Data numbers

[RDawkinsPhDMPH]

Something in the discussion yesterday got me thinking so I wanted to share some thoughts.

We use the data download from the CPAP machine when someone comes in or calls about a problem they are having. (And for insurance compliance data, of course.) This data gives us some insight into what the problem might be.

However, to accept the respiratory numbers (AHI, etc) from the CPAP machine without question is naive. (The numbers on pressures, time, leak, etc, are probably more reliable.) When we do a sleep study we use 5 channels of data to recognized respiratory events on the diagnostic night and 4 on the titration night. Your machine only has one. Plus, there is a bit of subjective interpretation involved also. For example, in looking at the data it is sometimes clear that the "event" on the screen is really a lost or degraded signal. Your machine cannot make that judgement. Also, after moving you will frequently have an "event" which may be real or, again, a lost signal. It takes subjective judgement to know whether this is an artifact or not. Your machine cannot do this. I you are restles one night because of some outside influence, say the cat or dog keeps getting on the bed, your machine might well show a higher AHI for that night. And, of course, your machine doesn't know whether you are awake or asleep when it calculates AHI.

[robysue]

RDawkinsPhDMPH,

You raise some good points---which underlie the manufacterers' statements that these numbers should be used for trending purposes only and that they really cannot be compared to the data gathered in the sleep lab.

But so-called "noisy" data is better than no data: If one of these full data machines is consistently reporting AHI numbers that are less than 2.0 (say), then that's a pretty good indication that the machine is doing its job of preventing most of the apnea patient's apneas and hypopneas. Likewise, if the machine is reporting AHI numbers well above 5.0 night after night for an extended time, that's a sign that something needs to be looked at because the apnea is likely not being treated sufficiently well.

And for patients, this data, noisy as it may be, can be a serious, positive motivating force for sticking with a therapy that is not really natural and not all that easy for some of us to get used to. I would very likely have thrown in the towel by the end of week 2 had it not been for the extremely good AHI numbers---which were less than 1.0---during that time frame. Now, was my AHI really less than 1.0 during that period? I don't know---after all, even the lab computed AHI is in many ways a rather gross estimate of the real problem that's going on in apnea. But given the data off my S9's LCD, it was clear that the CPAP was taking care of the vast majority of my apnea/hypopneas even though I, myself, was feeling far, far worse during those first two weeks than I felt before starting CPAP. Indeed, the way I felt during my first two weeks was the worst I have ever felt in my entire adult life. And in all seriousness, the only reasons I was able to stick with CPAP were a combination of the really good data coming from my S9 and a very sympathetic PA at the sleep doctor's office who was (and continues to be) very good at coming up with ways to address the worst and most intolerable side affects of CPAP therapy that I've dealing with these last 12 weeks.

And for patients who truly cannot sleep well enough in the lab setting for the sleep techs to get enough quality data out of a PSG, machines that can generate data, albeit "flawed" data because it only comes from one channel, are a real godsend in helping find out the real pressure the patient can both tolerate and benefit from. In my case, my CPAP titration study resulted in a pressure setting of 9cm based on the last 30 minutes of the night where the tech bumped the pressure from 7cm to 9cm, most likely because there was only 30 minutes before "lioghts on". But I was unable to tolerate 9cm for a variety of reasons. So the sleep doctor's PA ordered a week of auto-titration to find out where my 95% pressure level was. It came in at around 7.8cm and hence my prescription has changed from straight 9cm to an APAP range of 7--8cm. This slightly lower pressure still manages my apnea as well as the 9cm did (no statistically significant change in the S9's reported AHI) and is marginally more tolerable physically for me (based on both my daytime and nighttime symptoms). But my toleration of the pressure remains problematic, and I am now awaiting my DME getting my BiPAP shipped to them, which should happen by early next week. Now mind you, my titrated BiPAP pressure is based on a scant 17 minutes of sleep at the very end of the night---having spent several hours at a lower pair of IPAP/EPAP settings. So, while I'm hopeful I'll be able to tolerate the prescribed BiPAP pressures of 8/6, it will not surprise me if I need to have further work (in the form of auto-titration) authorized by the PA to properly tweak those prescribed pressures.

