new user would like to know more about stats

[john 070]

Hi All,

I have a Philips Respironics System One REMstar Auto....I'm really curious as to how it calculates the usage. For example, how does it even know when a day has passed by, what is its time stamp? I did find this explanation, "For this calculation, all usage within a 24-hour period is recorded." That describes the days at least x hours. So I understand that to mean 2 hours here, 6 hours there, is 8 hours for the day. But when does the day begin and end? Tried searching and can't seem to find the answer...thanks!

John

[redjoe]

I believe that when you first plug it in, the clock starts. The first 24 hours would be day 1, the next 24 hours day 2, etc. If you unplug it and then plug it back in, the clock starts over.

[john 070]

hmmm...but I always unplug it when not in use...just think if traveling it's unplugged, and if in a hotel, it's put away in its bag....I thought that unplugging doesn't change the count so to speak....edit: I did find a post on another forum saying it counts sessions, not days....so if you use it 1 hour, unplug it, or 1 hour, and leave it plugged-in but do not use again for 2 hours, it counts that one hour as a session and next time is a new day.

That would explain the odd results I think it stores. I mean it should be like baseball, you get a hit, or you strike out! The batting avg. is easily related to or envisioned...

[Mary Z]

My machine has an internal clock and turns over to a new day at 12 noon, it runs on a 24 hour clock.. I never unplug it unless I hae to- taking it on a trip for instance.

[john 070]

My machine has an internal clock and turns over to a new day at 12 noon, it runs on a 24 hour clock.. I never unplug it unless I hae to- taking it on a trip for instance.

In this case, does the machine actually display the time of day? Mine doesn't seem to have any reference to the time, which is why the "session" idea makes a little sense...

[Pugsy]

Respironics machines have an internal battery for the internal clock. Electric source doesn't alter the internal clock.
All the Respironics machines are set at the factory to GMT time zone so the "day" begins at Noon GMT time. A few of the System One machines allowed a GMT time offset in the clinical menu but in the System One machines that I have seen...there was none so it wasn't universal.
For those without the GMT offset...we are stuck with GMT time zone and we adjust to our time zone via software using our computer's internal clock with the software except for Encore Pro which won't use computer clock and we have to set the time zone within Encore Pro software settings.

The Respironics machines don't let us alter the time so it doesn't show the time or date.
ResMed machines do allow the users to change the time stamp and date.

[john 070]

Wow, so the whole session theory is totally wrong, huh? I mean where would the person get info that if you unplug, it's a new session (this could add up to multiple days in a day?), if unplugged over 2 hours, etc......well, if it's noon GMT, then you would really kill your stats if you say went on a 5-day business trip and didn't bring the machine, because that could be 0 hours of the last 30 days.....all interesting stuff and thanks for the info!

[robysue]

Hi All,

I have a Philips Respironics System One REMstar Auto....I'm really curious as to how it calculates the usage. For example, how does it even know when a day has passed by, what is its time stamp?

The PR System One's internal clock is set (at the factory) to Greenwich Meridian Time (GMT). There is no way to reset the clock time in either the patient set up menu or the clinical menu. GMT is used for all timestamps on the data. And there is some kind of small battery to keep the clock working when the machine is not plugged in

The machine updates the LCD data at noon GMT each day. If you live on the east coast of the US (Eastern Time Zone), in the winter time, NOON GMT = 7 AM Local Eastern Time. In the summer, when we are on daylight savings time, NOON = 8 AM Local Eastern Time.

Both Encore Viewer (and Pro) and SleepyHead use your computer's Time Zone setting to adjust the GMT timestamps to reflect your local time (including the DST offset when we're on daylight savings time). In Encore Viewer and Encore Pro, the software sets the 24 hour day to run Noon Local time to Noon Local time since that will make almost all patient's overnight sleep data fall into one continuous block on one "day" of data. In SleepyHead, the user can choose the time to start/end each day so that if the user sometimes sleeps through noon, the data is not split over two different days.

