I had the pleasure of meeting many new OHDSI friends at the Symposium that are also struggling to accurately map their EHR data to the OMOP CDM. I would like to form a working group that would meet to discuss these issues and then come up with conventions. Here’s a couple problems that come to mind:
The medication administration record (MAR) contains data for continuous infusions with varying amounts of drug given per minute or hour. Inotropes, sedatives, anesthesia, insulin, etc. are titrated to effect. Do we roll up a single medication into one Drug Exposure record? i.e. the patient received X amount of drug from the start to end date. Do we take each individual source record and create one record in the CDM? i.e. the patient received X amount of drug from 8am to 10am, then X amount of drug from 10am to 11:12am, the Drug X was held for 46 minutes, and restarted at 11:58 to 2pm and X amount was given. Do researchers need to know the varying dosages 5 mcg/kg/min or 10 mcg/kg/hr?
Another discussion that came up is how to correctly map visits to one of the visit_concept_ids and map face to face Visits to the CDM? EHR source data may describe a visit as a “surgery” and surgery may happen during inpatient or an outpatient visit.
I’ve noticed that in 4 separate EHR datasets I have worked with, ~15% of drug records only contain an internal identifier. These internal identifiers do have a text string, but mapping thousands of text strings using Usagi is a very time consuming process. Do others have this issue? Have others found a solution? Should we work together to identify a solution? And why the heck is this happening? These are all different types and classes of medications. Sometimes these aren’t true drugs, but are ordered like drugs (breast milk, needles, oxygen). But many times these are drugs that map to standard drug concepts.
If you are interested in joining, reply to this thread with your name, email and time zone. Once I get a list of emails and time zones, I will send out a poll to get these meetings started!
Tagging a few people I know are interested: @roger.carlson @Sgp6a @cukarthik @DTorok @Daniella_Meeker @Adam_Black @burrowse @samart3 @mgkahn @Andrew I am limited to tagging 10 persons per post. Please forward to others that are interested!