I’m not sure about the error under characterization, but could be related to that you were unable to generate the cohort. So I would start with getting a successful generation,
I think you should watch the videos found here: https://www.ohdsi.org/past-events/cohort-definitionphenotyping-tutorial/. These instructional videos talk about the philosophy of cohort definitions, what they are, and how they are defined in Atlas.
To summarize:
For example, I start with people > 60 years
60 years old, as of when? That’s the key to longitudinal data, people have certain characteristics (such as their age) at different times in their patient history. You might say ‘60 years as of 2016’. to do that in atlas, you could select the earliest visit per person who was age > 60. But I would take what you said about defining your population, and work it backwards: Who are those people taking warfarin? among those, who are a certain ethnicity? Among those, who are > 60?
Atlas cohort definitions are about identifying people who satisfy a criteria between two points of time. which means you don’t work in the mode of ‘Find all Men…then find all Hispanics, then find all who are age > 60’, none of these things specify any sort of time index.
In Atlas, you begin by identifying patient-level events that satisfy some criteria, and then from those, you can apply additional criteria to filter out people from the cohort. These events come from conditions, drug exposures, procedures, etc.
So, start with getting a cohort to generate. Then you can move on to characterization and ask questions like 'How many of these were > 60? How many were {ethnicity}. etc
