visit_occurrence start_date and start_datetime: hardcoded?

[eroell]

Hey,

Coming across the visit_occurrence table, I was wondering whether here

BlendedICU/blended_preprocessing/omop_conversion.py

Line 24 in 5d3a68f

self.ref_date = datetime(year=2000, month=1, day=1)

and then subsequently here

BlendedICU/blended_preprocessing/omop_conversion.py

Line 239 in 5d3a68f

visit_occurrence['visit_start_date'] = self.ref_date.date()

leads to all visit_occurrence entries to have this artificial date.

If yes, this would make this table oddly out of sync with other tables who retain their timestamps, such as observation or measurement, no?

Best

[elbg]

Thank you for your interest in the project. I changed institution and can no longer maintain it, but I'm happy to help if I can :)

TL;DR : everything is synced, every patient stay starts at the same arbitrary date !

Source datasets handled timestamping differently:

• some did only work with timedeltas (all patient admitted at time 0)
• other created a random admission date for patient anonymization

In our harmonization work, we used the timedeltas method for most data processing.

The OMOP standard requires dates, and not time offsets, so we decided to make all stays start at an arbitrary ref_date.

Tables are not out of sync, because as you can see on further apparitions of ref_date, every timeseries has a date that is ref_date + timedelta:

some examples

year of birth

measurements

drug_exposures

I hope this is of help,
Best regards,
Matthieu