Medical terminologies like SNOMED CT often provide codes for frequently co-occurring associations of findings and disorders, such as syndromes or diseases with sequelae. The current release of SNOMED CT still lacks a principled solution for representing these concepts, which was the reason for the IHTSDO project group "Event, Condition, Episode" to elaborate a well-founded approach based on criteria of formal ontology. The group analysed complex SNOMED CT terms and proposes a simple solution, which draws on the interpretation of findings, disorders, and diseases as clinical life phases. Co-occurrence, temporal relatedness and causal relatedness were represented by distinct modelling patterns in OWL-DL.
3.1
11.9
44.3
3.1
1.5
9.0
1.8
6.6
1.5
1.3
7.7
12.1
6,497
Σabs
951
11,974
1,627
257
259
408
162
243
353
584
23,039
While SNOMED CT is increasingly incorporating
principles of applied ontology and provides a description logics
(DL)
' after ' ' and ' ' caused by ' ' ue to' 'with' 'without' Σ ALL .0 .1 .6 .3 .0 .0 .1 .2 .3 .0 .0 .0 2.5 3.0 2.3 1.5 2.3 5.8 1.1 1.8 1.1 1.2 2.9 .0 .1 .0 .0 .0 .0 .0 .0 .0 .0 .2 .0 2.7 1.9 .0 .0 .0 .0 .0 .2 .0 .0 1.0 .6 4.7 9.6 .5 .0 5.4 5.8 .4 3.9 .4 .1 3.0 .0 1.3 3.1 1.2 .0 .0 .3 .0 .4 .0 .0 .5 nected to Neoplastic disease using due to. Another example is given by the concepts Dermatomycosis associated with AIDS (disorder) and AIDS with dermatomycosis (disorder), which appear to be duplicates. Whereas the former one uses the relation associated with for establishing a connection with the concept AIDS, the latter one is represented as a subclass of AIDS. This motivated the project group Event, Condition, Episode Model (ECE) of IHTSDO1, the organization that maintains SNOMED CT, to conduct a thorough investigation of this phenomenon and to suggest a solution that is in line with current principles of ontology development in SNOMED CT. 2
The ECE group decided to limit the scope of the
investigation to the SNOMED CT hierarchy Clinical Finding /
Disorder, following IHTSDO's current strategic directions in
the content development process
Typology and ontological analysis Our analysis yielded four patterns of coordinative expressions in the SNOMED CT Finding / Disorder hierarchy, as shown in Table 3.
Further analysis focused on the following questions:
• Which are exactly the entities that are denoted by the
concepts under scrutiny?
• Which temporal relationships have to be distinguished?
• What does causality mean and how is it linked to
temporality?
According to Schulz et al. (2012b), we interpret all finding /
disorder codes as Clinical Situations or Clinical Life Phases
(we will use the latter term and illustrate it by the suffix
"CLP") i.e. a patient's life phase during which a clinically
relevant condition is present. For instance, the SNOMED
CT concept EncephalitisCLP denotes the class of processual
entities of the type Life phase, in which some encephalitis
process is present in any temporal instant covered by this
life phase. Accordingly, HerniaCLP denotes the class of
processual entity of the type Life phase, in which the material
disorder Hernia is fully present. The advantage of this
interpretation is that we do not have to deal with hierarchies of
entity types of different ontological categories under the
same umbrella. To this end, BTL2 provides the defined
class Condition – the disjunction of Material object,
Disposition and Process
'has condition' some X The relation 'has condition' in BTL2 holds between a life phase and an entity that is constantly present during this life phase. Independently, we have to look at temporality, where we need to clarify what "following" and "co-occurring" exactly mean. In BFO2 we find the relation 'is preceded by', which is defined as relating two processes, one of which ends before the second one begins.
