Referrals in OpenMRS

Thanks Norman!

Iā€™m circling back to this thread. We would like to create any backend extras to support referrals. @burke, we would depend on you (others) to lead data modeling efforts.

Once this is modeled, we will figure out how to get done in coming months.

Of course, we would love to collaborate with any interested parties.

JJ

@jecihjoy @corneliouzbett @nkimaina @ibacher

We might be able to cover most use cases with the base order class (i.e., an order type that doesnā€™t need its own table).

Of course, weā€™d want to align with FHIR.

Do you have a list of attributes of ServiceRequest that weā€™d need to support for referrals?

@burke, I donā€™t think we can get away with just using the data model as is. Thereā€™s now way I can see to support where + what the patient is being referred to in the existing openmrs data model.

Iā€™m worried though that we would need to add in this table to core which could potentially take more time than we have.

Could we place in fhir2 module? Too hacky? Other ideas (orders extension module)?

Just skimming super quick here. While this will also require Core changes it looks more scalable to introduce order attributes, @burke / @jdick thoughts?

@jdick introducing order attributes wouldnā€™t have to be too overwhelming unless the deadline is next week of course.

Generally, considering how attributes have helped us enhance the data model in other areas, i would also vote with a YES to introducing order attributes.

Also to add more on this, and connecting with the last TAC discussion, the most inexpensive way to doing so could be:

  1. Make orders attributable. That will have to go in Core, no way around it.
  2. Introduce a referrals API in EMR API for the time being.

The latter would leave some room for error until we get things right. Of course this implies that youā€™d depend and require the EMR API module @jdick.

Cc @ibacher @mseaton

Iā€™m in favor of making Orders attributable, although without real implementations of the methods I complained about here (I opened a ticket to mark them deprecated). This would hopeful stop the need to have a TestOrder class (too late on that), a MedicationOrder class, a ReferralOrder class, etc.

The where could be put in orders.instructions and the what could be defined in orders.concept_id. For example:

CARDIOLOGY REFERRAL
Please schedule patient in Webuye Cardiology Clinic

Otherwise, Iā€™d favor anything that takes us in the direction of aligning our model with ServiceRequest ā€“ whether changes to core or changes to core initially introduced through a module.

Attributes are our equivalent to FHIRā€™s extensions (e.g., ServiceRequest extensions). They can be useful for implementation-specific needs and workarounds, but arenā€™t a good choice for standard functionality.

Chanced upon this thread. For referrals, couldnā€™t we enhance the ā€œordersā€ model to have ā€œlocationā€ info (textual & coded)? FHIR ServiceRequest has provision for that, Older ReferralRequest had ā€œrecipientā€.

bumping this. Any thoughts?

@dkayiwa @burke @jdick @mksd @ibacher

I continue to favor any changes that take us toward FHIRā€™s model.

While order attributes could allow implementations greater flexibility in associating additional information with orders, the core code should never be referencing specific order attributes. If, for example, we depend on attribute for a specific type of order within the core API, it should be part of the data model and not an ā€œorder attributeā€.

While I favor the design of a base Orders object to manage those features needed across all types of orders, Iā€™m not a fan of simply expanding Orders to have all the attributes for all order types. For example, dosing & administration information is needed for medication orders, but has no place on an order for a lab test or for a referral.

The model we currently have is:

  • orders ā€“ contains properties common to all types of orders
  • drug_order ā€“ adds anything beyond orders needed to support medication orders
  • test_order ā€“ adds anything beyond orders needed to support test orders

FHIR has conflated test orders (labs & procedures) and referrals into ServiceRequest. So, I would favor changes that take us the same direction. Specifically, adapting our test_order to become more like FHIRā€™s ServiceRequest.

For referral orders and including location info as mentioned in this thread, this could mean adding test_order.location_code (to align with FHIRā€™s ServiceRequest.locationCode) and either holding our nose and using TestOrder for referrals or renaming TestOrder to ServiceOrder (or even ServiceRequest) and adding TestOrder as a deprecated alias of it for backwards compatibility.

1 Like

TestOrder becoming ServiceOrder or ServiceRequest looks like another candidate item for 2.5.

Cc @grace @dkayiwa

I could be wrong but in my opinion ServiceRequest/Order sounds to me like what we have as the base Order class

Itā€™s a good point, @wyclif (miss seeing you!). The properties of FHIRā€™s ServiceRequest that might be specific to service requests (tests, referrals, etc.) and not belong on a base order would be:

  • bodySite
  • location (locationCode, locationReference)
  • occurrence
  • quantity
  • relevantHistory
  • requisition
  • specimen
  • supportingInfo

Yes ServiceRequest seems to map to TestOrder in OpenMRS, I guess what I was trying to say is that ServiceRequest is not an intuitive name for us to adopt in OpenMRS in place of TestOrder because it ā€˜suggestsā€™ a plain Order

@burke if we add a test_order.location_code, then it can serve intended destination for a lab test.
To model a Referral (to external hospital) request,. I prefer capturing as different order_type. which maps to FHIR ServieRequest.category. Instead of introducing additional tables (or changes to sub-types), it would be simpler to just add ā€œorders.location_codeā€ as optional (nullable).

And yeah, In OMRS, we have both ServiceRequest and MedicationRequest mingled.

  1. We could deflate orders and rest_order onto 1
  2. Stand drug_order on its own.

there is also the aspect of correlation to observation obs ā€¦ that would now to potentially need to have different references as new FKs.

I think that FHIR-wise the performer field is actually what youā€™d use to say ā€œget this done at lab XYZā€ (as a reference to an organisation) assuming that itā€™s referring to an external service. Location code seems to be a bit higher level (ā€œget this done at an external labā€). In either case, I donā€™t see anything in the OMRS data model we can use to capture this information.

@ibacher - speaking fhir - yes. more meant as reference to OMRS data model. the closest to Organization, HealthcareService or Location resources we have in OMRS is location, which i think can be well leveraging location attributes and tags.