinates me. It actually perturbs me greatly as I see these machines as a wealth of potential misinformation.

I've got to admit, I feel great since APAP. But I like data. Real data! Data you can believe in! Where have I heard crap like that before?

Candidly, part of the reason I posted this topic and embarked on this journey to understand the differences in reported metrics is that it seems there is some level of consensus that a desirable level of nightly results would be to have an AHI < 5.0. I also see some folks posting about having numbers consistently < 1.0 - which, in the surface, sounds TERRIFIC - **IF** those metrics are reliable and can be believed.

A grand example is found in this topic when it was suggested that the ResMed is providing better therapy based on the reported metrics.

I remain unconvinced. I am not normally a skeptic, though I *am* analytical and I *need* to see the logic and rationale and veracity behind the reported metrics.

None of this is meant to cast aspersions on any company. I am convinced (on faith, at this moment) that the S9 and the IP are both excellent devices and providing excellent results. I simply want to know more before I am personally willing to declare the S9 as the 'best in class'.

Kudos to DeVilbiss for providing the detailed information in their webinar series. I have also managed to find some bits and pieces from ResMed about their technology - though not nearly as much as for the DeVilbiss. Perhaps it is merely the lack of detail from ResMed that fuels my questioning - but given the results are so widely disparate, it only makes sense to search for some more definitive answers as to WHY. Of course, those answers may have major significance in application of each machine's features to optimize therapeutic benefit, and in future machine purchase decisions.

FWIW

As a S9 auto user, I too am wondering about the meaning of the data my machine reports, particularly since my only 14% of my events during the baseline study were apneas, but the overwhelming number of events the S9 reports are apneas.

Given the descriptions of how the S8 and M-series define apnea and hypopnea that are posted above, then yes, it is reasonable to assume that an S8 could report more apneas than the REMstar M-series Auto since the apnea definition for REMSstar Series M auto requires substantially more loss of airflow than the S8 Auto's definition. But in turn, that should also alter the Hypopnea Index as well: It seems to me that there is clearly a range of event severity that the S8 will label apnea and the M-series will label hypopnea. So perhaps some additional questions also need to be considered:

1) How much do the AHI (not just AI) vary between machines?

2) How does the ratio of apneas to hypopneas vary between the machines?

Perhaps with the S8's rather looser definition of apnea, it both over reports apneas and under reports hypopneas, with the resulting AHI index still somewhere close to a meaningful number?

And of course it would be nice to be able to figure out what changes to the S8's algorithm were made for the S9 auto. (And similarly, what changes were made to the M-Series for the REMSstar System One autos.

DHC wrote:

ozij wrote: DHC, some of the DeVilbiss events you listed are not what DeVilbiss calls "responders". And note that the DeVilBiss definitions for apnea and hypopnea are very different from the one's on Velbor's chart. Everybody has a low AI and a higher AI on a DeVilbiss, by definition (literally!), since for DeVilbiss, obstructive apneas only happen between a 90% and a 95% flow drop. Above 95% you're in "non-responsive" territory, less than 90% and you're in hypopnea territory.

A study comparing ResMed and Respironics previous model autos (Titration Efficacy of Two Auto-Adjustable Continuous Positive Airway Pressure Devices Using Different Flow Limitation-Based Algorithms) showed that a ResMed will go to higher pressure, and leave a higher residual AHI that will a Respironics. And that's when we're counting events in a PSG, not what the machines report.

I need to apologize in advance for apparently being dense, but I really do not understand your point about everyone using a DeVilbiss machine will have a "higher AI."

It is me who had to apologize. My sentence should read:

Everybody has a low AI and a higher HI on a DeVilbiss, by definition (literally!), since for DeVilbiss, obstructive apneas only happen between a 90% and a 95% flow drop. Above 95% you're in "non-responsive" territory, less than 90% and you're in hypopnea territory.

I''ll fix the original.

OK - so let me see if I have managed to make sense of the metrics, at least, a little.