A commonly accepted framework for describing temporal
relations is
X
Y Y Y X Y
Y Y X
Y
X takes place before Y Looking at the examples where we had asserted cooccurrence, we agreed to interpret "x co-occurs with y" as the disjunction of "x starts y", "x during y", "x finished y", and "x is equal to y". Let us take the example Hay fever with asthma (Fig. 2). We have three Clinical life phase entities: 'Hay fever with asthma'CLP, 'Hay fever'CLP, and AsthmaCLP. The possible temporal patterns result from any combination of Fig.2 left hand side with Fig.2 right hand side. All temporal instants of Hay fever with AsthmaCLP temporally coincide with some instant of 'Hay fever'CLP and some instant of AsthmaCLP. Finally, we will have a look at causality. Notwithstanding past and on-going philosophical debates about its nature, we consider the notion of causality as a primitive predicate, which is essential for medical reasoning and decisionmaking. Whether y follows x accidentally or because it is caused by x is seen as fundamentally different. There are important temporal implications of causality. It is a truism that an effect cannot precede its cause, or conversely, an effect has to follow its cause. Referring to the Allen calculus, "x causes y" would then be only compatible with "x takes place before y", "x meets y", "x overlaps y" as well as with (switching the arguments x and y) "y during x" and "y finishes x". All these relations have in common that the starting point of x precedes the starting point of y. 3.2
Proposal of modelling patterns In the following, we will propose modelling solutions for 'X with Y' concepts in SNOMED CT for the frequently encountered clinical patterns in Table 3. The simplest modelling approach for representing X with Y concepts in SNOMED CT is: 1. Both X and Y are co-occurrent, but with no causality between X and Y.
XCLP with YCLP, both simply asserted as co-occurring, and no
known causal/manifestational relationship implied:
XwithYCLP equivalentTo XCLP and YCLP
XwithYCLP denotes the class of life phases that are
characterised by the full presence of both the conditions X and Y:
XwithYCLP equivalentTo 'Clinical life phase' and
'has condition' some X and
'has condition' some Y
It can be shown that both definitions are equivalent, by
logical transformation or by a reasoner like
It can easily be shown that the same applies for complex
clinical life phase types with more than two conjoints, e.g.
in the case of the Tetralogy of Fallot
XcausedByY equivalentTo X and 'is caused by' some Y However, according to our interpretation, SNOMED CT disorder concepts are clinical life phases and the underlying conditions are not or are only indirectly available2. We therefore axiomatically extend the notion of causation and allow that a clinical life phase is causally related to another clinical life phase. To express this, we use the SNOMED CT relation 'due to'.
XcausedByYCLP equivalentTo XCLP and
'due to' some YCLP This is a simplification of the correct representation, which should be (using the BTL2 relation 'is caused by'): *XcausedByYCLP equivalentTo 'Clinical life phase' and 'has condition' some (X and 'is caused by' some Y) Due to the problem of referring to clinical conditions in SNOMED CT, in the modeling pattern we propose, 'due to' connects two CLPs that are related by the fact that the first one has a condition that is caused by a condition that defines the second one. E.g., all instances of 'Disorder of optic chiasm due to non-pituitary neoplasm'CLP imply an instance of 'Non-pituitary neoplasm'CLP (which implies an instance of 'Non-pituitary neoplasm'). We consider this approximation as sufficient for the reasoning services required. 3. X temporally follows Y. This does not specify that X is due to Y, although causality is frequently implied XfollowsYCLP equivalentTo XCLP and
follows some YCLP As mentioned before, the BTL2 relation is preceded by excludes the Allen relation overlaps. We argue that the relation follows should include overlaps, as it is common in medicine. E.g., Postvaricella encephalitis might include cases in which the varicella infection has not ended at the inception of the complication, viz. Encephalitis. 4. X is due to Y and both X and Y are co-occurrent Here we propose a combination of patterns 1 and 2: XdueToCooccurringYCLP equivalentTo
