NSRR staff
Boston, MA
0000-0002-0506-8368
Top Topics
Recent Topics
Margot,
Thanks for checking out the resource and for carefully looking through the SHHS variables. We are in the midst of cleaning up many of these variables - the indices themselves (like ahi_a0h3a) and the component variables (from the variable calculation). The calculations and labels are not quite right.
ahi_a0h3a
First and foremost, I think ahi_a0h3a is the variable you want, though the calculation should be changed to this:
ahi_a0h3a = 60 * (hremba3 + hroa3 + hnrba3 + hnroa3 + carbp + carop + canbp + canop + oarbp + oarop + oanbp + oanop ) / slpprdp;
We will make that fix in an upcoming dataset release, but you could calculate the variable yourself in the meantime. To address your points directly:
Hope this helps! Let us know if you have additional questions.
Mike
Glad this helped!
Based on my experience from other, older datasets, filtering based on "airqual" sounds logical.
"oahi" only contains obstructive apneas, while "ahi_a0h4" contains both obstructive and central apneas.
These two variables are very similar, save for the filtering:
https://sleepdata.org/datasets/shhs/variables/oahi https://sleepdata.org/datasets/shhs/variables/ahi_o0h4 (I notice the calculation incorrectly lists the 3% variables, but the values appear correct)
Variables like "oai0p" and "cai0p" also have filtering. The primary AHI/RDI variables used in SHHS1, like "rdi3p", do not appear to always have filtering present.
Winda,
The Ns are different on those variables because "oahi" had some sort of filtering applied to it based upon the quality of the sleep study signals (e.g. oximqual). "ahi_a0h4" did not have any filtering applied. I don't know the exact set of filters used on the older variables. "ahi_a0h4" was created more recently.
The scoring rules are laid out in the Manual of Procedures. Here's a page that describes respiratory event scoring: https://sleepdata.org/datasets/shhs/pages/mop/6-627-mop-scoring-respiratory-events.md
Let us know if you can't find sufficient description in those documentation pages.
I believe the answer is "Yes", but I am going to ask another member of the NSRR team to comment on "current definitions".
Thanks!
Hey Winda,
The AHI has been calculated a bunch of different ways, as you've probably noticed. In SHHS, something like RDI4P includes obstructive apneas, central apneas, and hypopneas, yet all the events require a >=4% desaturation.
For OAHI and other, newer variables (like the ICSD-3 set of AHI variables), the trend seems to be toward using "all apneas regardless of desaturation" (obstructive or central or both depending on the index of interest), instead of using a consistent desaturation requirement across both apneas and hypopneas.
I think the Medicare AHI definition and current AASM guidelines use this sort of approach.
Hope that helps!
Yes, that is correct.
Yes, there are many variables that describe sleep architecture.
TIMES34P is the "percent of sleep time in stage 3-4 sleep" (deep sleep).
My best guess is that SLP_EFF was filtered in SHHS1 to only use cases where the lights off/on and sleep onset data were reliable. SLPEFFP clearly contains impossible values (>100) in SHHS1.
In SHHS1, STONSETP represents the number of epochs from the beginning of the recording until the marking of sleep onset. SLPLATP is the number of minutes between lights off and sleep onset. Having a "STONSETP much larger than SLPLATP" might simply mean that the device was recording a lot of time before the participant got into bed (lights off). If SLPLATP is large this would mean the participant was lying in bed for a long time but not sleeping.
What do you mean by "sleep-restorability"? Something like "sleep quality"? Oftentimes sleep efficiency (SLP_EFF) is used as a measure for how well someone slept over the night. A higher sleep efficiency typically indicates you are waking up less during the night and that your sleep is more continuous (not fragmented).