Page tree

OBSOLETE

Introduction

A discussion paper from the Content Managers Advisory Group, entitled Discussion Paper - Allowance of Extensions to Modify Core Content has been submitted to SNOMED International management to resolve issues surrounding the interpretation of what national extensions are allowed to modify in the International release of SNOMED CT.  The discussion paper outlines a number of proposed clarifications to the interpretation of what is allowed and details the requirements for those changes.  This response proposes a series of policy statements that explicitly state what can and cannot be done to content of the International release by national extensions.

While this response is primality focused on the needs of national extensions, the general policies discussed here will apply to all extensions, regardless if they are directly dependent on the International release or another edition.  

Background

The primary concern expressed is the interpretation of Clause 4.1 of the SNOMED CT® AFFILIATE LICENSE AGREEMENT, which states: 

“Subject to clause 2.1.4, the Licensee may not modify any part of the SNOMED CT Core distributed as part of the International Release or as part of a Member’s National Release.”

Clause 2.1.4 states:

 “(The Licensor grants the Licensee, license to) modify the manner of formatting of the copy of the SNOMED CT Core distributed to the Licensee as part of the International Release or as part of a Member’s National Release”

 According to the discussion paper, there are two primary issues associated with this clause:

 "This restriction on modifications has been interpreted generally in two ways. 

  1. The RF2 distribution must not be modified, beyond appending additional rows. Overwriting data in the distribution files - such that an extension violates the append only model is not allowed (editor's addition). Additionally, the full history provided by the international release must provided, i.e. extensions may not omit anything;

  2. International components must not be modified.
    Overriding the international content through either:

    • the addition of new versions for international components within extensions; or

    • the addition of relationships to international concepts and changing their DL definition.

Restrictions covered by B have been shown to be impractical and prohibit proper quality terminology authoring."

Proposed actions allowed for National Extensions

The discussion paper proposes seven actions that national extensions may want to perform in the course of their content development and maintenance:

  1. Create new concepts.

  2. Fully define concepts they create.

  3. Classify terminology extensions.

  4. State additional IS A relationships against core (international) concepts.

  5. Retire (redundant) IS A relationships (not necessarily stated).

  6. Add additional defining (non-IS A) relationships to primitive international concepts.

  7. Retire content considered "inappropriate" - concepts, descriptions or relationships

The paper then goes into substantial detail justifying the need to perform each of these actions.  A request was forwarded to SNOMED International management for review and clarification of the interpretation of clause 4.1 and acceptance of the allowable proposed actions available to national extension managers.

This document attempts to clarify the feasibility of each of the proposed actions and provides justification for the acceptance or rejection of each, in order to make the interpretation of Clause 4.1 more clear.

Evaluation of proposed allowable actions for national extensions

Create new concepts

The current interpretation of Clause 4.1 with regards to creation of new concepts is that new extension concepts would necessarily be be subtypes of an international concept, and not be a supertype of an international concept following classification (if the extension does perform classification of its content).  This type of content, by its very nature would have no impact on the international content, as it would be distal in the hierarchy to all international content.  National extensions have found that, due to gaps in the current International release, it is often necessary to insert intermediate (i.e. non-leaf) concepts into the terminology to meet user requests as well as to provide additional navigational support.  These intermediate concepts may be either primitive or fully-defined.

In general. the identification of a need for an intermediate concept in the International release indicates a gap in content coverage.  Unless realm specific, requests for addition of the intermediate concept should be forward to SNOMED International for disposition by the editorial staff.  It is recognized; however, that there are instances where the addition of an intermediate concept may be required initially in an extension.  The primary motivation for the promotion of intermediate extension concepts is to support the needs for cross-border exchange of health data.  If intermediate concepts resided solely in national extensions, then countries would need to reconcile multiple extensions in order to get a full representation of the structured data they may receive.  Also, there is a potential issue with divergent definitions of similar content if it is solely maintained in an extension.

The creation of intermediate concepts necessarily results in changes to the relationships in the international terminology.  These changes can be of almost any type including: assignment of new parent(s), retirement of inferred or stated parents, inheritance of new/additional defining relationships, removal of inferred relationships by the classifier, etc. There are (at least) 2 ways in which an extension concept can become an "intermediate" concept in an edition:

  1. The creation of a new fully defined extension concept as a subtype of an international concept.  The extension concept's definition may result in it being classified as a supertype of an international concept. In this case, the transitive closure of the international edition stays the same, except for the addition of the new inferred |is a| relationships for the extension concept. The defining attribute relationships of the international subtypes stay the same; however, results in a redundant inferred |is a| relationship in the national edition.
  2. A new extension concept (fully defined or primitive) is created and used as the destination of a stated |is a| relationship from an international concept, meaning that the new extension concept is a supertype of an international concept. An example of this might be the creation of an "A or B" concept, where the relationships "A is a 'A or B'" and "B is a 'A or B'" are stated. In this case, the extension is changing the stated definition of an existing international concept (the subtypes), which is more likely to lead to significant consequences.

