Same night comparison of polysomnogram & SleepyHead reports

A few weeks ago "Lanky Lefty" (Jason at freecpapadvice.com) wired me up for a home sleep study. He's developing a home sleep study private practice and the main purpose of my study was to help him "dry run" his office setup protocols. This was not a typical home sleep test, it was a full-scale sleep study so it involved getting fully wired up as one would at a sleep lab.

I decided to use my VPAP that night which afforded the opportunity to compare the data produced by the polysomnogram recordings (using an Alice PDx recording device and scored manually by Jason) with the data produced by my machine and reported by SleepyHead, for the same night.

This would have been a more interesting and informative experiment if I had slept more and had some exciting events to analyze. There were only a few events to look at and compare. Nevertheless, Jason posted a short commentary about the study and a couple of side-by-side images on his website here: http://www.freecpapadvice.com/forum/vie ... f=19&t=775

As reported by SH, my S9 flagged one hypopnea and one clear airway event. In scoring the Alice data, Jason ruled out the hypopnea because the EEG data showed that I was awake. There was another event that was flagged by the VPAP and reported by SH as a clear airway event -- this same event was identified as a mixed apnea on the polysomnogram report. Jason posted images of this event as seen in the SleepyHead and the Alice files. Jason also scored a RERA event -- these are not flagged by the S9 machines.

There wasn't enough going on in this sleep study to draw any big conclusions, but it does illustrate something that our forum experts here remind us about frequently -- our machines don't know whether we're awake or asleep, they can only record our breathing. It's a good thing to keep in mind when we're looking at our SleepyHead data (or whatever software we're using -- because the software can only report what our machines are recording). As valuable as our data-capable machines are to us, we get limited information from them. And as my hypopnea-that-wasn't-a-real-hypopnea shows, sometimes the data is not accurate. It sure would have been interesting if there had been big clusters of events to examine and compare (though for the sake of my night's sleep I'm glad there were not).

Data comparisons aside, this was a fun and interesting experience. I slept at home, but except for the absence of video monitoring this was a full-scale sleep study with all the usual parameters being measured (Jason describes what the parameters are in his forum commentary). When I returned his equipment the next day I brought my SD card with me and we looked at some of the data together. I learned a lot. And Jason was great to work with -- I hope his business thrives.**

**Neither one of us got paid for our time and effort, but we did get the fringe benefit of a few laughs out of the deal. I had a 30-minute drive home from the office that evening, praying all the way home that I wouldn't get stopped and have to explain to some law enforcement guy why I was wired and wearing a white box strapped to my chest. I think most of Jason's patients/customers who live locally will choose to have him come to their home for the setup rather than going to his office.

Interesting.
Wish I had a chance to play with his new study. Night before last I had an ugly cluster of what appeared to be CSRish looking stuff even with my S9 Adapt. I don't remember being awake but I know that doesn't mean anything and it lasted for enough time to be an ugly cluster on the ResScan and SleepyHead reports. Didn't really bother looking at it real close other than to spot the CSRish breaths. Not totally CSR looking but close. I am still not all that comfortable when they aren't clear cut CSRs.

Since these types of clusters are very rare for me it's not practical to get all wired up at night because I might not have a similar cluster for a month.
It is a curiosity for me though as I like to know the "why" for things and it's a hard pill to swallow when I can't get any answers.

Pugsy wrote:Interesting.
Wish I had a chance to play with his new study. Night before last I had an ugly cluster of what appeared to be CSRish looking stuff even with my S9 Adapt. I don't remember being awake but I know that doesn't mean anything and it lasted for enough time to be an ugly cluster on the ResScan and SleepyHead reports. Didn't really bother looking at it real close other than to spot the CSRish breaths. Not totally CSR looking but close. I am still not all that comfortable when they aren't clear cut CSRs.

Since these types of clusters are very rare for me it's not practical to get all wired up at night because I might not have a similar cluster for a month.
It is a curiosity for me though as I like to know the "why" for things and it's a hard pill to swallow when I can't get any answers.

