OHDSI Home | Forums | Wiki | Github

Missing birth date

CDM Builders
#1

I was wondering how others treat missing birth dates. I can see 3 options myself:

(1) Use a birth date that falls within the range of birth dates in the rest of the database. This would not interfere with any analysis of the data, but it wouldn’t be clear that this birth date is in fact not correct.

(2) Use a bogus birth date, like e.g. 1900-01-01. In this case it would be clear that the birth date is not correct, but it could potentially skew any analyses involving birth date / age.

(3) Drop the record.

What are people’s thoughts on this?

#3

Hi @bertoverduin : The only required date-of-birth field for the CDM is YEAR_OF_BIRTH (you do not have to populate DAY_OF_BIRTH or MONTH_OF_BIRTH). If you truly do not have even year of birth, I would recommend to drop the person. Nearly all analytic use cases rely on some simple calculation of age, and absence of reliable information for that will cause problems in all of your analyses. I do not recommend creating bogus dates, all that does is dirty your data and make the life of all analysts harder (and likely cause downstream quality issues that will devalue the reliability of the evidence generated).

t