Questions for Respironics regarding their APAPs.

I'd really like to understand the advantages of different alogarithms for diffefrent conditions.

I have the Respironics APAP but my challamge is mostly Hypop's. I have "heard" that a different machine might give me better treatment.

I'd like to beter understand the differences and if a different machine might yield better results and a lower AHI.

I'd like to also understand why my AHI apprears to be lowest at a 5-6cm range when I was titrated at 9cm.

Tnx,

Tom

NightHawkeye wrote:

Titrator wrote:This is a built in feature, not a defect.

Oh, that is so funny. .

...that's exactly why those of us in the design realm kid around about built-in "design features". Did you hear that from a Respironics sales rep?

The thing that puts the lie to that answer is that the 10 minute gap doesn't always happen. It happens only about 50% of the time on my Remstar-auto, and when exactly it happens is unpredictable. If this were truly a designed in feature, then shouldn't it happen 100% of the time? Conversely, if it wasn't designed in, then it should never happen. Either way, there's a design flaw in the software. (I'm sure someone will take issue with that statement though.)[/url]

I will, Bill.

Titrator is exactly correct---as he most often is in my opinion. It's called "settling period". On the Resmed and Puritan Bennett machines, time alone is the sole criteria for the expiration of that "settling period". However, on the Respironics AutoPAP machine, that "settling period" is also derived by assessing whether a sufficient number of sequential respiratory cycles have occurred that do not fit high-probability patterns of wakefulness. So the "settling period" is fixed for Resmed and PB, but it can be variable by design regarding the Respironics algorithm. At least that's my own take reading the patent description. I honestly think you're mistaken by labelling this one as a bug and a lie. Simple "black box" analysis can be so deceiving on these machines.

Back to holiday weekend client-based work for me... .

Algorithm questions:

What kind of breathing event does the machine identify as as hypopnea, and how is it programed to respond to it?

Are the manufacturers aware of the fact that some people experience runaway pressure on the machine if the upper pressure isn't limited. Do they know which kinds of breathing patterns are more prone to that?

O.

roztom wrote:I'd like to also understand why my AHI apprears to be lowest at a 5-6cm range when I was titrated at 9cm.

Tnx,

Tom

Tom, to understand the pressure curve on the APAP one must understand how the individual AHI values are calculated at each pressure, and what it means. The AHI at 5cm for example is the total number of OA/H events recorded at that pressure (5cm) divided by the total number of minutes accumulated at that pressure (5cm). Notice also that the accumulated time at that pressure is relatively LARGE compared with higher pressures. That equates to a very LARGE denominator, and of course the AHI for that pressure (5-6cm) may be relatively small, as in "good", like golf scores. If you have MyEncore, notice the graph labeled "AHI vs pressure". By your comments I would presume there is a peak at some pressure higher than 5-6 that you consider worrisome. That implies that there are a number of OA/H events (you suggest most are hypopneas) that are recorded at that higher pressure. Also there are far fewer minutes accumulated at that higher pressure, hence a higher AHI calculated at that higher pressure.

Zoom back a level and ask yourself (as an xPAP user) what are you trying to improve upon? Is it the AHI? or is it really the total number of OA/H events, the total time spent in OA/H, the average time spent in OA/H, etc.? I think you would agree that it is the latter, and not the rather artificial AHI number. Given that, do you REALLY want to be on the LEFT side of the AHI curve where the "numbers are better" but the therapy is lousy? That's not to discount the meaning or significance of the "TOTAL AHI" number. But that's not what you're asking. Somehow I think that you may have interpreted the low AHI at the low pressure to mean that if you just set the machine as straight CPAP at, say, 6 your TOTAL AHI will be what the number suggests at the pressure of 6. What is missing is that the low pressure will fail to resolve those "events" that require higher pressures. Those won't go away just by removing those pressures from your machine. In fact those higher pressure events will now be "charged' to that lower pressure, and the calculated AHI at that pressure will now be higher, and not what you may have expected. You can try that as an experiment, say for one night set it as a straight CPAP at 6, or even 5, and see if the TOTAL AHI improves. (I feel like I'm suffocating at 5, and leave the lower limit at 6.5 just to breathe normal).

To put it another way, you are not allowed to play ONLY the par 5 greens, and just walk away from the par 7, 9, 12... whatever, just because your 'average hitting index" (AHI) is soooo good on the easy ones.

You were titrated at 9cm because on your sleep study that was the pressure required to "resolve" =>90% of the OA/H events. Even then it is likely that the AHI for a given low pressure (5-6) was better than the actual pressure required to do the job. If you look at the titration curve on your EncorePro "full details" report, I will guess the number is not much different from your SS titration. It is likely higher than the peak value in the row labeled AHI: on the "daily events per hour" chart. And certainly higher than the AHI on the leftmost pressure value. On an APAP you want to be within striking range of that titrated value, usually +/- 3cm. Here's a more benign experiment you could try: raise the lower pressure limit by only one cm, leave it there for a week, see if the TOTAL AHI improves or worsens. Theory would suggest it might improve.

Not sure if that clarified or confused the issue. I await your experimental results! .

