OHDSI Home | Forums | Wiki | Github

Age Group Characterizations with Feature Extraction

Is there a way to characterize the number of patients in a cohort who are in the pediatric (<18 years old), adult (18-65 years old), and elderly (65+ years old) age groups using Feature Extraction? That grouping may be more consistent with clinical practice. Thanks!

Hey @mattspotnitz, saw this discussion and your other post here: Absolute Age Group Counts vs. Percentages in Feature Extraction regarding characterization. Were you able to solve this? I don’t use Feature Extraction, but I created some OMOP tools to perform characterizations outside of tools like Feature Extraction. A tutorial for one such tool is here: Using OMOPCDMCohortCreator with R 🏴‍☠️ · OMOPCDMCohortCreator.jl I’d be happy to assist with running this with you and assisting in any way as to my knowledge, there isn’t a separate HADES package for running characterizations. Feel free to email me at jacob.zelko@gtri.gatech.edu or DM me either here or on Teams. :slight_smile:

@schuemie or @anthonysena do you have any suggestions how one might be able to modify the age groups for a study package?

@cukarthik - I think this should be possible using a custom covariate builder as described in: Creating custom covariate builders • FeatureExtraction. At a high level you could create a custom covariate builder to identify people in a given age range and add that to a list of covariate settings. The example function: Creating custom covariate builders • FeatureExtraction shows how you can include custom SQL and I believe you could add some arguments to the function to provide a min/max age range. Just keep in mind you’d want to have unique analysis IDs for each age range you define.

Thanks @anthonysena. I was hoping it was something easy like a parameter setting :slight_smile: We’ll take a look into the custom covariate builder.


I have also been following the instructions on the linked page (Creating custom covariate builders • FeatureExtraction) in an attempt to create custom covariates for a PLE study, but am struggling to apply them.

When you have managed to construct the age group covariate builder, I would be really interested in seeing your code as I am hoping that looking at another example may help clarify the process for me. Thank you very much.