Someone 'splain this........Lars4life - ok to read!

This has nothing to do with leak rate math or plastic bags or picking nits off or vacuum cleaners in heat, but I've been laughing so much at the witty posts here, and have been so in awe of the math displays... just had to share this:

I admit to the most embarrassing things about my lack of basic math skills --
viewtopic.php?p=9427

So... as you can see, it's the funny comments in this thread that kept me reading.

You all do a great job in every respect. I love the way you write even if I've been getting left wayyyyy behind on the numbers trail.

This, though, I do understand and agree with absolutely:

Pugsy wrote:When it really comes down to it though, 5 L/min variable in leak is really insignificant. The reports and software were never designed to be an exact reporting medium. Designed for trends and patterns to see what is going on in general. We are trying to fit a square peg in a round hole. Nit picking is time consuming, prone to give headaches and when all the nits are picked we still are left with imprecise numbers. They are close and really close enough to give the general idea as to what is going on and that is all it was really supposed to do.

rested gal wrote: This, though, I do understand and agree with absolutely:

Pugsy wrote:When it really comes down to it though, 5 L/min variable in leak is really insignificant. The reports and software were never designed to be an exact reporting medium. Designed for trends and patterns to see what is going on in general. We are trying to fit a square peg in a round hole. Nit picking is time consuming, prone to give headaches and when all the nits are picked we still are left with imprecise numbers. They are close and really close enough to give the general idea as to what is going on and that is all it was really supposed to do.

My problem with the "numbers" is not that they are fuzzy. My problem (as a mathematician) is that this critical information about how to interpret those fuzzy numbers is far too deeply buried for the average, but curious and moderately numerically-inclined patient/user to actually find.

For example: The graphs and verbal information provided about my Swift FX for Her's intentional leak rate include NO mention of +/- error rate---although I know (as a mathematician) that there is one---and the figure Pugsy cites sounds like it's a very reasonable guesstimate. Personally, I've been assuming all along that my "negative" unintentional leak rates are caused by nothing more than the fact that my real unintentional leak rate is pretty darn close to 0 and within the margin of error for the mask. Hence, factoring in the margin of error means:

• -5 L/m <= My real unintentional leak rate <= 5 L/m
  which implies
  15 L/m <= My total leak rate <= 25 L/m (at 4cm of pressure much of the time)

And even though I've not yet had such a night, this bit of arithmetic means that I'd not be worried (at all) if I started seeing flattish total leak rate lines up around 25 instead of my more typical 17 to 18. Shape of the curve is more important here than what the exact y-value is. Oddly enough I've found such +/- margin of error figures in reading OTHER Resmed mask user manuals, just not the Swift FX's.

But the BIGGER problem for users of the PR S1 is that the phrase Large Leak is NOT defined!
The documentation that came with my PR System One BiPAP Auto (both the user AND clinical manuals!), mentions that "large leaks" will show up in the 7 and 30 day averages in the on-screen data and says absolutely NOTHING about what constitutes a "large leak." My Encore Viewer and Encore Pro help browsers say nothing more than Large Leaks are flagged by a change in the Green Bar in the leak graph from darker, bright Green to a ligher, fuzzier green. And without (even a fuzzy) definition, there is no way to meaningfully interpret even the fuzzy data is provided by the machine either on its LCD or in Encore.

For example, I have NEVER, EVER been flagged with a "Large Leak" in some 100+ days of using the PR S1. Is my mask fit really that good that I've NEVER, EVER had even a short Large Leak in 100+ days of use?

Well, in the roughly 90 days of using the Resmed S9 AutoSet, there were isolated times where my unintentional leak rate went above the RedLine at 24 L/min for as long as 5 to 20 minutes or so. Not on a daily or weekly basis, but certainly a couple of times a month I'd have a brief large leak. Not important from a therapy point of view, but a large leak nonetheless. So this establishes a baseline that at least rarely I have leaks large enough to make the S9 flag them in ResScan as "large leaks."

