Is it possible to use “fake” cohorts (where the cohort table is contains real subject_ids and cohort start/end dates, but the cohort_definition_id is not linked to an Atlas-created cohort - e.g. there is no JSON specification for the cohort), and then use those in characterizations, incidence rate, pathway, PLE and/or PLP analyses?
Our use case is that we are using OHDSI for operational analytics on our clinical data, and there are many situations where we can’t fully define a desired cohort within Atlas, but we can create a list of patients using other means. We’d like to populate fake cohorts with that information, and then use OHDSI to further study them.
I’m seeing from PLE and PLP that I can specify createCohorts = FALSE. In those cases, if I populate my cohortTable with those fake cohorts, will PLE and PLE augment that table with the additional cohorts it needs (e.g. for negative controls)? In other words, should this (or some adaptation) work? I can certainly test whether it runs, but want to know whether this approach is likely to lead to invalid results (e.g. run successfully but yield wrong conclusions). If we do this, is there a numbering system we should follow (such as using 2B+ values for local concept ids)?
Separately, I’m sure I can’t use fake cohorts in Atlas (since they won’t be available in the GUI). Can someone point me to R code to run and visualize a characterization analysis outside of Atlas (again avoiding auto-creation of the cohort details)?
Lastly, would such fake cohorts work with Strategus?