OHDSI Home | Forums | Wiki | Github

Difference between Condition Era and Condition Group Era

Researchers
#1

Hi,

The CDM wiki defines a Condition Era as a span of time when a person is assumed to have a given condition by combining individual condition occurrences.

Next to that I came across the Condition Group Era, which in the FeatureExtraction vignette is defined as: one covariate per condition era rolled up to groups in the condition_era table […].

What is meant with rolled up to groups and how does it differ from the normal Condition Era?

#2

The logic of the Condition Group Era is defined here: https://github.com/OHDSI/FeatureExtraction/blob/master/inst/sql/sql_server/DomainConceptGroup.sql#L24, but basically it creates a concept ‘rollup’ using the CDM concept_ancestor (with additional rules around min_levels_of_separation) to create a parent-child relationship for SNOMED clinical findings. Once this hierarchy is established, the individual condition eras are ‘grouped’ together under the ancestor of the condition concept found int he condition_era table. For example, ‘Psoriasis’ maybe be a parent of the different sub-types of ‘psoriasis’ that are found in condition era, so a person with different diagnosis sub-types of Psoriasis will be counted only once in the ‘Psoriasis’ group, while without this grouping, this person would contribute a count to each psoriasis sub-type the person was diagnosed with.

2 Likes
#3

Friends:

Why wouldn’t we promote the Condition Group Era as a CDM artefact?

#4

How do you confirm that the descendants, in fact, were rolled up and grouped under the ancestor? My team and I desire this feature exactly. Instead of counting each descendant as a separate indicator variable, only assign a subject a 1 if they have any of the conditions. I have tried a few things (e.g. looking at the number of subjects with a non-zero covariate value with and without the ‘addDescendantsToInclude’ turned on and off).

I am not sure that this feature is working how you explained it. Please advise. Thank you.

#5

You would have to just use the vocabulary attached to the CDM to understand what sub-concepts were involved in the rollup.

In Atlas, when a characterization is generated, the prevalence of the ‘features’ (which could be conditions, drugs, etc) are tabulated, but there’s also a function to ‘explore’ a concept to see the concept hierarchy.

#6

Thank you for your timely response. A small nuance, but it looks like we were able to confirm that the descendants were rolling up. Thank you!

t