In modern medicine, how much effort should we be putting into creating
detailed archetypes of physical examination findings? When I learn't
medicine, 30 years ago, there was a still a huge focus on detailed
examination technique and documentation e.g. clubbing, pulse
description, detailed murmur description etc. Although cast down on
tablets of stone, a lot of this is very poorly researched and
validated, barely reproducable and is increasingly being bypassed by
the use of imaging or other testing.
As an ex-GP, I am certainly not disputing the value of clinical
examination, or recording the results, or indeed that some examination
findings have very specific computable utility.
I do wonder, however, how much effort we need to place in developing,
fr example, the pulse archetype, structuring all of its vagaries of
rythm and character, or is the Heart rate archetype sufficient with
some space to record some text and/or coded terms for 'pulse'
characteristics.
Ultimately it comes down to whether we ever need to query for
pulse/character = 'water-hammer' ? I am not seriously suggesting
ditching the pulse archetype but I am conscious that there are many
other similar examples where modelling the vagaries of examination can
be pretty difficult and the effort may be disproportionate to the
utility.
Flak-jacket ON
Ian
Dr Ian McNicoll
office +44 (0)1536 414 994
fax +44 (0)1536 516317
mobile +44 (0)775 209 7859
skype ianmcnicoll
ian.mcnicoll@oceaninformatics.com
Clinical Modelling Consultant, Ocean Informatics, UK
Director/Clinical Knowledge Editor openEHR Foundation www.openehr.org/knowledge
Honorary Senior Research Associate, CHIME, UCL
SCIMP Working Group, NHS Scotland
BCS Primary Health Care www.phcsg.org
Thanks for the input. Why question was genuinely one of enquiry rather
than that I have a strong opinion either way.
Also I think the issue of prioritstaion of which archetypes to cover
first is a little different. On the face of it, very generic
archetypes like medication should be a priority but actually they are
amongst the hardest to model internationally as they are often heavily
dependent on national terminologies, require a lot of research and
analysis to cover existing standards and approach, and clearly
meritvery slow and careful review. All of this is resource intensive,
something that right now, we do not have.
Thankfully there are some forthcoming European and international
initiatives that may give us some protected resource, both to restart
Archetype reviews and to expand the editorial team.
Regards,
Ian
Dr Ian McNicoll
office +44 (0)1536 414 994
fax +44 (0)1536 516317
mobile +44 (0)775 209 7859
skype ianmcnicoll
ian.mcnicoll@oceaninformatics.com
Clinical Modelling Consultant, Ocean Informatics, UK
Director/Clinical Knowledge Editor openEHR Foundation www.openehr.org/knowledge
Honorary Senior Research Associate, CHIME, UCL
SCIMP Working Group, NHS Scotland
BCS Primary Health Care www.phcsg.org
Excellent controversy alert, Ian. It helps me to remember always that EHR is not only about informatizing the medical record. It´s a new way of delivering care, a whole new world.Many observations were made to guide, compare clinical findings or build consensus. We clinicians had to have skills to recognize signs and symptoms, evaluate them agregated and make a decision on what to do. I. We clinicians learn in our semiology casset, that the artery dancing (carotid) was a sign of imcompetence of aortic valve. We used to see a lot of uncompensated patients in the infirmaries and I always remember to make the clinical rounds to learn to recognize those signs. But today we have means to diagnosis diseases before we can see any physical sign of aortic regurgitation. I know it was a artisan diagnosis done without any test help but it was only done when the disease was already severe.
I have a 14 years old dog, a poodle, two years ago he has loss of appetite, persistent cough especiallly with acitivity or stress.With these sign, together with some dempographic simple data, like his breed (a toy poodle) and a normal cardiac observation, and the support or a test a eccodopler who could demonstrate that he was in the early stage of a progressive degenerative deformity of the mitral valve, could aid the evaluation and act of the vet. He was medicated with furosemide and enalapril. He´s compensated fot the last two years and didn´t developed cardiac insuficience. Although the disease is progressive and non-reversible, we didn´t need any clinical information of a more advanced stage of the disease like ascistes, tiring quickly, fainting to do a diagnosis because technology helped us to treat him earlier. He´s a dog, and his vet is more informatized than many of doc´s office I´ve seen. And because of he´s adog that doen´s speak, I think the vets are more adavnced in adopting EHRs. They know the value of IT to help them to decide.
It´s also the same regarding to Scales. They were also built to help us to be moe accurate, to, based on epidemiologic and clinical observations, diagnose the gravity of a condition for many uses,like prescription, assesssment,comparison etc. But why today we still need to fill a Glasgow scale if the computer can help us, based on our clinical observations to assess it? Scales are data agregation and a value attributed to a particular aggregation of data. I think they belong to workflow and not to clinical modelling. By using artificial intelligence to do the data aggregation and evaluation, we could be more accurate and reliable than today.
I think we should concentrate ourselves on primary care and preventive medicine first. These other informations which were more frequent whne we didn´t have the diagnostic tests, will be more and more rare. They are good for a certain enviroment, like research…Like Stef said, only upon some important request.
I assume you meant Heart beat rather than heart rate. Heart beat and pulse are completely different entities and their characteristics not always identical. I certainly agree about the extent of the granularity to which any archetype requires to be developed.
When I was a student/junior doc we has to complete a detailed
'clerking sheet' on admission with a fair amount of tick-boxes for
detailed physical examination - is this still routine practice?
