Axiomatization of the CDM

[haydenbspence]

While writing validators for the CDM, I have not been able to find outside of the DataQualityDashboard tests a set of axioms that present accepted facts about the model beyond the guides and other documentation.

Even if the facts are obvious, this would be a great way to document the model further and provide some insight that may often be tucked away in tools or code.

I would suggest this format: proper title, followed by first order logic, followed by a plain English translation. Here are some examples:

1. Visit and Procedure Occurrence Synchronization:

   • ∀ visit_occurrence, procedure_occurrence (occurs_during(procedure_occurrence, visit_occurrence) → occurs_at_time(visit_occurrence, t) ∧ occurs_at_time(procedure_occurrence, t))
   • If a procedure occurs during a visit, both the procedure and the visit occur at the same time.

2. Consistency in Observation and Measurement:

   • ∀ observation, measurement (related(observation, measurement) → (observation_date(observation) = measurement_date(measurement)))
   • If an observation is related to a measurement, they must have the same date.

[cgreich]

@haydenbspence:

Couple points:

1. This is something for the Forum. It's not an issue of the CDM. Let's bring it up there. Therefore, I should probably not say the following here, but I'll do it anyway. :)
2. Generally, I like your idea. We should formalize the myriad of assumptions and axioms.
3. "If a procedure occurs during a visit, both the procedure and the visit occur at the same time" - Not sure how you define "during" and "at the same". Usually, the visit is a time interval, and the procedure is more like a point in time, and it happens any moment inside the interval.
4. "If an observation is related to a measurement, they must have the same date" - What are you alluding to? What Observation is related to what Procedure? Aren't there Observations observed after a Procedure?

[clairblacketer]

I think this is more THEMIS territory. @MelaniePhilofsky is working on cataloging conventions and the plan is to create a library similar to what @haydenbspence is describing.

[haydenbspence]

I agree! I will bring this up at the next THEMIS meeting and potentially move it forward there.