Information entities in medical records
Stefan Schulz
I had a closer look on IAO because I want to represent clinical
information as, e.g. encoded by HL7 V3, OpenEHR archetypes, or parts
of SNOMED CT (e.g. the "context model").
In their statements clinicians often convey negative, speculative, or
probabilistic information. There are diseases without diagnoses just
as diagnoses without diseases.
Using IAO I see the following problems:
1.
How to represent plans that are never realized:
For properly representing the SNOMED CT concept (ID:183983001):
- 'Plan of heart operation'
I would suggest
'Plan specification' and isAbout only 'Heart operation'
The problem is that in IAO 'Plan specification' implies 'isAbout some
Entity'
However, a characteristics of plans is that what they are about may
never happen. By using existential quantification, the IAO definition
requires at least one instance of the kind of thing the plan is
about.
2.
Is Diagnosis an information entity?
E.g., I want to represent "Suspected diagnosis of gallstones" along
the following lines:
DiagnosticStatement and isAbout only Gallstones and hasQuality some
Suspected
Again, I do not want to use an existential quantifier ('some') because
in the moment such a statement is recorded nobody knows whether the
patient has gallstones.
Any comments or suggestions?
Stefan
Alan Ruttenberg
>
> Dear all,
>
> I had a closer look on IAO because I want to represent clinical
> information as, e.g. encoded by HL7 V3, OpenEHR archetypes, or parts
> of SNOMED CT (e.g. the "context model").
> In their statements clinicians often convey negative, speculative, or
> probabilistic information. There are diseases without diagnoses just
> as diagnoses without diseases.
>
> Using IAO I see the following problems:
>
> 1.
> How to represent plans that are never realized:
>
> For properly representing the SNOMED CT concept (ID:183983001):
> - 'Plan of heart operation'
>
> I would suggest
>
> 'Plan specification' and isAbout only 'Heart operation'
>
> The problem is that in IAO 'Plan specification' implies 'isAbout some
> Entity'
> However, a characteristics of plans is that what they are about may
> never happen. By using existential quantification, the IAO definition
> requires at least one instance of the kind of thing the plan is
> about.
and we can say that this realizable entity is realized only in heart
operation.
Plan of heart operation is concretized as (is_realized_by only heart operation)
> 2.
> Is Diagnosis an information entity?
> E.g., I want to represent "Suspected diagnosis of gallstones" along
> the following lines:
> DiagnosticStatement and isAbout only Gallstones and hasQuality some
> Suspected
"suspected" of the DiagnosticStatement? Or is it of the Gallstones. If
the gallstones don't exist you have the same problem as you identify -
nothing for the suspected to inhere in. If they do exist, then they
are not expected. So I would omit the suspected. All diagnoses are
suspicious in the same sense.
Focusing on the gallstones, this is more thorny. There are two
approached that have been discussed. First that the relation is to a
type rather than an instance. I'm not particularly fond of that as I
don't like considering types as independent of their instances. The
other approach is to say that the gallstones that the diagnosis is
about aren't necessarily the ones in the patient. If the patient turns
out not to have gallstones then the diagnosis is about some
previously found gallstones not in this patient but about this
patient.
Other thoughts:
1) The Diagnosis isn't about gallstones, but about the symptoms and signs.
2) Diagnosis of gallstones: DiagnosticStatement and isAbout some
(Patient or (Patient and has_part some Gallstone))
-Alan
Stefan Schulz
> The thing the information artifact is about is a realizable entity,
> and we can say that this realizable entity is realized only in heart
> operation.
> Plan of heart operation is concretized as (is_realized_by only heart operation)
is_concretized_as some (is_realized_by only heart operation) ?
Then you say that plans are always concretized by sth (what ?) that is
not necessarily realized.
> Not sure I like the "suspected" - doesn't really work. Is the quality
> "suspected" of the DiagnosticStatement?
> Or is it of the Gallstones.
> If
> the gallstones don't exist you have the same problem as you identify -
> nothing for the suspected to inhere in. If they do exist, then they
> are not expected. So I would omit the suspected. All diagnoses are
> suspicious in the same sense.
one is very relevant in medicine, and therefore in the medical record
because the latter justifies much more an invasive or expensive
procedure than the first one.
> Focusing on the gallstones, this is more thorny. There are two
> approached that have been discussed. First that the relation is to a
> type rather than an instance. I'm not particularly fond of that as I
> don't like considering types as independent of their instances.
types such as prevents(A,B), the use of the universal quantifier is a
way to achieve the same effect.
> The
> other approach is to say that the gallstones that the diagnosis is
> about aren't necessarily the ones in the patient. If the patient turns
> out not to have gallstones then the diagnosis is about some
> previously found gallstones not in this patient but about this
> patient.
to refer to a gallstone instance would be to refer to "all gallstones"
(i.e. the scattered, mereological sum of all gallstones in the world).
This exists, just as the patient exists. The question is then whether
the two objects overlap. But this is a rather quaint solution.
>
> Other thoughts:
>
> 1) The Diagnosis isn't about gallstones, but about the symptoms and signs.
So this statement must be represented.
