Recently, we have posted the request for feedback on enhancements to logic profile for SNOMED CT.  Since two key features “hidden GCIs” and “property chain” will be implemented in the new concept models for drugs and substances, we would like to receive feedback from users who are maintaining these content in their SNOMED CT extensions.  The following are highlights for the planned changes in the July 2018 international release:

  • Substance and its modifications will be represented as siblings
    Currently, a substance subsumes its modifications in substance hierarchy. However, these “overloaded” |IS A| relationships have caused incorrect inferences. The new substance model will remove these overloaded relationships. For example,  

                387307005 |Calcium carbonate (substance)| and 387217003 |Calcium lactate (substance)| are no longer subconcepts of 5540006 |Calcium (substance)|. They will be represented as siblings of |Calcium|.

  • New attribute |is modification of| for modeling substances
    A new attribute |is modification of| will represent relationship between a base substance and its modifications to replace the overloaded IS relationships. For example,
          |Calcium carbonate| : |is modification of| = |Calcium|
  • Property chain for |has active ingredient| and |is modification of|
    The property chain will support correct inferences for drugs after the above changes in the substance hierarchy. For example, a |Product containing calcium| will still subsume a |Product containing calcium lactate| though these substances are not subsumed anymore. Please note, without this property chain, a product containing base substance will NOT subsume any product containing its modifications. A single OWL expression for property chain will be added in the OWL reference set of international release. 

                SubObjectPropertyOf(ObjectPropertyChain(|has active ingredient|  |is modification of|) |has active ingredient|)

  • Role of product is defined separately
    Role of product will be defined by hidden GCIs in separate axiom in the OWL refset. This will provide flexibility for defining a product that can have multiple roles in the international release or national extensions. 
  • Represent “Only” by active ingredient count
    Active ingredient count will be used for presenting product containing only one or more active ingredient. This will ensure fast classification and support of concept modeling for majority of product content.

The enhancements are represented by the OWL refset that will eventually replace the stated relationship file. The changes will have impact to extension content that is depended on current hierarchical relationships between base substance and its modifications in the international release. In particular, users who are relaying on DL classifiers need to use the OWL refset in order to get correct inferences.  Therefore, these extensions need to assess the proposed concept model changes and update their tooling to avoid potential impact and unexpected classification results. 

Please provide feedback on the Confluence page or in the document for SNOMED CT logic profile enhancements at: 

https://confluence.ihtsdotools.org/display/mag/SNOMED+CT+Logic+Profile+Enhancement

SNOMED CT Logic Profile Enhancements

We would appreciate if you can provide any comments before the October business meeting or attend the meeting to discuss with us.


Comments

    Add new comment