ResScan - leak info.

[chopper06]

Could someone please explain how to interpret the leakage data? What does median, 95th%, etc. stand for? How do you know if you are having lots of leaks or minimal leaks? I am confused........thanks!

[robysue]

Could someone please explain how to interpret the leakage data? What does median, 95th%, etc. stand for? How do you know if you are having lots of leaks or minimal leaks? I am confused........thanks!

The short answer is: Resmed machines report unintentional leak rates. In other words, any leak reported by a Resmed machine exceeds the expected, intentional leak rate for the mask. So in an ideal world, you'd like that leak rate as low as possible. But from a purely practical point of view, it's unreasonable to expect a leak rate of 0.0 L/min all the time, every single night. The Resmed engineers believe the leaks are not bad enough to affect the efficacy of your therapy or the accuracy of the data when the leaks stay AT or ABOVE a level of 24 L/min for AT LEAST 30% of the night. So as long as your leaks stay below the Red Line at 24 L/min for 70% of the total time the machine is running, they should not affect your therapy or the accuracy of the data. (And when your leak rate is AT or ABOVE 24 L/min for 30% of the night, Mr. Red Frowny Face shows up on the short version of the LCD's Sleep Quality report.)

So from a purely practical point of view: If your 95% is below 24 L/min, the Resmed engineers are happy with your leak rate.

But--many people are bothered by leaks that are not that large or not that long. And if leaks bother you, you need to deal with them regardless of their size. So I would say that if your 95% is below 24 L/min and your median is well below 24 L/min then you probably don't need to worry too much about leaks unless they are waking you up.


The long, mathematical answer is based on explaining the meaning of the statistical terms median and 95th percentile.

Loosely speaking, th 95th percentile leak rate is the number at which your leak rate stays AT or BELOW for 95% of the entire night. The median leak rate is the number at which your leak rate stays AT or BELOW for 50% of the night. The median is the default "middle" number shown in ResScan. (In SleepyHead 0.9 you have your choice of median or average for calculating middle numbers.)

To understand the difference between average, median, and the 95th percentile (aka 95%), numbers, it helps to consider a somewhat simplified example or two with (fake) numbers.

So let's suppose that on Night 1 the (raw) leak data looks like this:

• 10% of the night spent with a leak rate of 8.4 L/min
• 45% of the night spent with a leak rate of 9 L/min (50% of the total time winds up in this band since 10+45 > 50; so the median leak rate is 9 L/min)
• 30% of the night spent with a leak rate of 9.5 L/min
• 5% of the night spent with a leak rate of 10 L/min
• 8% of the night spent with a leak rate of 10.1 L/min (95% of the total time winds up in this band since 10+45+30+5+8 > 95; so the 95% leak rate is 10.1 L/min)
• 2% of the night spent with a leak rate of 15.3 L/min

The (weighted) average leak rate for the night is 9.354 L/min because:

• (0.1 * 8.4) + (0.45 * 9) + (0.3 * 9.5) + (0.05 * 10) + (0.08 * 10.1) + (0.02 * 15.3) = 9.354

In order to get a better handle on the meaning of median, 95%, and maximum pressure levels, it may help you to think about the data being made up of 100 distinct "events"--i.e. leak rate readings.

The data for Night1, then translates into the following string of 100 leak rate readings for the unintentional leaks:

• 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9.5, 9.5, 9.5, 9.5, 9.5, 9.5, 9.5, 9.5, 9.5, 9.5, 9.5, 9.5, 9.5, 9.5, 9.5, 9.5, 9.5, 9.5, 9.5, 9.5, 9.5, 9.5, 9.5, 9.5, 9.5, 9.5, 9.5, 9.5, 9.5, 9.5, 10, 10, 10, 10, 10, 10.1, 10.1, 10.1, 10.1, 10.1, 10.1, 10.1, 10.1, 15.3, 15.3

Now if you actually count the individual data points, you'll find that the first highlighted 9 is the 50th number on the list, and since there are 100 data points total, that means that half the data is AT or BELOW this number and hence the median of the data is 9. And you will also notice that other 50 data points are AT or ABOVE that median value of 9. (Technically speaking the median is the average of that highlighted 9 and the next 9 on the list---the 51st data point. We can safely ignore that technicality since we're focusing on the meaning of median rather than its formal computation.) The fact that the median leak rate is 9 L/min means that for 50% of the night, the leak rate was AT or BELOW 9 L/min and for 50% of the night, the leak rate was AT or ABOVE 9 L/min.

Again, if you count the data points, you will find that the 95th data point of our 100 data points is the highlighted 10.1, and hence the 95% leak rate is 10.1 since 95% of the data is AT or BELOW this particular numerical value. And you will notice that that there are 5 data points AT or ABOVE the 95% leak rate number. Hence for 95% of the night we know that the leak rate is AT or BELOW 10.1 L/min and for 5% the leak rate is AT or ABOVE 10.1 L/min.

The maximum leak rate is the last number on the list: 15.3. That means that at some point during the night when the machine sampled the leak rate, your leak rate was 15.3 L/min.

Hence the statistical data the night is:

• Median leak rate = 9 L/min
  95% Leak rate = 10.1 L/min
  Maximum Leak rate = 15.3 L/min
  (Weighted) Average Leak Rate. = 9.25

It's also important to realize that the individual statistics numbers by themselves don't always tell the full story. Let's consider another night with this data

Night 2:

• 40% of the night spent at 8.4
• 40% of the night spent with a leak rate of 9 (50% of the total time winds up in this band since 40+40 > 50; so the median leak rate is 9 L/min)
• 5% of the night spent with a leak rate of 9.5 L/min
• 5% of the night spent with a leak rate of 10 L/min
• 8% of the night spent with a leak rate of 10.1 L/min (95% of the total time winds up in this band since 40+40+5+5+8 > 95; so the 95% leak rate is 10.1 L/min)
• 2% of the night spent with a leak rate of 15.3 L/min

Again, lets re-write this as a string of 100 individual data points. The discrete data looks like this:

• 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 8.4, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9.5, 9.5, 9.5, 9.5, 9.5, 10, 10, 10, 10, 10, 10.1, 10.1, 10.1, 10.1, 10.1, 10.1, 10.1, 10.1, 15.3, 15.3.

Again, I've highlighted the 50th data point, the 95th data point, and the 100th data point. The highlighted data points again correspond to the minimum, the median, the 95%, and the maximum leak rates for Night 2.

The weighted average for Night 2 is 9.049 L/min since

• (0.4 * 8.4) + (.4 * 9) + (.05 * 9.5) + (.05 * 10) + (.08 * 10.1) + (.02 * 15.3) = 9.049

While the weighted average of both nights is different, both nights have:

• Median = 9 L/min
  95% = 10.1 L/min
  Max = 10.3 L/min

Also note that in night 1: 95% of the night is spent with leak rates AT or ABOVE 9 L/min and fully 45% of the night is spent with leak rates ABOVE 9 L/min even though median leak rate = 9 L/min

In night 2: 80% of the night is spent at with leak rates AT or BELOW 9 L/min and fully 40% of the night is spent with leak rates BELOW 9 L/min. And only 20% of the night is spent with leak rates ABOVE 9 even though median leak rate = 9 L/min

And a final note: While it is harder to construct, it is also possible for two sets of data to have the same minimum, (weighted) average, median, 95 percentile, and maximum numbers and have noticeably different data distributions.[/quote]

[chopper06]

Thanks for your help.......I appreciate it!