A clinical modelling conversation...

Dear Colleagues,

This email is jointly sent by the openEHR Clinical Knowledge Administrators, Silje and Heather.

Following recent email threads, we would like to establish some common understanding and expectations about the current clinical modelling effort and effect that we hope might stimulate a constructive and innovative conversation within the openEHR community about moving the clinical modelling work forward.

Let’s say that publication of a typical archetype takes four review rounds. Each review round runs for 2 weeks. If we had dedicated Editors who can turn those archetype reviews around immediately then we can take a draft archetype and publish it 8 weeks later. The reality is that there will be some lag times so the reality might be closer to 10 or 12 weeks, but we’re not talking 6 months or years.

Let’s also say that the typical Editorial time for each review round is 3 hours – an hour for an Editor to do the editing and one hour each for two Editors to facilitate the comments. So let’s add in one hour preparing an archetype for a review round and we have a total of 13 hours editorial time per archetype. Simpler archetypes and well-known scales or scores can be published in one or two review rounds. More complex ones like the adverse reaction archetype has taken tens of hours, possibly closer to a hundred, but worth the effort to get it right because of its importance in clinical safety. However we’re not talking unsustainable hours per archetype to get the majority published.

Within the typical standards environment where review of information models are done en masse in 3, 6 or 12 month cycles, our agile and dynamic approach to archetype review and publication is outrageously fast and requires only a modest budget. And the priorities can be driven by the implementer community.

We really need a different conversation happening about the archetype development process, one that recognises the efficient and value for money that we have put in place but is largely untapped, rather than complaining that the work to date is not complete enough, not focused on the right topics, not enough.

The practical reality is that by far the majority of the Editorial work is not resourced, so there is a limited strategic plan apart from the Archetype ‘Sprint’. Rather that the work is largely opportunistic:

  • dependent on archetypes that are volunteered as a result of real life implementations;
  • translations by those reusing archetypes in different geographical contexts; and
  • reviews occurring when people ask and then volunteer to participate in the process.

The Norwegian Nasjonal IKT work is a perfect example of this – so many of the archetypes published in the international CKM in the past couple of years are the direct result of the Norwegian priorities for content, driven and facilitated by the Norwegian Editors but with enormous value contributed by international input. Nasjonal IKT have effectively funded the majority of this work to support their national program, gaining the enormous benefit of international collaboration and input, and in return making available high quality archetypes for the rest of the international openEHR, and broader, community. They recognise this as a win-win situation.

The source of most of the limited funding that has recently been made available for editorial work in the international CKM is Norway’s Nasjonal IKT membership fees, which have been deliberately directed toward the international clinical modelling effort on request from Nasjonal IKT. A few hours a week of dedicated editorial time has already increased the international CKM activity manyfold in recent months. This includes timely responsiveness to community requests and contributions, for the very first time. It would be exciting to see this grow and expand through member organisations joining and specifically allocating some of their fees towards the clinical modelling effort.

With only modest, strategic resourcing, the collective benefit will be orders of magnitude larger than any single organisation can achieve by itself. The impact of this can extend way beyond the openEHR international community but to other standards organisations and digital health in the broadest sense.

Kind regards

Silje Ljosland Bakke and Heather Leslie

Clinical Knowledge Administrators for the openEHR CKM

Dr Heather Leslie

MB BS, FRACGP, FACHI, GAICD

M +61 418 966 670

Skype: heatherleslie

Twitter: @atomicainfo, @clinicalmodels & @omowizard

www.atomicainformatics.com

A few thoughts come to mind:

  • sets of archetypes could potentially be developed closer to completion by the grass-roots level, before submission to CKM, which would reduce editorial time, if better guidelines on development rules, patterns, etc i.e. the fabled handbook existed

  • consider a set such as for ante-natal care + birth + post-natal (6 weeks) - there might be 50 archetypes implicated here, with (we hope) at least half being generic (e.g. lab tests used in pregnancy are mostly not unique to pregnancy) - there is a lot of work here.