In the 100+ days using the PR S1, my largest leak was a 20 minute leak with a total leak rate at about 50L/min (twice the intentional leak rate of my mask at my pressures). I've also seen total leak rates this large for somewhat shorter periods several other times---maybe two times a month on average, which is about the same rate as "large leaks on the S9". And none of these leaks on the PR S1 has been flagged as a "Large Leaks" either on the PR S1 LCD or in Encore. So is it really possible that an unintended leak rate that is as large as the intended leak rate is really NOT a large leak on a PR machine?

That's the mathematician's question that bugs me: To properly interpret how much a patient/user needs to worry about the maxes in their leak rate curve ("Do I have Large Leaks that might affect the quality of my therapy?"), we need to know how PR defines "Large Leak. On the PR S1:

• Is Large Leak defined as a leak exceeding an absolute number---like the 24 L/min RedLine in ResScan? If so, what's the magic number?
  
  Or is Large Leak defined in some kind of relative way? If so:
  • Must the unintentional leak be x-percent of the intentional leak rate? And what is x?
    
    Or must the total leak rate be x-percent higher than the baseline leak rate for y minutes? And what are x and y?

robysue wrote:Is Large Leak defined as a leak exceeding an absolute number---like the 24 L/min RedLine in ResScan? If so, what's the magic number?

It was the same when I started with the M series. I looked and looked for that line in the sand but never found anything close.
It must be an ultra secret number.

I will be getting a PR S1 APAP sometime next week. I volunteer to be a guinea pig and play around with leaks to see if I can find the trigger point. I do know I have seen 80 L/min when pillows slipped out of nose and was in my eye and I didn't get a Large Leak trigger. I would be guinea pig now but I have no way to monitor data as I have lent out my backup APAP and card reader for someone to try to sort through some problems. I am suspecting it is going to be upwards near 100 for Total Leak number.

Regarding the plus or minus variables with mask leak paperwork. I know I have seen them in the past but now I can't find it in the literature. It was a tiny footnote beneath the graph explanations. I will poke around and see if I have any paperwork from old masks that show it. Maybe it is something they have done away with on that pressure/graph page. I know that when I first started therapy it was there on each mask brand/model I tried.

Pugsy wrote:I will be getting a PR S1 APAP sometime next week.

Wait. What? You mean like the one I have? Kewl.

DD

Drowsy Dancer wrote:Wait. What? You mean like the one I have? Kewl.

Yep, yep, yep

PR S1 APAP with AFlex. Gently used. Hubby was like "why, you have 2 perfectly good machines right now".....The quieter excuse didn't work since he says he can't hear the one I use now even if he hovers over me while I am asleep. So I had to use the "more data" excuse. It was an "I want" rather than "I need" which is usually all it takes in our household.

Maybe then I can make some sense out of waveform data that people ask me about. All in all, just curious.

It isn't the latest backlit one but I don't mind. I don't look at data on screen anyway and I have no doubt that any machine noise will be a non issue for me. The M series machine I use I can't hear either.

Pugsy wrote:

Drowsy Dancer wrote:Wait. What? You mean like the one I have? Kewl.

Yep, yep, yep

PR S1 APAP with AFlex. Gently used. Hubby was like "why, you have 2 perfectly good machines right now".....The quieter excuse didn't work since he says he can't hear the one I use now even if he hovers over me while I am asleep. So I had to use the "more data" excuse. It was an "I want" rather than "I need" which is usually all it takes in our household.

Maybe then I can make some sense out of waveform data that people ask me about. All in all, just curious.

It isn't the latest backlit one but I don't mind. I don't look at data on screen anyway and I have no doubt that any machine noise will be a non issue for me. The M series machine I use I can't hear either.

You won't regret it. You may not find the waveform data as much of a revelation as I do as a newbie (it's all a revelation to me), but I strongly suspect you will find it very interesting. You may have a less eventful sleeping life than I do at this point. At least I would hope so. There are some aspects of this that I am still trying to dial in after almost two months. I know, I know, the body isn't a clock, it's a garden (as my favorite doctor says).

