OHDSI Home | Forums | Wiki | Github

Handling date of diagnosis, date of symptom onset, body site of symptom onset -- for an ALS clinical dataset

Implementers
#1

Hello OHDSI, it was great to see everyone at the symposium!

I’m working on an ETL of an ALS clinical data repository for Answer ALS, and there are important details related to each person’s ALS diagnosis that I want to represent. These are:

  1. Date of diagnosis
  2. Date of symptom onset (this is distinct from diagnosis as onset symptoms may not be as widespread or severe enough to warrant ALS diagnosis at that time, or it may be understood as ALS onset only in hindsight)
  3. Body site of onset

Body site of onset values look something like this for our dataset (other ALS datasets usually have similar categories, but it can vary):

  • Generalized
  • Bulbar
  • Axial
    • Neck
    • Trunk
    • Limb
      • Upper
        • Left
        • Right
      • Lower
        • Left
        • Right

Ok, now my questions:

Dates:

I saw someone had a related question here (How to store date of onset and date of diagnosis for a disease?), and want to make sure I have the right approach. It looks like we can add ALS as a condition with the condition_start_date set as the date of diagnosis, put Date of onset of symptoms in the observation table, and use the fact_relationship table to indicate for which condition the date of onset of symptoms is for.

However, there was another reply in that thread which pointed out that fact_relationship might not be as useful for applying standard OHDSI tools. I’m hoping we can set a standard approach for most ALS research so that we can join multiple datasets together and use them to their full potential. So does this approach make sense, or should we find another route, like perhaps requesting to create new observation concepts for, for instance “Date of ALS Symptom Onset”?

One more thing I just realized is I’m not sure how to represent a date value in the observation table. Do I fill in the observation_date with the onset date and leave all value fields blank, or do I put a separate date into the value_as_string field, and if so in what format?

Body Site of Onset:

A similar question applies to site of onset, apply fact_relationship between Date of onset of symptoms and Body Site in the observation table, or add a new concept for “Body Site of ALS Symptom Onset”?

Note that Body Site has standard answers that don’t line up with our answers, so if we use the existing concept then I’m also wondering whether to just define our own answers using the observation table’s value_as_string free text field and ignore the existing answers, but I don’t know whether that’s a proper way to use the an observation concept.

t