Proposed policy

SNOMED International recognizes that SNOMED CT is not always comprehensive in its terminological coverage in particular domains and that there exist "gaps" in the hierarchies, where concept coverage is not adequate for extensions.  It is proposed that intermediate concepts may be added to extensions under the following guidelines:

  1. The addition of the concept does not change the transitive closure for the International Release, except to add additional rows that include the new intermediate concept). The transitive closure of the national edition must be a superset of the transitive closure of the international edition. This necessitates the retention of all transitive IS A relationships from the international lease.
  2. No stated or inferred defining attribute relationship of international content is inactivated in the extension.
  3. Inferred |is a| relationships should only be inactivated due to redundancy, but should still represent a valid relationship in the transitive closure. See "Retire (redundant) IS A relationships (not necessarily inferred)". 
  4. No stated IS A relationships from a international concept to an extension concept is allowed.
  5. Additional inferred defining relationships do not create crossovers (i.e. subtypes with attribute values that are supertypes of the values represented by the same attribute of the parent)
  6. Additional inferred IS A relationships of subtypes do not represent a change in meaning as represented by the Fully Specified Name. 
  7. All extension concepts that are the supertype of an international concept, following classification, must be submitted for promotion to the international release as soon as possible..

Fully define concepts they create

Full definition of intermediate concepts will in many cases result in the reassignment of international relationship due to inferences assigned from the classifier (see next section).  Likewise, any subtypes that are inferred under this concept may have their inferred defintions changed (depending on the modeling pattern applied to the extension concept). 

Proposed policy

  1. The creation of intermediate fully defined concepts within an extension is allowed, contingent on compliance with the proposed policy for creation of new concepts above.

Classify terminology extensions

The ability to classify extensions is an essential part of leveraging the description logics represented in the International Release for the proper maintenance of extension content.  However, in some cases the results of classification may change the inferred relationships of existing international content.  

Proposed policy

  1. In cases where the classification results in an inferred IS A relationship from the core to become transitive(i.e. redundant). In cases where the classification results in an inferred |is a| relationship from the core to become transitive (i.e. redundant), this redundant inferred relationship from the core should not be inactivated, but should remain redundant in the release.

N.B. This policy has implications on the 'Expression Constraint Language's "childOf" and "parentOf" operators (which return the direct children/parents based on non-redundant |is a| relationships). Rather than relying on the inferred relationships in a release to determine the results of a childOf or parentOf constraint, a pre-processing step would be required to remove the redundant inferred relationships from the release.

State additional IS A relationships against core (international) concepts

Missing stated IS A relationships for international content is usually the result of the existence of intermediate primitive concepts that editors are unaware of.  These intermediate primitive concepts are under review, but the volume makes it unlikely that all will be identified and resolved in the very near future.  Identification of these missing stated relationships by national extensions provides an opportunity for incremental improvement of the international terminology.   

Proposed policy

  1. Addition of new stated IS A relationships to international content is not allowed.
  2. Where a needed IS A relationship has been identified, a request to SNOMED International for addition of this relationship should be submitted as soon as possible.  These will be handled as high priority by the editorial staff. 

Retire (redundant) IS A relationships (not necessarily stated)

Redundant IS A relationship exist in SNOMED CT where variations in modeling style have been used and hierarchically related terms have been assigned as stated relationships. These are considered internal quality errors in the and are being incrementally addressed through the SNOMED CT Internal quality improvement program.  Many of these are already identified (See: https://dailybuild.ihtsdotools.org/qa/  Pattern 1) and are being addressed by SNOMED editors as resources allow.  Redundant relationships that impact extension structure should be reported to SNOMED International for prioritized disposition.  

Proposed policy

  1. Inactivation of International release stated IS A relationships is not allowed. 
  2. Relationships identified as redundant and impacting extension content should be reported to SNOMED Int'l as soon as possible.   
  3. Where retirement of stated relationships are needed, a request should be made to SNOMED International to correct the error.

Add additional defining (non-IS A) relationships to primitive international concepts

The incomplete modeling of international content resulting in a plethora of primitive content is well-recognized.  However, correction of these deficiencies within a national extension can lead to a number of unintended consequences including: 1) the impact on the classification process and the potential change in meaning of descendant concepts (e.g. loss of subtypes) 2) eventual remodeling of the primitive concept in the International release could lead to conflicts between the national edition and the International release.

