Procedure without procedure

As a followup on the "Disorder without disorder" discussion.  We received the following from the UKTC:

 I think we’d be clinically OK to map them to the vanilla # or intracranial injury going forward. The whole ‘without mention of’ construct is deeply unsatisfactory anyway – it just means that the coder examining the paper record, not the patient, couldn’t find any evidence that the patient also had e.g. a spinal injury, and therefore not that they definitely did not. So strictly clinically its an intrinsically ambiguous statement. I would have thought we could/should therefore retire them all to AMBIGUOUS with a solitary MAYBE to the relevant vanilla #/injury code which should currently be its parent:

This is better than retiring them to e.g. LIMITED with the usual WAS-A to their prior parents, because some of them have several parents not all of which are really appropriate or useful candidate substitutes for reasoning. Also better than retire to INAPPROPRIATE with REPLACED_BY, because historically that construct is used rarely and when it is its quite close semantically to DUPLICATE/SAME_AS, which definitely doesn’t apply here. The relationship between the inactive concept and its active substitute is definitely one of a lossy map, and so we’d need to be using one of the lossy map flavours of historical relationship, which in turn limits which inactive concept bucket we can put them in.

For the existing historical data to be mapped to SNOMED, capturing forever within the computer processable coding that the patient did not also have either a spinal cord lesion or a skull fracture or an intracranial injury is IMHO unlikely to be of practical value. There are very few inference algorithms that specifically check for the an explicit statement that some clinical phenomenon is absent. Absence of clinical phenomena is more usually tested by confirming the absence of any relevant presence code, not the presence of a specific absence code.

And going forward for new data capture, I don’t think anybody is going to object to only having the basic # codes to use. If for some reason they want to also explicitly record that some related complication or comorbidity was NOT present, then they need to be made to use a more generic mechanism for doing that anyway as a second coded entry.

Comments from other members of the CMAG related to the topic of "Disorder without disorder" indicated that they had a few of this type of concept in their extensions.  Most felt that the content more appropriately belongs in the situation hierarchy.  Plan is to inactivate and recreate IF they are requested from members. 

Jim

If I might comment, as an observer, on three statements made above:

1.  "...There are very few inference algorithms that specifically check for the an explicit statement that some clinical phenomenon is absent. Absence of clinical phenomena is more usually tested by confirming the absence of any relevant presence code, not the presence of a specific absence code..."

2.  "...Most felt that the content more appropriately belongs in the situation hierarchy..."

3.  "...they need to be made to use a more generic mechanism for doing that anyway as a second coded entry..."

The author of statement 1 may well be right (not sure whether it's you or a quote from the UK) when referring to automated reasoning, but human clinical decisions (I *thought*) are often very strongly influenced by explicit significant negation. Supporting its representation isn't a luxury - it's a requirement, and I'd argue that at times it's a requirement to support indivisible 'X without Y' statements.

Now things may be muddied by the presumed classification 'without mention of' origins of the # concepts, but if the FSN is the 'arbiter of meaning in SNOMED CT' then to my reading these concepts mean WITHOUT, not 'without mention of'. If statement 2 holds, there is precedent  - concepts (with comparable FSNs) in which the presence and absence components are formally represented using the mechanism that SNOMED CT currently offers:

433807000 | History of occlusion of cerebral artery without cerebral infarction (situation)

371622005 | Elevated blood pressure reading without diagnosis of hypertension (situation)

If the proposal is to inactivate the "# WITHOUT LOC/spinal injury" content...

a. What will be the fate of existing X WITHOUT Y situations?

b. Do you know that the suggestion in statement 3 is acceptable to the SNOMED CT community. Language isn't well-behaved, and the value of some clinical utterances is dependent upon them being preserved as single units. How would 'History of occlusion of cerebral artery without cerebral infarction' be fragmented whilst preserving its combined semantics? Have SNOMED CT's users agreed that they are content to fragment their phrases in the way suggested?

Kind regards

Ed

I would agree with Ed on this. I think that the "plain vanilla" construct can be interpreted as "without mention of" while explicitly stating the absence of a condition has a different and arguably clinically necessary meaning. I am attaching a paper of Stefan's. See the section on exclusion. Another mechanism for representing exclusion possible with the current SNOMED description logic and outside of the situation hierarchy is what he refers to as reification i.e. cerebral artery infarction with absent cerebral infarction. This would require the creation of many negated primitives so it is likely not that practical.

1472-6947-8-S1-S9.pdf

Bruce

Bruce and Ed,

Thanks for your comments.  There were a number of reasons that we (the EAG) had recommended the inactivation of these concepts, noted in the minutes of the last meeting.  Due to the issues that these concepts cause us, I think it is important that we have a definite solution, i.e. the status quo is not acceptable.  I will add this back to the EAG agenda for the next call.  My current opinion is that the very limited use of these terms by members may not justify their continued maintenance.

One area that I think we agree on is that terms using "without mention of" are essentially indistinguishable from their parent.  I don't think I would take the step that Ed suggests that what they really mean is "without". In the open world, no statement does not mean absence.

As for the proposed existing fate of X without Y (situation), there are currently 15 concepts that fit that pattern, 7 of them are "unilateral X" concepts that are scheduled for inactivation, so there is not a big impact there.

We should also consider Procedure X without procedure Y as there are many use cases especially for the radiology domain where without contrast needs to be specifically specified (118 such concepts in SNOMED CT).

Bruce

I missed that. I should have scrolled up.

Thanks both for the responses.

I've looked over the related minutes and still don't see a clear message here. Certainly sampling the table above only a minority have ever carried a description (& never the FSN) which uses the "without mention of" construct. However based on a judgement that these are classification-derived, "without mention of" is what these concepts actually mean, and therefore:

1. these concepts will be inactivated.
2. a plan has been suggested to "... recreate IF they are requested from members".

In step 2, what would you expect to be requested - 'X without Y' concepts or 'X without mention of Y' concepts? 

If you actually mean 'X without Y' concepts (and I presume you do) then the  concepts could have been left alone in the first place, since the FSN form 'X without Y' is apparently acceptable.

Is there a step missing? Would you plan to change the FSNs of the above set to 'X without mention of Y' at the time of inactivation so that future users will better understand the decision making process/sequence, and more easily distinguish them from superficially similar future 'X without Y' requests?

Ed

Commenting on "X disorder without Y disorder" here related to the discussion within the CMAG. 

Has there been consideration to simply making sure "X disorder" is represented in the clinical findings content and then representing the absence of "Y disorder" as a statement in the situation hierarchy.  This would allow post-coordination of "X disorder without Y disorder" for those who require the combination.  From a interface terminology perspective we would simple map to both "X disorder" and "absence of Y disorder" which supports the various use cases.  As well, this allows the capture of other negative clinical findings that may be deemed to be significant.   

Thanks! Jeff

The majority of X absent concepts are in the findings hierarchy rather than the situation hierarchy. If we think it is appropriate to keep these as findings then Jeff's suggestion might work.