In reviewing some recent requests for amputation procedures, I have run up against a conundrum that I would like input on. There are currently several procedures representing amputations of lateralized body structures, but without specifying the precise laterality. Back in 2015-16 we inactivated concepts that specified "unilateral" procedures as being both ambiguous and a threat to patient safety if added to an EHR. While the existing procedures that do not specify laterality act as organizational groupers, they too seem to pose the same threat to patient safety.
We have received a number of requests for lateralized amputation procedures which are correctly classifying under these ambiguous groupers, but highlight the potential issue with these concepts. Example:
46028000 |Amputation of hand (procedure)|
895522007 |Amputation of left hand (procedure)|
895473002 |Amputation of bilateral hands (procedure)|
895523002 |Amputation of right hand (procedure)|
895473002 |Amputation of bilateral hands (procedure)|
Many of the existing laterality agnostic procedures are long-standing within the terminology, having been extant since the initial 20020131 release.
The questions I have related to the importance of having groupers like this in the age of ECL vs. the patient safety risk of using these in a medical record. Is it more important to have a hierarchical structure vs. a flatter structure that requires more precision in the use of the terminology?
With disorders, this does not seem to be as big an issue, but with procedures, it strikes me as being significant. As SNOMED is meant to be a clinical terminology, i.e. used in contemporaneous recording of data, ambiguous terms in procedures could cause problems.
3 Comments
Jim Case
From Feikje Hielkema-Raadsveld
In the Dutch hospitals, laterality is recorded as a separate data item for both diagnoses and procedures. So we use those laterality-agnostic procedures extensively. Yes, for patient safety you would expect the physician to record the laterality - but it does not follow that laterality must be precoordinated into the procedure.
Matt Cordell
A lot of clinical systems use a second field to qualify the laterality.
Two big use cases where surgical concepts are used in Australia don't require lateralised sites.
Requiring surgical procedures to all be lateralised leads to the combinatorial explosion, that historically has been avoided. (Ligament repairs etc. How many subtypes of 118949002|Procedure on extremity| need to be retired and replaced by to specific lateralised variants?)
I agree that you don't want to amputate the wrong body part, but if we get rid of all the "groupers" and only leave adequately specific concepts the extreme end of that is just a flat list of the leaf concepts (no hierarchy).
As great as ECL is, widespread adoption is still some way off, especially for existing systems. It requires the use of dedicated terminology servers or vendors to implement an ECL engine.
Ed Cheetham
Presumably the "threat to patient safety" posed by non-lateralised procedures relates to their potential contribution to instances of "wrong site surgery" (as an example of a "never event").
As such my suggestion would be that SI considers how SNOMED CT can play its part in a more complete risk management approach rather than simply looking to suppress/remove particular content from the reference data.
It is also not clear what the scope of the concern is. The suggestion seems to be that disorders are of lesser concern than procedures (but no evidence is given to support this - and what about, for example, specimens) and the nature of the in-scope procedures is left open. Is it only amputations, or does it include any 'destructive' or 'excising' surgical action, or is it, in fact, any action at all?
Likewise, is the concern limited to bilateral symmetrical structures or could it include any class of body structure which can be specifically described or can be referred to in a more abstract way (teeth, vertebrae, ribs etc.)?
I imagine that retaining non-lateralised content is also of considerable value for text-entity matching in NLP use cases where explicit mention of sidedness may not be available or required.
It is also worth noting that, even if we limited scope to amputations and considered only the set of concepts specified by...
...of the 120-or-so international concepts returned nearly half are found at least once in the UK primary care usage data for 2022/3 (the most frequently used being the 'doubly abstract' 371186005 |Amputation of toe|).