macewa wrote: I went back to bed at 3:33 and that is when I had all kinds of probems getting the mask to fit and feel right. Almost that whole hour from 3:33 until 4:20 I was awake, struggling with the mask. Then I fell back asleep waking up at 7:09 am.
My question is, since I had that awake time and the machine was scoring stuff while awake, would my actual AHI be lower if that time was removed? or does that really make a difference? My numbers are slightly higher last nights chart than the night before.
When I do rough calculations based on what I can see in the posted data, this is what I come up with put together in a nice table for those who interested just the results without looking at the math:
Code: Select all
Index SH computation Adjusted computation Percent Change
AHI 16.99 15.91 6% Decrease
OAI 7.92 8.86 12% Increase
CAI 4.44 2.27 to 2.63 40-50% Decrease
HI 4.63 3.86 to 4.32 7-17% Decrease
Now, how did I get those numbers? For those interested in the math behind the numbers, I've given the details.
Part I: Seat of the pants: Which numbers change and by how much?
The time elapsed between 3:33 and 4:20 is 47 minutes, about 3/4 of an hour.
Most of the events during that 3/4 hour of "non-sleep" were centrals. And there's a bunch of them. My first guess is at there about 16-20 or so events during this time. Looking closer at the data, it seems to me like there are about 18 events scored from 3:33 to 4:20. Two of them are OAs. There's somewhere around 5-7 Hs. The remaining 9-11 events during this time period are all centrals.
Take out that 3/4 hour and the associated 9 centrals, and the CAI should go WAY DOWN since it appears that about half of all the centrals recorded during the evening occur during this 47 minute period.
But there's another SOMEWHAT LARGER group of events (mainly OAs) between 4:45 and 5:45 when you were asleep. (Judging by the AHI graph, I'd guesstimate about 20-25 events in that hour.) And if you eliminated the 47 minutes of "non-sleep" from the data, you still have all these events, but the denominator of the AHI fraction is decreased by about 47 minutes---about 3/4 of an hour with only two OAs.
So the OAI should go UP. (But not as dramatically as the CAI should go DOWN)
As for the Hyponeas: They seem to be more evenly spread throughout the night and it's tough to tell if there are 5 Hs or 7 Hs (or perhaps 8?) in that 47 minute period of "non-sleep". That's probably not enough to noticeably increase the HI (on the one hand), but may not be enough to not noticeably decrease the HI (on the other hand).
Number sense says the HI should be affected less by throwing out the 47 minute period of "non-sleep" than either the CAI or OAI is affected.
And your total AHI? My best guess on eyeballing the data is that there are a FEW MORE OAs in that second cluster than there are CAs in the hour of "non-sleep". And in both clusters, there seems to be a few more events than the over all AHI = 16.99.
My seat of the pants guess is that the AHI with and without that hour of "non-sleep" are pretty close to each other, but the distribution of events is far different.
Part II: Some detailed, but not hard, arithmetic is used to figure out the adjusted indices
I can make a somewhat closer analysis of the data, but it's still not precise since I can't zoom in for exact counts of events that occurred during that 47 minute period of "non-sleep". Tell me those numbers, and I can easily compute exactly what the indices would be if you just throw out the 47 minutes of "non-sleep" data. The data is good enough, however, for me to make intelligent approximations of all the indices.
So to start, let's recall that to my eyes it looks like there are about 18 events scored from 3:33 to 4:20. Two of them are OAs. There's somewhere around 5-7 Hs. The remaining 11-9 events are all centrals. (If there were 5 Hs, then there were 11 CAs; if there were 6 Hs, then there were 10 CAs; if there were 7 Hs, then there were 9 CAs)
Adjusted AHI
We're throwing out 18 events during the 47 minutes of "non-sleep". That means there's 88 - 18 = 70 events scored when you were most likely genuinely asleep. The 88 events were scored in 5:10:47 of mask time; let's call it 5:11 of mask time for convenience. So:
- The SH AHI = (88 events)/(5:11 mask time) = (88 events)/(5.18333 hours of mask time) = 16.98, with that 0.01 discrepancy caused by round-off error, both mine and SH's.
