Unknown is a result
When a record has no value in the required window, the honest answer is “unknown”. Counting it as “does not meet” removes people who might be eligible; counting it as “meets” admits people who might not be. Keep Unknown visible and route it to a person.
Absence of a record is not absence of a condition
For exclusion criteria, a dataset can’t prove that something never happened. Showing “no record found” — distinct from a confirmed negative — tells the screener what still needs checking.
Units and code systems
The same analyte can arrive in different units (HbA1c in % or mmol/mol) and the same diagnosis under different code systems. Convert where a defined conversion exists; otherwise mark the value as Unknown and surface it at import.
Check coverage before you run
Before a cohort run, look at how many records have a value for each criterion in its window. Low coverage predicts a large “data gaps” group.