OHDSI Home | Forums | Wiki | Github

Age_group classification

I’m curious how OHDSI classify age_group on CDM.
In Korea, we have Korea-CMS data, and many age groups are classified elaborately
as 0~10 - first / 11~20 - second / 21~30 -thrid group.
How do you classify age_group on OMOP_CDM?

Thank you

The FeatureExtraction package, which is used in various OHDSI tools including ATLAS, currently implements 5-year age bins, so 0-4, 5-9, etc.

If there’s a need we could add another analysis that uses different age groups, although I expect it will be hard to agree on which grouping :wink:

1 Like

Thank you @schuemie!

As a statistician, I would like the finest grouping, e.g. by year … and then model as a continuous variable, allowing for non-linearity in effect … are there objections to this more refined approach?

FeatureExtraction also allows age to be output as a continuous variable, amongst others for the reason you cited.

It is possible to better define numerical data with skewed distribution using quantile discretization. The method cuts continuous data in to categories (quantiles) which can help the classification model to better separate between classes.

Please find more information in the link below.