[Pugsy]

Encore Pro defaults to GMT time zone. Pro doesn't change that time zone to make it work with local time zone by reaching the computer clock. Instead the User has to go into the preferences and make the time zone change within the software package. If you don't...your reports are split and look whacko.
Encore Viewer uses the computer clock.
SleepyHead uses the computer clock.

Pro...uses GMT time per the machine's settings and we have to modify the time zone in Encore Pro.

[redjoe]

My comments were based on the observed behavior of my Intellipap. It is made by a different company (based in the US), and used by a much smaller percentage of people, than the machines used by the other posters here (whose manufacturers are not US-based). Maybe that explains the difference, or maybe I'm just confused (I'm open to that possibility).

[john 070]

After some more observation, surely enough, the stats change at 7 AM EST! Here's another mystery, so as to not load the question, the 7 day count and avgs move up, but not the 30 day. Not even the >4, when each session is >4. Any thoughts on the 30 day numbers?

[robysue]

After some more observation, surely enough, the stats change at 7 AM EST! Here's another mystery, so as to not load the question, the 7 day count and avgs move up, but not the 30 day. Not even the >4, when each session is >4. Any thoughts on the 30 day numbers?

Once you have more than 7 days worth of data, the 30 day averages have more data being used to compute. Hence one night's worth of data is much, much less likely to change it. For lots of detailed information on how those 7-day and 30-day averages for AHI (and leak and pressure) are computed, see my blog

Now as for why the 30 day >4 number doesn't change very often. When this number changes depends on WHEN the low usage day(s) occur. Once you have ONE day with less than 4 hours of usage with at least 30 days of data, the 30-day >4 number will remain at 29 for a full 30 days--provided no additional low usage days occur. For example, consider this string of usage data for the first 60 days of usage where the 30-day >4 numbers are listed in parentheses:

1: 4.5 (1)
2: 5.0 (2)
3: 4.7 (3)
4: 3.5 (3)
5: 5.0 (4)
6: 7.3 (5)
7: 5.2 (6)
8: 7.1 (7)
9: 4.5 (8)
10: 3.9 (8)
11: 4.1 (9)
12: 5.2 (10)
13: 6.1 (11)
14: 1.2 (11)
15: 5.8 (12)
16: 6.9 (13)
17: 4.7 (14)
18: 5.3 (15)
19: 7.2 (16)
20: 4.5 (17)
21: 5.2 (18)
22: 7.1 (19)
23: 5.8 (20)
24: 8.6 (21)
25: 5.7 (22)
26: 5.7 (23)
27: 3.7 (23)
28: 4.2 (24)
29: 2.8 (24)
30: 4.2 (25)
31: 4.3 (25) based on days 2--31
32: 5.7 (25) based on days 3--32
33: 5.8 (25) based on days 4--33
34: 4.8 (26) based on days 5--34
35: 4.9 (26) based on days 6--35
36: 7.5 (26) based on days 7--36
37: 4.7 (26) based on days 8--37
38: 9.2 (26) based on days 9--38
39: 5.3 (26) based on days 10--39
40: 3.2 (26) based on days 11--40
41: 6.2 (26) based on days 12--41
42: 5.2 (26) based on days 13--42
43: 8.2 (26) based on days 14--43
44: 4.6 (27) based on days 15--44
45: 5.4 (27) based on days 16--45
46: 8.3 (27) based on days 17--46
47: 5.9 (27) based on days 18--47
48: 6.2 (27) based on days 19--48
49: 5.4 (27) based on days 20--49
50: 4.5 (27) based on days 21--50
51: 5.9 (27) based on days 22--51
52: 5.3 (27) based on days 23--52
53: 5.4 (27) based on days 24--53
54: 8.3 (27) based on days 25--54
55: 4.6 (27) based on days 26--55
56: 4.5 (27) based on days 27--56
57: 5.2 (28) based on days 28--57
58: 5.4 (28) based on days 29--58
59: 4.8 (29) based on days 30--59
60: 4.5 (29) based on days 31--60

And so on.

