Where/how to store derived elements?

[kryptec]

We develop case identification algorithms on EMR data and would like to store the results in the CDM. The ideal place would seem to be the condition_occurrence table, but there doesn't seem to be a good way to store metadata about how the condition was derived there.

If it's an ICD code, something taken directly from a problem list, or a chunk of text from a note then it can go into the condition_source_value field. However, for something like a machine learning algorithm that relies on multiple lab values and NLP results, than it's unclear on how to include this. We could insert some remarks into the condition_source_value_field, but that feels like a hack.

Essentially we just want to flag that this condition row was generated algorithmically, and add a reference to the algorithm and version.

I'm sure this community must have already encountered these issues, so I'm wondering if there are some best practices on how to deal with them?