Question about OutOfBoundsDatetime error (Year 5117) in pharmacy table for MIMIC-IV v3.1

[u-nsiq]

Hello,

I'm currently building a local SQLite database for MIMIC-IV v3.1 using the official import.py script from the buildmimic/sqlite folder.

The script fails with an OutOfBoundsDatetime error when processing the pharmacy.csv file. The traceback points to a specific timestamp value of 5117-07-25, which is outside the valid range for Pandas' to_datetime function.

I've read the official documentation on "Date shifting," which states that dates are "randomly distributed in the future" for anonymization purposes. A date nearly 3,000 years in the future seems to be an extreme outlier, and I'm wondering if it's a known data artifact or a placeholder value (e.g., for an "indefinite" event).

I have a few questions for the community:

Has anyone else encountered this specific 5117 year value in pharmacy.csv?

Does a value like this fall within the expected range for the standard date-shifting process, or is it indeed a placeholder?

What is the recommended best practice for handling this? Is modifying the import.py script to use errors='coerce' (converting the value to NaT/NULL) the standard approach, or is there a better way to preserve the original information?

Thank you for any insights or suggestions!