  • It might be a better approach if development teams were to try to develop whole packages to a reasonable level, rather than just submitting single archetypes and wait for results of review

  • whole package generally would be based on some process, care pathway etc, not just a data-oriented view. E.g. pregnancy; chemo+ monitoring; etc- if the fabled handbook of patterns and criteria for good archetypes existed, more editors could be trained.

  • is there any reason not to have just more people on the editorial group, e.g. 10?

  • is it time to agree a set of major clinical sub-specialties (< 20) and designate an owner for each one (i.e. an editor; some editors could own more than one area)?

  • we possibly need to distinguish two layers of archetypes, which would potentially change how editorial work is done:

  • generic all-of-medicine archetypes:

  • vital signs

  • many signs and symptoms

  • a reasonable number of labs

  • general purpose assessment / evaluations, i.e. Dx, problem description etc, many things like lifestyle, substance use

  • ?all of the persistent managed list types: medications, allergies, problem list, family history, social situation, consents, etc

  • the specialties, for each:

  • specific signs and symptoms

  • specific physical exam

  • specific labs

  • specific plans- more than one relationship between specialty archetypes and generic ones is possible, e.g. some are just new; some are formal specialisations in the ADL sense.

My guess is there is a number of issues to consider. Whether any of the above are the main ones I don’t know.

  • thomas
(attachments)

image001.jpg

Hi,

Imho the grass-roots editors need to (as you write)

  • A set of generic patterns to start with. Patterns for any: Observation, Evaluation, Order, Action, for use in documentation of Medical aspects and Administrative aspects about the Patient System.
    Including generic concepts like (Diff) Diagnosis Lists, Episodes, (Family, Social, …) History, Problem List, Orderlist, ActionList, …
  • Guidance, handbook, on how to use the patterns
  • Model one set of coherent archetypes dealing with a clinical domain/speciality (such as: medication, physical exam eyes, ENS, breast exam, clinical pathway, …)

The generic Patterns need to be created, maintained by IT-modelling experts.
The medical professions need to model and maintain the Archetype packages.
Local healthcare providers will create Templates to be used in their context.

Gerard Freriks
+31 620347088
gfrer@luna.nl

Kattensingel 20
2801 CA Gouda
the Netherlands

A few thoughts come to mind:

  • sets of archetypes could potentially be developed closer to completion by the grass-roots level, before submission to CKM, which would reduce editorial time, if better guidelines on development rules, patterns, etc i.e. the fabled handbook existed
  • consider a set such as for ante-natal care + birth + post-natal (6 weeks) - there might be 50 archetypes implicated here, with (we hope) at least half being generic (e.g. lab tests used in pregnancy are mostly not unique to pregnancy) - there is a lot of work here.
  • It might be a better approach if development teams were to try to develop whole packages to a reasonable level, rather than just submitting single archetypes and wait for results of review
  • whole package generally would be based on some process, care pathway etc, not just a data-oriented view. E.g. pregnancy; chemo+ monitoring; etc- if the fabled handbook of patterns and criteria for good archetypes existed, more editors could be trained.
  • is there any reason not to have just more people on the editorial group, e.g. 10?
  • is it time to agree a set of major clinical sub-specialties (< 20) and designate an owner for each one (i.e. an editor; some editors could own more than one area)?
  • we possibly need to distinguish two layers of archetypes, which would potentially change how editorial work is done:

Level three is the Template level for the local context.

Hi

Regarding

A set of generic patterns:

There is for example a generic CLUSTER to be used for examinations: https://ckm.openehr.org/ckm/#showArchetype_1013.1.2017

In the international CKM there is also a variety of other ENTRY archetypes to be inspired by. I strongly believe there is a lot of other good archetypes out there, but have never been uploaded to the int’nat CKM. That’s a pity.

Guicance/Handbook:

I had an idea (with my colleague Hallvard Lærum) to write the book “The Ultimate And Final Guidance To Make Perfect Archetypes – The Draft”. J Unfortunately, that’s not possible. The variety of concepts are far to broad, and the knowledge is ever expanding. But it’s a good idea to make a basic guidance – available online. Who will finance this? I’m sure we can make it.

On the other hand, there has been training courses in Norway, UK and Germany during the last years, and they could (and should?) be available as streaming online. Again: Who will finance and do this?

