POLL: Typical AHI for ResMed Users

[Muffy]

The whole kit-and-kaboodle... Bummer!

No real surprise I guess.

Well, from the HIPPA compliance perspective, it's actually great. 2.1 is password-protected and has a timeout for inactivity, and data encryption would really complete the package. But yeah, no more sorting through the list looking for VBs.

Muffy

[-SWS]

Why is this poll relevant? We all know that different machines register apneas differently.

I think the premise of the poll is that we don't all know that yet---at least not the newcomers. Newcomers to Resmed machines need to know that HI is scored differently from one manufacturer to the next, but that the different brands provide great treatment more often than not.

OK, I thought about changing the poll, but all I really wanted to see was a spread of numbers and perhaps some comments from ResMed users.

A simple HI curiosity/comparison peek is as good a reason as any for an informal poll IMHO.

Did I miss a turning point when this message board converted from allowing informal threads with polls like this to requiring purely scientific polls instead? We can have both can't we?

[-SWS]

The whole kit-and-kaboodle... Bummer!

No real surprise I guess.

Well, from the HIPPA compliance perspective, it's actually great. 2.1 is password-protected and has a timeout for inactivity, and data encryption would really complete the package. But yeah, no more sorting through the list looking for VBs.

Muffy

Well, HIPPA compliance sure sounds likely to me---perhaps even in a roundabout way. Their new online Encore solution is clearly going to require some major interdiction measures to be HIPPA compliant: http://encoreanywhere.respironics.com/screenshots.aspx

Once Respironics understandably heightened security to make their Internet-based Encore records HIPPA compliant.... it probably made good sense to employ the same database security measures across the board---including card-based Encore 2.1. Can't say that I blame them.

[Muffy]

In an analysis designed for the expressed purpose of giving the exacting statistician Velbor a stroke, if we take 2kat's numbers:

2kat Poll

and compare it with the -SWS VB Poll:

-SWS VB Poll

the relationship of rogue HI (AIs are constant and therefore need not be considered) to VB% is demonstrated:



and the precise generalization can be made that the Variable Breathing Mode of the Respironics algorithm "weeds out" those artifactual hypopneas associated with "troubled wake".

Muffy

[jnk]

I think the trend may be for manufacturers to take the desires of their real customers, the B&M DMEs, into consideration by taking further steps to hide data from patients, thus requiring patients to go through their local DMEs to get any detailed treatment-efficacy info and making things even harder for the online sellers all at the same time.

Boy, do I hope am particularly wrong about that, though!

Speaking of wrong, is it OK if the next time someone claims RM doesn't treat anything over 10 cm that I then claim that RR doesn't treat anything over 8 cm? (j/k)

[-SWS]

and the precise generalization can be made that the Variable Breathing Mode of the Respironics algorithm "weeds out" those artifactual hypopneas associated with "troubled wake".

That's possible IMO... However, so far I don't see anything in the Respironics patents describing that kind of activity.

On the contrary: all the Respironics patent descriptions instead show the VB control layer sitting beneath the H layer regarding priority. The patent descriptions always show the higher control layers usurping control away from the VB layer to service detected A or H---with absolutely no exceptions mentioned in any of the patent descriptions.

Still, a patent description is not a design document. So what you theorize may very well be happening. It's reasonable IMO to speculate that if the VB control layer can accurately distinguish wake with high detection sensitivity and specificity, that the designers may have decided to discard hypopneas during that suspected wake.

However, I don't happen to think that's an inescapable conclusion:

Highly erratic breathing is guaranteed to present a HUGE challenge to all the manufacturers, regarding the derivation of their respective recent-flow baselines. Hypopneas for the APAP machines are inherently but a disadvantaged comparison during highly erratic breathing. Those single-data-channel based hypopnea measurements are but a severely disadvantaged flow-data comparison---of wildly moving volumes or peak flow against fluctuating baselines that were also derived from those same recent wild flow-data movements. All the manufacturers are guaranteed to take different patent-protected approaches as they derive their recent-flow baselines during those most difficult baselining and comparison circumstances. Those differing patent-protected baseline approaches----in the midst of that sampling and data adversity presented by highly erratic breathing---are virtually guaranteed to yield different hypopnea detection sensitivity and specificity scores, depending on the strengths and weaknesses of each mathematical method employed.

So while I think it's reasonable to speculate that any manufacturer might decide to discard wake-based hypopneas, I don't see that being mentioned anywhere in the Respironics patent descriptions. And I think there are other possible explanations for that data phenomenon as well.

