Traumatic Injury of [body structure] (disorder) - v1.0

Yongsheng Gao ,

Could you add template language to this template?

Yongsheng Gao , now that I look at it, this one is restrictive in that it only allows for 1 RG so we can exclude complex cases and focus on the simple issues. That  might reduce the number of mismatches for the more general template as the last report I ran this morning had over 1900 mismatches.  

I am still concerned about the allowance for 2 DUE TO relationships as we do not have guidance on how these should be used and may cause inconsistency in modeling even though the traumatic event DUE TO is required.  We do not have any concepts modeled with 2 DUE TOs at this time that I am aware of.

Jim Case , Yongsheng Gao . I think I should create at a separate template for traumatic injuries due to procedures. This would have a due to traumatic event and some temporal relationship to the procedure. If we want to include a due to relationship to the procedure as well, then we would need two due to relationships. As an aside, we do have several concepts that have >1 inferred due to relationship.

Bruce

Bruce Goldberg  are any of those multi-DUE TO concepts modeled with two stated DUE TO relationships?  May just  be a by product of missing remodeling.

Jim Case There are 30 concepts with 2 or more stated due to relationships.

<<64572001 |Disease|: [2..*] 42752001 |Due to|=*

searchResults_AUTHORTEST-225

Yongsheng Gao,

Thank you for your analysis.  We are aware of these anomalies and are in the process of fixing them.  Having a mechanism to identify those that are not in the right place is very useful.  I have already remodeled pathologic fractures and am awaiting a review by the EAG to accept the proposed restructuring of fractures. 

419945001|Traumatic injury due to event (disorder)| is scheduled for inactivation as the vast majority of traumatic injuries will be modeled using an event, so this is no longer necessary.  Bites are currently subtypes of traumatic event, so right now we are getting some "duplicate" inferred parents.  Most of the traumatic injuries from accidents are being remodeled to classify under Accidental injury, which is now a subtype of Traumatic injury.

I would like to restrict the use of two DUE TO relationships except under very specific conditions which we would need to document in the editorial guide.  I have not yet worked on the Injury due to procedure area, but recognize the issues.

Yongsheng Gao, It is hard to find a single, consistent definition of traumatic injury but based on most  damage resulting from energy transfer from an excessive external physical force to a subject. Some sources limit the external forces to blunt and penetrating trauma while others include chemical, heat and cold exposure as well as absence of vital requirements such as anoxia.

We had discussed the modeling of 417746004 |Traumatic injury (disorder)| without adding a requirement for an ASSOCIATED MORPHOLOGY = <<37782003 |Damage (morphologic abnormality)|. This was to avoid the need for an additional RG that would be needed for traumatic morphologies that are not subtypes of 37782003 |Damage (morphologic abnormality)|.  This leads to an issue when trying to define non-traumatic injuries, which are current defined (using 417163006 |Traumatic AND/OR non-traumatic injury (disorder)|) with as ASSOCIATED MORPHOLOGY = 37782003 |Damage (morphologic abnormality)|.

My suggestion is that we bite the bullet and require the second RG to ensure that these non-damage morphologies classify under Traumatic injury.

I question what the intent of not placing "Injury due to procedure" under traumatic injury. The approach that we have taken for traumatic injuries, requiring the use of << 773760007 |Traumatic event (event)|, would result in those concepts modeled using a DUE TO Procedure not classifying under Traumatic injury.  If the intent is that trauma that occurs during a procedure is of a different type than other types of trauma, then we might need some external consensus on that perception.  There are a large number of concepts that refer to damage to various body structures during or after a procedure.  Currently, some concepts as simple as "needle stick injury" have classified under Traumatic injury for many years. 

In recent remodeling of complications of termination of pregnancy, we have approached this as DUE TO = Traumatic event; <<TEMPORALLY RELATED TO = Termination of pregnancy.  This results in the classification under Traumatic injury.  Would this be a reasonable general approach for concepts currently named Injury DUE TO procedure? i.e. make them Complications?

Jim Case I am confused when you refer to an issue when trying to define non-traumatic injuries, which are currently defined (using 417163006 |Traumatic AND/OR non-traumatic injury (disorder)|) with an ASSOCIATED MORPHOLOGY = 37782003 |Damage (morphologic abnormality)| and then state that we require a second RG to ensure that these non-damage morphologies classify under Traumatic injury. Did you mean that you would want to classify non-traumatic injuries under 417163006 |Traumatic AND/OR non-traumatic injury (disorder)|) and this would thus require a role group containing 37782003 |Damage (morphologic abnormality)| in addition to another role group containing the primary morphology? If so, this makes sense to me. 

