Detailed Clinical Modelling for EHR Development and deployment and for H...

Clinical (archive)
1

Dear Eric, Tom,

I will answer in the sections.
In a message dated 8-11-2008 10:31:58 W. Europe Standard Time, eric.browne@montagesystems.com.au writes:

Regarding coded ordinals, ADL provides for both the ordinal value and the
corresponding text. If we use the Barthel Index archetype ( e.g.
http://www.openehr.org/svn/knowledge/archetypes/dev/html/en/openEHR-EHR-OBSERVATION.barthel.v1.html
) as an example, then you can see that both the ordinal ( assessment score
) and the corresponding meaning (both in English and in Dutch in the above
example ADL) are included.

I know the example: Sam made the ADL out of our document. This is fine, but still needs work to keep it combined according the new ISO standard on datatypes which is in sync with 13606

Again, taking Barthel Index as an example, the ability to derive the total
score from the individual ten assessments as a formula in the model is
attractive since it documents the meaning of the total score explicitly in
the model. This is possible with ADL.

I see also in Toms example that it can be part of the adl file, but my question is also how and where to include it in the archetype editor and what settings need to precede in order to allow specifying this.

One problem is in adding such capabilities into the user interface of the
editor. Ideally, I suspect clinicians designing archetypes would like to
be able pull down a sum() function from a list of functions and reference
each of the 10, say, contributing scores used to build the total Index
score.

Yes, that would be very helpful.

Generalising such capabilities is not trivial. Where would one stop with
building formulas? What if you wished to reference data items outside of
the current archetype? How complex should the expression builder be? Have
you a list of such requirements?

No we do not have a list, but we do have examples:

  • sumscore of 2-n underlying variables (Barthel, Apgar)
  • sumscore of 2-n underlying variables leading to score on a subscale plus sumscore of several subscales into a scale total (minimental state exam)
  • logical / mathematical operations (body surface and BMI direved from lenght and weight and doing the trick against a formule
  • some simple logics, like if a and b and not c, then d.

The second problem, is that it is then not possible to translate such
formulas into a formalism of your choice - certainly not into HL7 RIM
based models as you might wish, since, as far as I know, it does not have
the cability, and it violates the fundamental V3 principle that all
knowledge is expressed in the foundation RIM components and cited
vocabularies.

This is a not correct assumption. HL7 v3 classes have the attribute of derivation expression which we widely use to express the calculations and logics and which can work with Arden syntax. Arden would perhaps be a first more complete set. Other options include several standard formulas like average, sum, median etc as in excel or simple statistical packages.
HL7 v3 has some internal principles, but also facilitates external knowledge, vocabularies etc. See in particular the different message models on decision support.

Sincerely yours,

dr. William TF Goossen
director
Results 4 Care b.v.
De Stinse 15
3823 VM Amersfoort
the Netherlands
email: Results4Care@cs.com
phone + 31654614458
fax +3133 2570169
www.results4care.nl
Dutch Chamber of Commerce number: 32133713

2

Hi William,

Please correct me if I misunderstand your text.

It seems to me that your complaints about openEHR are really things that
aren't (and probably shouldn't be) in the archetype editors. However,
when you make HL7v3 comparisons, you talk about capabilities of the
model and not capabilities of any particular tools.

So, I am curious as to which level you want to make critical review on;
tools, or model?

As an aside:
As far as Arden Syntax MLMs are concerned, I believe that their
limitations are widely known (though the EGADSS project did make strides
in abstracting the curly braces issue). I believe that future looking
decision support systems are looking to a better method (such as
ProForma or GELLO) of encoding clinical guidelines. Especially since
GELLO is a HL7 standard.

Cheers,
Tim

3

the above are already catered for in ADL, e.g. (in the following I have used ‘a’, ‘b’, ‘c’, etc instead of the usual long paths):

simple_sum: d = a + b + c

or to be more correct, depending on whether any of the items can be missing:

simple_sum: (exists a and exists b and exists c) implies (exists d and (d = a + b + c))

multiple levels of such scores are done in the same way, with invariants giving each first level sum, then more invariants summing the level 1 sums.

body_mass_index: path_to_bmi = path_to_weight / path_to_height ^ 2

In general, you have the following operators:

  • arithmetic: +, -, *, /, ^

  • relational: <, >, <=, >=, =, !=

  • boolean: not, and, or, xor

  • exists

  • for_all

  • parentheses: ()
    Other functions should be included one day, although the current grammar does not handle them: log2, logn, sin, cos, tan, and so on. (Note that functions like average, sum and median can be expressed using basic operators).

  • thomas

4

Thomas, Thimothy,

Thank you both for the further explanation.

Given your replies, I think the issue on tools is partly our unfamiliarity in its use, and partly features that are not available. While working on more DCM and their formalisation to ADL more questions will come up I think.

Thimothy,

Yes you are rigth, Gello is currently used for Decision Support work in HL7. I meant to explain that in the HL7 v3 Act attribute ‘derivation expression’ it is perfectly possible to use examples of external from HL7 RIM expressed calculations, mathematical operations and rule expressions like Arden, like Gello.
In the perspective of DCM, I believe that we in the early developmental stages will have to deal with small / simple examples like the BMI example Thomas provided.

Sincerely yours,

dr. William TF Goossen
director
Results 4 Care b.v.
De Stinse 15
3823 VM Amersfoort
the Netherlands
email: Results4Care@cs.com
phone + 31654614458
fax +3133 2570169
www.results4care.nl
Dutch Chamber of Commerce number: 32133713

5

I would have to agree with Tim in that the question of 'what can the
tools do?' and 'what can the formalism do?' need to be differentiated. I
understand of course that if you are a tool user, you might not easily
guess what the formalism can do if parts are not yet implemented.
Nevertheless, for standardisation purposes we need to be aware of both.

- thomas beale

Williamtfgoossen@cs.com wrote: