Inactivation of Active Duplicate Descriptions in the Same Hierarchy

While we agree that this content has issues, we disagree with the proposed solution, which addresses only the symptom—validation failures in the managed service—rather than the root cause (duplicate concepts). These validation failures should be seen as indicators of deeper problems, not something to bypass. Simply appending additional text (such as semantic tags) to satisfy validation rules undermines the purpose of a preferred term.

As noted in the brief, these duplicates are the product of replicating EDQM content in SNOMED CT - resulting in the creation of disjoint subhierarchies rather than maintaining and updating existing content. This may have been done for convenience, avoiding delays to the drug model project.

A more detailed response submitted via email.

"...These validation failures should be seen as indicators of deeper problems..." - I agree.

I have previously raised a concern with the (at that time relatively 'new') 736474004 | Has dose form intended site attribute and value range, stating:

"I understand the justification for the attribute ‘has_dose_form_intended_site’, but why it needs its own set of ‘intended site’ values is a mystery. Taking ‘urethra’ as an example, SNOMED CT now has three non-comparable ways of referring to the urethra in the context of administering a medicine:

For the procedure 432907007 | Administration of substance via urethral route (procedure) we have:
410675002|Route of administration (attribute)|=90028008|Urethral route (qualifier value)|
405814001|Procedure site - Indirect (attribute)|=63202008|Structure of urethral lumen (body structure)|

And now, for the dose form of any suitably administered product:
736474004|Has dose form intended site (attribute)|= 738988002|Urethral (intended site)|

This latter value is unrelated to either of the values available in the act of administration, thus weakening any SNOMED CT value proposition in medicines administration checking."

I added:

"To SI’s credit devices will use body structures as the values of ‘has_device_intended_site’. This means that there will be a common value mechanism for site modelling.

For example, looking at 708621008 | Insertion of stent into urethra (procedure) we will have:
405814001|Procedure site - Indirect (attribute)|=13648007|Urethral structure (body structure)|

…and via the referenced stent’s device modelling:
836358009|Has device intended site (attribute)|= 13648007|Urethral structure (body structure)|

Both use the same class of value (body structures) and therefore checking for congruence between intended and actual administration might be feasible using SNOMED CT."

As Matt says, the validation tests detect superficial problems (matching terms) which in turn point to something deeper (conceptual redundancy or near redundancy). Simply suppressing the term duplication signals with distinct suffixes (which, I would argue, result in terms that are unlikely to be anybody's in-use 'preferred terms') fails to address more fundamental problems in the data.

Ed

I have also provided a response via email, on behalf of the NHS England Pharmacy Terminology Team who have reviewed this BN.

For visibility, a couple of questions we have asked:

Regarding the text:

‘Validation implemented into the authoring environment for SNOMED International and Managed Service extensions…’

Q. What was the basis for the validation being implemented? Was it due to an internal finding or prompted externally from a SI member?

Regarding the text:

‘…which may lead to incorrect implementation with the potential for resultant clinical risk.’

Q. Is there evidence of this leading to incorrect implementations, and if Yes, what happened?

Paul