WRT whether all injuries due to procedure are traumatic, I don't think so. Examples of non-traumatic injuries due to procedures that immediately come to mind are strokes due to anesthesia and necrosis due to an air embolism.

Bruce Goldberg,

Based on conversations you and I and Yong have had, we had remodeled Traumatic injury simply as a disease with a DUE TO traumatic event.  This enabled any morphology to classify as a traumatic injury, even if it was not a subtype of damage.  417163006 |Traumatic AND/OR non-traumatic injury (disorder) on the other hand is modeled simply as a disease with an ASSOCIATED MORPHOLOGY = <<damage. If, to make this consistent between traumatic injury and 417163006 |Traumatic AND/OR non-traumatic injury (disorder), we would need to add back the ASSOCIATED MORPHOLOGY of damage to traumatic injury, which would then require that we use two RGs to sufficiently define those concepts with a morphology that is not a subtype of Damage (which was what we had originally done).  If we do not do that then all we can do is remove the ASSOCIATED MORPHOLOGY relationship from 417163006 |Traumatic AND/OR non-traumatic injury (disorder), which simply makes it a primitive subtype of Disease and would be  nearly impossible to curate as we would not have a consistent way of determining what is an injury and what is not.  We discussed on EAG that many injuries that could be classified as non-traumatic, are in fact traumatic in some respects and non-traumatic is a negation that we cannot replace simply by using Spontaneous event. For example, a pathological hip fracture due to a misstep. 

So I am leaning towards keeping the 417163006 |Traumatic AND/OR non-traumatic injury (disorder) concept modeled as is (although the FSN is a bit misleading as I don't see how an injury can be both traumatic and non-traumatic at the same time, so it should be Traumatic OR non-traumatic injury).  We would also add ASSOCIATED MORPHOLOGY = Damage, back to Traumatic injury and then add the second RG where needed when the morphology is not a subtype of damage.

I am really sorry that this has gone back and forth, but with each new sub area of injury, we run into other problems that cause us to rethink prior decisions.

Why not model Traumatic or non-traumatic injury using GCIs?

Bruce Goldberg,

Are you suggesting:

This is a representation of the current modeling?  It seems inconsistent that non-traumatic injuries are modeled with Damage and traumatic injuries are not.  What I was trying to say in my comment above is that we should model all injuries as a mandatory <<damage and where the morphology of a traumatic injury is not a subtype of damage, and additional RG representing damage be added.  So if we are going to do GCIs, then I think it should look like:

This does result in some challenging classification results, which we may be able to resolve.  An alternative is to create an intermediate primitive grouper of "Non-traumatic injury".  

I was thinking of the GCI model in your 2nd screen shot.

That would work I think.  Testing requreid

I have added the GCIs and the results are acceptable. 

This is different from the GCI that I suggested. We only need to have a single GCI for 'due to' traumatic event. The condition associated with damage would be the normal axiom. In the first screenshot, the second GCI can be changed to a normal axiom. 

Then, we do not explicitly model traumatic injury by two role groups. The reason is that the second role group for 'damage' would be inferred from the GCI approach in the grouper concept. The key benefit is that we only need to model traumatic injuries by 'due to' traumatic event. The 'damage' would be inferred if the associated morphology is not a subconcept of damage morphology. If the morphology is a subconcept of 'damage', then the single role group is sufficient and the additional 'damage' role group would be redundant and won't be inherited.  

Yongsheng Gao ,

So are you suggesting that the ASSOCIATED MORPHOLOGY for the GCI be "Morphologically abnormal structure" but in the normal axiom it would be "Damage"?  Like this:

In testing this, without the notation of the normal axiom as an equivalent class, Traumatic and Nontraumatic injury loses all of its subtypes except for two.  All of the nontraumatic subtypes become subtypes of disease.  We also run into the problem where we have a non-damage morphology (i.e hemorrhage) that is specified as non-traumatic, e.g. 405576001 |Nontraumatic hemothorax (disorder)| and we have issues with traumatic injuries that do not specify any morphology, but do have a causative agent.  That one, however could probably be resolved by just adding a Damage morphology relationship.  In any case, the simplified GCI version for Traumatic and/or nontraumtic injury does not seem to give the expected result.  You can look at the result in https://authoring.ihtsdotools.org/#/tasks/task/QIJAN21/QIJAN21-533/edit

Note: By making the normal axion SD, it resolved the issues I was having and it looks much better now.  FInal modeling looks like: 

I did not add a morphology to the GCI for those cases where one was not specified as adding "Morphologically abnormal structure" did not seem to be worth the effort.

