Reading OSCAR leak stats ?

Does this look like an okay leak stat report? There's that 95% figure again which I still do not understand. Is that an okay % of time over threshold? I do understand that <25 on part of the OSCAR report is an okay leak score. I have leak problems and they seem correlate more with good/bad sleep than the AHI score.

Hard to tell by looking only at these numbers. Leaks should be 24L or below for the machine to be able to compensate for them. 96% means that 95% of the time you were at this level OR BELOW.

If the leaks are not bothering you, I wouldn't worry about it. Resmed says anything above 24 liters is excessive. However, the machine can handle at least 34 liters and still adequately treat your apnea.

ejbpesca wrote: ↑

Sun Dec 25, 2022 9:56 am

Is that an okay % of time over threshold? I do understand that <25 on part of the OSCAR report is an okay leak score. I have leak problems and they seem correlate more with good/bad sleep than the AHI score.

If the leaks bother you, they're not OK, not matter what the fromal threshold si.

The table you posted shows that you're spending a growing percent of your time above the leak threshold - so clearly something has to be taken care of.

LSAT wrote: ↑

Sun Dec 25, 2022 1:13 pm

Hard to tell by looking only at these numbers. Leaks should be 24L or below for the machine to be able to compensate for them. 96% means that 95% of the time you were at this level or below.

I get the 95% of the time at that level, it is the or below that makes the level number meaningless to me. If 95% of the time I may have been anywhere below the level indicated, what does that tell me?

ejbpesca wrote: ↑

Thu Dec 29, 2022 3:11 pm

If 95% of the time I may have been anywhere below the level indicated, what does that tell me?

95% numbers are just numbers and don't mean much of anything. So it doesn't tell you a lot and you just have to learn to accept the fact that the "or below" part of the definition is just words. 95% numbers...be it pressure or leak...are just numbers. They aren't the holy grail of anything like people want to make them out to be.

ejbpesca wrote: ↑

Thu Dec 29, 2022 3:11 pm

LSAT wrote: ↑

Sun Dec 25, 2022 1:13 pm

Hard to tell by looking only at these numbers. Leaks should be 24L or below for the machine to be able to compensate for them. 96% means that 95% of the time you were at this level or below.

I get the 95% of the time at that level, it is the or below that makes the level number meaningless to me. If 95% of the time I may have been anywhere below the level indicated, what does that tell me?

95 percentile, or 95% for short, is a statistics term. It may be easiest to illustrate what this means with a couple of sets of discrete set of data.

The following list contains 100 numbers on it, arranged in increasing order. The 95th number is highlighted.

0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 2, 3, 4,
5, 7, 8, 8, 9, 9, 10, 10, 10, 10,
10, 10, 10, 10, 10, 10, 11, 13, 15, 18

That 95th number on the list is a 10, and that means that the 95% for this set of numbers is 10. In other words, 95 percent of the numbers on this list are at or below 10. And at most 5 percent of the numbers on the list can be strictly above 10. (In this particular example, only 4 of the numbers on the list are strictly above 10 because the 96th percentile and the 95th percentile are both equal to 10.). You should also notice that 57 of the numbers on the list are actually 0's. So the majority of numbers on this list are significantly less than the 95% number. But that does not always happen. Consider this list of data that also has a 95% equal to 10:

0, 1, 2, 3, 4, 5, 6, 7, 8, 10,
10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
10, 10, 10, 10, 10, 11, 12, 13, 15, 18

Again, 95 percent of the numbers on this list are at or below 10. And at most 5% of the numbers on the list are strictly above 10. But 86 of the numbers on this list are equal to 10. So in this case the 95 percentile data point equals the vast majority of the data on the list.

In the context of the Resmed leak data, if your 95% leak rate is below 24 L/min (the Resmed definition of "Large Leak"), then you can be sure that the time spent in seriously large leaks is less than 5% of the entire night. From a practical point of view, that means that if your 95% leak rate is below 24 L/min, you only need to worry about leaks if they are waking you up because any genuine large leaks did not last long enough to adversely affect your therapy.

Also in the context of Resmed leak data, if your 95% leak rate is well below 24 L/min---as in your 95% leak rate is in the single digits---that indicates that the total length of time spent in moderate sized leaks (say leaks with a rate between 10 and 24 L/min) is less than 5% of the night.

And finally in the context of Resmed leak data, if your 95% leak rate is at or above 24 L/min, that indicates that it is well worth looking at the leak data for the night in Oscar's daily data to see whether you had really large leaks for a bit more than 5% of the night or whether you were leaking like a sieve most of the night. If your median leak rate is anywhere close to 24 L/min, then you also need to look carefully at the detailed daily data to see just how bad the leaks are.

Does that help answer your questions?

ejbpesca wrote: ↑

Fri Dec 30, 2022 7:22 am

robysue1 wrote: ↑

Thu Dec 29, 2022 5:47 pm

Does that help answer your questions?

Yes. Thank you. I passed stats class with a B, but left more confused than I entered.

I'm sure there is some esoteric, but practical, stats applications of the 95 percentile as shown in example with number groups, but as far as users of CPAP machines go, it would be better to state leaks in a simple quantitative format like events shows in OSCAR. At least it is not shown by a whisker box.

But leaks are often not discrete, short-lived events. When folks have serious leak problems that must be addressed, it is possible that they are leaking at a reasonably high rate all night long. Of course this will be obvious if you look at the leak graph rather than just the statistical numbers in the left side bar or the summary data.

In my own opinion, the 95% leak rate is useful in demonstrating that leaks are not a problem that is adversely affecting the quality of xPAP therapy: If your 95% leak rate is below the definition of "Large Leak", then you only need to worry about leaks if they are waking you up.

On the other hand, if your 95% leak rate is in "Large Leak" territory, more investigation needs to be made. On the Resmeds, the first indication that large leaks are lasting long enough to adversely affect the quality of the xPAP therapy is the dreaded Mr. Red Frowny Face. He shows up if your unintentional leak rate is above 24 L/min for at least 30% of the night. (In stats talk, that means the 70% is at or above 24 L/min.)

The median leak rate shown in Oscar is also useful to look at when trying to figure out whether your leaks are large enough and long enough to adversely affect xPAP therapy: The closer the median leak rate is to 24 L/min, the more likely that your xPAP therapy is being seriously compromised by excessively long large leaks.