Thank you for making the code public, excellent work!
I just have one small question, as the paper suggests, the data for mortality task will be processed with the following criteria
To avoid potential data leakage between mortality and the preventative events immediately preceding it, we only include the chart events within the first 24 hours after ICU admission as the input for the mortality prediction task.
However, the experiment used tables like DIAGNOSES_ICD and PROCEDURE_ICD, which don't seem to have a timestamp data in the table. How can I filter out ICDs that are 24 hours after admission?
Hi,
Thank you for making the code public, excellent work!
I just have one small question, as the paper suggests, the data for mortality task will be processed with the following criteria
However, the experiment used tables like DIAGNOSES_ICD and PROCEDURE_ICD, which don't seem to have a timestamp data in the table. How can I filter out ICDs that are 24 hours after admission?
Thank you very much!