Other opinions?

[-SWS]

Speaking of wrong, is it OK if the next time someone claims RM doesn't treat anything over 10 cm that I then claim that RR doesn't treat anything over 8 cm? (j/k)

That works as equivalent misinformation for me.

[jnk]

Other opinions?

My overly-simplistic take is that if one brand counts every certain-percentage drop below a certain-formulated baseline as a hypopnea and doesn't attempt to filter how it reports that information with other factors, whereas another brand factors in things like recovery breaths and treatment pressure in how it reports, the filtered brand will report less than the unfiltered one. Or is that just blowing smoke?

Here is part of the Hypoventilation Rule [OPTION] for adults from 2007:

"At this time, there is insufficient evidence to allow specification of sensors for direct or surrogate measures of PaCO2. Both end-tidal CO2 and transcutaneous CO2 may be used as surrogate measures of PaCO2 if there is demonstration of reliability and validity within labrotory practices. At this time, there is inusfficient evidence to allow specification of a duration of hypoventilation though the duration should be sufficient to account for the effects of response time of the sensor used and to exclude brief changes that reflect sensor artifact."

Whatever that means. Sounds to me like home machines check for it about as well as the PSGs do.

[-SWS]

the filtered brand will report less than the unfiltered one. Or is that just blowing smoke?

As in filtered/unfiltered cigarette smoke? Good play on words, jnk!

I haven't looked at the breath recovery issue yet. But I agree with what I think your general point is: I happen to think the different brands necessarily entail a variety of different design-judgments, regarding implementation specifics such as mean-flow data normalization, event scoring thresholds, and event discounting/filtering---especially in the midst of both baselining and temporal-variable trending adversities presented by highly erratic patient flow signals.

Or inversely stating my data view for the sake of an entertaining analytical paradigm shift: I agree that the manufacturers' proprietary HI scoring methods should yield far more similar HI outcomes during non-variant or less variable flow data moments---versus sustained periods of highly erratic flow data.

Those latter moments of wild data movement---when baselining attainability itself becomes highly challenged---are exactly when we should expect maximum HI scoring disparity from differing proprietary methods.


_________________________________________________________________________________________________________________________

the relationship of rogue HI (AIs are constant and therefore need not be considered) to VB% is demonstrated:

<insert separate Resmed AHI and Respironics VB% survey results here>

...and the precise generalization can be made that the Variable Breathing Mode of the Respironics algorithm "weeds out" those artifactual hypopneas associated with "troubled wake".

When you get a chance can you explain to my pea brain how comparing a sample of Resmed AHI against yet a different and independent sample of Respironics VB% somehow statistically demonstrates that Respironics is discarding hypopneas during the wake stage?

[jnk]

If a hypopnea is a particlular type of variation in breathing, and a brand has a fairly strict set of criteria that a variation has to meet in order to be reported as a hypopnea, I assume the logical place to file the variations that didn't meet the criteria would be a big drawer called "variable breathing." Or where else would those hypopnea-wannabes get filed?

That would be merely a guess based on my assumptions from my understsanding of the definitions of the words. But that would be somewhat in line with the Muffpothesis, right?

So would the trick be to find enough people who had used both to see if there is any correlation individually between VB percentage on a RR and hypopnea count on a RM? If so, should we all just loan one another our different brands of machines for a while and see what happens?

[X55]

I was on a loaner Respironics machine for 2 full weeks before I was provided my Resmed S8. The Respironics data had my AHI at between 1-2 per night (on average), and now with the S8 I've had for 2 weeks:

-using the same mask and connector tube
-using the same humidity setting
-using the same pressure and EPR setting
-getting the same perceived quality of sleep/feeling rested the next day

I'm seeing AHI's averaging just a touch over 5 each night.

I realize that my results are not "scientific" as the humidity of one machine at 1.5 may be different than another, but I feel the results should be close enough.

[-SWS]

If a hypopnea is a particlular type of variation in breathing, and a brand has a fairly strict set of criteria that a variation has to meet in order to be reported as a hypopnea, I assume the logical place to file the variations that didn't meet the criteria would be a big drawer called "variable breathing." Or where else would those hypopnea-wannabes get filed?

Respironics claims that the VB control layer exists because: 1) data trending becomes difficult during erratic breathing, and 2) the APAP pressure treatment algorithm itself would also become erratic.

