For CDM v6 consolidate visit occurrence and visit detail into a single table

[don-torok]

Add a new column, parent_visit_occurrence_id, to visit occurrence and eliminate the visit detail table.

[gowthamrao]

I am in support of this idea (infact this was the original proposal) I like less tables. But it may become a major change.

The reason we split it this way, was because the CDM work group did not want to mess up a core CDM table.

1. Performance, @chrisknoll said (I think) that recursive reference to self i.e. parent_visit_occurrence_id is a performance problem.
2. We have visit_detail_id in the clinical event tables that need to be deleted. E.g. condition_occurrence

We may also want to add a concept_id field that what level of the ancestor is the current record - something similar to concept_ancestor. Or, we need to create a visit_ancestor table that is similar to concept_ancestors design

[gowthamrao]

I also think the proposed TREATMENT table by the oncology workgroup maybe then absorbed into this visit_occurrence table, because

treatment is a span of time that a person gets care, represented by multiple events
Visit is a span of time that a person gets care, represented by multiple events

[clairblacketer]

I would prefer to leave VISIT_DETAIL as-is for now. In the initial proposal @cgreich laid out many good reasons why we need the two tables. Additionally, some organizations just converted to CDM v5.3, which is where the VISIT_DETAIL table was first added, and I don't think it would be smart to remove it in v6.0 undoing the work that has already been done.

[clairblacketer]

Closing this but @dtorok feel free to reopen if you would like this to become a full proposal

[cgreich]

Actually: Is anybody using VISIT_DETAIL?

[clairblacketer]

Yes! I populate and use VISIT_DETAIL regularly. I even have standard logic that builds VISIT_OCCURRENCE records from VISIT_DETAIL

[gklebanov]

Yes, we populated in all of our ETLs too. But I think Christian's question is not whether we populate it, but if it is being used in studies?

[cgreich]

Right. How do we populate, and does that feed a use case. Should we review?

[cukarthik]

We populate it. We don't use it in OHDSI studies, but are starting to use it internally for other analysis on the inpatient side.