ResMed does not score hypopneas as events, hence, their AHI would leave out the 'H' part of the equation. Correct?

To some extent that would be offset by ResMed setting their apnea detection levels to a lower point than DeVilbiss (75 % reduction in ventilation for ResMed versus 90 % for DeVilbiss).

The metrics shown in my opening post were for AHI - and this would explain at least some of the difference in the metrics between ResScan and SmartLink. Essentially, the events the IP is presently detecting that fall between 50 % to 75 % reduction in flow rate would NOT be reflected in the AHI reported by ResMed's ResScan.

This would support ozij's comments that ALL users of DeVilbiss machines will have a higher AHI than using a ResMed machine.

When I had my sleep study done, all recorded events were hypopneas. That study was conducted using Respironics equipment, hence, I suspect it applied the Respironics definitions found upthread - making it more difficult to compare current results with the original sleep study - although, it looks like the Respironics metrics and the DeVilbiss metrics are similar in what they capture and report.

Interestingly (to me anyway) is I just went to review the ResScan reports and Help system pertaining to their reported AHI metric. I did not find a written description of the equation, but just noticed in ResScan the report says "AHI & AI" and then "Apnea Index". Notably, it did NOT describe it as 'Apnea Hypopnea Index" (wiki/index.php/Apnea_Hypopnea_Index).

So it seems what has become clear is that AHI as reported from ResScan is NOT the same metric/calculation as AHI reported by DeVilbiss - and probably not the same as AHI reported by Respironics.

I am not clear on how ResMed uses events determined by monitoring Flow Limitation - though the DeVilbiss webinar described the metric and how it was captured. The Flow Limitation, along with Snores, are predictive indicators that allow an active response BEFORE actually experiencing an Apnea (or Hypopnea). Does ResMed count those and include them in any metrics? Something I plan to research further.

Am I on the right track - or ?

DHC wrote: Candidly, part of the reason I posted this topic and embarked on this journey to understand the differences in reported metrics is that it seems there is some level of consensus that a desirable level of nightly results would be to have an AHI < 5.0. I also see some folks posting about having numbers consistently < 1.0 - which, in the surface, sounds TERRIFIC - **IF** those metrics are reliable and can be believed.

Yes, I am one of those people who routinely gets AHI's of 0.3, 0.2, even 0.0 one night! I certainly do not trust that - even though my AHI was only about 14 to begin with. I think 80% of my events when tested were hypopneas, which seem to be subjective in how you want to score them.

Yes, I feel better, but could feel even better than that - so I wonder if the people who make these things have them reporting lots of BS - basically. As I've posted in another thread today, oximetry data I did last night seems to suggest that I am waking up fairly often as my O2 goes up to waking levels (99%) instead of sleeping levels of about 94%.

I really wish there was a reliable way to see if I am not being led around by my nose (pun intended) with all this data that may possibly be very irrelevant to reality. You buy a medical device and tend to believe it, but who is to say it is really telling you the truth? The numbers you posted at the top of the thread indicate that these machines are telling you quite possibly what they want you to hear.

All this undermines the reality of self-empowered treatment. If we are following BS and adjusting things accordingly - then possibly we are all wasting a lot of time. I've never seen anyone post O2 data that really followed what their machine was telling them. Usually there are so many events to sort out that the graphs cannot be compared well. It's not just O2 de-saturation that is important anyway.

How do we really know if we are not waking a lot more times than our machines would lead us to believe? I've tried to figure out a way to "home test" some parameter that indicates arousals - but nothing really can do that. At first I was suspicious of mouth leaks - but now I am terribly suspicious of all of the data itself. Short of an EEG ... what is there available? But the discrepancies between machines and the fact that many (if not the greater part) of CPAP users report not a significant amount of relief suggest we may be led in the wrong direction often.

jonquiljo wrote:

DHC wrote: Candidly, part of the reason I posted this topic and embarked on this journey to understand the differences in reported metrics is that it seems there is some level of consensus that a desirable level of nightly results would be to have an AHI < 5.0. I also see some folks posting about having numbers consistently < 1.0 - which, in the surface, sounds TERRIFIC - **IF** those metrics are reliable and can be believed.

