What exactly does the 95% stat mean?

Can someone explain what exaclty is meant by the 95% statistic. Resmed defines it as:

The Pressure value displayed is the 95th centile level of
the air pressure delivered during your previous session of sleep. This means the pressure you received for 95% of the night was at or below this level.

I am trying out APAP therapy with a min of 9 and a max of 13. Last night my 95% pressure was 11. Does this mean that I spent 95% of the night at 11? Or, is this the upper limit I bounced off of during the night? I guess I am confused by Resmeds language "at or below this level".

JIM

Jim - it means that you spent 95% of the night at or below 11.
It means that you were above 11 for 5% of the night.

Mindy

Thanks for the quick reply...I am just wondering how much higher it went. Could it be significantly higher? I'm guessing I need software and a card reader to find that out.

I guess I am also wondering if I should set my minumum and maximum a little wider to begin with, and then narrow it after I get more data.

I was titrated at CPAP 9, but have have seen much better AHI levels at 10 and 11. After reading several posts I thought I would try the APAP portion of my machine to help me see where my best sleep comes, and then go back to the CPAP set to my 95% reading. However, now I am not sure. Given other recent posts, I wonder if I should stay with APAP...seems to be some good benefits.

Only one night under my belt though, so we'll have to see. I am open to any thoughts...

soulrhythm wrote:Thanks for the quick reply...I am just wondering how much higher it went. Could it be significantly higher? I'm guessing I need software and a card reader to find that out.

Jim,

That 95% (or in the case of Respironics, 90%) is, statistically speaking, a way to define the characteristics of the curve and the significance. Think of a bell-shaped "normal" curve. 95% significance means the whole area under the curve except 2 1/2% at each of the "tails" of the curve. Thus if your AHI is below 5 and leak rate and shape are good (i.e. you don't have large leaks and leak graph doesn't look like a whole series of mountains and valleys), it would tend to indicate that the "95% pressure" is just fine. If your AHI is above 5 and/or leak rate doesn't look good, you'd probably want to increase the top end of your range.

Mindy

soulrhythm wrote:I guess I am also wondering if I should set my minumum and maximum a little wider to begin with, and then narrow it after I get more data.

I was titrated at CPAP 9, but have have seen much better AHI levels at 10 and 11. After reading several posts I thought I would try the APAP portion of my machine to help me see where my best sleep comes, and then go back to the CPAP set to my 95% reading. However, now I am not sure. Given other recent posts, I wonder if I should stay with APAP...seems to be some good benefits.

Only one night under my belt though, so we'll have to see. I am open to any thoughts...

It may depend on your age and cardiovascular health. Medical researchers has not come up with a definitive answer but there are indications that CPAP may be better than APAP for those with cardiovascular risk factors.

Mindy

mindy wrote:

soulrhythm wrote:Thanks for the quick reply...I am just wondering how much higher it went. Could it be significantly higher? I'm guessing I need software and a card reader to find that out.

Jim,

That 95% (or in the case of Respironics, 90%) is, statistically speaking, a way to define the characteristics of the curve and the significance. Think of a bell-shaped "normal" curve. 95% significance means the whole area under the curve except 2 1/2% at each of the "tails" of the curve. .... snip ....
Mindy

95th percentile has nothing to do with bell curve or tails.

95th percentile is a value for which all other values in a set of values is equal to or less than that value ... so you had it right on your first post.

Think of it this way ... you have 100 values (ie. measurements or observations). If you sort all 100 of them from smallest to largest, then the 95th value is the the 95th percentile.

You cannot determine how high your pressure went above the 95th percentile unless you have the software ... all you can determine is that it was somewhere between the max pressure setting and above the 95th percentile for 5 percent of the time.

DreamStalker... since he had the apap set at a top pressure of 13, it wouldn't have gone above that, no? So the highest he could have gone for that 5% of the time would be somewhere between 11 (his 95% number) and 13 (his top number.

Of course, it could have been ANYthing between 11 and 13, even 11.1. Without the software, you can't tell that, but you could tell it would have to be at or below 13, right?

It could have even maxed out at 11.

You don't know just given the 95 %ile number. You just know 5 % of the night it was at or above 11. A graph of the pressure would tell more.

The pressure sure isn't a normal distribution - agree with DreamStalker there.

Dreamstalker is right. Resmed is counting "minutes at pressure" and then calculating the 95 percentile for those calculations. There is no way of knowing what went on beneath the 95 percentile: whether you spent half your total time a a pressure of 9, 25% at 10 and 20% at 11, or whether you spent 20% at 9, 30% at 10 and 50% at 11. Both results mean you need a constant pressure of 11 at least.

If you want to use Redmed's info for setting up your machine in straight PAP then the value at (or below) which the machine found you spending 95% of your time is a good one. Though I wouldn't base my decision on one night.