Ian
Dr Ian McNicoll
office +44 (0)1536 414 994
fax +44 (0)1536 516317
mobile +44 (0)775 209 7859
skype ianmcnicoll
ian.mcnicoll@oceaninformatics.com
Clinical Modelling Consultant, Ocean Informatics, UK
Director/Clinical Knowledge Editor openEHR Foundation www.openehr.org/knowledge
Honorary Senior Research Associate, CHIME, UCL
SCIMP Working Group, NHS Scotland
BCS Primary Health Care www.phcsg.org
doesn't the addition (mooted of the new hybrid CLUSTER / ELEMENT
type take care of this nicely, i.e. if you create a node as being of
this type (I called it VALUE_CLUSTER I think in a previous post) then
you can safely leave the details till a later round of development, and
when that happens, the changes made then won't break any paths or other
detals of the previous version, nor any data.
The world is fractal, and so is its information ...
I have a 14 years old dog, a poodle, two years ago he has loss of appetite, persistent cough especiallly with acitivity or stress.With these sign, together with some dempographic simple data, like his breed (a toy poodle) and a normal cardiac observation, and the support or a test a eccodopler who could demonstrate that he was in the early stage of a progressive degenerative deformity of the mitral valve, could aid the evaluation and act of the vet. He was medicated with furosemide and enalapril. He´s compensated fot the last two years and didn´t developed cardiac insuficience. Although the disease is progressive and non-reversible, we didn´t need any clinical information of a more advanced stage of the disease like ascistes, tiring quickly, fainting to do a diagnosis because technology helped us to treat him earlier. He´s a dog, and his vet is more informatized than many of doc´s office I´ve seen. And because of he´s adog that doen´s speak, I think the vets are more adavnced in adopting EHRs. They know the value of IT to help them to decide.
brilliant example!
It´s also the same regarding to Scales. They were also built to help us to be moe accurate, to, based on epidemiologic and clinical observations, diagnose the gravity of a condition for many uses,like prescription, assesssment,comparison etc. But why today we still need to fill a Glasgow scale if the computer can help us, based on our clinical observations to assess it? Scales are data agregation and a value attributed to a particular aggregation of data. I think they belong to workflow and not to clinical modelling. By using artificial intelligence to do the data aggregation and evaluation, we could be more accurate and reliable than today.
I think this is also a very good observation - I think of scales as a classification tool, not ‘observations’ per se - and as such, they are indeed meant for informing workflow, rather than making detailed assessments.
It goes to the heart of some of the key people, process and technology challenges modern healthcare faces.
Firstly there is a lot of tradition taught at medical school ( or there was in my time!) without a lot of evidence for some of it.
I can tell you that process of medical clerking at the bedside can be as inefficient today as it ever was..
Example.. in my ED the juniors take the history and examine at the bedside (10-20mins) , then go back to the desk to write it up on paper (10-20mins).
I suggest they all at least use a clipboard, so they can document as they go. Few enough do..
Bearing in mind our EDs process 200,000 patients per year in 1 English city.. think of the inefficiencies in this aspect of healthcare alone..! http://frectal.com/book/healthcare-an-introduction/healthcare-challenges-at-the-frontline/
Secondly, just as there is a natural balance between the complex and the complicated, the art and the science in medicine, there is a balance to seek in narrative versus structure in the patient record.
I cite 3 patients to illustrate
a) rr40, hr 130, GCS 12, ph 7.1, Glu 35
b) lady who after argument with neighbour about their fence, took an overdose of 20 paracetamol 8 hours ago
c) patient who was so desperately frustrated with work that he squashed a sandwich into his face
No specific patients , but the 3 illustrate the mix we see in healthcare
a) Diabetic Ketoacidosis (assume could be understood by some of you from the numbers)
b) Mix of mental health and medical management dilemmas
c) mental health issues
They illustrate the range from
a) easy to structure
b) mix of narrative & structure
c) all the value is in the narrative, ie nothing could be/should be structured.
Toms point abou the fractal nature of information is also key.
I'd suggest that archetypes could/should start with the minimum of structure+ allow room for narrative routinely.
In any event the only way to prove the right mix is iterative learning from exposure at the frontline..
A good approach. Very practical. I am all with you. I could also add,
when looking at the narrative in the user interface perspective, that
a user interface could react to specific terms when they are put in by
the user. Some of these particular terms could be able to quantify or
otherwise structured, and the user interface could present a way for
that. I have seen this method in some demonstrations of GUIs. Like
when you write 'body temperature', the user interface would react and
ask for a number to store in the EHR and than continue to record the
narrative. One demonstration actually had small context sensitive drop
down menues generated on the actual context sensitive word within the
text input field.
I think this is a beatiful way of moving narratives to structure, as
far as it is at all possible. So in conclusion, the user would start
by 'clicking the structured data' as far as we can foresee it and then
writing the narrative down but with a context sensitive user interface
which would structure whatever is possible, as far as possible. As
time goes by, the user might fancy the structure even to begin with,
be more engaged in structuring what is on the user's mind to get more
overview of the care process step by step. The structure will ask for
'order sets', 'clinical pathways', 'standardized care plans' and other
names of structured artefacts. With the means of archetypes, I
believe.
We are in my team at the moment looking a lot at the feedback from
structured and not so structured data, as both are clues to medical
knowledge artefacts and health care plans, which in the next step will
affect the reuqest for further structured input and facilitate the
overview of both patient data as in conditions and previuos
treatments, and connecting this to known ways of medical treatments.
of the new hybrid CLUSTER / ELEMENT