I’ve been troubled for a while by a view sometimes put forward to the effect that codes like “Amputation of finger” or “Trapped at scene of incident” are not required in SNOMED because, if entered into a patient record, they would clearly be medicolegally underspecified: surely you know exactly which fingers were amputated, and so both can and should use the available more detailed codes specific to each possible finger. The grouper isn’t needed .. and so we can conveniently excuse ourselves from worrying about what the term “finger” even means.
I was on a call earlier today that illustrated perfectly why this train of logic is not correct: in an emergency telephone triage setting, it is entirely sufficient to both only know and so to only record that the patient says they have chopped off an unspecified digit, in order to be able to progress rapidly to the next important triage question: how much blood are you losing? Indeed, it would actually be clinically counterproductive - even unsafe - to take the time at this point in the process to pin down and record exactly which fingers were and were not involved.
But the triaging clinician still needs to be able to record and exchange what they did know and establish, to the level of detail this was or was not requested, made available and clinically relevant. Thus, a code for “Amputation of finger” isn’t a grouper whose sole purpose is to support data extraction over populations. Its a perfectly valid primary data capture code at the level of an individual patient. Valid and clinically useable primary data capture does not - can not - only happen using the most detailed leaf nodes.
This also reminded me of an example put forward by Angelo Rossi Mori in the 90s, illustrating how the exact same surgical procedure would need to be coded with similarly variable precision at different stages of the patient journey. The clinic booking note might say “Reattachment of finger”. The Theatre Preparation system might change this to “Reattachment of amputated RIGHT finger” so that the surgical instrument table was placed on the correct side of the bed. And the post surgery note might be “Reattachment of partial amputation of right middle finger”. All are correct.
Hi Jeremy, this certainly shows why a grouper concept can be necessary. Different levels of detail can have their separate use cases and we should be careful before we remove a set of grouper concepts entirely.
In the finger/digit discussion though, we have two grouper concepts which, although distinct, are very close in meaning:
digit of hand, i.e. digitus manus I, II, III, IV and/or V
finger, which can be interpreted in different ways:
By orthopedics, surgeons, most English clinicians (?) as digitus manus II, III, IV and/or V
By patients, paramedical specialties (?) as digitus manus I, II, III, IV and/or V
So when a patient calls and says that someone has accidentally amputated a finger - are you even that certain it’s not a thumb? Will the triage act differently if it is a thumb? Or would the grouper concept for digit of hand, with a synonym of ‘finger and/or thumb’ support the use case just as well?
So to me, this discussion was not: do we need a grouper concept? But: do we need two grouper concepts whose meaning overlaps so strongly?
I think many, possibly most, anglophone patients would also interpret “finger” as exclusively II, III, IV or V ! Except when by convention we all tacitly understand from the wider context that I is also included.
As in (to a child): How many fingers do you have on each hand?
Correct Answer: Five
But also (to same child): Point to ANY of the fingers on your left hand
Wrong answer: child points at thumb.
So when a patient calls and says that someone has accidentally amputated a finger - are you even that certain it’s not a thumb?
Pretty much yes. If they’d amputated their thumb, they’d say so. If they had amputated I, II and III in the same accident, they would say “I’ve amputated my thumb and two fingers”. Reporting it as “I’ve amputated three fingers” would, I think, cause cognitive dissonance in the vast majority of native English patients and clinicians.
Personally, I think we do need two sets of concepts (including groupers) for the inclusive meaning of “finger” as apparently understood by most non-anglophone speakers and cultures but that would I think be cognitively strongly dissonant to most anglophones, and the existing exclusive set that is not only cognitively familiar to the anglophone world but that also corresponds to a globally clinically useful and familiar category (including to Dutch orthopods!) … but that is unfortunately dissonant to everybody else, especially when labelled as “finger”.
Leaving aside the one-time cost of standing up the second set, the more interesting questions include whether and in which releases the word “finger” should be offered within synonyms, and how clinicians across all cultures and linguistic backgrounds can be assisted in searching over and using both codesets correctly (ie not picking an exclusive code when the inclusive meaning was intended, and vice versa). A range of technical solutions spring to mind, mainly involving realm language refsets to suppress “finger” descriptions across all inclusive codes in anglophone countries and the opposite everywhere else.
I agree with the discussion so far, but I think if we attempt to represent both definitions in the terminology itself, even using a GCI, we may end up doing a disservice to the end-user community.
An approach that we have not considered, or at least that I have not heard discussed, is to select the most accurate modeling approach based on anatomical references that are considered to be authoritative sources for SNOMED and then supplement with an appropriate refset. Granted, this option adds slightly more maintenance overhead for each monthly release, but the optimization for end-users may make it worth the effort.
This path was taken with the lateralized body structure refset.
One option is that we select to use the exclusive definition and keep the finger and thumb structures and definitions as they stand. We can then create an inclusive supplemental refset definition that can be used by implementors and researchers.
About 10% of the people will complain that the classification isn’t correct and 90% will be sufficiently happy. I am beginning to theorize that when we have questions about grouping concepts, if we can’t come up with a classifiable solution, we need to leverage other features provided by the extensive terminological construct that is SNOMED CT.
The question then becomes, is there a more pragmatic approach that we can take that will ensure the accuracy of the terminology while providing implementors and translators with the tools and groupings that they need?