How are the Ventilator modes sampled?

[jimmsmit]

Hi everyone,

we noticed that the MIMIC-IV dataset has a significantly lower frequency of logged modes—approximately one entry every 3-4 hours, compared to hourly entries in the AmsterdamUMCdb, for example. Additionally, we observed that the intervals between logged modes in MIMIC-IV were often exactly 2, 3 or 4 hours.

Given this, I was wondering if you could clarify how the ventilatory mode data in MIMIC-IV was generated.
Specifically,

• were the modes logged by the ventilators downsampled in some way?
• is a new mode is always recorded immediately upon a switch, or is it possible that a mode change in reality occured hours before it is actually logged in the dataset?

Thank you in advance! :)

Kind regards,

Jim Smit (PhD candidate, Erasmus Medical Center)