Proposal for inclusion of “lateralized” content to the International release of SNOMED CT

In Canada, we welcome this change of direction to include laterality in SNOMED CT core for more than the body structure.

I think the modularization of such content would make sense for those who would not want to maintain such content if not used.

On the other hand, If this is the approach selected, clear communication of optional content will be required as well as implementation guidance provided, so that implementers are made aware that lateralised content is available and maintained separately from 'main core' content and they are guided in the 'how to' implement this so there are no glitches, especially when using extension content.

The other challenge is the amount of changes this decision will require on existing content.

Can we have an idea of the extent of changes that will be required on current content? This will most likely require prioritisation not only for current content but for things that are pending to be added as per the related tracker.

What will it mean for project where there is usage of content that currently represent a lateralised concept? I think this requires written documentation as well.  

Another example of the kinds of impact this may have: in the FHIR resource Condition there is a bodySite attribute which allows lateralized Body structures. If Body structures no longer are lateralized, an additonal attribute would have to be created to allow SNOMED CT to be used in this way.

I have discussed this in the NRC in the Netherlands. I did not yet mixed myself in the discussion, because our opinion is that laterality does not belong in a precoordination, so we won't use it in a precoordination. But it does have impact on for example data-analysis, they have to transform the precoordinations to postcoordinations. I do understand from the paper that in postcoordinations the laterality still is related to bodystructures, which I will encourage. 

We do think when laterality will be used in precoordination, we agree with the preferred option in the paper.

Some additional remarks/questions:

• The executive summary would provide a clearer overview of the content if it contained something like 'The committee proposes to extend the domain of the laterality attribute to include clinical findings & procedures'
• Why use screenshots from Protegé instead of the Compositional Grammar?
• If this will be implemented, I think you should supply more information about the consequences with respect to data analysis. Explain when and how the transformation should be performed?
• You certainly need to present the new definition of the laterality attribute more clearly, e.g. in a separate textbox. Domain, range, type
• Is it to become a transitive property?

For the record, this is the position I articulated during our telecon earlier this week.

1) The proposed solution is not ideal, but seems to be a reasonable trade-off among the various options.

2) NRCs should be provided ample time and proper guidance to retrofit their content

3) The proposed solution should be evaluated in terms of authoring cost compared to e.g., generalizing lateralized anatomical entities

4) Quality Assurance processes should be put in place throughout this important content migration

I'm finally getting around to providing a response from down under on this.

When the proposal was floated in October I missed the revised modelling element, and thought it was just about relaxing the restriction of including pre coordinated content.
I think allowing laterality is good*, but the revised modelling is problematic. We've had feedback from stakeholders that existing implementations, particularly those that use the existing model to postcoordinate will be impacted. Ofcourse, there's always always going to be impacts - but this one seems more like a shift of the problem?
 
I'd understood the "acceptance as required" to be include "precoordinated lateralised anatomy"  as required, not creation of >30,000 new anatomy sites.

My view of the perils of precoordination are more when it's applied automatically/systemically, and you end up with things like "Atomic power plant malfunction in watercraft, water skier injured" or "Poisoning by carbon monoxide from aircraft whilst in transit, without accident to aircraft, occupant of spacecraft injured" (actual concepts - I'm not even sure what sort of scenario that second on is). Anyway, in these cases, it seems a bunch of terms were generated according to some pattern for all possible permutations, sometimes resulting in nonsense. Nobody wants that.

The counter argument is when precoordinating too much, unless it's done comprehensively you can end up with "gaps".

When drawing a line on precoordinated content - where is it drawn? This issue is just about left/right. But there's all sorts of boolean variants - open/closed, chronic/acute etc. And many more that aren't boolean. e.g. The myriad of agents that could be causative agents for infections, poisonings or allergies....

As it is, I don't think the existing model is "broken" - lateralizing anatomy - so no need to change it.
I'm not sure of the maintenance concerns, "Left thumb" could be fully defined, would this assist.
And gross anatomy is a fairly finite (and comparitively small) universe , so that puts some limit things.
(I haven't caught up on the latest of the anatomy proposal, so apologies if that explains things more)

I think option 2 (which also supports option 1 as postcoordination) is preferable. And again, just as required.

The proposed option (4) I think might be useful for other areas where precoordination is an unbound risk.
Things like:  Nucleic acid assays - where there may be variants for all the organisms, and no model exists (no precedent) for fully defining something like "Candida species DNA" - and getting accurate classification. But that's beyond the current proposal. 

Matt,

thanks for your comments.  We have gotten a fair amount of comment that asks us to re-evaluate Option 2 over Option 4.  As we stated at the meeting in October, there was not intention of creating new lateralized content algorithmically for hierarchies that use FINDING SITE, just, as you suggested, when it is needed.  However, if a "disorder of right X" is requested for a lateralized structure, we felt is was consistent to then add the "Disorder of left X".  The main concern is whether we should preemptively add laterlaized anatomy concepts and if so, how many of the potentially 50-60K new concepts should we add in advance.  

I am going to be providing a more comprehensive list of pros and cons of options 2 and 4 in London.  Look forward to more heathly discussion.