There’s a CVD risk assessment tool I’m working on which prepopulates clinical info from GP software. This includes diagnoses, smoking status and checklist for certain medications. Note that some of the underlying info might be coming from previous visits (e.g. problem list type) but also can be newly entered as a result of GP’s assessment. Now, regardless of what happens in GP software, when it is transferred onto this tool (whether automatically prepopulating and/or manual entry) are these Observations or Evaluations? Note that the GP does not make any further clinical judgement here, just rephrase existing data for a different purpose. My gut feeling is the former (Observation).
I know this is tricky and has been brought to this list many times here but thoughts? Masters?
Good afternoon,
I use MedTech32 almost everyday. The recent CVD risk assessment is incorporated with diabetes risk assessment but not 3 month diabetic check-up one. CVD risk assessment actually adjust risk and recall patient for 3 monthly, 6 monthly, annual or every 5 yrs, best on the data that we input in this new form (observation data). Few old things are automatically populates (such as previous BP the most recent one and medication information). The other blood test factors (HDL, LDL, total cholesterol and more are new results and we enter at the time of patient assessment. It again adjust risk for recall. There are wizard to enter data automatically, while referral patient to other provider.
We can do more modification it looks though.
Because it is a statement about an inference about characteristics of a possible ongoing process inside the patient system.
And not a state in a process that can be observed via our senses at a point in time.
E.g. The measured/observed systolic blood pressure is a state of a cardio-vascular process inside the patient system.
The measurement of the systolic blood pressure is an Action type archetype.
Both the blood pressure Action and Observation are linked and each have a state model attached to it.
An inference about a process inside the patient system is modelled using the Evaluation pattern.
We can not observe a process in principle, only states at a point in time when we perceive observables and record the observation.
Evaluation is about inferences about characteristics in general of the process and risk is one of the characteristics we can attach to that process and make inferences about.
I think it would be a composition with observation and evaluation archetypes.
In my experience on surveillance program, your form could be built up
these items.
* OBSERVATION
History of medication, smoking
* EVALUATION
Diagnosis, GP assessment status
(INSTRUCTION/ACTION)
Medication(GP prescribed/prescribing)
The risk itself is an Evaluation and can be used to store data about the risk.
The procedure/method to do the calculation is not an artefact that will be stored, but referred to in the Evaluation.
Using a RM and an AOM it must be possible to specify in a Composition type of artefact the components (queries) and the procedure/method/algorithm/calculation, etc.
Provided we have a standardised way to express the needed logical operations, calculations, etc.
Just out of the Risk analysis box see eg. ISO 14971:
Took me some time to figure it out, so I share it just in case:
Risk consists of:
- A hazard
- a probability that this hazard will typically occur
- the severity of the harm that this hazard will cause on humans or non-human subjects
- probability together with severity will describe the risk
some parts of this may well be observations, others will be evaluations
Greetings from Vienna,
Stefan Sauermann
Program Director
Biomedical Engineering Sciences (Master)
University of Applied Sciences Technikum Wien
Hoechstaedtplatz 5, 1200 Vienna, Austria
P: +43 1 333 40 77 - 988
M: +43 664 6192555
E: stefan.sauermann@technikum-wien.at
a probability that this hazard will typically occur
the severity of the harm that this hazard will cause on humans or non-human subjects
probability together with severity will describe the risk
Should we add the temporal factor?
E.g. the chances in a defined period of time.
When you mean that the constituting elements of a RISK are Observations and Evaluations, you are right.The RISK itself is documented by these components.
E.g. RISK= High Risk of Something to occur in a certain time period.
The composite is an Evaluation.
The result expressed in the Evaluation is a semi-quantitative result consisting of the constituting Observations and Evaluations.
HIGH RISK= High probability of Something to occur in a certain time period and causing a severe harm to persons or other subjects
LOW RISK = something will never occur and even if it does it will not harm anyone or anything
Greetings,
Stefan Sauermann
Program Director
Biomedical Engineering Sciences (Master)
University of Applied Sciences Technikum Wien
Hoechstaedtplatz 5, 1200 Vienna, Austria
P: +43 1 333 40 77 - 988
M: +43 664 6192555
E: stefan.sauermann@technikum-wien.at
Hi Stefan, whenever you assess someone as having a risk, you are doing something similar to diagnosing: i.e. you are comparing some input data items about the patient (high BPs over last 3 months, smoker, 55years old etc) to some knowledge base that says that these things taken together constitute a risk for hypertension, due to some population studies. Any probability etc is derived from that knowledge base. So these are all evaluations. - thomas
Some of the latest versions of existing archetypes might suit your purpose – the most recent are currently on the NEHTA CKM and we are looking to transport them across to the openEHR CKM.