NOTE: Since there was a < 4 hours day back on day 40, the 30-day # days > 4 will remain at 29 until day 70 as long as the usage is > 4 per night.

[john 070]

Ok, watched the stats a little more. Didn't use the machine one night, so that's a big fat zero. The 7 day avg. went up, the 30 day stayed the same.

Seriously, I know how to calculate 1 yr, 3 yr., 5yr, 10yr. and YTD on any stock or mutual fund. The 30 day has stayed the same for at least 5 days when I started looking at it. Also, the >4 for 30 day is not changing either.

The only thing I can think of is that it's like unemployment, they are not looking at the quarter where you got fired or laid-off for your basis. They go back and choose 4 quarters that potentially ended 6 mos. ago.

What I'm thinking is this--if you got wind that your insurance etc. wants the data next Mon., you could start using the machine 12-24 hours a day and really bump the stats up. If they're not even looking at today through next Mon., they will get more valid stats as even though you have a window to hand it over, yet nothing you can do can alter them in your favor. I'm just guessing because I see what I see. How could I not use the machine for a day, and yet the 7 day goes up? Obviously, the new day that entered into the calc. was not the 0 usage day.

[archangle]

One truly evil thing is that there's no way for the user to correct the time stamp on the machine. You have to take it in and have it serviced. The clock time on the machine drifts over time, too. There's no provision for the user to replace the clock battery.

[robysue]

Seriously, I know how to calculate 1 yr, 3 yr., 5yr, 10yr. and YTD on any stock or mutual fund. The 30 day has stayed the same for at least 5 days when I started looking at it.

I think you are comparing apples and oranges here. The 7-day AHI and 30-day AHI are NOT moving averages of the daily AHI numbers. The 7-day AHI and the 30-day AHI are computed as a simple averages using 7 and 30 days of data:

7-day AHI = (number of events scored in the last 7 days of data)/(number of therapy hours accumulated in the last 7 days of data) rounded to ONE decimal place

30-day AHI = (number of events scored in the last 30 days of data)/(number of therapy hours accumulated in the last 30 days of data) rounded to ONE decimal place

A detailed example of how nightly numbers tend to affect the 7-day AHI in only minor ways is on my blog page at: http://adventures-in-hosehead-land.blog ... 0-day.html

The 30-day AHI is very, very stable once you have more than 30 days of data. There are two reasons why this true.

First, for most of us, long term our nightly AHIs tend cluster in a rather narrowish range with a few outliers both directions. For example, my AHI is almost always between 1 and 2, with most of my nightly AHIs clustered between 1.3 and 1.7. And so it's no surprise that my 30-day AHI on my S1's LCD is almost always 1.5. If I have an exceptionally good string of 4 or 5 days with AHIs less than 1.0, it will temporarily drop to 1.4. If I have a bad streak (head cold, allergies, etc.) with a string of days with AHIs > 2.5, it will temporarily go up to 1.6.

Second, mathematically speaking that 30-day AHI is remarkably stable because so little of the data the data actually changes each night: You are typically replacing 1/30 = 3.33% of the "old data" with "new data" AND for the most part, the replaced "old data" and the "new data" usually don't look that different from each other. And this means that it's almost impossible to change that 30-day AHI by replacing the oldest data with the newest data. Mathematically speaking, the usual time that 30-day AHI is changed is when enough small changes in the 100ths place accumulate to cause the number to round differently in the 10ths place. And this typically represents a real, long term trend in the AHI gradually increasing or decreasing over time.

An example might help.

Let's suppose that the 30 day AHI is 2.6 and let's suppose that 2.6 was computed based on:

551 events scored in 210 hours: AHI = 551/210 = 2.624 = 2.6 rounded to 1 decimal place.

