During the promotion of content related to phenotypes from the Community Content area to the International release, it was observed that there were a large number of "Abnormality of X (body structure)" concepts. All of these were modeled with an associated morphology of 49755003 |Morphologically abnormal structure (morphologic abnormality)|. In some cases, however, in some cases it is unclear from the use of the term "Abnormality" in the FSN whether these actually represent a structural or functional abnormality.
Examples where it is clear that the term represents a structural abnormality:
897432004 Abnormality of limb bone (disorder)
896855001 Abnormality of tip of finger (disorder)
Examples where it is unclear whether the abnormality refers to structure or function:
896984006 Abnormality of renal collecting system (disorder)
897434003 Abnormality of pyramid of medulla oblongata (disorder)
Is it necessary for these more vague terms to include the word "structural" or "functional" to the FSN to designate the specific type of abnormality? If so, should it be editorial policy to add the word "Structural" to the FSN even in cases where it can be reasonably assumed to be structural in nature? We have a number of existing concepts that use this terming pattern:
54873004 |Congenital structural abnormality of orbit proper (disorder)|
128599005 |Structural disorder of heart (disorder)|
15980831000119102 |Congenital structural abnormality of right eyelid (disorder)|
However, this is not universally applied where the abnormality is implied to be structural.
12 Comments
Jeremy Rogers
I lean toward the FSN always including "structural" whenever the underlying modelling is going to limit the nature of the abnormalities that can be classified below it to those with a morphological/structural abnormality present. For consistency of subsequent lexical indexing, this naming convention should therefore also apply even when it would be reasonable (or clinically conventional) to discount the possibility of the semantic scope of the term also including "functional abnormality in the presence of grossly normal morphology".
There are, however, many cases where clinicians use the terms "X abnormality" or "Abnormality of X" in the sense of there being either a structural or a functional abnormality, or both simultaneously. For example the term "cardiac abnormality" by convention covers all of valvular stenosis, ventricular hypertrophy, arrhythmia and congestive cardiac failure. A "pancreatic abnormality" could be a tumour or a functional problem secreting digestive enzymes. "Bladder abnormality" can mean any of a tumour, stone, delayed emptying or detrusor instability.
Where such abstractions exist but are agnostic to whether the abnormality is morphological or functional, and where they are in common clinical use, the case for them to persist in SNOMED's taxonomy as groupers is strong (though how to model them may be an 'interesting question'). But they raise a rlated naming convention question: should they be left as "X abnormality", as per usual clinical terming convention, or should the FSN at least be explicitly "functional or structural abnormality of X"? My preference would I think be the latter.
Jim Case
WRT your last paragraph, thre are times when a functional abnormality can occur either in the presence or absence of a structural abnormality, so a top-level concept that embraces all of the scenarios, i.e. structural abnormality without functional abnormality, structural abnormality with functional abnormality, or functional abnormality alone would suggest an FSN of "Structural and/or functional abnormality of X". This can be modeled using GCIs with the functional abnormality represented by an INTERPRETS/HAS INTERPRETATION relationship group. However, functional abnormalities are often currently classified in SNOMED CT as clinical findings, rather than disorders. An example of an exception to this is 102882004 |Abnormal placental secretion of chorionic gonadotropin (disorder)|.
Does this mean that current Abnormal X/Abnormality of X disorders are necessarily structural in nature? I think this would be true only if the current modeling represents them as such. There is often implied meaning based on the intent of the term as it was added to SNOMED CT, but many of these are legacy concepts where the intended meaning may not be available. One can only assume based on a combination of the FSN and the modeling and the observation that no one has complained that there is an incongruence between them.
Matt Cordell
I think we should keep the concepts the non-specific grouper concepts. I worry in an effort to make everything absolutely unambiguous SNOMED will end up not reflecting real world language and move away from open world semantics.
But like Jeremy says, modelling this distinction such that they subsume as expected is an interesting problem...
I guess it depends if :
There's a mix of findings and disorders here too by the looks..
There's concepts like
Sometimes it looks like Abnormal just means "disease.
Jim Case
I agree that if a concept is modeled in a way that it does not specify that structural or functional abnormality, that leaving it as a grouper is appropriate and we should try to model these using GCIs to allow for both structure and functional concept to classify under them; however, as noted above, many of these vague groupers are modeled as disorders, whereas functional abnormalities are most often modeled as clinical findings. This would lead to a large wholesale change to the taxonomy that might not be of benefit to users.
Jeff Pierson
Overall, I agree with Matt's comments.
Coming from the genomics space, there is likely an intented context for these phenotypic terms, which not only appears to be structural but could be congenital.
While SNOMED could massage the FSN's and tweak the modeling, I think it would be helpful to understand how these terms have been used and the intented meaning.
Can we get more background information or is there already a source?
John Snyder
I agree with the comments stated thus far by Jeff, Matt, and Jeremy.
To expand the scope, abnormality could also represent a change in content and/or volume (e.g. 9630001000004104 | History of semen abnormality (situation) |. While I think it could be difficult to identify all the different permutation implied by the use of the term "abnormality" in the FSN, more fully specifying the definition as being structural, functional, etc. may help to mitigate the amount of ambiguity in the largest majority of the content.