Yes, I am one of those people who routinely gets AHI's of 0.3, 0.2, even 0.0 one night! I certainly do not trust that - even though my AHI was only about 14 to begin with. I think 80% of my events when tested were hypopneas, which seem to be subjective in how you want to score them.

Yes, I feel better, but could feel even better than that - so I wonder if the people who make these things have them reporting lots of BS - basically. As I've posted in another thread today, oximetry data I did last night seems to suggest that I am waking up fairly often as my O2 goes up to waking levels (99%) instead of sleeping levels of about 94%.

I really wish there was a reliable way to see if I am not being led around by my nose (pun intended) with all this data that may possibly be very irrelevant to reality. You buy a medical device and tend to believe it, but who is to say it is really telling you the truth? The numbers you posted at the top of the thread indicate that these machines are telling you quite possibly what they want you to hear.

All this undermines the reality of self-empowered treatment. If we are following BS and adjusting things accordingly - then possibly we are all wasting a lot of time. I've never seen anyone post O2 data that really followed what their machine was telling them. Usually there are so many events to sort out that the graphs cannot be compared well. It's not just O2 de-saturation that is important anyway.

How do we really know if we are not waking a lot more times than our machines would lead us to believe? I've tried to figure out a way to "home test" some parameter that indicates arousals - but nothing really can do that. At first I was suspicious of mouth leaks - but now I am terribly suspicious of all of the data itself. Short of an EEG ... what is there available? But the discrepancies between machines and the fact that many (if not the greater part) of CPAP users report not a significant amount of relief suggest we may be led in the wrong direction often.

>>All this undermines the reality of self-empowered treatment.<<

I am not so sure the manufacturers or the doctors are all too interested in patient's engaging in "self-empowered treatment" - witness the controversy stirred up by member 'Calist' elsewhere in this forum. [Note: I prefer to think of myself as a 'customer' (NOT 'patient'] of the doctor and they are nothing more than a paid consultant]

Further, I noticed this at ResMed's site regarding their algorithm (this page -- http://www.resmed.com/us/clinicians/tre ... clinicians) :

Autotitration devices differ greatly from one manufacturer to another. Each manufacturer uses a different algorithm that responds to different parameters, so you will find greater variance between two autotitration devices than you will between two CPAP devices.

Clinicians should ask each manufacturer for validation of their algorithm.

Note the emphasis on the "clinician" to request validation of the manufacturer's algorithm - which may simply be due to the fact I found it in their 'Clinicians' tab on the website.

Still, I would like it if more information were made readily available to us - the end users of the devices.

FWIW

my results look almost identical to your Intellipap. The lowest I have ever had was 1.5 and that only happened once. I have been using it since March 2010 and have slept all night and every night with it 100% of the time.

DHC wrote:ResMed does not score hypopneas as events.....

I have both some As and some Hs in my composite AHI score from my S9 on most nights.

DHC wrote:When I had my sleep study done, all recorded events were hypopneas. That study was conducted using Respironics equipment....

If you didn't have a full Polysomogram (PSG) sleep study done, which certainly sound like the case, you will be chasing shadows with trying to line up your sleep study "numbers" with anything much.

DHC wrote:Am I on the right track - or ?

If you are not starting from the basis of a PSG, then you are definitely "or"

Cheers,

Bill

DHC wrote: OK - so let me see if I have managed to make sense of the metrics, at least, a little.

ResMed does not score hypopneas as events, hence, their AHI would leave out the 'H' part of the equation. Correct?

I'm not sure I understand the question/statement. But my understanding is that an S8 might score a change in breathing as a "hypopnea" that an S9 might not. But it is not a matter of either one being more "accurate" than the other, as I understand it--the data is mostly just for judging whether a change to therapy improves things or not. The home-treatment-machine-generated data isn't so much for diagnostic purposes or for direct comparison to sleep-study numbers or to other brands or series.