It is recognized that there may be cases where proper structuring of a national extension requires the full definition of currently primitive international content and that the immediate needs of an extension do not coincide with the timing of International releases. The potential dangers of content divergence in national extensions by allowing definition of international primitive content must be recognized by extension managers prior to undertaking such an effort. It must also be recognized that allowing the definition of international concepts to be changed in an extension makes cross-border, cross-organisation exchange/management/querying of health records more difficult than it already is.

Proposed policy

  1. When needed to support national extension requirements, the addition of defining |non- IS A| relationships to existing primitive international content is allowed, providing:
    1. the additions are consistent with the current International Machine readable concept model
    2. the additions are submitted to SNOMED International as soon as possible for inclusion in the International release

Retire content considered "inappropriate" - concepts, descriptions or relationships

It is recognized that there are components within the International release that are erroneous. These is a current issue tracker IHTSDO-724 - Retire inappropriate synonyms, that will address the presence of inappropriate synonyms. The existence of other components deemed inappropriate by a national extension are often context or realm dependent.  The inactivation of international release components within a national extension can result in substantial deviation from the International release and thus an alternative representation of SNOMED international content.  The integrity of SNOMED as an internationally consistent terminology would thus be compromised. 

Proposed policy

    1. Inactivation of components (concepts, descriptions, stated relationships) originating in the International release, by a national extension, is not allowed.
    2. Components in the International release identified as inappropriate for a national extension may be excluded from the national release through refsets.  Components in the International release must not be inactivated simply because they are deemed inappropriate for national use. 
    3. The identification of issues in the international edition that impact a national extension must be added to a Content Request to escalate its priority for resolution by the SNOMED International editorial team  

  


 

 

  • No labels

10 Comments

  1. I believe that it is essential that extensions to be able to modify behavior created by inheriting core content for many reasons. In stating this belief, I am not saying that extensions should be allowed to masquerade in a way that it is not clear the origin of that modification through use of appropriate identification of the author (organizational or otherwise) of a change, and how it relates to the original content. 

    Although I support the general ideas in this proposal (sanctioning the need for extensions to make changes), I do not believe this policy is sufficient because it is limited to just National Release Centers, we need a clarification in interpretation of what it means to modify the core that applies to all affiliate licensees. I also do not believe that the interpretation of the license agreement to mean that extensions cannot cause deviations from what the core content alone provide is accurate. The consequences of interpreting the license agreement to mean that extensions cannot make necessary changes to ensure proper functioning of derivative works is unfortunate. 

    Another part of the affiliates license agreement is that all affiliates are to be treated the same... Martin Severs made it very clear to me the reason why there is only one form of the license... Not one for vendors, another for providers, a third for a release center, and a forth for a private terminology provider. The affiliate license for a country is the same as the affiliate license for a vendor, or for a health-care provider. The idea that a national release center may make changes, that a different licensee cannot, is contrary to this idea. Policies need to apply to all affiliates equally. In particular, national release centers that do not care for patients, and do not indemnify users from harm that may result from the use of the release center's content, should not have special rights that are not available to organizations that actually provide care, and are at risk (financially and morally) for the correctness of the care they provide. 

    One example of necessary conditions where policy should require an affiliate to make modifications to the core is to fix any errors that they identify which may lead to type 1 or type 2 errors in a system using SNOMED. It seems kind of silly that on one hand, the policy is currently interpreted that you are not allowed to make changes, even to correct type 1 or type 2 errors, yet the affiliates license agreement does not indemnify against harm brought by an affiliate who might rigidly follow this bad policy. It seems to leave the organization providing care on the hook to either: 1. Violate policy and modify the core when errors are identified that impact their systems, or to 2. Knowingly deploy an unsafe system and risk patient harm and tort/malpractice liability. And all the other organizations in the chain of culpability have an indemnification clause that does not shield the provider. When called to testify at a trial concerning harm to a patient, would you want to say you acted according to scenario 1, or scenario 2 above? 

    Any affiliate should have the ability to inactivate a component in their extension, for example to inactivate an inaccurate stated relationship—such as an improper parent that might cause a decision support rule not to fire? This would be a a type 2 error (false negative), with a potential for ISO 62304 Class C consequences (Death or SERIOUS INJURY is possible), depending upon the nature of a the affected rules...