Really Pugsy? You don't like not having the answers? We would never have guessed that about you!

Actually I'm right there with you. It's "too bad" my OSA is so well treated on VPAP that I rarely have events -- I thought it was a real gift to Jason that I at least gave him a hypopnea because I almost never get anything beyond one clear airway as I'm waking up in the morning.

On the other hand, I did get some information that was helpful to me -- I've wondered since I started PAP if I have the alpha-delta sleep seen in people with chronic fatigue and fibromyalgia because I have a pre-existing chronic fatigue type of condition (and I never had a diagnostic sleep study at Kaiser, only the Watch-PAT screening). I did not sleep a lot that night, but I did have some time in N3 and no alpha intrusion was seen. Good to have that ruled out. Even if it doesn't shed light on why my sleep is so non-restorative, it narrows down the field of directions to look in.

kaiasgram wrote:There was another event that was flagged by the VPAP and reported by SH as a clear airway event -- this same event was identified as a mixed apnea on the polysomnogram report.

Actually, I believe your VPAP made the correct assessment. The waves on the thoraco-abdominal channels that are attributed to effort (thus supposedly qualifying that apnea as mixed) are running at a rate of 60 and closely matched to heart rate, thus making them ballistocardiographic artifact (and more evidence that the apnea is purely central). Note the little blip between the next 2 breaths following the apnea.

Sludge wrote:

kaiasgram wrote:There was another event that was flagged by the VPAP and reported by SH as a clear airway event -- this same event was identified as a mixed apnea on the polysomnogram report.

Actually, I believe your VPAP made the correct assessment. The waves on the thoraco-abdominal channels that are attributed to effort (thus supposedly qualifying that apnea as mixed) are running at a rate of 60 and closely matched to heart rate, thus making them balistocardiographic artifact (and more evidence that the apnea is purely central). Note the little blip between the next 2 breaths following the apnea.

Now I'm curious -- what's the average inter-scorer reliability for polysomnography?

kaiasgram wrote:Now I'm curious -- what's the average inter-scorer reliability for polysomnography?

LOL! Well I would hope a lot more than 50% (although we haven't seen the hypopnea yet)!

A closer review of the data provides clues as to what actually happened:



There was phasic REM with some pretty dense eye movements. There can be significant respiratory variation during those periods, and increasing respiratory effort was seen. The driving force disappears, so now you are left with the need to have a pause, so this is a REM-related central and should be considered normal phenomenon.

Also highlighted are assorted ballistocardiographic artifacts and a very coarse (1.0%) oximetry channel.

Hey Sludge, thanks for taking a look at the images and for the explanation of what you're seeing. (I had to look up "ballistocardiograph" and I'm still fuzzy on "coarse oximetry channel" but mostly I followed. )I wish there had been more "stuff" to look at and compare.

I was still curious about inter-scorer reliability studies, so I did a bit (a very little bit) of online research and found a book Sleep Apnea: Current Diagnosis and Treatment edited by Randerath, Sanner & Somers. I read a small section on ISR. Seems it's not so easy to get a really high percentage of 'agreement' unless polysomnographers have trained/worked together. They also noted that ISR differs for various parameters of polysomnography such as sleep stages, events, leg movements, arousals, etc. For example, inter-scorer reliability tends to be higher for the scoring of leg movements "because the definition of leg movements is rather simple."

All pretty interesting -- thanks again for taking the time to give your take on it.

kaiasgram wrote:Seems it's not so easy to get a really high percentage of 'agreement' unless polysomnographers have trained/worked together. They also noted that ISR differs for various parameters of polysomnography such as sleep stages, events, leg movements, arousals, etc.

The AASM ISR Program says you need to have an overall score of 85% for 4 parameters (sleep staging, respiratory events, arousals and PLMs) in a 200 epoch selected record to pass.