The machine is bascially saying: "Listen, there was a lot of time during the night that I had to raise the pressure up to 11 in order to keep you breathing smoothly. Above 11 was only 5% of the night, but you better pay attention to below that".

As for the understanding the range: if someone hits the top too much (more than 5% or 10% of the time), they have tor raise the top of the range - not true in your case, Jim.

If you don't necessarily hit to the top, but your AHI is high, you have to raise the lower range; once again, that is not true of you case, Jim.


O.

soulrhythm, the other thing you need to know to make some sense of all this, is that the Resmed data capture algorithm records a pressure and leak rate value each minute, so you get 60 values per hour (etc).

30/07/2007 12:42:00 AM 12.6
30/07/2007 12:43:00 AM 12.8
30/07/2007 12:44:00 AM 12.6
30/07/2007 12:45:00 AM 12.8
30/07/2007 12:46:00 AM 13
30/07/2007 12:47:00 AM 12.8
30/07/2007 12:48:00 AM 12.8
30/07/2007 12:49:00 AM 12.8
30/07/2007 12:50:00 AM 12.8
30/07/2007 12:51:00 AM 12.8
30/07/2007 12:52:00 AM 12.6
30/07/2007 12:53:00 AM 12.6
30/07/2007 12:54:00 AM 12.8
30/07/2007 12:55:00 AM 12.8
30/07/2007 12:56:00 AM 12.6
30/07/2007 12:57:00 AM 12.6
30/07/2007 12:58:00 AM 12.6
30/07/2007 12:59:00 AM 12.6

(and so on... pressure value only shown in the above real data extract from Autoscan 5.7)

If you sleep for 8 hours (480 minutes) the 95th percentile value represents the value at or below which you spent at least 456 minutes of your 480 minute sleep time.

As to what happened in the other 24 minutes.... that's what you need the software to be able to determine!

Cheers,

Bill

Wow. Thanks for all the info. That is helpful. Last night I tried lowering the minimum to 7 and had a higher AHI 5.6 (leak rate was .26). Not sure if it is related though because I am also sleeping in the basement because we have company here. So I will have to triy this over the next week and see.

Is the AutoScan 5.7 Software from ResMed? Is it still available?

Yes, it is probably still available if you search for it. However, ResScan is the newer Resmed software and either will work for you. Search for both, find the best price and let that determine which one you opt to buy.

soulrhythm wrote:Wow. Thanks for all the info. That is helpful. Last night I tried lowering the minimum to 7 and had a higher AHI 5.6 (leak rate was .26). Not sure if it is related though because I am also sleeping in the basement because we have company here. So I will have to triy this over the next week and see.

Is the AutoScan 5.7 Software from ResMed? Is it still available?

Another thing to keep in mind is that the numbers provide more useful information when there are more of them … in other words, one night’s worth of measurement or observation values will not be as useful as one or two weeks worth of these numbers.

Our forum host sells the software and card reader https://www.cpap.com/simple-find-cpap-products/software



Also, just to clarify my previous post.

The only relationship that a percentile has with a normal distribution is that the 50th percentile (also known as the median) is equal to the arithmetic mean (often referred to as the average). If one sorts the set of values from smallest to largest, the middle value of that sorted set is the 50th percentile and if it happens to equal the “average”, then the set of values can be said to have a “perfect” normal distribution (when plotted it looks like a symmetrical bell-curve). “Perfect” normal distributions are theoretical and therefore depend mostly on the precision (significant figures) used in the measurements. So like porno, one usually knows it is a normal distribution when one sees it. The AHI values observed by the machine should have a normal distribution or very close to it.

Because of the way the APAP algorithms work, pressure values (ie. measurements and observations) from an APAP most likely belong to a “skewed” probability distribution (when set up correctly). If the APAP min/max pressure settings are properly set, the values should be skewed (weighted) towards the lower APAP pressure setting. As Bill’s 18 minutes of sample data show, the pressure varies and is skewed towards the lower 12.6 values with a median value of 12.8 and a mean value of 12.7 … because the mean is less than the median, it is skewed or weighted towards the lower values. If the pressure values were skewed towards the higher max pressure setting (ie. mean was greater than the median), you would likely want to raise the max pressure setting (assuming that your AHI values do not go up). If the pressure values were not skewed and instead looked more like a normal distribution, you would most likely want to raise the min pressure setting (assuming it does not cause intolerable discomfort).

So Mindy was not completely wrong with her statement, pressure values can have a normal distribution (though not likely) ... however if they did it would have nothing to do with the 95th percentle.

… and that is today’s statistics lesson for all you Turkey stuffed OSA patients …

Soulrhythm, they put you in the basement when company comes?