Online official certification courses could also be available online, preferably through openEHR Foundation. Any money for this anywhere?

Today, there is both this clinical list available for asking questions. Could be used more pinpointed to concrete archetype design questions.

Also there is a Slack channel, where modelers can post questions and discuss design issues.

In my experience, making good archetypes is nothing that can be done in solitude. You need input from the community, and deep knowledge of existing archetypes to be able to make reasonable new ones. I’m afraid of leaving too much to local medical professionals.

Vebjørn Arntzen

Hi All

The value and use of structured data in health care has long been debated: openEHR allows for arbitrary levels of structuring and reuse. Many of the larger companies, due to commitment to database technologies, and AI proponents arguing for natural language processing and fuzzy indexing. There is little evidence either way, but it is clear to me that structuring is the easier and safer way (certainly for the foreseeable future). I have seen very little advance in systems over my career, and most that has come has been as a result of detailed information being captured and stored in a retrievable way.

Given our community’s buy-in to the pro-structure approach, Heather and Silje have some valid proposals for us all. Anyone involved can see the advantage to quick turn arounds sharing development over a large pool of interested parties. It is not so obvious to others, who enjoy the quarterly gatherings and the narrow set of demands that arise from a (very) few people in a room. However, rather than take national program objectives as our driving force (which it has often been), I think we should add a layer to the debate; that is, how do we prioritise efforts to agree on data structure. The positive outcomes (health, process, reporting) that arise from having data structured would appear to be important. I have been working on Microbiology results lately and reporting around these results. The aggregated reporting that is possible is really limited only by the structure of the data, not just in microbiology but all the patient, medication and logistic factors you want to report against. Adding one element to one aspect of the data can massively affect the power of reporting for instance.

High fidelity data is often critical in providing patient care in ways that are difficult to foresee and which are implemented differently all over the place. Hepatitis B infectivity and immunity is a current example of this for me. Results of lab tests need to be structured, as do vaccination data and time processed to see what is going on. This is relatively easy for a hepatic specialist, but very difficult for a generalist without good records and reports. It probably warrants a small case study. I raise this because we could raise particular scenarios as priority areas of work rather than data itself.

Other examples I have come across are exposure to and treatment of syphilis and managing advanced Kidney Disease. Are there common areas of process that struggle without good data? What are the scale of these process, do they required shared care to be effective? How many of these exist. Antenatal care is another example but efforts to describe this as workflow have been particularly unsuccessful – do we need AI and highly structured data?

I am interested in other people’s ideas.

Cheers, Sam

(attachments)

image001.jpg

(I tried to respond the following Friday, but it seems not to have been sent/distributed to the list. If it actually happened, I apologize for the double up…)

Hi

Regarding

A set of generic patterns:

There is for example a generic CLUSTER to be used for examinations: https://ckm.openehr.org/ckm/#showArchetype_1013.1.2017

In the international CKM there is also a variety of other ENTRY archetypes to be inspired by. I strongly believe there is a lot of other good archetypes out there, but have never been uploaded to the int’nat CKM. That’s a pity.

Guicance/Handbook:

I had an idea (with my colleague Hallvard Lærum) to write the book “The Ultimate And Final Guidance To Make Perfect Archetypes – The Draft”. J Unfortunately, that’s not possible. The variety of concepts are far to broad, and the knowledge is ever expanding. But it’s a good idea to make a basic guidance – available online. Who will finance this? I’m sure we can make it.

On the other hand, there has been training courses in Norway, UK and Germany during the last years, and they could (and should?) be available as streaming online. Again: Who will finance and do this?

Online official certification courses could also be available online, preferably through openEHR Foundation. Any money for this anywhere?

Today, there is both this clinical list available for asking questions. Could be used more pinpointed to concrete archetype design questions.

Also there is a Slack channel, where modelers can post questions and discuss design issues.

In my experience, making good archetypes is nothing that can be done in solitude. You need input from the community, and deep knowledge of existing archetypes to be able to make reasonable new ones. I’m afraid of leaving too much to local medical professionals.

Vebjørn Arntzen