Backlight feature is pretty useless anyway.

Oh, yeah:

robysue wrote:That's the mathematician's question that bugs me: To properly interpret how much a patient/user needs to worry about the maxes in their leak rate curve ("Do I have Large Leaks that might affect the quality of my therapy?"), we need to know how PR defines "Large Leak. On the PR S1:

Is Large Leak defined as a leak exceeding an absolute number---like the 24 L/min RedLine in ResScan? If so, what's the magic number?

Or is Large Leak defined in some kind of relative way? If so:
Must the unintentional leak be x-percent of the intentional leak rate? And what is x?

Or must the total leak rate be x-percent higher than the baseline leak rate for y minutes? And what are x and y?

+1 to that, although I would call it the "logical person's" question as I am not a mathematician, just logical.

DD

Drowsy Dancer wrote:+1 to that, although I would call it the "logical person's" question as I am not a mathematician, just logical.

Tis very true. While I do understand the frustration that Robysue feels (been down that road when I started cpap) I tend to look at things a bit differently now. If I can get an answer fine.. If I can't get an answer after expending considerable time searching, I accept the fact and move on to try to help people sort through issues that I can maybe help. Sort of like picking your battles. Learned a long time ago to not bother to pick the ones I never had a chance in winning. Got a lot fewer headaches with that attitude.

Is RobySue wrong with what she says? Gosh no, she is very right and I agree 100 percent with what she says.
It's just I don't let it bug me anymore. No more headaches trying to sort through minute numbers that always will vary anyway. I work with what I have and do the best I can with that. Tis just a difference in the way we look at things. Nothing more, nothing less.

Found this statement in Swift LT mask literature. Not the plus/minus thing that I remember seeing but hey, I am old and forgetful sometimes. Could have swore at some time or another I saw actual numbers....

The mask contains passive venting to prevent rebreathing. As a result of manufacturing variations, the vent rate flow may vary.

Pugsy wrote:

Drowsy Dancer wrote:Wait. What? You mean like the one I have? Kewl.

Yep, yep, yep

PR S1 APAP with AFlex. Gently used. Hubby was like "why, you have 2 perfectly good machines right now".....The quieter excuse didn't work since he says he can't hear the one I use now even if he hovers over me while I am asleep. So I had to use the "more data" excuse. It was an "I want" rather than "I need" which is usually all it takes in our household.

Maybe then I can make some sense out of waveform data that people ask me about. All in all, just curious.

It isn't the latest backlit one but I don't mind. I don't look at data on screen anyway and I have no doubt that any machine noise will be a non issue for me. The M series machine I use I can't hear either.

FWIW, just got a replacement Sys1 auto, not the new backlit one, my new machine is quieter, so there is some variation within models. hubby even says it's quieter

Gotta run off to class momentarily.

The mathematician angle is the larger problem we mathematicians face with the idea of innumeracy in our culture: the idea that it is just fine to not understand numbers and data at a truly basic level. And the fact that the PR machines create this data labeled Large Leak with no plausible definition of what constitutes a Large Leak is a working example of (corporate) innumeracy at its worse: We (PR) won't let the users in on how our engineers define this (rather important) concept because the mathematics behind it will frighten them out of their pea sized brains. Best leave it totally undefined altogether.

For more details about the problem of innumeracy in American culture, find and read a book by John Paulos called Innumeracy. It's second edition is ten years old, but sadly not much has changed. A lighter read that makes the same points is Paulos's A Mathematician Reads the Newspaper.

And getting back to McSnoresalot's original post and how it ties into PR's fostering of innumeracy on us users of the System One: Recall that McSnoresalot posted his leak data from two recent nights and it looks like this:

McSnoresalot wrote:

Note that there are two brief periods flagged by Encore Viewer as Large Leaks in the second graph. In actually printing off these graphs and using a real ruler because my eye-to-hand coordination was not working sufficiently well with the computer's mouse, it appears that McSnoresalot's UNINTENTIONAL leak rate (as guessed by the PR S1) was a bit above 50 L/min during those periods.

