30 seconds usually far enough back. Sometimes it's worth looking at the whole minute before the event that you are wondering about. But I'll be honest, I typically prefer to see about 2 minutes back before the event occurred so that I get an understanding of the overall picture of what was going on at the time the event got flagged.Tec5 wrote: ↑Fri Dec 09, 2022 10:42 amThe breath that was snipped (that apparently is associated with the a 0.04 flow limitation occurred 20 seconds before the hypopnea flag.
It's quite reasonable to use all the data in the 10 second interval for evaluating a hypopnea event, but to go back to an observation of data well in advance is unreasonable.
How far back in time do you folks want to go?
Hypopneas can last more than 10 seconds. And I have found that the time flags on the hypopneas are usually far less accurate than the fact that something should be flagged. Sometimes the flag is also placed on top of the recovery breaths rather than at the end of the hypopnea itself. Why? I don't know, but programming these things is always going to still have some bugs where stuff doesn't always line up exactly correctly.The the period this hypopnea flag should be examining should be from 01:09:10 to 01:09:20, not something that happened ten seconds sooner. (or do you disagree?)
Here is a marked up screen shot from my own data of a hypopnea that is most likely a post-arousal false positive hypopnea, and it's easy to wonder why the H is drawn where Oscar draws it:
Note that the H flag occurs after the first recovery breath. Why? I don't have the foggiest idea. Now the machine flags this as a 10-second hypopnea---we can see that when I use the mouse to hover over the hypopnea and Oscar "highlights" the breathing being flagged in this next screen shot of the exact same post-arousal hypopnea:
But that the breath right before the "official start" of the hypopnea is clearly both flow limited and the smallest inhalation in the whole bit of displayed breathing, and in my opinion, the hypopnea starts with that inhalation rather than the 10-second window that is officially "flagged" as the hypopnea. Why isn't this hypopnea scored as a 15 second event with that first flow limited and smallest inhalation included in the area flagged as the hypopnea? I don't know.
Yes, it can take more than 10 seconds---most simply because a hypopnea can last more than 10 seconds. Moreover, scoring the end of the hypopnea is also a programming decision: How big does the recovery breath have to be? Do you need two of them? Those are also programming choices that only the Resmed programmers know the actual coding decisions that were made and only the Resmed programmers understand what kinds of things can creep into the flagging data when something barely meets the criteria.(I guess an alternative theory might be that the hypopnea detecting algorithm takes more than ten seconds to decide)
It's important to remember: Resmed does not claim 100% accuracy in the flagging of events. They know there are false positives and false negatives. They know the flags don't always line up completely correctly. But statistically, the algorithms are good enough to have validity for trending data: If the machine AHIs are reasonably low for several weeks or months, it's a good bet that the actual number of real events that would be scored on an in-lab PSG are also quite low and the OSA is being well treated.
Questions I have for both you and Wondering1: What is the point of your analyzing the particular hypopneas in question? Are you trying to understand whether a particular hypopnea is a false positive or real? Or are you trying to make the case that the algorithms for scoring Hs are so inaccurate that we just shouldn't trust the things the machines flag at all? Or something else?
I ask, because neither one of you is willing to accept the answer: Programs based on sophisticated numerical analysis are never 100% accurate and errors can and do creep in when those algorithms are faced with data that is right at the edge of their abilities to "make decisions" based on programmed, numerical definitions. And that if you keep that point in mind, it's actually pretty easy to see why the machine could decide to flag the hypopneas you have presented in this thread.
