mrueschman

I'm glad that helped. Thank you for the kind words. I will share your links with other members of the team and bring it up at our next group meeting.

We are familiar with CDEs and have toyed around a bit with linking NSRR data back to established CDEs. What you describe in your aims for HIV/AIDS sounds very similar to what we have contemplated for studies/trials of sleep.

The data dictionary CSV files are publicly available for each dataset. You'll find them in the Files area under the datasets folder, e.g.

• https://sleepdata.org/datasets/shhs/files/datasets
• https://sleepdata.org/datasets/chat/files/datasets
• https://sleepdata.org/datasets/heartbeat/files/datasets
• etc.

The data dictionary files you see (domains, forms, variables) represent what you get from running the Spout export.

Tags as you describe do not exist, though we do have some data harmonization enhancements in mind for the future.

Thanks for your interest in the site!

I will make a note in the documentation regarding this, but it's possible that only the sleep period was thoroughly reviewed for alignment between the beat detector and the EKG signal. One of our technicians reviewed a couple MESA studies and found that beats very early in the recording (e.g. the first 15 seconds) did not match as consistently as beats further on into the recording. In looking at mesa-0001 it seemed like the very first beat in the R-point file was mismatched, though the following beats looked OK. We were comparing the "seconds" column in the R-point file to our visualization of the raw EKG signal (measuring the time from recording start to the apparent R peak).

Let me know if you come across issues as you dig in further.

Hi,

The SHHS/MrOS datasets may have some data on Parkinson's diagnosis and medication use, e.g. https://sleepdata.org/search?search=parkinson

Filtering to those subjects may allow you to find some PSG records of patients with Parkinson's.

Mike

I think EDFbrowser and Polyman both have EDF header editing capabilities. These tools will also do checks similar to those in the "edfize" tool in regards to whether or not the EDF header is written to the EDF specification.

We have come across plenty of sleep software manufacturers who do not adhere to the EDF specification, which is unfortunate. I assume that is what you are dealing with here. It's easy to see from the quick edfize command you ran that the date is not conforming to the "dd.mm.yy" requirement. Maybe one of the tools will be able to revise the header, rewrite the EDF, and produce something that functions in your other workflows.

Thanks for inquiring and for using the NSRR site.

I believe the NEPSY variable you want is nep10b_nepsy. The means appear to follow the patterns shown in Table 2 from the NEJM article. The Ns are slightly different due to limiting the sample in the manuscript, but the magnitude of change is consistent when looking by treatment arm.

I'm not sure stroke or hypertension included in the CVD Outcomes definition. From the CVD Outcomes Protocol:

It should align with the component variables (e.g. MI, PTCA, CABG, etc.), though I cannot guarantee 100% agreement. This work was done by the SHHS Coordinating Center before finalizing the datasets for BioLINCC and NSRR.

One option might be to create your own "new" sort of "any CVD" variable.

Thanks for checking out the site.

Everything from the Health Interview is self-report. There was no confirmation, diagnostic testing, nor medical chart review done to compare with what the participant marked on the questionnaire.

#1 - I'm guessing "COPD" is a term less people had discussed with their doctors. SHHS1 was almost 25 years ago now so I'm not sure how these pulmonary diseases were conceptualized and presented to patients back then.

#2 - I think you'll want "asthma15"/"asth1215" for SHHS1 and "hi201d"/"hi201e" for SHHS2. In SHHS1 we only know if they ever received a diagnosis and whether they had an asthma attack in the last 12 months (asth1215), which you might use as a proxy for "having asthma at the time of the PSG".

#3 - The Health Interview questionnaires should have been done around the time of the actual PSGs in the SHHS visits. So, I expect it's on the order of days. (Comparing "stdatep" and "formdate_hi" could be informative; at a glance they look to match as I expect.)

#4 - Unfortunately all we have is self-report. I expect the parent cohorts have more definitive information about testing, diagnoses, and ongoing presence/treatment.

Here are the parent cohort sites where I would start:

• Atherosclerosis Risk in Communities (ARIC): https://sites.cscc.unc.edu/aric/
• Cardiovascular Health Study (CHS): https://chs-nhlbi.org/
• Framingham Heart Study (FHS): https://www.framinghamheartstudy.org/

The NSRR does not have ongoing relationships directly with these cohorts. I expect each one has different policies and procedures regarding access to supplemental data.

Here's some more information on the ECG data present in the SHHS dataset:

Baseline ECGs were performed by the parent cohorts prior to the PSG. All Field Sites performed a standard resting 12-lead ECG with the participant supine. Ten seconds of data were acquired simultaneously from each lead (I, II, III, aVR, aVL, aVF, V1-V6) to a Marquette MAC PC or MAC II system. At each site, a paper copy of the ECG is produced and filed. Minnesota coding of the ECG data is performed on all participants in ARIC, CHS, and SHS. Minnesota coding may be performed for other sites as well. Currently, NYCC ECG data were interpreted by a clinical cardiologist, and FHS ECG data were interpreted by the clinic physician. Minnesota codes provided by the Field Sites were then used to derive the 34 variables in the ECG dataset.

My guess is that the parent cohorts tracked additional details (e.g. ongoing presence, treatment) about each participant's AF, though these details did not trickle down into the SHHS datasets since these were secondary to the original outcomes of SHHS.

PS. It's Mike, not Mark!