For standard OMOP Version 5 implementation, there is a vocabulary_id type of 
'Drug Type' in CONCEPT table that includes codes such as :
"randomized drug", "Prescription dispensed in pharmacy", "prescription 
dispensed through mail order", "prescription written", "medication list entry", 
"inpatient administration", and a few others.  This code, as you mentioned 
below, is what is used in the drug_type_concept_id field in DRUG_EXPOSSURE.

The way I'd handle this if done in OMOP v5 is to add custom rows into the 
CONCEPT table for each of the categories below, and map to those.  I would 
expand the surescripts dispensed to include dis-ambiguation between pharmacy 
and mail, then you can use the CONCEPT_RELATIONSHIP  table to map the custom 
term with a relationship_id = 'is a' between the "SureScripts Pharmacy 
Dispensed" and "Prescription dispensed in pharmacy" as a standard OMOP term, 
which allows roll-up and use in other implementations but allows you to retain 
granularity in your local instance.

I agree, PCOR doesn't seem to have the granularity at present to approximate 
the above solution.

michael

From: Russ Waitman [mailto:[email protected]]
Sent: Monday, December 14, 2015 12:43 PM
To: [email protected]; <[email protected]>
Cc: Nathan Graham; Jiaxi Quan; Shawn N. Murphy; Lesley Curtis; Daniella Meeker; 
Matheny, Michael; Duke, Jon David; Patrick Ryan; Keith Marsolo; Michael Kahn; 
Charles Bailey; Abel N Kho; Klann, Jeffrey G.; Kevin Haynes; Grannis, Shaun J; 
Tamara McMahon
Subject: Tagging for Surescripts fills, claims, and nulls with i2b2 modifiers 
and looking for OMOP/PCORnet analogs/harmonization

Hi GPC-DEV, PCORnet, OMOP colleagues (I have met recently), and the i2b2 
community,
Nathan Graham, Jiaxi Quan, Tamara McMahon, and I are looking over the data that 
we're now getting from Surescripts into Epic here at KUMC.

They turned on the feature here in the Spring and now we have 14.6 million 
events on 184,000 patients since May of 2014.

As we look at the data,
3.4 million were marked as fills, 5.9 million as claims, and 5.3 million were 
null.

In our data, nothing links back to a parent order in Epic.

We're initially thinking of tagging these facts with a modifier.

"SureScripts Dispensed"
with then 3 sub-modifiers:
"SureScripts Dispensed (Claim)"
"SureScripts Dispensed (Fill)"
"SureScripts Dispensed (Unknown)"

(Note:I want to loop back with folks at Epic/Surescripts regarding what null 
means).

PCORnetCDM

We looked at the PCORnet CDM to see if there were any fields we need to load in 
CDM 2 or 3 related to attribution and CDM3 PRESCRIBING table has


RX_BASIS

RDBMS Text(2)

SAS Char(2)

01=Dispensing
02=Administration
NI=No information
UN=Unknown
OT=Other

Basis of the medication order

PCORnet


But that really doesn't let you know if the source of the info was a fill or a 
claim.

I know the SCILHS people are mapping to CDM 3 now
https://github.com/hms-dbmi/i2b2-to-PCORNET-CDM

but don't seem to see any source modifiers for i2b2 in SCILHS that cover how 
they are loading the PRESCRIBING.RX_BASIS field

Jim Campbell in GPC had noted that the RX_BASIS field in the PCORnet CDM was 
inconsistent with the guidance on most EHR dispensing records.

OMOP

Now when we look at OMOP v5
https://github.com/OHDSI/CommonDataModel/blob/master/OMOP%20CDM%20v5.pdf

It looks like they code the source of the med in a Drug_exposure table using 
the drug_type_concept_id

drug_type_concept_id Yes integer A foreign key to the predefined concept 
identifier
in the Standardized Vocabularies reflecting the
type of drug exposure recorded. It indicates how
the drug exposure was represented in the source
data: as medication history, filled prescriptions, etc.

But, I couldn't easily find a standard vocabulary used by OMOP that would 
specify the types of dispense record.  I searched  http://omop.org/VocabV4.5 
for "fill" and didn't find anything.  It seemed to mainly hold ATC and RxNorm 
which to my knowledge don't specify the terminology to attributing fill versus 
claim versus order, etc.

If anyone has a standard vocabulary for OMOP that's commonly used, we'd ideally 
like to see if we can move in the direction of harmonizing the attribution of 
these medications facts.

BTW, I presented the nominees for the PCORnet Data Committee and hope the names 
should be announced at the face to face PCORnet meeting Wednesday.  So, stay 
tuned for more info on the PCORnet Data Committee shortly,


Russ Waitman, PhD
Director of Medical Informatics
Assistant Vice Chancellor for Enterprise Analytics
Associate Professor, Department of Internal Medicine
University of Kansas Medical Center, Kansas City, Kansas
913-945-7087 (office)
[email protected]<mailto:[email protected]>
http://www.kumc.edu/ea-mi/
http://informatics.kumc.edu<http://informatics.kumc.edu/>
http://informatics.gpcnetwork.org - a PCORNet collaborative



_______________________________________________
Gpc-dev mailing list
[email protected]
http://listserv.kumc.edu/mailman/listinfo/gpc-dev

Reply via email to