OHDSI Home | Forums | Wiki | Github

Empty cohorts when running package, despite positive cohort generation reports


I’m writing with regards to an error experienced trying to run an ATLAS PLE package. The package fails to generate cohorts with any population (in the log file: “Target and comparator cohorts are empty”). However, results from the Atlas Cohort Generation reports show that there are patients and records in the data which should be appearing in these cohorts. Does anyone know what could be causing this? The log file is attached below.

log1.docx (165.7 KB)

Perhaps you have some additional requirements in your PLE specifications, such as restricting to some specific calendar time, or requiring a minimum amount of continuous observation prior?

In this case, doesn’t the JSON file for a cohort also contain that information? So if the ATLAS Cohort Generator is showing patients remaining in the cohort after each inclusion rule, it seems logical they should be truly in the cohort even taking into account additional requirements.

Additionally, when running through with the edits synthesizePositiveControls = FALSE and
packageResults = FALSE, the resulting CohortCounts.csv shows patients in both the CohortEntries and CohortSubjects columns for each outcome.

The CohortCounts.csv shows the counts on the database server. The extra restrictions that CohortMethod applies when downloading from the server are not yet considered in CohortCounts.csv. They also do not appear in the cohort definition JSON, because this is something that CohortMethod enforces (when downloading the cohorts).

If you manually load one of the CohortMethodData objects (using loadCohortMethodData()), you can call getAttritionTable() to see where the subjects are dropped.

Thank you!

For some reason, running the following code continued to generate empty sets but it seems somehow the sets aren’t dropping due to criteria, but rather starting empty…does this then seem likely to be an issue with the OMOP mapping of the database?