Only looking at temporal aspects:
The risk to die within 10 years when you are 100 years old is HIGH
The risk to die within 10 years when you are 30 years old is LOW
The risk to die within 100 years for each average person is HIGH
The risk to die within one day for the average person is LOW
The risk to have prostate cancer until 40 years of age LOW
The risk to have prostate cancer after 70 years is HIGH
Adding severity:
The risk to die of prostate cancer after 70 years is LOW (Most man have prostate cancer at this age but most never die because of it)
How is RISK used in healthcare?
Almost never it is calculated with mathematical precision.
It is expressed in a semi-quantitative way: Low, Normal, High, etc.
It would be nice when archetypes can express that in a sensible way.
It will be impossible to use the DV-Data Type for this.
Reason? It is not expressive enough because the semi-quantitative result needs to document more.
It needs inclusion and exclusion criteria because there is no generic definition of Low, Normal, High.
Each context is unique and opinions change over time.
Each and everybody has to define the criteria in the local context. (Speciality, organisation, disease/condition, healthcare provider even patient specific at one point in time)
Completely agree with controversy in RISK evaluation.
I have similar experience on SEVERITY evaluation.
openEHR-EHR-problem-diagnosis archetype has severity metrics,
but it does not fit for various evaluation criteria.
I specialized to have a 'severity detail' slot to apply various severity
criteria for many diseases.
As Risk evaluation depends on cases, we need to develop cluster to
express such risk evaluation models on demand.
Hi Heather, all others – many thanks for your responses.
I agree that risk is an evaluation – no question with that. However I reckon I couldn’t explain the issue I had clearly:
My question was when documenting a patient’s health information at a point in time base on past diagnoses, medications and smoking status not necessarily assessed by the same physician, should these be observations or evaluation? I’m not really asking about risk assessment – in that case the physician (or decision support tool) is considering past observations etc. against a knowledgebase and making a clinical judgement. That should clearly be an evaluation
To give an example: a medical student is reading problem list, medication list and smoking status from patient’s EHR and putting on a form to present to a physician – is the stuff on that form Observation of Evaluation? Heather I take your point that in order to reuse data it is best to stay faithful to original models…
So that makes me think whether we find a smarter way of dealing with Observation vs. Evaluation types. Kind of multiple inheritance where, depending on a particular usage, they can inherit contextual parts of either Observation or Evaluation. I can see this will save a lot of hassle and still enable us to reason using the ontological Clinical_Entry classes. Probably same is true for Instruction/Action types as well… Anyway this is probably even a bigger discussion!
In EN13606 Association we think that all artefacts must be derived from (specialised) from one generic pattern.
The draft document SIAMS defines this, plus more.
All artefacts inherit this generic pattern.
It is just one change of one attribute in the pattern that changes it from constraints on the generic ENTRY class used as Observation to one that is an Action or Instruction or Evaluation.
And all use the same generic pattern, but depending on different requirements they use other combinations of possibilities in the pattern.
What they have in common, will not change.
An other attribute defines the usage of the artefact for either documentation, querying, presentation.
This brings back flashbacks to the 'data quality' discussion we had on this list about 5 years ago. The conclusion then was that we first had to learn to walk before we could run. Hopefully today we're learning to run.... (can't wait untill we're learning to fly:-) ).
Personallly i still think that any RISK or SEVERITY evaluation if completely worthless unless that evaluation AT contains a detailed protocol describing the criteria (preferably based on evidence based literature) on which that evalution is made. To come back to Gerard remarks, that also would mean that it decribes that the outcome high means that the risk for the event x in that particular situation is greater that y%.
Personallly i still think that any RISK or SEVERITY
evaluation is completely worthless
You may want to define "worth" to put this into context.
unless that evaluation AT
contains a detailed protocol describing the criteria
(preferably based on evidence based literature) on which
that evalution is made. To come back to Gerard remarks, that
also would mean that it decribes that the outcome high means
that the risk for the event x in that particular situation
is greater that y%.
Statistician's wet dreams, not clinical reality. It
typically doesn't apply to a specific patient just so.
In the EN13606 Association in our SIAMS document we developed a generic semantic patterns that drives all artefacts,
One of the sub-patterns is to document a semi-quantitative result.
This semi-quantitative result is to document all ‘severity’ type of things.
Always these semi-quantitative things (Severities) are defined in a specific context.
So each semi-quantitative semantic sub-pattern allows one to include inclusion and exclusion criteria for each entry in the semantic ordinal sub-pattern.
See: http://www.en13606.org/wiki/index.php?title=File:Artefact_SubPattern-Semantic_Ordinal_v1.png
And yes, these sub-patterns always are Cluster Models.