    Although escalated priority for resolving particular kind of issues is a reasonable policy, the time from reporting an issue, until it can be reviewed, prioritized, possibly acted upon, published, integration tested, and subsequently deployed, is significant—possibly epic. Allowing local action to serve the best interests of the patients—and the organizations and individuals that serve those patients—seems to be the only prudent policy. 

    I think a prudent policy is that all changes are allowed, but that the changes must be clearly kept distinct from the modules they extend. 

  2. With regard to the particular language in the license agreement, here are the key clauses I could find: 

    The Licensor grants the Licensee, subject to the terms of this License Agreement, a perpetual (subject to revocation in accordance with clause 5), worldwide, non- exclusive, non-transferable license for the term of this License Agreement to:

    ....

    2.1.2 create Extensions and Derivatives from the International Release and use and modify those Extensions and Derivatives; 

    ....

    2.1.4 modify the manner of formatting of the copy of the SNOMED CT Core distributed to the Licensee as part of the International Release or as part of a Member’s National Release; and

    ....

    4.1 Subject to clause 2.1.4, the Licensee may not modify any part of the SNOMED CT Core distributed as part of the International Release or as part of a Member’s National Release.

    4.2 Subject to any express and specific statement to the contrary in the documentation distributed as part of the International Release, the Licensee may not modify any of the documentation (including Specifications) or software (unless provided in source code form) distributed as part of the International Release.

    ....

    So from the above, the specific license restriction is that a licensee "may not modify any part of the SNOMED CT Core". I do not interpret this to mean that the integration of the core with other extensions has to behave in the same way as the core would without those extensions. Rather I believe that the license is silent with regard to what an extension may do, as long as the distributed core is clearly identified, and not modified. 

    1.  "the license is silent with regard to what an extension may do"

      This was probably the initial driver for starting these discussions, and most (if not all) extensions are already doing changing core relationships (whether or not anyone realises it). 
       Correction of "errors" aside. Prohibiting changes to core relationships, effectively prohibits extensions fluently creating defined content. (limits extensions to primitive concepts, or the lucky sufficiently defined one, that doesn't subsume anything).

      Generally I probably almost agree with the proposed policy above, however - I'm not sure limiting the changes to "addition of non-|IS A|" provides guaranteed protection. There's lots of ways, damage can be caused....

      Also agree with part b of the policy, but we also need to sort out the promotion process/tooling. Because at the moment it introduces a maintenance burden for extensions.

  3. Keith,

    Thanks for the insightful comments. I want to be clear that the primary purpose of this policy is to provide more flexibility for extensions (I have added a section in the intro on the scope to include non-national extension managers as well), as well as to provide guidance for extension to correct the various errors that extension managers have identified over the years.  I understand the concern for immediate correction or identified errors that impact patient safety and SNOMED International wants to provide as rapid a remediation as possible within its current operational structure. 

    To that end we have instituted some pretty robust SLAs that have improved our turnaround of content requests, such that non-project related content requests are being resolved within one editing cycle in over 98% of the cases for the last two releases.  With the proposal here to make patient safety errors identified within the terminology the highest priority, I would go out on a limb and say that 100% of those would be addressed in a single editing cycle.  However, this is totally contingent on extension providing us with these identified errors as quickly as possible.  

    This document is in its early incarnation and comments such as yours and those we hope to receive from other extension managers will hopefully help us to gain consensus quickly, as this is an issue that has been too long unaddressed.

    1. Jim, 

      I recognize that the policy is going toward more flexibility, and I fully support this direction. And, I am very happy to learn of improved SLAs and success getting content addressed within a single editing cycle.

      I'm still concerned that a single editing cycle commitment (6 month turnaround) is still insufficient, and even improving to 3 months, 1 month, or similar will not fully address the need, and recognition that organizations may need to take action, and to distribute that action within their medical center, their larger enterprise, or to their customers and collaborators may be required. 

      I think requiring notification for these circumstances is wise, prudent, and proper, but prohibiting direct action in these circumstances is an overreach of policy that is not supported by the license. 

    1. Matt

      So sorry, but is there any alternative way that Dion's paper can be posted? I've tried a number of times and a number of different ways to follow this link without success.

      Kind regards

      Ed

  4. Matt,

    Thanks for this.  We are scheduling a full two hour discussion at the London meeting for this topic and hope to resolve many of the contentious issues related to the conflict between the needs of extensions and maintenance of the integrity of the International release.  As an extension manager myself, I appreciate the concerns that have been expressed.

  5. Jim, Matt,

    Like Ed, I have tried unsuccessfully to get access to this paper. Please could someone who can get access post it to this group directly or upload it to this page.

     

    Many thanks

    Paul

  6. I have added a copy to the EAG agenda.  You should be ble to get to it from there