When a value does not fit the parent domain
Synthetic data. Not submission-ready. Not a copy of official IG mapping tables. Licensed vocabularies are not republished here.
The problem
The source has a column the standard domain table does not have a home for: a device identifier, a rater comment, a local flag. Two failure modes show up in warehouses: the column is dropped, or it is forced into a standard variable that already has a meaning. Official CDISC SDTM material describes supplemental qualifier datasets (often referred to as SUPP– / SUPPQUAL) as the pattern for additional attributes of a parent domain row. Read the IG for the version you are on; this page is only the engineering shape.
A working example
Synthetic parent AE-like row: SYN-001 / AE-001, verbatim “mild headache after first dose”. Suppose the source also has collected_via = 'phone follow-up' — an invented local attribute, not a standard variable.
Illustrative supplemental row (not a copy of official SUPP columns, but the usual ideas): it points back to the parent row (study, subject, domain, the parent’s sequence or id) and then carries:
- an identifier for the extra attribute (a text name or a coded name from official CT, which this site will not invent a list for),
- the value of that attribute (
phone follow-up), - any units or origin you actually have,
- enough provenance to get back to the source column.
If you later add a second extra attribute, that is a second supplemental row, not a second parent AE. The parent grain stays one event; the supplemental grain is one qualifier of that event.
The gotcha
Supplemental data that cannot join back to exactly one parent row is decorative. If the parent sequence is renumbered after a reload, hanging SUPP rows will attach to the wrong event. Persist a stable source identifier as well as the SDTM sequence.
Putting the extra value in a comment field on the parent “because it is easier” hides it from any process that only reads standardized columns, and it can overwrite a comment that was already meaningful.
Do not invent a new “standard” column on the parent domain because it would make a star-schema join nicer. Compatibility with official domain structure is the point of using SDTM at all.
When not to use this
- The value does belong in an official parent variable. Use that variable, with official CT if required.
- The value is actually a new finding or a new event. That may be a new parent row (or a different domain), not a qualifier.
- You do not have a mapping specification. Supplemental datasets are easy to fill with junk.
The example attribute is synthetic. Official SUPP-- behavior should be taken from the CDISC SDTM IG version you use (via the CDISC Library). No official SUPP table is reproduced here.