GiorgioLimoncella
commented
2 years ago All pregnancies in the table should already be linked to the records that compose it. The structure of the dataset "sample_from_pregnancy" is as follows: for each pregnancy you have a first row indicating the result of the algorithm (also containing the pregnancy_id, which was generated by the algorith) and listed below all the records that composed the pregnancy.
Then all rows that consecutively do not have pregnancy_id are part of the same pregnancy, and refer to the previous pregnancy_id.
However sometimes it happened that the file did not come generated correctly, let me know!
DSThayer
The pregnancy sample CSV file the instructions tell us to use for validation doesn't include the person_id, which makes it impossible to validate pregnancies, since they can't be linked to other records.
The equivalent .Rdata file does have person_id, but it doesn't group all pregnancy records with the same pregnancy_id. Only one record per pregnancy has pregnancy_id assigned.
Understanding which pregnancy a record was assigned to by the algorithm is important for validating it--especially for error/edge cases that may occur, such as pregnancies that are close together or overlapping.
Note: I'm using the UMC-Utrecht version of the algorithm, because I think that's what I was supposed to be doing for CONSIGN. Let me know if this was wrong. I didn't seem able to raise the issue there.