It also appears that McSnoresalot's UNINTENTIONAL leak rate was above 50 L/min for a LONGER period of time on the first max bump on the leak graph for that second night that occurs from about t=1.6 to about t=1.9 (in other words, the bump just before t=2.) So what gives? Why are the two shorter periods with unintentional leak rate > 50 L/min flagged, but the longer period with unintentional leak rate > 50 not? Does it have to do with the fact that the flagged leaks are miniature maximum bumps on a longer leak at 40 L/min? Or something else?

Since I have no other meaningful way of defining the phrase Large Leak, I will temporarily define it myself in terms of the (possibly irrelevant) Resmed Red Frowny Face standard: Unintentionally leaking at a rate of 24 L/min for at least 25% of the night is enough to make the Resmed engineers believe that CPAP therapy is likely being compromised by the size of the leaks. Using this standard, McSnoresalot would CLEARLY have seen Mr. Red Frowny Face on the second night since his unintentional leak rate exceeds 24 L/min for all but the first hour according to my eyes. On the first night of the given data, McSnoresalot's unintentional leak rate is above 24 L/min for a total of about an hour and a half out of about 6 1/2 hours of mask time: That's pretty close to the line between when Mr. Red Frowny Face and his more pleasant cousin Mr. Green Smiley Face show up. So I wouldn't want to make a guess as to which one would show up.

And so to my mathematician's mind we get back to the same important fact: The way PR has chosen to present the Large Leake data is clearly confusing---to the point of being meaningless---to McSnoresalot and plenty of other users of the System One---including, quite frankly, myself. Fortunately, McSnoresalot knows about Cpaptalk. But suppose he didn't: I contend that the average, non-Cpaptalk reader would take these figures and numbers and graphs and incorrectly and innumerately focus in on the following things:

• * The fact that the leak line stays well below that Dark Green Line (and that's gotta be good, doesn't it?)
  
  * There's only a tiny bit of light green compared to the large amount of dark green (and that's gotta be good, doesn't it?)
  
  * That miniscule reported 3.0 minutes in Large Leak is really, really small (and that's gotta be good, doesn't it?)
  
  * That % of Night in Large Leak = 1% of NIght is the OBVIOUS CLEAR PROOF there's no problem because 1% of ANYTHING doesn't matter does it?

And that poor hapless, non-Cpaptalk PR S1 user would incorrectly and innumerately conclude that he has absolutely NO problem with leaks. So at the next follow up with the doc, the conversation goes like this:

• Doc: How are you feeling?
  
  Patient: Still having daytime sleepiness and exhaustion. I don't feel like I'm sleeping that well either.
  
  Doc: Any problem with leaks?
  
  Patient: Nope. The machine says I'm not leaking.
  
  Doc: Then just give it more time then.

So pugsy, I get your point: Once we get our therapy optimized (and I'm still working on that) we don't need to sweat the fuzziness of the numbers and the lack of definitions of those numbers quite so much. And we learn to live with the uncertainty that's inherent in any form of real data. But the problem with how PR chooses to report the Leak Data is not a minor inconvenience: It seriously distorts an average user's sense of whether there is (or more precisely) is NOT a problem with mask leakage. And taming the mask and leakage is critically important to obtaining decent quality therapy.

robysue wrote:But the problem with how PR chooses to report the Leak Data is not a minor inconvenience: It seriously distorts an average user's sense of whether there is (or more precisely) is NOT a problem with mask leakage. And taming the mask and leakage is critically important to obtaining decent quality therapy.

Very true and I never said it was unimportant. Only that I can't do anything about it so I choose to deal with what I can do something about. I do the best I can with what I have available and go from there.

Why Respironics chose to do it this way is way beyond my comprehension. Obviously no one with any common sense was consulted. Unfortunately that tends to be the status quo for a lot of what is done in the sleep apnea marketplace.

I'm afraid I just have my Don Quixote hat on today: Too much tilting at windmills. *sigh*