However:

• 85% is not the overall average. Depending on the difficulty of the record, the average among all scorers is greater to substantially greater than 85%.
• In practice, a scorer needs to review a 900 epoch record about 4 times (depending on their technique, and this needs to be done in about an hour (although a lot of the Blowfish on the other forum claim 30-45 minutes. Whatever.)).
• Anyway, one is only spending a second or two per review area, so agreement among scorers is going to vary, especially among ambiguous portions of the record.
• However, if a bunch of "experts" are going to sit around and discuss a record until they reach "consensus" (which is what they do in ISR to decide the "Gold Standard"), the agreement will not "approach" 100%, it will "be" 100%.
• Consequently, that event got stared at for a LOOOONNNNGGGGGG time, so "IMHO", that's not a "difference of opinion in ISR", that's missing the boat entirely.
• In an attempt to bring focus to variation in ISR based on ambiguity, calling the above event "mixed" would be like doing ISR where 100 fire trucks full of firemen drive past a building engulfed in flames. I seriously doubt if 15 trucks would drive by oblivious, but I'd certainly have some serious misgivings about the abilities of those who did.

Further, the ResMed algorithm appears to be quite sensitive at picking out centrals (i.e., if CSAD calls it a central, it's a central)(at least in this study)(which was done in ResMed Central):

The CSAD algorithm falsely classified 10 apneas as obstructive that were scored as central and 2 apneas as central that were scored as obstructive on polygraphy. The majority of these were post-arousal central pauses where, while cardiogenic flow was present, the airway was presumed to be partially collapsed due to a lack of tone.

http://www.resmed.com/us/assets/documen ... -paper.pdf

Sludge wrote:The AASM ISR Program says you need to have an overall score of 85% for 4 parameters (sleep staging, respiratory events, arousals and PLMs) in a 200 epoch selected record to pass.

However:

• 85% is not the overall average. Depending on the difficulty of the record, the average among all scorers is greater to substantially greater than 85%.

To try to underscore this point, if the national average for stage scoring of an AASM ISR Record is 97.5%, and you're doing 85.0%, then the appropriate evaluation should not be "Oh look, I'm within the boundaries for good ISR!", the correct interpretation is "You suck!"

kaiasgram wrote:I did a bit (a very little bit) of online research and found a book Sleep Apnea: Current Diagnosis and Treatment edited by Randerath, Sanner & Somers. I read a small section on ISR. Seems it's not so easy to get a really high percentage of 'agreement' unless polysomnographers have trained/worked together. They also noted that ISR differs for various parameters of polysomnography such as sleep stages, events, leg movements, arousals, etc. For example, inter-scorer reliability tends to be higher for the scoring of leg movements "because the definition of leg movements is rather simple." All pretty interesting -- thanks again for taking the time to give your take on it.

Thanks for telling us about your 'private' sleep study, sort of a mega-home study if I read you correctly! I'm a big fan of home-studies but they often don't provide enough data to really zero in on complex problems. Hopefully this will address that issue.

I'll try to track down that book. I suspect, like so many 'diagnostic methodologies' currently available in this modern age, there are wide and sometimes alarming degrees of interpretation. I'm often amazed how several of my doctors disagree on the diagnosis and treatment of my aliments, ie. stay on statins - get off statins; cut back on Vit D - take more Vit D, etc, etc. AND BY ALARMED, I MEAN REALLY ALARMED, mostly because I am the guinea pig in this game of medical roulette.

Sir NoddinOff wrote:I'll try to track down that book. I suspect, like so many 'diagnostic methodologies' currently available in this modern age, there are wide and sometimes alarming degrees of interpretation. I'm often amazed how several of my doctors disagree on the diagnosis and treatment of my aliments, ie. stay on statins - get off statins; cut back on Vit D - take more Vit D, etc, etc. AND BY ALARMED, I MEAN REALLY ALARMED, mostly because I am the guinea pig in this game of medical roulette.

Got a couple hundred to drop on a sleep apnea book?

Thanks for posting, SNO. This isn't the first time I've encountered a second opinion that differs from the first one, and I'm sure it won't be the last. At times it can be frustrating and confusing (happily not in this case), but on the other hand, medical science would never advance if everybody agreed with everybody else all the time.