new term - diagnostic criteria

[GoogleCodeExporter]

Consider adding a class labeled ‘diagnostic criteria’ to represent specific 
sets of criteria that are agreed upon by the medical community to be sufficient 
and/or necessary conditions for diagnosing a specific disease, syndrome, or 
disorder.  It should be a subclass of ‘data item’.

The class 'diagnosis' is good, but it does not allow the inclusion of 
diagnostic criteria in connection with specific diseases (or syndromes or 
disorders).  In developing the diagnosis portion of the Neurological Disease 
ontology (ND), I ran into problems because I wanted to include a list of all of 
the relevant diagnoses and to define them, in part, in terms of the currently 
agreed upon criteria for making a diagnosis of such and such type.  At least 2 
problems arose as a result:

1.  Not every diagnosis must be made (much less will, in practice, be made) 
according to the accepted diagnostic criteria for justifying the relevant 
assertion.  Thus, including diagnostic criteria in the definition for a 
diagnosis is inappropriate.

2.  In creating children of 'diagnosis' for neurological diseases, it seems 
appropriate to make them all sister classes with no (or at least very few) 
sub-hierarchies; however, the inclusion of diagnostic criteria in the 
definitions lends itself to the assertion of a clear set of sub-hierarchies of 
diagnoses.  This conflict has generated negative feedback on this portion of 
ND.  Here is an example of this conflict:

'diagnosis of mild cognitive impairment'
     'diagnosis of possible Alzheimers disease'
          'diagnosis of probable Alzheimers disease'
               'diagnosis of definite Alzheimers disease'

These diagnoses are made by clinicians all the time.  According to the 
diagnostic criteria perspective, a diagnosis of definite AD is a diagnosis of 
probable AD, which is a diagnosis of possible AD, which is a diagnosis of MCI.  
This is because the criteria for possible AD includes all of the criteria for 
MCI plus something extra.  The criteria for probable AD includes all of the 
criteria for possible AD plus something extra.  Similarly, the criteria for 
definite AD includes all of the criteria for probable AD plus something extra.  

Yet, from a diagnostic assertion perspective, this sort of hierarchy is wrong 
because the diagnostic statement, "Patient X has probable AD." is an instance 
of the class 'diagnosis of probable AD', but is not an instance of the class 
'diagnosis of possible AD'.

Another problem with the diagnostic criteria perspective is that it allows a 
diagnosis of a disease (e.g. AD) to be a subclass of a diagnosis of a syndrome 
(e.g. MCI).  This seems to imply that at least some diseases are syndromes, 
which is wrong.

CONCLUSION: These considerations show that diagnoses and sets of diagnostic 
criteria aren't the same thing (which, in hindsight, should have been obvious 
to me in the first place).  Thus, I think that OGMS should add the class 
'diagnostic criteria' or perhaps 'set of diagnostic criteria' as a subclass of 
data item.  My initial proposal for its definition is:

"A set of criteria that has been agreed upon by the medical community to be 
sufficient, necessary, or sufficient and necessary to diagnosis a patient whose 
clinical picture satisfies these criteria with the disease, syndrome, or 
disorder in question."

A further question regards how we should represent whether an instance of a 
diagnosis was made according to the relevant set of diagnostic criteria or not. 
 Should this be a relation of a specific type that sometimes holds between 
instances of diagnoses and instances (or perhaps classes) of diagnostic 
criteria?  Does there need to be a further distinction between a diagnosis that 
is made as a result of considering the accepted diagnostic criteria and a 
diagnosis that is made without regard for the accepted diagnostic criteria but 
is, nevertheless, in compliance with the accepted diagnostic criteria?

Original issue reported on code.google.com by Alexande...@gmail.com on 13 Jun 2012 at 8:11

[GoogleCodeExporter]

I have something similar, which I have called "diagnosis guideline" and defined 
as "A directive information entity that specifies how to establish a diagnosis 
based on clinical findings". I also have "diagnosis criteria" which are 
described in the guideline (for example, generalized hives is a major 
dermatologic criteria for anaphylaxis)

I subclass diagnosis guideline with (in my case) "Brighton case definition", 
which is the parent of "Brighton case definition for generalized convulsive 
seizure as an adverse event following immunization". The concretization of this 
case definition is realized during a "generalized convulsive seizure as an AEFI 
according to the Brighton collaboration" process.

