We are currently working on the forum. For the short-term, all forum content will be in read-only format. We apologize for the interruption and look forward to collaborating with you shortly. All the best in your research!

Numeric fields with unknown values

I see OC has a way to add "null" values to selection type fields. But what about Numeric fields? is there a way for a user to specify the value is "unknown"? also what about handling values like "<1"? do folks just create lab CRFs with String fields instead of numeric? (if so how would one then handle creating rules to check ranges etc?)



  • lindsay.stevenslindsay.stevens Posts: 404 ✭✭✭
    via Email
    Null flavours are a CDISC thing. We don't usually use them. That kind of
    metadata usually goes in discrepancy notes since its probably worth
    recording but not useful for analysis.

    It also causes problems with column data typing e.g. the null flavour NASK
    is always a string but it might be recorded for an INT single select. This
    is possible because the item_data.value column is a string.

    Data conventions handle the <1 type values. Sometimes they're irrelevant
    and get cleaned to missing. Sometimes they are relevant and we make a new
    helper field for it. For example a single select with:

    - detected and quantifiable (specify value)
    - detected and unquantifiable
    - undetected

    As mentioned in the other thread we aren't in the habit of checking lab
    ranges, unless it's to check for a value relevant to the study (e.g. test
    result indicates lack of response to treatment).
  • gpinkhamgpinkham Posts: 37
    thanks Lindsay.. We definitely do range checks and we also do CTCAE gradings.. so it is helpful when the lab values are numeric.. we handle this in our home grown EDC.. just another thing to add to my list..
This discussion has been closed.