Feikje Hielkema-Raadsveld
Browsing through the terms in our national hospital thesaurus of diagnoses, there are close to 500 concepts that mention afwijking (abnormality), but only a handful terms that specify that the abnormality is functional and 0 that specify it is structural. After struggling to form an opinion, I have only formed more questions:
Jim Case
The concepts in question originate from HPO where many of the term simply state "abnormality of X" without an accompanying text definition to qualify whether the term represents either structure of function. These were modeled into SNOMED terms by an external group that added the ASSOCIATED MORPHOLOGY = "Morphologically abnormal structure" relationship. In some cases, there are terms in HPO that act as parents for structural and functional subtypes.
As Matt Cordell suggested, where there is no HPO definition, we could keep these agnostic as to structure or function by removing the morphology relationship, but then they may become equivalent to "Disease of X", where such a concept already exists.
For those concepts where it is clear from the definition or the structure of the HPO taxonomy that a concept refers to a structural abnormality, should SNOMED align with the HPO description? They use the following patterns: "Abnormal morphology of X", "Abnormality of X morphology", Structural abnormality of X" (rare), "Abnormal X structure" (rare)? Or should SNOMED adhere to its own terming (i.e. Structural abnormality of X)?
SNOMED has not used terms to distinguish between structure and function abnormalities in the vast majority of cases (e.g. 253339007 |Right atrial abnormality (disorder)|; parent 128599005 |Structural disorder of heart (disorder)|) but has modeled these almost exclusively as structural.
To answer Feikje Hielkema-Raadsveld question, because HPO uses "Abnormality of X" terms occasionally as groupers and occasionally as structural concepts, we do now need to make the distinction where it is known. WRT to existing content, there are a little over 400 Abnormality of X terms that would need to be reviewed in light of the promotion of HPO content.
WRT to Jeremy Rogers comment on the agnostic terms, as agnostic groupers concepts would need to be termed as "Structural and/or functional abnormality of X" to avoid the notion of an exclusive OR. Or they could be left vague, which would work if all of the existing content was made explicit to the underlying model where an associated morphology was present.
Jim Case
In looking at these concept in more detail in HPO, they are nearly all referenced as morphological abnormalities. When the concepts were created in SNOMED, this notion was left out of the FSN: e.g. SNOMED - Abnormality of cochlea (disorder); HPO - Abnormal cochlea morphology.
There are a few that have grouper concepts representing both structure and function, and these will be identified. WRT to this particular issue, we have a number of concepts representing congenital structural abnormalities and as HPO is focused on phenotypes and genetic diseases, it would be safe to assume that these would most often be congenital; however, they do not always manifest at birth.
We will be looking to add these concepts as "Structural abnormality of X" for precision for HPO terms that refer to morphology. For those HPO concepts that group both functional and structural abnormalities, they will be left in that form, e.g. HPO - Abnormality of the gastrointestinal tract has subtypes "Abnormal gastrointestinal tract morphology" and "Functional abnormality of the gastrointestinal tract"
Monique van Berkum
Sorry for joining this conversation so late. I agree with Jeremy that if an "abnormality" is only structural it should be specified in the FSN and otherwise "abnormality" means "structural or functional". I also agree with Matt that the modeling will be an interesting problem and that sometimes it looks like Abnormal just means "disease".
For example, with respect to Jim's statement that, "For those HPO concepts that group both functional and structural abnormalities, they will be left in that form, e.g. HPO - Abnormality of the gastrointestinal tract has subtypes "Abnormal gastrointestinal tract morphology" and "Functional abnormality of the gastrointestinal tract":
If we consider the 3 concepts:
Only “Structural abnormality of gastrointestinal tract” can be sufficiently defined in SNOMED (using |Associated morphology|: |Morphologically abnormal structure (morphologic abnormality)|).
How will we sufficiently define Abnormality of X (e.g. gastrointestinal tract)?
How will we sufficiently define Functional abnormality of X (e.g. gastrointestinal tract)?
Interestingly, HPO often equates “Abnormality of X” with “X disease” . For example, in HPO, "Abnormality of the gastrointestinal tract" has synonym Gastrointestinal disease. https://hpo.jax.org/browse/term/HP:0011024.
IMO, this won't be easy to resolve. In SNOMED, there is also significant existing overlap between the use of anomaly, abnormality and malformation. See 276654001 |Congenital malformation (disorder)| which has synonyms of Congenital anomaly and Congenital abnormality and is a child of Congenital Disease.
Jim Case
This was in interesting issue when initial testing on models to support the more general Abnormality of X concepts as it exposed issues with other areas of the hierarchy. In these cases there are actually three aspects that need to be considered:
Structural abnormality of X
Functional abnormality of X due to a structural abnrmality
Functional abnormality of X (i.e. functional disorder) in the absence of a structural abnormality.
Currently, SNOMED conflates the last two of these in many cases, but it is clear that they are not the same. While I think it is possible to make the distinction and actually create sufficient defintions for many of these, it will require additional testing, which is out of scope for this issue, which is focused on terming
Jim Case
In the short term, it appears that the group agrees that if the meaning of a concept is a structure abnormality, then the FSN should specify that. For the majority of the HPO concepts in question, it is clear from either the synonyms associated with the terms or the text definition whether the concept refers to a structural abnormality.
While it is recognized that this is not consistently applied to existing content in the International release, the scope of the issues with existing content is outside of this specific discussion topic. It is recognized that clear editorial guidance will be needed to be applied to existing content and a ticket will be raised to develop that guidance and apply it to current content. Modeling of grouping concepts will require additional evaluation and discussion due to the inconsistent representation of functional abnormalities as findings or disorders.