I am still unclear as to how to relate the clinical finding (e.g. "hives 
finding") with the actually process "hives" (is that needed?) and then how to 
assert this is a major dermatological criteria for anaphylaxis and need to work 
on this part of AERO (maybe something along the lines of "major dermatological 
criteria for anaphylaxis according to Brighton" as a subclass of clinical 
finding?). If this sounds like something close to your topic I would be happy 
to talk further and see if we can find common grounds to accommodate both our 
needs.

Original comment by mcour...@gmail.com on 3 Aug 2012 at 6:22

[GoogleCodeExporter]

Note from Barry: 
Remember all terms need to be singular nouns. So either
diagnostic criterion
or 
(e.g.) set of diagnostic criteria

Original comment by mcour...@gmail.com on 7 Aug 2012 at 5:21

[GoogleCodeExporter]

Melanie, I think that our goals and general approaches are very much in line 
with one another and I welcome the opportunity to work together to find a 
common solution to our needs.

So far I have just created the class 'set of diagnostic criteria' along with 
about 60 subclasses of it for specific disease diagnoses in the Neurological 
Disease Ontology (ND).  I made 'set of diagnostic criteria' a subclass of 'data 
item' and defined it as "A set of criteria that has been agreed upon by the 
medical community to provide sufficient, necessary, or sufficient and necessary 
warrant for a clinician to reach a particular diagnosis concerning a patient 
whose clinical picture satisfies the relevant criteria."

I considered creating the class 'diagnostic criterion', which would serve the 
same purpose as your 'diagnosis criteria'.  The general idea was that each set 
of diagnostic criteria would have specific diagnostic criterion classes as its 
members.  Difficulties arose though when I attempted to represent things such 
as:
  -  how a single criterion can belong to multiple sets,
  -  the degree of importance of a particular diagnostic criterion for a particular set of diagnostic criteria, and
  -  the fact that a single criterion could have a different degree of importance in each set that it is a member of.

Having a single criterion belong to multiple sets is relatively unproblematic 
as long as multiple parentage is permitted, but what exactly is a diagnostic 
criterion?  Keeping in line with 'set of diagnostic criteria' as a 'data item', 
it seems that 'diagnostic criterion' should be some sort of data item as well.  
But a criterion qua data item is not the sort of thing that a patient can have. 
 A patient can, however, have a disorder or a genetic makeup or undergo a 
pathological process, and these things might be said to satisfy particular 
criteria.  This is one approach that might be used to partially represent the 
connection between a patient and a physician's diagnosis of that patient, but I 
think that this is more complicated than is needed.

Rather, I am currently leaning toward an approach that represents a diagnostic 
criterion as a relation instead of a class.  There are several reasons to 
prefer the relation approach including:

  1.  Some criteria will concern qualities that inhere in a patient (e.g. elevated temperature), some will concern processes that the patient (or part of the patient) participates in (e.g. vomiting), others will concern material entities that are part of or contained in the patient (e.g. a virus or bacterium), and so on.  The fact that these entities encompass a variety of disjoint classes makes it difficult to represent them all as the same sort of thing (i.e. a diagnostic criterion).  This sort of concern is also a major issue with representing signs and symptoms as classes.  I believe that the current plan is to represent 'sign' and 'symptom' as relations in OGMS for exactly this reason.

  2.  Creating the relation 'has criterion' (or something similar) to link a particular set of diagnostic criteria to instances of the processes, disorders, qualities, etc. of patients, eliminates the need to create additional classes in the ontology.  If we choose to create a class 'diagnostic criterion', then we will be forced to create a diagnostic criterion class for each criterion and will still need to connect these criterion classes to the instances in patients that satisfy them (perhaps by using a 'satisfies' relation); however, by using the 'has criterion' relation', there will simply be the set of diagnostic criteria, the instances in the patient, and the relation between them.

  3.  We can create subclass relations of 'has criterion' to represent the degree of importance that a particular feature has for each set of diagnostic criteria.  The relations 'has major criterion', 'has primary criterion', 'has supporting criterion', 'has minor criterion', 'has secondary criterion', 'has necessary criterion', or 'has sufficient criterion' might be candidates for representing this.  For example: 
  -  'set of diagnostic criteria for norovirus' has_major_criterion 'vomiting'
  -  'set of diagnostic criteria for pregnancy' has_supporting_criterion 'vomiting'
In this way, we can enhance the information that is represented in an ontology 
without needlessly multiplying the number of classes or assertions that it 
contains.

Original comment by Alexande...@gmail.com on 17 Aug 2012 at 9:45