Don't understand how Resmed calculate AHI

[sym]

I have been using the resmed vantage on constant pressure mode for a few days. I looked at the data collected each morning, but the data does not make sure to me.

For example, I got AHI=1.7, AI=0.1 and HI=1.6 for last night after 6 hr of usage.

If AI=0.1 is the average for 6 hr, does that mean the machine determined I have a total of 0.1 x 6 = 0.6 apnea event last night?

What does a "total of 0.6 apnea event" mean?

[Guest]

I have found a similar problem in the ResMed Autoscan software. It appears that there is an inconsistent method of rounding calculations to the nearest tenth. If I had the source code of the software it would be easy to find the problem ,but without that we must be satisfied with an approximation of the correct numbers.

I am absolutely certain that there is a math calculation error in both the ResScan software and the firmware on the machine. Its really a minor problem, though and is not of any practical concern.

Sign me "Retired engineer and computer programmer"

[bgalb]

Oops! I meant that EITHER the software or the firmware has a calculation error.

[SleepySandy]

I don't have a Resmed machine, but I can tell you that you can't have fractions of events.

Your AI is 0.1. If your AI is greater than 0, the minimum number of apneas you could have had is 1. So take 1 divided by 6 and you'll get 0.16666. I assume the machine truncates the number to 0.1. Or, if your usage isn't exactly 6 hours the machine may have rounded. In either case you had 1 apnea during the night.

Your HI is 1.6. 1.6 (HI) * 6 (hours) = 9.6. Since you can't have fractions of an event you would have had 10 events. Doing the math the other way - 10 divided by 6 - you get 1.66666, which could be truncated or rounded, depending on the exact value of your usage time. In any case, you had 10 hypopneas.

[JimW]

If I had the source code of the software it would be easy to find the problem ,but without that we must be satisfied with an approximation of the correct numbers.

Not source code, but you can export .csv files, which detail events and the times at which they occur. I've found that AHI, HI, and AI calculated from these event logs don't match up with the machine and software calculated numbers, but they're not too far off.