patrickthoral
commented
1 week ago @jimmsmit Hello Jim, due to disk space requirements originally only validated data was stored. In practice, nurses were required to validate the data at least every hour and after major changes (i.e. during desaturation, hypotension, etc). Most of the time nurses preferred to validate on the hour (06:00, 07:00). All none validated data were removed, unless validation was 'forgotten' on the hour. In that case, the archive stored at least the values on the hour. At one time we decided to keep everything (validated and unvalidated), which for the source database implied every minute. So basically, you have two sets of patients: those with hourly validated data (and some additional samples), and those with the 'complete' minute data.
Depending on your use case you may need to determine whether one or both sets are sufficient.
Hope this helps.
Dear All,
We are currently conducting a study on transitions between ventilator modes in ICU patients, using the AmsterdamUMCdb database.
I was wondering how (or if) the ventilation modes in the AmsterdamUMCdb have been processed. We noticed that the frequency with which modes are logged can vary significantly between patients—sometimes once per minute, sometimes once per hour. Additionally, the timestamps (in cases where the mode is logged hourly) are often on the hour (eg, exactly at 20:13, 21:13, 22:13, 23:13 etc.), which raises the question whether, in case of a change in modes, this is always logged, or could appear 'later' in the dataset as the new mode does not appear in the dataset at the exact moment it was chosen, but at a set time (ie, one hour after the previous log).
Have the modes been downsampled before being included in AmsterdamUMCdb? Could the large difference in frequencies be due to different ventilator devices?
I look forward to hearing from you! 🙂
Best regards,
Jim Smit