Therefore, there is no easy way to judge efficacy of one machine over another for a particular patient by merely comparing the home-machine-generated numbers--the brands differ too much in how aggressive they are in reporting, or "scoring" the results of what they are meant to prevent. There is no conspiracy (in the traditional sense) in that, though. Ways of scoring, just like ways of treating, get patented, so manufacturers have reasons to describe both aspects in a proprietary way.

The important thing, to me, is to understand that questions about scoring are not necessary directly related to questions about how the machines treat, since, for example, all reporting ResMeds can report "hypopneas" but do not actually react to them, as such, but instead react to the precursors to obstructive hypopneas. And the fact that a "hypopnea" is reported does not necessarily mean that a real hypopnea was in any way missed. It may have been a perfectly normal change in breathing that an aggressively scoring machine would score as a "hypopnea" and a less-aggressively-scoring machine would not score at all. ResMed autos attempt to react to the shape of the flow curve, the flattening, not so much the percentage reduction in flow (for a less-than-apnea reduction, that is). But that is only a question about how the machine chooses a level of pressure. On the other hand, what the S8s report is, as shown in the Velbor chart, based on the percentage reduction in flow, not so much shape (although obviously those things are related). At least, that's how I understand it.

DHC wrote: Essentially, the events the IP is presently detecting that fall between 50 % to 75 % reduction in flow rate would NOT be reflected in the AHI reported by ResMed's ResScan.

My understanding is that the events would likely be reported as "hypopneas" by an S8. But as far as I know, ResMed has not released the scoring criteria used by the S9. And shame on them for that, in my opinionated opinion. Or, then again, maybe they have, and I have just missed it. I do that sometimes.

billbolton wrote:

DHC wrote:ResMed does not score hypopneas as events.....

I have both some As and some Hs in my composite AHI score from my S9 on most nights.

DHC wrote:When I had my sleep study done, all recorded events were hypopneas. That study was conducted using Respironics equipment....

If you didn't have a full Polysomogram (PSG) sleep study done, which certainly sound like the case, you will be chasing shadows with trying to line up your sleep study "numbers" with anything much.

DHC wrote:Am I on the right track - or ?

If you are not starting from the basis of a PSG, then you are definitely "or"

Cheers,

Bill

Cheers Bill,

True enough - I too see hypopneas recorded in the ResScan graphs. Because I see what appears to be conflicting information, it is not yet clear to me how ResMed calculates their published AHI. Do you know ?

In terms of your reference to my original sleep study, you may be quite right that it was not a "full Polysomogram." The ability to correspond the results of my subsequent nightly therapy to the original sleep study was, as I wrote, only one of the reasons for pursuing this journey of enlightenment (about the manufacturer's algorithms. The OTHER reason is this question of how a manufacturer determines its AHI seems rather fundamental to this entire process. Evidence the CPAPWIKI (wiki/index.php/Category:CPAP_Definitions) which defines AHI thusly:

AHI (Apnea Hypopnea Index) is the number of apneas and hypopneas per hour. Or an index for sleep apnea. 5-20 mild, 21-50 moderate, above 50 severe..

In my apparently inferior sleep study, the AHI numbers were published both pre and post CPAP usage. The mere fact that AHI is meaningful as an output of the sleep study would seem more than a little significant - don't you agree? If you do agree, then it follows the construction of the index ought to be something the end users of these products would be allowed to know. After all, if (and this is a BIG *if*, as I do not know ..... yet) some insurance companies might be willing to reimburse for "severe" sleep apnea (based on AHI, among other metrics) - but would NOT be willing to reimburse a "mild" condition - then it is significant - OR - as I am recently learning ... if a life insurance policy considers "severe" sleep apnea differently than "mild" sleep apnea to determine eligibility or price of acquiring a policy - once again, these AHI values and their construction takes on great significance. If insurance companies are not basing decisions on the relative severity of a condition and applying AHI as a part of their equation, then this concern may be moot. Do you know?