When we subtract the 18 or so events that occurred during from the "non-sleep" time AND the 47 minutes of "non-sleep" from the mask time, we compute the adjusted AHI as follows:
- Adjusted AHI
= (70 events)/(5:11 mask time - 0:47 minutes of WAKE)
= (70 events)/(4:24 of sleep time) = (70 events)/(4.4 hours of sleep)
= 15.91
Note that difference between an AHI = 16.99 and an AHI = 15.91 is not that significant. (The relative difference between these two numbers is about 6% of the size of the numbers themselves.)
To try to see what would happen to the CAI, OAI, and HI in takes a bit more detective work and some arithmetic.
The Adjusted OAI
We need to figure out how many of the OAs occurred during the 4.4 hours of SLEEP. Here's how we can do that: The OAI for the
whole night is listed as 7.92. Since the machine ran for 5:11 = 5.18333 hours, we can get an approximation of the number of OAs recorded during the night by a simple multiply:
- #OA = (7.92 OAs per hour) * (5.18333 hours) = 41.05 = 41 (the 0.05 is due to round off error)
Of these 41 OAs, it appears that TWO of them occurred during the 47 minutes of "non-sleep". Hence about 39 of the OAs occurred during the 4:24 of real sleep. And so the Adjusted OAI would be computed as follows:
- Adjusted OAI
= (39 OAs)/(4:24 of sleep)
= (39 OAs)/(4.4 hours of sleep)
= 8.86.
Now recall that the SH OAI = 7.92, and the adjusted OAI is 8.86 a difference of just under 1 OA/hour. In other words, the adjusted OAI is about 12% higher than the OAI shown by SH based on the full night of data.
The Adjusted CAI
Similarly, we can compute an Adjusted CAI by first finding the number of CAs that were recorded during the
whole night:
- #CAs = (CAI) * (mask time in hours) = 4.44 * 5.18333 = 23.01 = 21 (the .01 is round off error)
Now recall that it looks like during the 47 minutes of "non-sleep" the machine recorded something like 9--11 CAs. Hence the number of CAs that were recorded during the REST of the night is somewhere between 10 = 21 - 10 AND 12 = 21 - 9. So the ADJUSTED CAI for the rest of the night lies between:
- Lower Bound for Adjusted CAI = (10 CAs)/(4.4 hours of sleep) = 2.27
- Upper Bound for Adjusted CAI = (12 CAs)/(4.4 hours of sleep) = 2.73
Now if we compare the range for the Adjusted CAI (2.27 to 2.73) to the SH CAI = 4.44, you'll notice that the adjusted CAI is around 50% to 60% the size of the SH CAI. In other words, throwing the "non-sleep" period out reduces the CAI by around 40% to 50%.
The Adjusted HI
Similarly, we can compute an Adjusted HI by first finding the number of Hs that were recorded during the
whole night:
- #Hs = (HI) * (mask time in hours) = 4.63 * 5.18333 = 23.99 = 24 (the -.01 is round off error)
Now recall that it looks like during the 47 minutes of "non-sleep" the machine recorded something like 5--7 Hs. Hence the number of Hs that were recorded during the REST of the night is somewhere between 17 = 24 - 7 AND 19 = 24 - 5. So the ADJUSTED HI for the rest of the night lies between:
- Lower Bound for Adjusted HI = (17 CAs)/(4.4 hours of sleep) = 3.86
Upper Bound for Adjusted HI = (19 CAs)/(4.4 hours of sleep) = 4.32
Now if we compare the range for the Adjusted HI (3.86 to 4.32) to the SH CAI = 4.63, you'll notice that the adjusted HI is between 83% to 93% the size of the SH HI. In other words, throwing the "non-sleep" period out reduces the HI by around 7%-17%.
