COHORT_* tables

gregk · April 28, 2020, 8:28am

@Christian_Reich @Patrick_Ryan - OMOP CDM v6 question

COHORT and COHORT_ATTRIBUTE tables were dropped from v6 - why? We used these tables to store cohort information if cohorts were distributed with that data. Some data sets, for example FlatIron, come with cohorts of people pre-computed. Another possible use case is to store arms in clinical trial data.

One note - CDM cohort_* tables are different from ATLAS cohort_*, these are static and only would be used to store cohorts from data sets.

Your thoughts?

Christian_Reich · April 28, 2020, 2:16pm

Probably an question to @clairblacketer. But in brief: Cohorts are artifacts of the analytics, not part of the data. The data describe what happened to the patients over time.

Cohorts from datasets? What do you mean?

gregk · April 28, 2020, 2:20pm

in this case, some data sets come with cohort information already provided and this info should be stored somehow.

A good example is clinical trial data, it comes with facts about arm assignments e.g. cohorts of people (arms). Again, something that already happened and should be captured.

Christian_Reich · April 28, 2020, 2:32pm

I see. Well, that would be a different use for the cohort tables. Even though clinical trials call randomized patient populations “cohorts” as well. The Clinical Trial WG should submit what they actually need. Because there are no conventions and no stipulations right now for that use case.