#145: transform (ETL) GPC i2b2 data to PCORNet CDM
--------------------------+------------------------------
Reporter: dconnolly | Owner: campbell
Type: enhancement | Status: assigned
Priority: major | Milestone: drn-basic-query
Component: data-sharing | Resolution:
Keywords: | Blocked By: 109
Blocking: 160 |
--------------------------+------------------------------
Changes (by ngraham):
* owner: ngraham => campbell
Comment:
Dr. Cambell, at HackathonTwo, you agreed to supply a specific example of
where that "path" approach to building the CDM breaks down (involving the
SNOMET-CT hierarchy). I may not be capturing it accurately, but I tried
to create a very simplistic example. Your specific (and clinically
relevant) example will likely be a lot more helpful.
== Details ==
Dr. Campbell from UNMC brought up a case involving the SNOMED-CT hierarchy
where he believes using i2b2 paths (current design approach - comment:3)
breaks down for building the CDM (see [http://listserv.kumc.edu/pipermail
/gpc-dev/2015q1/001117.html his e-mail on the gpc-dev list]).
As I understand it, there are SNOMED-CT concept paths that depending on
the parentage have different children. In other words, CODE:1234 may have
different codes associated with it depending on where it is in the
hierarchy.
For example, consider the following:
{{{
\i2b2\grandparent1\parent1\CODE:123\CODE:456\
..
\i2b2\grandparent_ABC\parent_DEF\CODE:123\CODE:789\
}}}
A researcher wants to find all patients with CODE:123 and CODE:789.
Knowledge of SNOMED-CT tells them that by selecting CODE:123 they should
also get all patients with CODE:789 (as CODE:789 is a child of CODE:123
the second path above).
However, their i2b2 search used the first path above and therefore get
patients with CODE:123 and CODE:456 instead.
In another example, say the following path was used in i2b2:
{{{
\i2b2\grandparent_UVW\parent_XYZ\CODE:123\
}}}
In this case, the researcher _only_ gets patients with CODE:123 (when,
again, given their knowledge of SNOMED-CT they should also have gotten
patients with CODE:789).
As I understand it from Dr. Campbell, a code may have 55 (or more)
different places in the hierarchy each with their own set of children.
--
Ticket URL:
<http://informatics.gpcnetwork.org/trac/Project/ticket/145#comment:19>
gpc-informatics <http://informatics.gpcnetwork.org/>
Greater Plains Network - Informatics
_______________________________________________
Gpc-dev mailing list
Gpc-dev@listserv.kumc.edu
http://listserv.kumc.edu/mailman/listinfo/gpc-dev