Now let's further suppose that the oldest data came from a "bad" night that had 20 events scored in a 6 hours of data---a night with an overnight AHI = 20/6 = 3.333 = 3.3 rounded to one decimal place. And the new night is a better than average night---let say you have 11 events in 8 hours--a night with the overnight AHI = 11/8 = 1.375 = 1.4. You'd think the 30-day AHI would go down, right? But the new 30-day AHI is computed based on:

(551 - 20 + 11 events)/(210 - 6 + 8 hours) = (542 events)/(212 hours) = 2.557 = 2.6 rounded to one decimal place

Also, the >4 for 30 day is not changing either.

To see why the >4 for 30 days can fail to change for several days, see my previous post: In the "fake" data I created, the 30-day >4 number remains at 26 for a total of 10 days and it remains at 27 for a total of 13 days. Once you have have 30 days or more of data, the 30 day >4 number only changes under two circumstances:

• You use the machine for less than (or equal to) 4 hours AND exactly 31 days ago you had a compliant day. In this case you are replacing the compliant day from 31 days ago with today's non-compliant day, so the number of compliant days in the last 30-day period drops by one. Hence the "30 day >4" number decreases by one.
• You use the machine for more than four hours AND exactly 31 days ago you had a non-compliant day with less than four hours of usage. In this case you are replacing the non-compliant day from 31 days ago with today's compliant day. Hence the "30 day >4" number increases by one.

The only thing I can think of is that it's like unemployment, they are not looking at the quarter where you got fired or laid-off for your basis. They go back and choose 4 quarters that potentially ended 6 mos. ago.

What I'm thinking is this--if you got wind that your insurance etc. wants the data next Mon., you could start using the machine 12-24 hours a day and really bump the stats up. If they're not even looking at today through next Mon., they will get more valid stats as even though you have a window to hand it over, yet nothing you can do can alter them in your favor. I'm just guessing because I see what I see. How could I not use the machine for a day, and yet the 7 day goes up? Obviously, the new day that entered into the calc. was not the 0 usage day. :?:

Are you talking about the 7-day AHI? That can easily go up with a 0 usage day being entered as the new day of data. Let's look at a detailed example that shows how the 7-day AHI can go up after a 0-hour day.

Let's suppose the last 8 days of data look like this, where Day 1 is the most recent day (with the zero hours day as the last day).

Day 1) 0 hours; 0 events
Day 2) 5 hours; 8 events
Day 3) 5 hours; 7 events
Day 4) 6 hours; 10 events
Day 5) 4 hours; 6 events
Day 6) 6 hours; 9 events
Day 7) 6 hours; 8 events
Day 8) 6 hours; 6 events

Before the 0-hour day, the 7-day AHI is equal to:

Old 7-day AHI = (8+7+10+6+9+8+6 events)/(5+5+6+4+6+6+6 hours) = 54/38 = 1.421 = 1.4 (due to rounding)

After the 0-hour day, the data for Day 8 drops out of the computation replaced by the 0-hour day, and the 7-day AHI is equal to:

New 7-day AHI = (0+8+7+10+6+9+8 events)/(0+5+5+6+4+6+6 hours) = 48/32 = 1.5

Note that the 7-day AHI went up because the day that we DROPPED from the data used to compute the 7-day AHI figure had an overnight AHI = 1.0. In other words, it was helping to pull the overall 7-day average down to 1.4. And when we discarded that data, the new 7-day AHI was figured on the six days with data and the average went up.

And as far as compliance for insurance purposes is concerned: They look at the raw usage data, not the statistical "averages" and numbers that shown on your machine's LCD. In other words, they download the nightly usage data for the period you've had the machine and they count the actual days where the usage was > 4 hours. And if that total exceeds their threshold, they pay. And if it doesn't, you have a paperwork nightmare on your hands if you are struggling to improve your compliance because you don't want to give the machine back. And sad to say, they probably don't look at the AHI data at all---in other words, neither the DME nor the insurance company seems to give a hoot about whether the machine is actually reducing your AHI or not. They only care if you are using it enough for the insurance company to pay the DME.