Reject a calendar duration from a start date held as text

Goal: write how long before the first dose each adverse event began as ISO 8601 duration text in years, months and days (AEPREDUR).

Input: adverse event records, one per event (AESEQ), with the event's start date as the case report form collected it and the subject's date of first dose.

Variables:

• AEPREDUR would hold the calendar time from the event's start to the first dose, so 26 August 2007 to 27 September 2007 gives P1M1D. Only two dates give a calendar duration, and the collected start is still text that nothing has read as a date, so the run is rejected before any data is read and no artifact is accepted.

How to fix

Declare the start a date where the input is read, so each value is checked as a complete date as the records are read:

input:
  AE_RAW:
    path: input/ae_raw.csv
    types: {AESEQ: int, AESTDTC: date, RFXSTDTC: date}

When a start may be collected to the month or the year only, keep it as text, take its date in a column of its own with to_date, and name that column as the start; a partial start then has no date and no duration:

- name: AESTDT
  type: date
  label: Start Date of Adverse Event
  derivation:
    to_date: {source: AE_RAW.AESTDTC}