-
We have 30 concepts explicitly mentioning deficiency that might be impacted.
-
We don’t see these changes as having any impact on our users at this point in time.
-
Having discussed the proposed changes, we had some thoughts and questions.
It’s mentioned that avoiding the observable entity hierarchy is the simplest way to deal with the issue. However, we have had several comments from Swedish users that they would like more content in the observable entity hierarchy so not avoiding it, if it is the most logical way to model, would be a double win for us.
Concerning the difference between clinical findings and disorders we agree that it sometimes is complicated to understand the difference. According to the editorial guide disorders “Necessarily have an underlying pathological process”. It seems the underlying pathological process is often implied when modelling disorder concepts but to differentiate the disorder from the finding, could this instead be made explicit in the modelling by expanding the use of the attribute Pathological process? And if we define a deficiency disorder as a pathological state of health due to / caused by a deficiency wouldn’t using “due to”, or “associated with”, be logical to connect the clinical finding to the disorder? And how would disorders caused by a quantitative deficiency be separated from disorders caused by a functional deficiency of the same substance if “deficiency of” was used?
Reaching into the judgement/interpretation ‘words’ found in << 272520006 |Degree findings| and promoting a couple of them to the status of attribute names seems entirely arbitrary and a break from any earlier (however hazy) SNOMED design approach.
For each, what is the ‘scope’ of the selected attribute names? The briefing note asks Members whether ‘deficiency of’ will impact on extension content but fails to define ‘deficiency’. I’d guess there is some convergence on meaning (in English, anyway – I cannot comment on any non-English translations), but what is to be the fate of other, words/concepts beneath 272520006 |Degree findings| with similar meaning? This figure shows the synonymy (purple) and hypo/hypernym (blue) relationships between these words as declared in WordNet 3.0 (opening the svg with the ‘Raw’ option reveals Synset tooltips). Along with this screen-grab of a home-brewed UpSet plot showing sample synonymy assertions within SNOMED CT’s own data, we can see that ‘deficiency’ is just one of several words of very similar meaning. Should, for example occurrences of ‘insufficiency’, ‘lack’ or ‘reduced’ be considered for modelling in the same way as ‘deficiency’ or will we have, by design, multiple different ways depending on what ‘thing’ is deficient?
And what of the other words beneath << 272520006 |Degree findings|? Are we to expect a progressive introduction of similar attributes (‘slowness of’, ‘enlargement of’, ‘lowness of’ etc.) as these requirements are encountered? And if not, why not?
The alternative seems to be a doubling down on SI’s growing reliance on the ‘interprets/has interpretation’ pattern (indeed this is actually still the proposal for the ‘within reference range’ variant). IMHO this (outside, perhaps, its original use in taming ‘ability’ findings) has never been a good solution, not least because of a persistent arbitrary split of semantics between each selected observable and corresponding values (the same finding could be defined by a whole spectrum of observable + value pairs, making equivalence detection inconsistent and unreliable.
However the discussion provoked by this briefing note also highlights how poorly the modelling to date has addressed the challenge of ‘property types’. Property type are important, indeed they should be an inescapable element of ‘property observations’ (as can be seen in the modelling of the recently developed LOINC/SNOMED data). However property types are largely neglected in SNOMED CT itself: of the 33000+ findings modelled with at least one ‘interprets’ relationship in the November 2025 data, only 2200+ refer to an observable which has a modelled property type (using just 77 observables, modelled with just 12 property types).
Returning to ‘deficiency of’ and ‘excess of’: these are also silent on the property type of interest. Surely it is, for example, the ‘mass concentration’ or ‘catalytic activity’ of a chemical or enzyme that is ‘deficient’, not just the ‘substance’ itself? The proposal is also heavily skewed towards biophysical requirements, and therefore excludes psychological and social ‘deficiencies’ etc. – suggesting that they are handled just fine by the currently available modelling options.
As such, my preferred solution would be closer to that outlined by John Snyder @jsnyder, augmented with direct modelling (to the finding itself) of the ‘property type’ that is ‘deficient’ or ‘in excess’ (along with, where needed, other epistemically neutral features currently only found in the observables model) rather than, if at all, indirectly via an ‘interprets’ relationship. Such an approach would be isomorphic with other direct finding=>substance associations, simpler to explain and allow much simpler query constructs for analysis purposes.
This approach would require a long-overdue and proper analysis of what has_interpretation=values would be suitable for each property type. Currently (AFAIK) there are no rules that link these two roles. Dynamic or interdependent model constraints have been looked at in the past but do not form part of SNOMED’s published formalism (although they may be available in internally-maintained templates) and perhaps now is the time to explore how this can be done. Failing to tame this area of modelling leaves us with yet more arbitrary judgements, including last year’s appearance of 758637006| Anatomic location | as the property type for 628001000005106 | Position of placenta in uterus |. This, in turn, allows the finding 445122007 | Low-lying placenta | to be modelled with 363713009 | Has interpretation | = 62482003 | Low |. The same property type is used to model many cancer-specific observables (such as 1285429003 | Anatomic location directly invaded by nephroblastoma |) – surely this latter group would expect to take actual anatomical structures as values, not relative values such as ‘high’ or ‘low’?
I would like to thank everyone who comments on this topic. The comments will be used to develop a revise briefing note taking into consideration the comments made here.
Apologies for my tardy reply! We will be affected somewhat, as we about have 40 concepts in our extension related to deficiency. However, we would be glad with improvement on this content area.
When translating, we have struggled often with unclear distinctions between:
- deficiency (finding)
- deficiency (disorder)
- deficiency syndrome (disorder)
Much like the short stature discussion Matt mentioned. Often a valid distinction can be made, but because it cannot be modelled, plenty of inconsistencies and concept duplication has crept in. Any remodelling of this branch should include a plan to resolve these inconsistencies. So I agree with the comments above that this proposal requires further work.
A revised briefing note addressing comments recieved was reviewed and endorsed by the Editorial Advisory Group January 27, 2026.