> 2) Diagnosis of gallstones: DiagnosticStatement and isAbout some
> (Patient or (Patient and has_part some Gallstone))
DiagnosticStatement and isAbout some Patient
and therefore all kinds of diagnostic statements
Personally , I would prefer
Diagnosis of gallstones equivalent to
DiagnosticStatement and isAbout only
(Human and locationOf some Gallstone))
It could be made more precise by introducing a relation isVerifiedBy
then:
Diagnosis of gallstones equivalent to
DiagnosticStatement
and isAbout some human
and isVerifiedBy only Gallstone
This allows the definition of
False diagnosis of gallstones equivalent to
DiagnosticStatement
and isAbout some human
and not isVerifiedBy some Gallstone
and
Confirmed diagnosis of gallstones equivalent to
DiagnosticStatement
and isAbout some human
and isVerifiedBy some Gallstone
-Stefan
Alan Ruttenberg
<stsc...@uni-freiburg.de> wrote:
> Alan,
>
>>
>> The thing the information artifact is about is a realizable entity,
>> and we can say that this realizable entity is realized only in heart
>> operation.
>> Plan of heart operation is concretized as (is_realized_by only heart operation)
>
> is_concretized_as some (is_realized_by only heart operation) ?
yes
>
> Then you say that plans
plan specifications - information artifacts
> are always concretized by sth (what ?) that is
> not necessarily realized.
Yes. A plan is a realizable entity and all realizable entities are
such that they many not be necessarily realized. The reason we leave
the concretization type unnamed is because we don't want to get in to
what sort realizable entity the plan is. Out of scope - more cognitive
science.
>> Not sure I like the "suspected" - doesn't really work. Is the quality
>> "suspected" of the DiagnosticStatement?
>> Or is it of the Gallstones.
>
> It is of the Diagnostic Statement.
>
>> If
>> the gallstones don't exist you have the same problem as you identify -
>> nothing for the suspected to inhere in. If they do exist, then they
>> are not expected. So I would omit the suspected. All diagnoses are
>> suspicious in the same sense.
>
> The distinction between a highly speculative diagnosis and a probable
> one is very relevant in medicine, and therefore in the medical record
> because the latter justifies much more an invasive or expensive
> procedure than the first one.
OK. So I wouldn't consider it a quality of the diagnosis, but rather a
part of it.
>> Focusing on the gallstones, this is more thorny. There are two
>> approached that have been discussed. First that the relation is to a
>> type rather than an instance. I'm not particularly fond of that as I
>> don't like considering types as independent of their instances.
>
> I think that in a formalism that does not allow predications over
> types such as prevents(A,B), the use of the universal quantifier is a
> way to achieve the same effect.
>
>> The
>> other approach is to say that the gallstones that the diagnosis is
>> about aren't necessarily the ones in the patient. If the patient turns
>> out not to have gallstones then the diagnosis is about some
>> previously found gallstones not in this patient but about this
>> patient.
>
> But then the diagnosis would be about other gallstones? The only way
> to refer to a gallstone instance would be to refer to "all gallstones"
> (i.e. the scattered, mereological sum of all gallstones in the world).
> This exists, just as the patient exists. The question is then whether
> the two objects overlap. But this is a rather quaint solution.
Well, yes, quaint might be a reasonable thing. I think information is
always about something - it doesn't come out of thin air. Dipping
(dangerously) into psychology, it may not always be the intention of
the communicator to convey what it is about, not even consciously
available to the communicator.
>> Other thoughts:
>>
>> 1) The Diagnosis isn't about gallstones, but about the symptoms and signs.
> But the physician writes something down like "suspected gallstones".
> So this statement must be represented.
I've been puzzling over how to think about canonical representations
and perhaps this problem is related. For example, with FMA, which is
"canonical" what is the relation of a heart to the canonical heart?
Here we might have the same schema - canonical pathologies, for
example gallstones. Then the assertion is one of relating the patient
to the canonical pathology.
>
>> 2) Diagnosis of gallstones: DiagnosticStatement and isAbout some
>> (Patient or (Patient and has_part some Gallstone))
>
> This would subsume
> DiagnosticStatement and isAbout some Patient
Good point. Equivalent, actually.
> and therefore all kinds of diagnostic statements
>
> Personally , I would prefer
>
> Diagnosis of gallstones equivalent to
> DiagnosticStatement and isAbout only
> (Human and locationOf some Gallstone))
Leaves the possibility that the DiagnosticStatement is about nothing.
> It could be made more precise by introducing a relation isVerifiedBy
>
> then:
> Diagnosis of gallstones equivalent to
> DiagnosticStatement
> and isAbout some human
> and isVerifiedBy only Gallstone
>
> This allows the definition of
>
> False diagnosis of gallstones equivalent to
> DiagnosticStatement
> and isAbout some human
> and not isVerifiedBy some Gallstone
>
> and
>
> Confirmed diagnosis of gallstones equivalent to
> DiagnosticStatement
> and isAbout some human
> and isVerifiedBy some Gallstone
What your proposal for the definition of isVerifiedBy?
-Alan