XCLP and YCLP and 'due to' some YCLP 2 In the case of fully defined disorder concepts, the condition would correspond to the combination of location with morphology, inside the role groups • • It looks uncommon that the same class YCLP appears both as a superclass and a class related via the relation due to. Taking the example 'Hernia with intestinal obstruction'CLP: All life phases of this type are both Hernia life phases and Intestinal obstruction life phases, and they are related, additionally, to a second Hernia life phase (which is a different one but is assumed to refer to the same hernia object). This second life phase is, actually, one that precedes the inception of the complication, in this case the intestinal obstruction. 3.3
Special cases There are other cases that we have excluded from our typology, but which, nevertheless, deserve consideration: • Terms of the type 'Abscess of urethral gland due to Neisseria gonorrhoeae'. Here, the right hand side of the particle "due to" denotes a material agent, not a process. In BTL2 this would be expressed by the relation 'has agent' and could be modeled in the following way (note that in this case, the relation 'has condition' is a paraphrase of the role group relation in SNOMED CT): XwithAgentACLP equivalentTo XCLP and
'has condition' some ('has agent' some A) Associativity. It can be shown that the following rule always holds and can be easily reduced to pattern one: XwithYCLP and ZCLP equivalentTo XCLP and YwithZCLP equivalentTo
XCLP and YCLP and ZCLP E.g. 'Diabetes mellitus with hyperosmolar coma'CLP superficially appears to be an X with Y pattern, but is more appropriately an example of X with Y with Z (Diabetes mellitus, Hyperosmolar state, Coma). There is no nesting.
On this basis, more complex chained sequences according to pattern four are possible using the 'due to' relationship. This might be represented as: 'Diabetes mellitus with hyperosmolar coma'CLP equivalentTo 'Diabetes mellitus'CLP and ('Hyperosmolar state'CLP and
'due to' some 'Diabetes'CLP) and (ComaCLP and 'due to' some 'Hyperosmolar state'CLP) 4
Table 4 gives an overview of the relations used in the proposed models of our approach and their mapping to the Allen relations. As the BTL2 relation 'is preceded by' does not allow overlap, it seems too strict. We prefer the relation follows, which makes the minimal assumption that the beginning of Y is later than X. This is also the assumption of the causality relation, which appears as a subrelation of follows. The proposal to include an overlap pattern for temporally following again blurs the distinction between pattern two and pattern three. This might be acceptable if we do not want to distinguish sequelae from other types of complications. When investigating definitions of sequelae under the concepts Sequela (finding) or Sequela of disorders (disorder), we found chronic or residual conditions that are complication of acute conditions that occur after the acute disease or injury phase. Sequelae can also be the result of the treatment of the primary condition. There is no time limit on when a late effect can occur; the residual condition may come directly after the disease or condition, or years later. This is a little vague in terms of whether the inciting condition is still present when the complication commences. In case there is a requirement to represent sequelae (late effects) as distinct from e.g., immediate complications, it might be worthwhile to define sequelae as not overlapping with their cause, and for this case to indeed use the BTL2 relation 'is preceded by'. √ √ √ √ √ √ √ √ √
Y is due-to
X √ √ √ √ √ Limitations identified and resulting tasks will be addressed by the ECE working group in the future: • To evaluate if the proposed patterns are generic and can be applied throughout SNOMED CT, especially to concepts in the Event and Procedure hierarchies. • To prove theoretically and empirically that the proposed patterns do not produce unexpected classification results, especially as a consequence of the simplification by asserting 'due to' between situations and not between the underlying conditions. • To check the impact of the new models on classification time. • To extend the approach to negated conditions ("without" ) by scenarios that extend DL expressiveness or that represent negations as primitives. The impact on reasoning behaviour will also be investigated. • To propose adjustments to the SNOMED CT naming conventions in the light of the new model. 5 CONCLUSION AND FURTHER WORK The proposed patterns have been prototypically implemented in SNOMED CT and have achieved better semantic clarity and consistency in terminology creation and maintenance. The formal analysis of temporal and causative relationships has been proved to be useful for determining the patterns.