NSRR staff
Boston, MA
0000-0002-0506-8368
Top Topics
Recent Topics
I got some input from a couple of our sleep scorers. The SHHS1 desaturation events, as you're finding, are not always present and may not always line up as expected with associated respiratory events. This issue has been recognized by other users of the SHHS1 data (pre-NSRR days) and the explanation points back to the desaturation data from the original scoring having been affected (damaged?) during conversions to newer versions of the Compumedics Profusion scoring software. The SHHS1 studies were first scored in an MS-DOS based package and eventually converted into Windows-based versions of Profusion in the 2000s. Evidently, Compumedics changed the way their scoring software handled the desaturation data and at that point these issues likely crept into our raw data, which then made its way into our EDF and XML files that were generated roughly 8-9 years ago.
Presently, our best suggestion for users is to have them derive new desaturation event data from the raw SaO2 channel. We may pursue doing this ourselves, or perhaps we may publish a tool to assist in this process.
Thanks again for raising the issue here and sorry for the confusion. I added a note to the SHHS documentation regarding the problems with desaturation events.
Philip,
Thanks for bringing this issue to the forum. I'll ask someone who is familiar with the SHHS1 sleep studies to take a look and share their thoughts.
Mike
I am going to email you a link to a dataset momentarily. We can continue the conversation there.
It dawned on me after replying yesterday that we have night-by-night adherence data for the PAP group, so we should be able to link this with the dates of follow-up sleep studies. We do not have nightly data from the oxygen concentrators, unfortunately.
I will follow-up with Dr. Gottlieb to see if he has any additional ideas. I will report back here in the next 1-2 weeks.
To be sure, you are looking for a dichotomous variable that indicates whether the participant is either 1) using or 2) not using their assigned therapy on the night of the follow-up sleep study, correct? Is there anything else you need along these lines?
Chris,
I asked some of our PSG experts here about this finding. The quick take is that the HRate channel is derived from within the scoring program, Compumedics Profusion, and is not often used. We don't have an immediate understanding about why the derivation breaks 50% of the way through this subset of CCSHS studies. We may explore further.
We will add a proper Montage and Sampling Rate Information page of documentation for CCSHS in the near future. We will note channels like HRate and others that are derived and were not part of the actual collection/scoring montage.
Any heart rate analyses should start by using the ECG channels that are present in the CCSHS EDFs.
Thanks,
Hey Mark,
To your second question: We didn't have EEG channels on the HeartBEAT Embletta studies, so we don't have sleep staging. Here's the montage information.
As to the first question, the randomization occurred after the screening/baseline sleep study and baseline clinic visit, so participants were not on any study treatment when the baseline data were collected. For follow-up, the study procedure was to have participants wear their assigned therapy on the night of the follow-up sleep study. That said, we of course had a non-trivial number of participants who were not adherent (in both Oxygen/PAP arms) who likely had elected to stop using the assigned therapy by the time follow-up rolled around.
From what I can tell, participants were not asked about whether they used their assigned therapy on the night of the follow-up sleep study, so I think this information will be elusive. Let me know if you want me to ping Dr. Gottlieb and one of our other staff who helped with the sleep scoring and therapy adherence tracking about this topic.
Winda,
Citing the data source website like that seems fine. Additionally, for each dataset we list recommended citations. For SHHS, we ask that you cite the primary NSRR paper (#1) along with two of the primary SHHS papers (#s 2/3).
Taken from: https://www.sleepdata.org/datasets/shhs
When citing this dataset please use:
Thanks and best of luck with the submission!
We didn't create healthy subject indicators, but you could use a variable like ahi_a0h3 to create a categorical variable (e.g. AHI<5 vs. AHI >=5).
The Medical History variables are also commonly used when grouping subjects by health status. Same for the CVD Outcomes variables.
Let me know if I can help further. Thanks for checking out the NSRR!
Thanks for the kind words - glad you're enjoying the resource!
I don't know the answer to your question, but I pinged a couple members of the NSRR team who should be more familiar with this area.
Sarmad,
No, accelerometer-level data have not been posted. We may consider this for the future.