OHDSI Home | Forums | Wiki | Github

Swedish Provider & Care Site Mapping

CDM Builders
#1

Two Swedish questions in one day!

We are working to map some Swedish data to OMOP. The first data set we have is prescription claims for drugs picked up at a pharmacy. Each prescription comes with multiple pieces of information about who prescribed it and where it was prescribed. We’d like to preserve as much of that as possible.

Specialty codes themselves are no real problem. There are 3 of them, but we can fairly reliably identify the primary (most used) and map on that.

The information of greatest interest is the type of clinic/hospital/other where the drug was prescribed. At its simplest form, it’s either a public or private place of medical care, something likely common on other countries. There is another layer beneath that, both for the setting itself as well as public or private. The settings are fairly similar to typical American concepts for organization but more detailed. For example, there are multiple types of physician’s offices or clinics that each are intended to see a different type of patient or patient for a different reason.

For public vs. private, the public funding may be national, county, city, or something else. Private also has a few divisions.

Researchers here would like to be able to identify the source of these prescriptions as that is very useful for funding and economic analysis. It could also help detect practice differences.

Does OMOP have any convention for dealing with public and private? That’s the most generalized case and one likely most useful for other countries.

Is there any mechanism for a more thorough mapping of these various prescribing channels unique to Sweden? Is that best maintained in source value fields or does some type of concept or vocabulary exist that could accommodate it? What is the best practice?

Thanks for any insight. We are excited to be moving in Sweden to the point that we have specific things to address!

#2

@dmyers:

Generally, the OMOP CDM is intended to project the healthcare experience of the patient, not the institutions or administrative systems. However, the structures we have where you could place such information is PAYER_PLAN_PERIOD in combination wtih the COST table and PROVIDER in combination wtih the CARE_SITE:

The PAYER-PLAN-PERIOD contains a source_value of the payer. You could create your Swedish payers. The problem is that they only work in conjunction with COST. Do you have cost information? If not, you may have to create dummy records with 0 swedish kronas for each.

The PROVIDER has a specialty_concept_id field, and the CARE_SITE has a place_of_service_concept_id field wich you could use for the prescriber’s information. The concept space is now drawn from the US system. If you need additional concepts let us know, and we add them for you.

#3

Are there different sources of data when the drug is prescribed/dispensed
from different settings? Or it is one dataset that just has a ‘setting’
flag? In either case, it seems to maintain this provenence, you could to
utilize the DRUG_TYPE_CONCEPT_ID field to distinguish inpatient vs.
outpatient records and dispensing vs. prescribed, but the specific location
and context of the care site (e.g. public/private) would be captured in the
tables as @christian_reich outlined.

#4

@Patrick_Ryan: The dispensing splits should be similar to the US. This database won’t have medications given during a hospitalization or infusion-things. By definition, it is drugs picked up at the pharmacy. More or less the same as PBM prescriptions, if you ignore the occasional nastiness with Part B Medicare drugs getting split between UB/1500 forms and PBM claims.

@Christian_Reich: We do have cost information, but you make a good point about the experience of the patient. We had been thinking to use the PAYER_PLAN_PERIOD to cover changes to the national benefits (i.e. max out of pocket was X kronor from 2005-2009, Y kronor from 2010-2016 and so on). Regardless of some of the micro-funding distinctions for the facilities, all Swedes are under the same national plan.

We have provider specialty information and had planned to use that exactly as we understand OMOP is set up. We even have a preliminary mapping of Swedish specialties into standard concepts (mostly just a language translation).

I am thinking the best plan is probably to leave it in the realm of source value fields rather than open a mess of new concepts. I do think the distinction of public/private is an interesting one and would seem to have an impact on patients, but this may not be the right place to explore it.

Thanks both for the quick input!

t