So Variable Breathing is not a hypopnea wanna-be state or lesser flow-gradient state: rather a hypopnea's most salient attribute is flow reduction while Variable Breathing's most salient attribute is excessive peak flow standard deviation (apples and oranges---at least the fruit is variable). So those two dissimilar attributes are really not comparative gradients of each other a la hypopneas versus hypopnea-wannabes. One describes flow reduction relative to a baseline, and the other describes high variability from one breath to the next---making it inherently difficult to establish that baseline.

That would be merely a guess based on my assumptions from my understsanding of the definitions of the words. But that would be somewhat in line with the Muffpothesis, right?

The Muffpothesis states: 1) that the Respironics VB control layer is selectively discarding hypopneas during moments of wake, and 2) informally observed or projected HI discrepancies in cases of high VB% precisely support that conclusion.


None of the Respironics patent descriptions claim that their VB control layer---or any control layer---selectively discards hypopneas during wake moments. Rather, those moments of highly erratic breathing are bound to yield the greatest discrepancies among various differing manufacturer methods. That kind of disparity between some manufacturers is a likely function of highly erratic breathing itself being extremely difficult to reliably signal-process and statistically trend.

I don't think that a hypothesis where the VB control layer selectively discards wide-awake hypopneas is an inescapable conclusion---at least not simply because HI scoring discrepancies happen to exist among those two manufacturers during difficult-to-trend moments of erratic breathing.

[jnk]

So if one brand continues to report assumed hypopneas during difficult-to-trend moments and another brand does not, is there any chance that one difference in the reporting approaches, if that is really true, could relate both to the differences in the hypopnea estimates and to the existence of the second brand's VB% category?

jeff (also known as dsm II)

[-SWS]

So if one brand continues to report assumed hypopneas during difficult-to-trend moments and another brand does not, is there any chance that one difference in the reporting approaches, if that is really true, could relate both to the differences in the hypopnea estimates and to the existence of the second brand's VB% category?

My opinion is that the existence of the VB category is not for that purpose---and the existence of the VB category is neither inherently advantageous nor inherently disadvantageous regarding hypopnea detection and scoring.

Rather the individual hypopnea detection and scoring methods themselves are simply not going to yield identical measurement sensitivity and specificity---with moments of erratic breathing (i.e. "wild data" moments) being when the greatest disparity SHOULD manifest.

With that said, what makes the Muffpothesis a very attractive hypothesis is that it's a darn good design idea IMHO. Maybe even a little too clever. That design premise is technically feasible only if the manufacturers can accurately discern wake breathing from a slew of other variable breathing conditions.

But then again, Respironics now claims to be able to differentiate EEG associated RERA's with nothing more than an APAP data flow channel...

[Muffy]

There's always the possibility Respironics went with a proprietary data structure---perhaps even as an interdiction measure in part.

Well, I tried the basic stuff and all I get is hieroglyphics. There are some protections geared toward HIPPA compliance, so I wonder if the data itself is encrypted.

AHA! Finally got it!

þ"óÍÿ¦ßX üÅæ›_©ßX ü%æ›ÿU¿ùÍÌ7©ùæWtið—™o~y— ¹ùæ—uiðW˜oþ—.
þJóÍÿÜ¥Á_e¾ùŸº4ø«Í7ÿc— ùæèÒà¯5ßü÷] üuæ›ÿ®Kƒ¿Þ|óßvið7˜oþ›.
þFóÍÝ¥Áßd¾ù¯º4ø›Í7ÿe— ‹ùæ¿èÒào5ßüç] ümæ›ÿKƒ¿Ý|óŸviðw˜oþ“.
þNóÍÜ¥Áße¾ùº4ø»Í7ÿa— ùæ?èÒàï5ßüû] ü}曯Kƒ¿ß|óïvið ˜oþ.
þAóÍ¿Ý¥Á?d¾ù·º4ø‡Í7ÿf— ÿˆùæßèÒà5ßüë] ücæ›­KƒÜ|ó¯viðO˜oþ•.
þIóÍ¿Ü¥Á?e¾ù—º4ø§Í7ÿb— ÿŒùæ_èÒàŸ5ßüó] üsæ›®KƒÞ|óÏvið/˜oþ™.
þEóÍ?Ý¥Á¿d¾ù§º4ø—Í7ÿd— ÿŠùæŸèÒà_5ßüã] ükæ›KƒÝ|óviðo˜oþ‘.

Clearly, there's no way one could feel completely rested if their þNóÍÜ¥Áße¾ùº4ø»Í7ÿa was ßüó] üsæ›®Kƒ.

If nothing else, I now have the board's biggest collection of WingDings.

Muffy