Since it seems a great many of the forum participants here are dedicated to learning as much as they can in order to play a direct role in managing their outcomes, it is only logical that those types of people will pursue answers to questions that play a significant role in their improved health.

I find it curious, at least, that so few people have provided any definitive information about ResMed's algorithms. The one thing we *do* know is that ResMed reported AHI is much different - and notably LOWER - than DeVilbiss (at least). There is *something* going on to create that disparity. Is it - as one person suggested - that the therapy received with the S9 unit is so vastly superior to the IP that it will result in a 10x reduction in events? Possibly - though it seems VERY unlikely for a number of other reasons.

So where does the disparity lie? Is it the algorithm? It certainly isn't my lack of a "full Polysomogram," so the answer must lie elsewhere. Do you know?

DHC wrote:OK - so let me see if I have managed to make sense of the metrics, at least, a little.

ResMed does not score hypopneas as events, hence, their AHI would leave out the 'H' part of the equation. Correct?

Wrong.
Resmed does not change pressure in response to hypopneas. But is scores them. The S8 reports hypopneas prolifically, the S9 seems to report less.
You have to keep three things in mind:

• The way a machine gather data about your breathing pattern
  The way a machine analyses the data it has gathered
  The way a machine responds to the anaylysed results

APAP machines differ on all three of those.

The metrics shown in my opening post were for AHI - and this would explain at least some of the difference in the metrics between ResScan and SmartLink. Essentially, the events the IP is presently detecting that fall between 50 % to 75 % reduction in flow rate would NOT be reflected in the AHI reported by ResMed's ResScan.

It makes no sense to discuss the metrics when the pressure is different. For all you know, had the DeVilbiss arrived at the pressure the Resmed arrived at, your results my have been similar.
What you have are 2 APAP machine reading differently, analysing differently and responding differenlty. As a result, you have them supplying different pressures, and -- not surprisingly -- you are getting different results.
You are spending time at higher pressures on the Resmed than on the DeVilbiss -- this in itself could be a cause of the differnt AHI -- and you seem to ignore that fact.
If you want to be scientificSet both machines at the same pressure, and then look at the AHI.

This would support ozij's comments that ALL users of DeVilbiss machines will have a higher AHI than using a ResMed machine.

Let me try to explain again: DeVilbiss' default event definitions are such that most users switching to that machine will find that suddenly many their previously reported apnea are now reported as hypopneas -- even if the machines are set at the same pressure.

When I had my sleep study done, all recorded events were hypopneas. That study was conducted using Respironics equipment, hence, I suspect it applied the Respironics definitions found upthread - making it more difficult to compare current results with the original sleep study - although, it looks like the Respironics metrics and the DeVilbiss metrics are similar in what they capture and report.

The person anlysing your study is not supposed to apply an autopap's definition of apneas and hypopneas. They are supposed to apply the definitions suggested by sleep medicine professionals -- those are not too standard either, but they take desaturations into account -- which the machines do not.

So it seems what has become clear is that AHI as reported from ResScan is NOT the same metric/calculation as AHI reported by DeVilbiss - and probably not the same as AHI reported by Respironics.

Right.

DHC wrote:it is not yet clear to me how ResMed calculates their published AHI. Do you know ?

The same way everyone else does AI+HI = AHI

DHC wrote:I find it curious, at least, that so few people have provided any definitive information about ResMed's algorithms.

There are numerous lengthy threads here that have discussed the various scoring algorithms from the various vendors in great detail... you need to do some work with the search facility.

DHC wrote:So where does the disparity lie?

It seems to me it lies largely in your lack of understanding of what any xPAP machine is capable of scoring solely from air flow analysis, which is the only basic mechanism available for a flow generator (there is a huge amount of discussion about that topic here too). Newer models, like the S9, usually have much better ways of analysing air flow than older models.

A PSG scores AHI on the basis of a wide range of different and concurrent factors/approaches, so your lack of PSG data as your baseline for understanding what any xPAP does is definitely not going to help!

Cheers,

Bill