Bruce Goldberg ,

I think we have the basic traumatic injury issues worked out.  Since you have created a separate template for Injury DUE TO procedure, do we still need the additional DUE TO in this template as well as the DURING relationship?  Those seem to be specific to the procedure injuries and to my knowledge, none of the traumatic injuries (aside from those occurring during a procedure)  use anything but a DUE TO <<Traumatic event.

Yongsheng Gao , Jim Case

As brought up during today's authors call, not all traumatic injuries are due to a spontaneous event.

I am still having a problem with nontraumatic injury. In Jim's model a nontraumatic injury is defined as a disease with a morphology of damage (or a subtype of damage). If a nontraumtic injury concept has a morphology which is not a subytype of damage and is not due to a spontaneous event, then we are again stuck. We ,may need to reconsider an idea I had presented in the past of moving some existing morphologies under damage based on what is in the ICD-11 injury chapter.

Hi Yongsheng Gao.  An issue was raised that pathologic fractures which are widely considered to be nontraumatic do not occur spontaneously but rather are the result of an external force applied to diseased bone. The force may not be traumatic as pathologic fractures can occur with normal daily activities such as standing and would not result in a fracture of normal bone. A proposed model is to model these with the pathologic fracture morphology and no due to event relationship. Jim Case has sent out a poll to EAG to vote for one of three alternate models.

Yongsheng Gao , Based on Jim Case' s model for Traumatic or nontraumatic injury is this what you are thinking?

Yongsheng Gao,

I am not so concerned with trying to determine what is traumatic or non-traumatic given the approach.  If a concept states non-traumatic in the FSN, then we would certainly ensure that the parent would be the Primitive "Non-traumatic injury" parent.  If it does not state it, with the current GCI model anything that is damage would classify under "Traumatic and/or non-traumatic injury".  The reason for the default interpretation of fracture is due to the historical position of fracture in the taxonomy.  The use case suggested, when someone does not know if a fracture is traumatic or non-traumatic, to me at least, seems rare.  There was another use case suggested for spinal cord injuries caused by space filling lesions such as tumors or hematomas, but I am concerned that we would be going down the path of having to add the primitive parent to a very large number of concepts if we were to try and make it a policy that every concept that has a damage type of morphology must have "Non-traumatic injury" as a parent unless it is traumatic.  Another scenario is where the morphology is not a subtype of damage, for example, I would not add the non-traumatic injury parent to concepts that had a morphology of hemorrhage.

I was confused by the last sentence "

I would not add the non-traumatic injury parent to concepts that had a morphology of hemorrhage." but I see that in context you mean 

I would not add the non-traumatic injury parent to concepts that do not specify nontraumatic in the FSN that had a morphology of hemorrhage.

I have modeled Non-traumatic injury (disorder) in a task. Once approved and promoted it can be used to revise these templates.

<<64572001 |Disease|: 42752001 |Due to|=773760007 |Traumatic event|, 370135005 |Pathological process|=* returns 256 concepts.

I've just looked at this discussion, and the one over here (Re: 2020-09-09 Editorial Advisory Group Conference call ) And it seems like a lot of things are being conflated here, and unnecessarily complicated.

• "Traumatic AND/OR non-traumatic injury" - isn't this just "Injury"? What injuries aren't included in this set? Is this concept necessary? Smells like a grouper...
  (Though I've seen it used in relation to brain injuries, I suspect because some people might not intuitively consider non-traumatic disorders injuries? I'm guessing)
• The word Traumatic seems to be a red-herring and unhelpful. Most of these things are only "traumatic" because an injury occurred. The associated events themselves are not necessarily "traumatic events".
• "Traumatic" had generally originated from the morphology, but the discussion here seems to be introducing traumatic events too, and as a necessary attribute.
  23957004|Friction blisters of the skin| - is wearing ill-fitting shoes a traumatic event?
• A fracture is a fracture. Pathological fractures generally still require some force mechanism, the force required is just much lower (bump vs hit by a car).I'm not sure how you'd model it, but it seems like it would be the same morphology (or maybe subtype of), with perhaps an associated pathological disorder?
• Perhaps just have "Physical injury \[due to event\]". The event is optional, and NOT necessarily "traumatic".
   "Fractured rib", "Fractured rib due to car crash", "fractured rib due to sneeze"...

