Convention to map conditions with sparse dates and transformations on the dates

CDM Builders
1

We have data taken through yearly questionnaires. Specifically, part of the questions are asked remotely, while some are answered by a nurse performing evaluations on the patient. For each questionnaire, I have the year and month it was done, and inside it, there are several questions such as:

  • When was the patient diagnosed with a condition
  • Is the patient currently diagnosed with a condition
  • Observations/findings/measurements on the patients

I have several questions about the conventions for mapping some of these aspects to OMOP CDM:

  1. Regarding episodes, visits, and observation periods, should all questionnaires be grouped as one episode, each questionnaire as a visit_occurrence, and the observation period just an episode, and observation_period that lasts only for the day of the questionnaire?
  2. How should we annotate in the CDM the transformations done on the date data? For example, how can we encode dates that were assumed to occur on the first day of the month in cases where the day was missing?
  3. What duration should I put for conditions when I have only for questionnaires if currently have it in every questionnaire? Should I assume it lasts between questionnaires? For example, if the patient says they had hypertension in June 2019, did not answer in September 2020, reports having it in August 2021, and reports they do not have it in December 2021, what should we put as condition_occurrence.condition_start_date and condition_occurrence.condition_end_date? Several instances where they last that day of the questionnaire only or one that lasts throughout?

Looking forward to your thoughts and suggestions!

2

@Francisco1:

Yeah, questionnaires are pretty special real-world data types, and the standard Closed-World assumptions do not work very well, meaning, we don’t get facts when they occurred and we cannot interpret the absence of such records or the time where we don’t capture data.

Pretty much. It’s a one-day observation period, where everything happens. Unless some of the questions cover time in a different way.

Just write those days.

No duration. Just write the CONDITION_OCCURRENCE record and leave the condition_occurrence_end_date empty.

3

Thank you @Christian_Reich !