OHDSI Home | Forums | Wiki | Github

New proposal for the Observation Period description

(Maxim Moinat) #1

An observation period is currently described as:


Table Description

This table contains records which define spans of time during which two conditions are expected to hold: (i) Clinical Events that happened to the Person are recorded in the Event tables, and (ii) absense of records indicate such Events did not occur during this span of time.

However, in many datasets this does not hold. Non-reimbursed and over-the-counter drugs are a good example. Another example are registries that only collect specific variables. Newcomers might doubt, not add the observation periods, and then discover some OHDSI tools don’t work without observation periods added.

Therefore I propose to change the definition to:

“… (i) Clinical events that happened to the Person and are expected to be collected in the context of the dataset are recorded in the Event tables …”

Any thoughts before making this a Themis proposal?

(Melanie Philofsky) #2

I like the proposal. The diversity of source data systems, level of granularity and ability to answer research questions varies significantly throughout the network. Where will dataset specific information be stored?

(Christian Reich) #3

Yeah. That’s the reason registries really don’t work that well. They are not observational data of the kind we assume, which is they don’t collect what happens, but only what is looked for. The entire ATLAS functionality is based on it. That’s where we have the greatest success.

(Chris Roeder) #4

That’s a subtle, but important part. Someone may thank you later!

I didn’t read “expect” the way you meant it the first time. Maybe just “…and are collected in the context of the dataset”. or “…and are required to be collected in the context of the dataset”.

Then also the next part? “Absence of records indicate either the events didn’t happen, or they weren’t recorded.”