[RDawkinsPhDMPH]

I agree with you entirely as to the utility of the data. And, yes, there is a lot of night to night variability in AHI and the sleep lab is a one night snapshot. That's why one of the questions they ask you in the morning is: "How did that compare to usual?" ... which, of course, is also rather soft data.

Unfortunately, some insurance companies think an auto titration is sufficient. To assume a machine set "wide open" will tell you what pressure the person needs is simplistic. Although rare, I have seen people "crash" on CPAP so I have a real problem using an auto machine on someone without ever first observing them in the lab.

But in cases where the person slept very poorly on their titration night, or had claustrophobia problems, we sometimes get them accustomed to the CPAP with an auto, set based upon what data we did get, before bringing them back for a retitration.

[M.D.Hosehead]

First of all, thanks for taking time to post your thoughts.

Correct me if I'm wrong about this. Your OP implies that the machine is more likely to produce a false positive event report than a false negative one. In other words, the reported AHI, if there is an error, will most likely be higher than would have been obtained in a sleep lab, where artifacts can be excluded.

[RDawkinsPhDMPH]

That would be a reasonable inference, although I'm not sure ... if artifacts are counted as events whether you are awake or asleep the AHI would be biased high ... if, however, you had normal respiration and no artifact while awake but a high AHI while asleep, and a significant amount of time awake on the CPAP, then the bias could be low

[HoseCrusher]

Let me see if I understand what is being said here...

The AHI-machine may not be accurate. So when people use machine values to determine what pressure to use, they may end up over pressurizing if the machines tend to over estimate their AHI values. Higher pressures can cause other problems, so when people are left to determine the best pressure on their own, they can actually make things worse...

I know this is a little simplified, but I think you get my gist.

[RDawkinsPhDMPH]

I get your gist. But I'm not sure we can go that far. As robysue pointed out, the data is good for trending. That is why I believe in a lab titration specifically with the intent of using an autoset machine. Then, with that information, the data is definitely useful, which is why we only recommend such machines. What I'm saying is, the numbers themselves may be biased and probably are not exactly correlated with lab numbers. So the exact meaning of the exact, absolute value is unsure. Within that context, we could debate what effect it MIGHT have on how the machine adjusts pressure and how someone might use the data without the background of a titration.

[NotMuffy]

However, to accept the respiratory numbers (AHI, etc) from the CPAP machine without question is naive. (The numbers on pressures, time, leak, etc, are probably more reliable.) When we do a sleep study we use 5 channels of data to recognized respiratory events on the diagnostic night and 4 on the titration night. Your machine only has one. Plus, there is a bit of subjective interpretation involved also. For example, in looking at the data it is sometimes clear that the "event" on the screen is really a lost or degraded signal. Your machine cannot make that judgement. Also, after moving you will frequently have an "event" which may be real or, again, a lost signal. It takes subjective judgement to know whether this is an artifact or not. Your machine cannot do this. I you are restles one night because of some outside influence, say the cat or dog keeps getting on the bed, your machine might well show a higher AHI for that night. And, of course, your machine doesn't know whether you are awake or asleep when it calculates AHI.

That is why I believe in a lab titration specifically with the intent of using an autoset machine. Then, with that information, the data is definitely useful, which is why we only recommend such machines. What I'm saying is, the numbers themselves may be biased and probably are not exactly correlated with lab numbers. So the exact meaning of the exact, absolute value is unsure. Within that context, we could debate what effect it MIGHT have on how the machine adjusts pressure and how someone might use the data without the background of a titration.

Well, "IOHO", these generalizations range somewhere between misleading to untrue.

We believe that while a bare-bones Night Event Summary may not give us the data we need to evaluate a complex case, if we review Detailed Data from a machine with one of the newer algorithms (S9 or SysOne), we pretty much have everything we need to know.