Can anyone clarify how the 'Rewritten' value in the statusdescription column impacts the data? It appears impossible to ascertain the final "approved" sequence of events in a particular fluid delivery. The time sequence of multiple rewritten columns from a particular linkorderid seem to have overlapping time segments with no way to determine which ones are the accurate (Read final set accepted by the clinicians) time / rate / amount.
For what it's worth for anyone else who stumbles upon this question and want an answer, see https://github.com/MIT-LCP/mimic-code/issues/47#issuecomment-282734892
As far as we know, all rewritten rows should only be used for audit purposes, i.e. these rows correspond to drug doses never being given to patients. We are considering removing this audit trail in the future as its use in research seems limited and it only acts to create a confusing table.
So in my own analysis, since I care about what the patient received, I exclude those that have
Rewritten in the column
statusdescription. However, I also only include
cancelreason = 0. The corresponding code for cancelreason is unclear, but if there's any reason for cancelling it then in my analysis I do not include it.