-Ric

A settling period for the time when I'm falling asleep, but no asleep yet, makes a lot of sense to me - a PB420E user.

I don't want my treatment statisitice cluttered up by misidentified events, that are the result of my tossing and turning before I fall asleep. I know, from reading the manual that SL3 will not record anything for the first 15 minutes, and that's fine with me.

Bill, you seem to imply that your machine has 10 minute gaps which I understand as stopping and starting without any active involvement by yourself. This is not what Ric is describing: Also, my machine lags 10 minutes before it starts to accumulate usage data. For example, if I use the machine for 12 minutes, it records only 2 minutes. Anything less than 10 minutes "didn't happen". It is the last two minutes that will get recorded according to Ric, not your "nagging question of why does the machine record exactly two minutes of data preceding the ten minute gap?"


If my understanding is correct, and your machines stops recording just like that, without your having turned it off, then your machine is probably faulty.

Unless it has suddenly decided that you're awake.

If that is so, I have another algorithm question:

Does the machine ever decide on its own to stop recorcing/reporting data because it has identified the breathing pattern as that of wakefullness?

O.

ozij wrote:Bill, you seem to imply that your machine has 10 minute gaps which I understand as stopping and starting without any active involvement by yourself

<snipped>

If my understanding is correct, and your machines stops recording just like that, without your having turned it off, then your machine is probably faulty.

Bill, a couple of questions, if you don't mind...

Do you occasionally turn your machine off and back on before the nightly session is over? For example, to go to the bathroom? Or for any other reason?

I'm sure you've checked to see that the power cord to your machine is seated firmly.

ozij wrote: Bill, you seem to imply that your machine has 10 minute gaps which I understand as stopping and starting without any active involvement by yourself. This is not what Ric is describing: Also, my machine lags 10 minutes before it starts to accumulate usage data. For example, if I use the machine for 12 minutes, it records only 2 minutes. Anything less than 10 minutes "didn't happen". It is the last two minutes that will get recorded according to Ric, not your "nagging question of why does the machine record exactly two minutes of data preceding the ten minute gap?"

I thought Bill was talking about Ric's "settling period" question. I didn't realize that Bill had ten minute gaps in the middle of sessions. If he does, then your question, Ozi, and Rested Gal's question each go a long way to discovering exactly what is happening with Bill's Encore reports.

1) Ozi's Question is: if there are ten minute gaps in the middle of sessions, is it because the algorithm has detected a high-probability occurrence of wakefulness based on patterns in sequential respiratory cycles? (in this case a mid-session "settling period" might occur)

2) Rested Gal's question is: Could it be that Respironics users are powering down their machines at night to go to the bathroom, lower pressure, etc. Could it also be that people are unknowingly experiencing brief power-down situations related to the AC power itself. (in this case multiple "settling periods" just may occur through out the night based on multiple power-up occurences)

I also want to go back to the patent description for the first time in more than a year and review my own initial thought that "sequential respiratory cycle" determines a variable length of settling period, versus simply ten minute fixed sessions of "settling period". Giving it more thought, it sounds as if that sequential respiratory cycle assessment (the probability-assessment for wakefulness) may simply be criteria for whether a ten-minute "settling period" should be deployed in the middle of a sleep session. So I'd like to even overturn myl own initial assumption that settling period can be variable in the Respironics case (I still tentatively think it may be variable---perhaps on an infrequent basis only----based on the possibility of back-to-back sequences of proabable wakefulness detected, each followed by machine-imposed "settling periods")

Highly complicated, but interesting stuff. .

rested gal wrote:Bill, a couple of questions, if you don't mind...

Do you occasionally turn your machine off and back on before the nightly session is over? For example, to go to the bathroom? Or for any other reason?

Always glad to answer questions, RG. Yes, in the earlier chart, I believe that I got up twice during the night. However, as you can see in the chart (same chart as earlier) I got up at hour 4:40 and when I turned the machine back on it provided 2 minutes of data followed by a 10 minute gap in data.



Here's another one which illustrates the ten minute gap when I went to bed and also after getting up close to hour 5.

rested gal wrote:I thought Bill was talking about Ric's "settling period" question. Embarassed I didn't realize that Bill had ten minute gaps in the middle of sessions. If he does, then your question and Rested Gal's question goes a long way to discovering exactly what is happening with Bill's Encore reports.

You were entirely correct to begin with SWS. The ten minute gaps only occur at the beginning of a session, that is to say, when I power the machine ON. I used to get up several times during a night which accounts for all the ON/OFF cycles in the data. Whenever the ten minute gap showed up, the pattern was always the same: exactly two minutes of pre-gap data, then exactly ten minutes of gap, then normal recording.

Your explanation about the settling time could be absolutely correct if the machine consistently provided the gap immediately after being turned ON. However, as is apparent from just the couple of sessions I've provided here, that is not what the machine does. Sometimes the gap occurs, sometimes it doesn't.

When I first posted about this months ago, I asked if anyone else saw this and received positive responses to the question which means that it isn't unique to my particular machine. And from what Titrator said, even Respironics acknowledges that the gap occurs, so it isn't my machine at all.

