In looking at the nsrr/shhs1 and shhs2 datasets, I see that there are variables such as ndes2ph that gives you the count of desat events in each record at different % decreases. Am I correct in assuming that 1) those are what was used to calculate ODI, and that 2) there is no annotation of where all those events occurred in each record? The xml annotations include "SpO2 desaturation" events but now I'm understanding that those are just SpO2 levels associated with hypopnea and/or arousal events, and not pure desat events independent of airflow. Thanks for the help!
The series of ndes*ph variables may be tricky because they (can) include desaturations that occurred in wake time.
As to your assumptions:
That said, SHHS is a bit tricky and presents difficulties with the above. We have lost data from the SaO2 channel over time due to the studies being converted from Compumedics Replay (original MS DOS scoring program used 20 years ago) to Profusion at some point. Only desaturations that were associated with other events were retained, for whatever reason. In newer datasets (e.g. CHAT) the XML files will contain desaturation events that occurred independent of other events.
Let me know if you'd like more clarification -- I can call upon some of my colleagues who know more about SHHS' history and the utility of these variables. Thanks for inquiring!