GELLO v.1 as a candidate openEHR CDS language - message reduced to <40kb, RESEND

Decision Support (archive)
1

Dear CDSSers

I am a newbie on this list but have followed and worked with archetypes
for a little while now. I would like to talk about the strongly typed,
OO, declarative CDS language called GELLO. This is a HL7 standrd that
wasn't really ever implemented. OpenClinical gives a fair treatment of
the topic. Our group (Medical Objects) has implemented it after some
work on resolving ambiguities in the BNF and the development of various
editors and interpreters such as a GELLO v.1 editor/ interpreter; an
ADL editor (that saves to ADL 1.2, basic xhtml and en13606), a GLIF
editor, an existing Snomed- CT server of ours and a virtual medical
record (vMR) layer.

We use GELLO almost as a scripting language to calculate values on
nodes, to initialise values, to allow node visibility to preserve screen
real estate and to calculate reference ranges and autocoments for
pathology systems. It is useful for querying (and processing) data
retrieved from EN13606 archetypes which we have persisted in HL7 v2 and
abstracted to a simple vMR. An example of this would be in a full blood
count (or CBC) lab archetype where many elements of type
quantity/physical quantity are derived from other values.Indeed we
retrieve patient deomgraphic data from the PMS for things like age and
gender in 'invariant' ELEMENT node calculations. We propose and have
implemented functionality in the GELLO class hierachy for SCT validation
and implies methods, as well as general property retrieval on Snomed
Descriptor calasses which are a type of Coded Value.

A pathology system built on this technology is real, deployed and used
every day by an Australian pathology lab. It uses about 40 lab
archetypes which I would like to contribute in openEHR ADL 1.4 form to
the CKM. While I am waiting for the new pathology archetypes to
specialise or comment on, I just don't know how to do our existing
functionality with AQL. Is A-Path capable , what is CLIPS?

for example eGFR (a derived value which indicates renal function) eg:

                    Let o:observation = parameter[1]

                    Let Gender = Patient.gender
                    Let Age: Integer =
                    patient.age_at(factory.absolutetime("0").now())

                    // get the creatinine level value from higher in the
                    tree
                    Let Cr: Real =

                      if o.isdefined() then
                         o.find_observation('1.1.9').value_asPQ().value
                       else
                         0.0
                       endif

                    //do the agorithm and return the PQ

                         Let x: Real = Cr*0.0113
                         Let y: Real = Age

                      Let Result: Real =
                          If y > 17.0 then
                    186*x.power(-1.154)*y.power(-0.203)
                          else 0.0
                          endif

                       Let FemaleResult: Real = Result*0.742

                    if Gender = 'M' then Factory.PhysicalQuantity
                    (Result, 'mL/min/1.73m2')
                    else Factory.PhysicalQuantity (FemaleResult,
                    'mL/min/1.73m2')
                    endif

  Ok its a it ugly but hey I did it and I am a clinician.

You may have noticed the term GLIF earlier in this post - this higher
level guideline vehicle uses GELLO and archetypes for an Australian Govt
funded demonstration project around the management of lymphoma, a
complex use case again.

happy to discuss further. I think GELLO is a candidate along with AQL
and A-Path. Perhaps they complement each other. The BNF for GELLO v1.x
is available from the www.medical-objects.com.au website .

Best regards

Peter Scott
GP and health informatician

2

Hi Peter,

A pathology system built on this technology is real, deployed and used
every day by an Australian pathology lab. It uses about 40 lab
archetypes which I would like to contribute in openEHR ADL 1.4 form to
the CKM.

Great news.

what is CLIPS?

CLIPS is considered the gold standard for forward-chaining inference
engines.

I suggest you get the documenttion and read the Users Guide first:
http://sourceforge.net/project/downloading.php?group_id=215471&filename=documentation_624.zip&a=94460201

You may have noticed the term GLIF earlier in this post - this higher
level guideline vehicle uses GELLO and archetypes for an Australian Govt
funded demonstration project around the management of lymphoma, a
complex use case again.

Though InterMed & DSG:
Decision Systems Group (DSG), Harvard Medical School in association with
the Clinical Decision Support Technical Committee and Clinical
Guidelines Special Interest Group of HL7 and (in early work) the
InterMed Collaboratory (USA).

have kind of been out of business since funding dried up a few years
ago.

I had an email exchange with Dr. Bob Greenes last week and he told me
that there is renewed interested in GELLO & GLIF. His words were "stay
tuned". I don't know if you are working with the HL7 DSS group but
maybe they could use some of the improvements you have made?

This is a very important area to move ahead with real computability in
health care.

Cheers,
Tim

3

Hi Tim thanks for the sourceforge link.
Yep we are on the HL7 DSS committee. Andrew McIntyre has presented on
GELLO v.1 (and GLIF) at the last two or three working group meetings,
and are looking at ballotting some minor changes to the BNF, examples
and of GELLO v.1 midyear. So we are contibuting to a standards process.

Regards

Peter Scott