Now, SWS, it's easy enough for me to accept your explanation that the algorithm isn't working for the first ten minutes, but if that were true, then the data would be consistent and the gap would always occur. It doesn't. The only question in my mind is whether Respironics deliberately wanted the gap to be present or whether the gap accidentally occurred.

If Respironics wanted the gap to be there, then the design flaw is that it isn't always there. If the gap is accidental, then that is the design flaw. The explanation given by Titrator is fallacious one way or the other. Like many other things in life, in explaining this, they can't have it both ways.

Regards,
Bill

This is representative of my data and as you can see it pretty much matches Bill's to a T.

Just reread his post and look at my graph, something is just not right.

Bob

NightHawkeye wrote:The explanation given by Titrator is fallacious one way or the other.

Oh, that is so funny. .

No disrespect intended, NightHawkeye. But just for the record, Titrator answered Ric's question correctly. There was nothing fallacious about it.

NightHawkeye wrote:Your explanation about the settling time could be absolutely correct if the machine consistently provided the gap immediately after being turned ON. However, as is apparent from just the couple of sessions I've provided here, that is not what the machine does. Sometimes the gap occurs, sometimes it doesn't.

If the algorithm is pattern and probability based (it is), then there is a probability based explanation that keeps the "bug jury" still in session:

Infrequent patterns that are none-the-less highly probable of sleep ("non-wakefulness" patterns) might plausibly and very occasionally prompt that algorithm to suspend that initial "settling period". Now admittedly any pattern-and-probability based algorithm unfortunately entails most APAP user-trials falling on the good/long end of that probability stick, while a few will necessarily fall on the short end of probability's stick regarding pattern detection and event response.

Looking at your data and explanation, Bill, I see that there may very well be a bug here. However, at this point I'm still not convinced that simply because the settling period is there most of the time and it's not there some of the time, that alone is sufficient criteria to call this a definite bug. Not without knowing exactly what's going on in the algorithm as well as what is happening across a highly diverse signal and condition patient population----and the designers' objectives, constraints, and even strategies regarding that vast and highly diverse body of patient circumstances.

Bill, do you think it's feasible under any circumstances that a pattern-and-probability algorithm might elect to suspend an initial settling period? My point with making definitive statements relying on "black box" analysis only, is that if we don't know the vast mitigating circumstances and considerations, then the probability of analysis error on our part significantly increases. Or putting it another way, the probability for erroneous black box assumptions and analysis goes up significantly with the overall complexity of the system or issue at hand. And this issue is highly complex.

It will be nice to hear from Respironics on this issue. And, Bill, thanks for a great discussion.

Looking at those charts I'd be most inclined to initially suspect data reporting error also. However, I also think it's entirely feasible that the algorithm might use probability to decide whether to issue a "settling period". I see absolutely no reason why a "settling period" might not be either issued or suspended (for either "beginning-session" or "mid-session") based solely on probability of wakefulness. Technically feasible from my own view. Very much so.

And yes, I too think Titrator answered Ric's question perfectly.

NightHawkeye wrote:

rested gal wrote:Bill, a couple of questions, if you don't mind...

Do you occasionally turn your machine off and back on before the nightly session is over? For example, to go to the bathroom? Or for any other reason?

Yep, I did write that. Thanks for the answer.

I didn't write the following, though -- this quote attributed to me is from a post by -SWS. But that's ok.

NightHawkeye wrote:

rested gal wrote:I thought Bill was talking about Ric's "settling period" question. I didn't realize that Bill had ten minute gaps in the middle of sessions. If he does, then your question and Rested Gal's question goes a long way to discovering exactly what is happening with Bill's Encore reports.

Re this 10 min gap. I well recall Bill reporting this situation and the responses.
But, please refresh my memory re which machine was producing this 10-min gap.

IIRC it was the Bipap AUTO & not the Remstar AUTO ?

In other words, IIRC it is one model that displays the gap ?.

Tks

DSM

Sorry, RG. I meant to attribute the quote to -SWS. oops. .

Actually, DSM, it is the Remstar-auto that produces the gaps. I don't see anything like that with the BiPAP-auto, just a nice consistent output data stream without gaps.

I am amazed that this small glitch has attracted so many folks to defend the algorithm so profoundly. I'll repeat what I said earlier, now that I know that it's a problem with the machine it doesn't bother me in the least.

However, since you guys are so determined to defend the Respironics algorithms, as if they are perfection incarnate, I'll offer up one more proposition for you to defend.

From a human factors perspective, what in the world would be the advantage of having an exactly ten minute long data gap randomly show up 50% of the time exactly two minutes after the machine has been turned on. I hardly think there is one. After all, the machine is still blowing air at a certain pressure. Why not report what the machine is doing? I would propose to you that reporting data showing that the machine is turned OFF, when instead the machine is very definitely turned ON, is an unnecessarily confusing factor in interpreting the data. We already know that Respironics has had to defend against questions about it. I am also supposing that, since Titrator was the one reporting the Respironics response, those questions which needed to be answered came directly from sleep professionals.

Regards,
Bill