CareTeams - why do they apply to a single patient?


I’m looking at using HL7 FHIR to map the UK Stroke Sentinel National Audit data to HL7 FHIR objects to try to integrate electronic record storage to the national audit process. ( for anyone interested in what the project is and what it has helped deliver to date). I’ve currently applied for Health Foundation funding to develop this, and one of my deliverables will be a mapping for the current UK Stroke dataset to HL7 FHIR. So sorry if more questions follow when we delve into the intricacies of mapping the data - but hopefully it will really test out the spec!

This morning I’ve been looking at care teams. If you have a stroke, you pass through a number of clinical teams, each of whom play various roles: some provide Hyperacute Stroke Unit care, some provide inpatient stroke rehabilitation, some community-based rehab etc. We can define team by its purpose. An organisation may be responsible for several different teams undertaking several different purposes.

One of the things the national audit does is report on the performance of these teams. Its current custom CSV-format based, tabular audit data collection manages this quite clumsily, and this looks to be an area where pulling data using HL7 FHIR as the messaging format from EHRs and potentially the same data structure for storing the records for the audit makes a lot of sense.

I was hoping CareTeam would solve these problems - superficially it certainly looks like that. A group of practitioners would be associated to the CareTeam, the CareTeam would have a name which would identify that team for the audit.

So it surprised me to see that CareTeam has a 1-1 mapping with Patient. A CareTeam would, in my mind at least, look after many Patients. The Team members would be Period-based, but the team name would be carried through. Clearly you can make a CareTeam object for each stroke patient, but that a) would cause a lot of duplication, and b) would risk fragmenting the data.

You can’t use Group for this function because Group doesn’t have a suitable mapping to EpisodeOfCare or Condition to describe what the Group’s role was.

Could anyone explain the rationale for CareTeam being 1-1 mapped to Patient? Or indeed, have I missed the point of the object and is there another object which might deliver the same thing?

Best wishes,


The expectation was that groups of practitioners responsible for delivering care would be handled using Organization. If you have a use-case for a group as a whole asserting a Condition, you could submit a change request for that to be added as an option to Condition.asserter. They would be supported as the EpisodeOfCare.managingOrganization. They can also be a member of a CareTeam taking on varying responsibilities.

Group would never be used because Groups are not actors, they are recipients or subjects of care - they’re intended for things like herds of cows, cages of chickens, all females between 40 and 65 in a particular geographic region, etc.

If you wanted to make a case for a CareTeam being able to be patient-independent, it would need to be a situation where the standing team had specific people with specific roles within the team. That’s not a use-case that’s been raised previously. If that is the situation, we’d need a better idea about exactly how such a team would be referenced in the various other resources.

Hi Lloyd,

Thanks for the reply. I think from your description the latter is exactly the use case I’m thinking of. I wonder if it helps if I describe the use case in more detail because it perhaps does challenge the existing model (which is good). Sorry for the long message - just making sure to share the detail on the use case - that way if there is a better way of modelling it then I’m open to suggestion!

I guess I may be throwing up an interesting use case because I’m trying to work on a process of mapping stroke care to HL7 FHIR objects at individual patient care level on the one hand, but with the idea that we can undertake a national audit process with the results on the other hand.

I’m predominantly looking at UK NHS teams - so ‘organisation’ in my head is the NHS Trust responsible for that - which can be very large (ie. 5,000 individual employees or so in a medium-sized organisation). Organisation may be responsible for multiple clinical teams.

I’m coming at this as a physician but also as a member of the audit programme.

For the National Stroke Audit, we look at individual team performance, and we look at transfers of care between teams. gives you the outputs of the audit, and hopefully I can describe the clinical pathways.

Picking an existing set of teams, London Northwest Healthcare Trust runs Northwick Park Hospital in London. This has a Hyperacute Stroke Unit (HASU), an Acute Stroke Unit (ASU), and they will have a number of community rehab teams (probably one but possibly more than one).