Ideally, the mechanism for these would predominately be captured in the information model.
A 44465007|Sprain of ankle| is only a traumatic injury because it's a sprained ankle, and it hurts!
Stepping on a ball the wrong way, a misstep off the sidewalk, playing touch football - are not in themselves traumatic events.

Basically, I think we're talking about here are Physical injuries, caused by internal/spontaneous or external mechanisms. Rather than "traumatic and non-traumatic injuries".

• 202964000|Non-traumatic rupture of Achilles tendon (disorder)| vs 275335003|Ruptured Achilles tendon - traumatic (disorder)|
  
• 35644004|Nontraumatic hematoma of testis (disorder)| vs 262922005|Traumatic hematoma of testis (disorder)|

Side note:

|Sprain of ankle| is a subtype of |Traumatic AND/OR non-traumatic injury|.
But only through it's ancestor |Sprain of ankle and/or foot| - courtesy of the morphology.
It's other parent, Traumatic arthropathy of ankle, is not...

Well argued, and I agree completely.

This over interpretation of traumatic injury has been at EAG and ECE as Bruce has explained and the Injury hierarchy is being remodeled to reflect a more restricted approach to the assignment of traumatic events to those disorders that only occur DUE TO trauma (e.g. open wounds) or those concepts that expressly state traumatic in the FSN.  Bruce is correct that 417163006 |Traumatic AND/OR non-traumatic injury (disorder)| should be renamed (and will be).  With the inactivation of traumatic abnormality morphologies, we have much more flexibility in managing which terms are specifically traumatic and which are agnostic to mechanism of injury.  Fractures have thus been moved out from traumatic injury to child of 417163006 |Traumatic AND/OR non-traumatic injury (disorder)|.  Aside from the general classification grouping of disorders as injuries, no other changes to inferred parents are being affected (other than improvements that come from more consistent modeling).   

Bruce Goldberg and Yongsheng Gao ,

Given the remodeling of Traumatic or non-traumatic injury, I think we can extend this template to support more complex cases by simply changing the cardinality of the RG from 0..1 to 0..n (in some cases there may be up to four RGs for lateralized traumatic disorders).  I do not know how to update the template language to accommodate this change but will make the change in the table.

Jim Case , Yongsheng Gao. I would like to get consensus on the FSN for traumatic injuries: Traumatic X morphology X vs. X due to trauma (e.g. Traumatic hematoma vs. Hematoma due to trauma?)

Bruce Goldberg,  I would think that in line with our terming in other areas, "X due to trauma" would   be the proper FSN, with "Traumatic X" as the PT.

Hi Jim Case, I have updated the template language according to the model. But there are a lot of misalignments in the report. For example,

283580009 Pellet wound of lower limb (disorder)

"Cardinality mismatch: N/A
Relationship Group mismatches:
A{ [I1] 116676008 |Associated morphology (attribute)| -> 59091005 |Open wound (morphologic abnormality)|, [I1] 246075003 |Causative agent (attribute)| -> 64867003 |Pellet gun missile, device (physical object)|, [I1] 363698007 |Finding site (attribute)| -> 61685007 |Lower limb structure (body structure)| }
A{ [I2] 42752001 |Due to (attribute)| -> 45224000 |Struck by missile (event)| }"

The above model is aligned with the template. I am not sure why it is reported as a mismatch.  Peter G. Williams Could you please have a look at the report? docs.google.com/spreadsheets/d/1nkfs_xNRVenDQd-p7iiTk4oxQ8xjvTM0nALt5Bf74jQ/edit

Hi Bruce Goldberg, I have updated the cardinality and its format. The cardinality for the main role group has been restricted to 1..1 for identifying misalignments. Could you please check the misalignment report and determine if further changes are needed for this template? 

Cheers,

Yong

Yong, please remind where the misalignment report is located?

Thanks,

Bruce