Interweave Implementation Guide
0.1.0 - ci-build
Interweave Implementation Guide - Local Development build (v0.1.0) built by the FHIR (HL7® FHIR® Standard) Build Tools. See the R4 profiles here.
An important, and sometimes challenging, aspect of defining a profile is agreeing on the most appropriate code list to use for certain concepts. FHIR of course provides a starting point - and for many of the shorter and simpler code lists this is helpful and uncontroversial. However some of the more complex code lists are classified only as “example” or “preferred” by FHIR - and making the right choice for these can be much more difficult.
This page therefore elaborates in more detail on the choices and decisions made regarding these more challenging code lists.
In making these decisions, the following sources of code lists have been considered:
FHIR STU3 and Care Connect - this is our starting point and default choice
FHIR R4 and UK Core - sometimes the code lists evolve and are improved in these later versions. This is relatively rare, but worth watching for. Where it happens then the changes are generally significant, useful, and relatively uncontroversial to pre-adopt. As of this writing (Feb 2022) then UK Core is still incomplete and under development.
PRSB and NHS Data Dictionary - whereas FHIR is an international standard, these standards are more focused on the UK. Sometimes therefore they contain a more relevant list which can be used in preference. In some cases this decision has already been made by Care Connect and/or UK Core - in which case it is simple to adopt. However in other cases the “world of FHIR” appears to be at odds with the “world of NHS” - and these cases are more difficult to resolve.
A further complication is the need to ensure that both Health AND Social Care are catered for. All of the above standards are historically “Health” focused, but they have both begun to make some consideration of Social Care more recently. Sometimes one is more advanced than the other in this regard - and so this also needs to be taken into consideration.
NB: Of course many of the above lists end up referencing SNOMED codes, see https://termbrowser.nhs.uk/?
This is relatively uncontroversial and basically refers to the “setting”.
We will use the standard FHIR list. The main issue is that this seems rather focused towards acute healthcare. It is however classified by FHIR as “Extensible”, and so we would be open to requests for additional regional codes to be added to the list, as need arises.
This is a tricky field, as the FHIR definition is not all that clear. However it seems to be mostly about categorising the type of place where the encounter took place.
308467007: Seen in Establishment - this provides a short list of 22 top level codes which cover a good range of care settings eg “Seen in clinic”, “Seen in own home”, “Seen in supervised accomodation”, etc.
However… some of these expand into a much longer list. For example “Seen in clinic” expands into a list of 118 different types of cinic.
Our understanding is that the purpose of this field is to describe the type of place where the encounter occurred (not the service provided), and so it is not required to expand down to the level of the specific clinic type.
Based on the greater richness of the above SNOMED lists, plus their mandating in UKCore then these will be used - as per the defaults in CareConnect and UKCore
This is an important field that covers what everyone really wants to know about an Encounter - what the service was.
Our conclusion is that this is an important field, with good consensus around the above SNOMED refset. As it is only available in R4, it is proposed to add an extension to pre-adopt it.
This provides information about the purpose of the Encounter. (There are some subtleties here - eg the reason is why you came, the Service Type is the clinic you want to, and the diagnosis is what was learned).
FHIR R4 also renames this field to “Reason Code” and adds the ability to also have a “Reason Reference” which points to a full dataset about the Condition. To some extent this dilutes the need for a “Reason Code”, but only in R4, and only in a fully mature implementation which offers all of these extra FHIR Resource types. The use of a basic reason code on the Encounter itself therefore still seems valid.
Our conclusion is that this is a useful field, but with important additions to the code lists for emergency and social care in R4. We will therefore pre-adopt the FHIR R4 code lists.
Gives a high-level summary of the Service Type.
This is potentially a useful summary field, but unfortunately “broken” by CareConnect. Therefore we cannot use for now.
This is essentially the same field as Encounter Service Type. However:
Proposed to replace the FHIR default with the above SNOMED refset, as it looks a better list, and is then consistent with Encounter
This is about the type of practitioner needed (eg for scheduling):
Strangely enough, the list of “SDS Job Roles” which CareConnect (mis)uses for the Service Category would actaully be rather good here. But for some reason no-one is using it?
Proposed to replace the FHIR default (which is not very useful anyway) with the list of SDS Job Roles
This is the same field and same considerations as Encounter Reason.
Proposed to pre-adopt the UK Core mandated approach, incorporating additional reference sets for Social Care. (A question remains however as to the role of the overlapping but slightly different PRSB preferred list)
Use the standard FHIR code list, but pre-adopt R4 list to get the additional Social Care entry
Proposed to pre-adopt the UK Core mandated approach, incorporating additional reference sets for Social Care. This is also well-aligned with PRSB
This could be especially relevant for Community and Social Care, to understand if there are any follow-up actions once a patient leaves the initial care setting
Proposed to tentatively adopt the FHIR example list, acknowledging that although not great it appears to be the only option. Remain open to potential extension or replacement in future
However a more formal version of the list can be found via the TRUD ODS dataset https://isd.digital.nhs.uk/trud/users/authenticated/filters/0/categories/5/items/341/releases
The first part of the (very large) download file provides the list from which this is taken:
CodeSystem name="OrganisationRole" oid="2.16.840.1.113883.2.1.3.2.4.17.507"
Adopt this list of organisation types as provided by ODS / TRUD
Given that the FHIR list is mandated then there is not much choice. Remain open to requests for extending with additional values
Tentatively adopt the FHIR example list, however understanding that this is an active area which may need revisiting in future
This is clearly a difficult and confusing area, with no clear standard defined. Having sucessfully started using the “EHR Composition Types” there therefore seems no strong driver to move away from this immediately. However this remains an area to monitor - and implementers need to be aware that there may be a need for a mapping / migration exercise in future if a different list achieves clear prominence. Sites who wish to be future-proofed could also choose to provide more than one of these codes.
This is mandatory in FHIR and defines in more detail the type of diagnostic report. The coding of diagnostic reports is known to be a difficult topic, and at this stage a single set of codes is difficult to prescribe. It is therefore suggested that the following are preferred, most favoured first:
SNOMED - CareConnect (and UKCore) defines the use of SNOMED coding based on the use of 371525003 - Clinical procedure report
NICIP - this code list covers only imaging. However within that domain it has been defined by NHS Digital and mandated for use by the Information Standards Board. See https://digital.nhs.uk/services/terminology-and-classifications/national-interim-clinical-imaging-procedure-nicip-code-set. Note that the NICIP codes include a maintained standard mapping to SNOMED
LOINC - the default mapping in FHIR is to LOINC codes for diagnostic reporting
Local Codes - outside of the imaging domain there is significant variation in coding and, whilst mapping to one of the above standards is preferred, it may not always be feasible.
TODO - finally confirm approach - potentially allowing for capture of more than one of these?
NB: In all cases display text must be provided so that, regardless of coding, the type of report can be understood by a user