If you have a stroke, you would be admitted to their HASU for the first few days until stable. A HASU is usually based on a ward (or wards for extremely large services), but for audit purposes you would want care defining by the multidisciplinary team of individuals rather than the location. When stable you would step down to the ASU for further inpatient care and rehab (unless you do so well you go straight home). In some Trusts (such as Northwick Park) this is within the same organisation; in others (such as our own) it’s a different organisation. There will be overlap of practitioners within the teams - so some consultants may provide HASU and ASU care for example, others might just provide ‘on call’ services to the HASU. When you recover, you would be discharged to a community Early Supported Discharge therapy service to deliver therapy at home. That may be integrated to and managed by the same organisation, or it might be commissioned by another provider or provided privately (or people might have NHS and private care simultaneously for example).

So a HASU team might have around 8 consultants, a transient team of junior doctors (which is OK - they can be mapped with PeriodStart and PeriodEnd dates), a static team of physiotherapists, occupational therapists, nurses, speech and language therapists, psychologists, dieticians, orthoptists etc.

An ASU team for a regional hyper acute stroke centre might share some but not all of those people with the HASU, and it may contain people who only do the ASU. Where the team lives within a different organisation they may have none of the members the same.

Both those services are inpatients, so if we are saying the modelling might be need refinement and whilst we’re asking questions: should a CareTeam be able to be optionally associated with a Location set? This might allow different organisational insights such as if we have recorded the ward nurses as part of the care team (which we should) we could estimate how many ward nurses are associated with the HASU. Here is a link to a piece of national news that was generated from the audit a couple of years back by Ben Bray one of my colleagues: - so in stroke care knowing nurse staffing levels is really important at national strategic level This data would be used to inform and drive improvements in HASU nurse staffing numbers.

That study could be done with increased granularity if the Care team contains all the staff nurses, and the group is linked to set of Locations of physicalType ‘bd’.

For the national audit, insights into stroke care come from following these patients across pathways and providers as the actual clinical pathways do vary somewhat from team to team. For example not all HASUs provide thrombectomy - so you might have to go from HASU A to HASU B in another organisation for procedure and back to HASU A. That needs mapping because nationally we want to see that, credit both organisations with the care they are providing, but also be able to track which HASUs are providing care for which patients.

So the audit will want to pick out all the patients managed by a particular ‘team’ in a particular auditing period. My thoughts had been that if we modelled each team as a CareTeam, with potentially multiple teams sitting under the umbrella of an Organisation. Members of that CareTeam will have Encounters with the patient.

If you look at the existing audit ( performance reporting is based on ‘team based’ performance and ‘patient based’ performance. So for the first we are measuring what care a team provides. Hence my thoughts were that this query, if CareTeam maps to Patient as a one-to-many, is just going to be SELECT queries with a predicate starting ‘(CareTeam == MyTeam) && (PeriodStart == X) && (PeriodEnd == Y)’ We also need to provide ‘whole pathway’ performance so we would also look at the whole journey of patients who spend any period of time under a particular team - that data is used to study outcome measures for example.

Sorry - I know that’s a hugely detailed email, but hopefully gives the insights to the modelling.

So yes I guess it raises the questions of:

  1. Have I got the modelling right that CareTeam might actually be the right object to describe each multidisciplinary team, and if so does that argue the case for a 1-many mapping to Patient?
  2. If CareTeam is the correct model for that, what are the thoughts re. being able to map Location to CareTeam to get an idea of their underlying infrastructure (such as number of beds)?

Best wishes,


If you look at the definition of Organization, you’ll see that it’s much more flexible than you might have thought. It’s any group of people who can take collective action or responsibility. And that seems to be your purpose. Note that CareTeam isn’t something that can “act” or take responsibility - none of the resources reference it that way. It’s more a means of tracking who’s involved and what their respective responsibilities are. If you have an Organization that represents one of your “stroke units”, you could then link individual Practitioners to that Organization using PractitionerRole (including what their respective type of involvement was and what time-period they were so-associated. More importantly, Practitioners can perform actions in the context of their role with the stroke unit and, on occasion, you can indicate that actions were performed by the stroke unit as a whole.

I’m not seeing anything in your use-case that can’t be met by (and met better by) Organization. (But it’s entirely possible I’m missing something :>)

Per, the concept of a non-patient centric care team has been discussed in the past. The recommendation to use Organization has some limitations as outlined in the aforementioned Zulip chat. For example, the Organization resource doesn’t contain or reference the practitioners/members, which means that you can’t have a “contained” organization and updating the “team” means updating the members/practitioners to create that link.

Thanks lloyd. I suspect mapping to Organisation is going to be a better fit!