As part of the review of 1263452006 |Anesthesia and/or sedation procedure (procedure)| hierarchy, and after revision of several topics previously commented, we can see a clear need for properly classify those procedures in SNOMED CT. While there are a few different classifications from different authors with slights differences, I'd like to ask for the current and more widely accepted classification in order to properly remodel such concepts.
Here, a broad classification for General anesthesia from the British Journal of Anaesthesia (2002) which may serve as a starting point for discussion:
Currently in SCT, 27372005 |Regional anesthesia (procedure)| is a subtype of 386761002 |Local anesthesia (procedure)|.
Although the objective of this discussion is to start with the broad classification of Anesthesia procedures, you can also review the discussion about Nerve block vs Local anesthetic nerve block (procedure) which was documented previuosly in a tracker document (https://confluence.ihtsdotools.org/display/IHTSDO1/Content+Project+Tracker+Documents?preview=%2F18777631%2F18777817%2FProject+1_Artf221500-RestructuringofLocalanesthesia_Inception_20141202.docx) from 2014.
Q: Are Local anesthesia and regional anesthesia siblings?
Q2: Should 27372005 |Regional anesthesia (procedure)| be a subtype of 386761002 |Local anesthesia (procedure)|?
Q3: Should Sedation procedures and Analgesia procedures be classified under 50697003 |Administration of general anesthetic (procedure)|?
Q4: Current grouping of 373266007 |Anesthetic (substance)| and 372614000 |Sedative (substance)| in SNOMED CT is updated or should it be corrected?
Feel free to write your answers in this space.
6 Comments
Steven Dain
Victor Medina Andrew Marchant Patrick McCormick My opinions
Q1: I'm OK with that
Q2: No
Q3: No, should be siblings of the top Anaesthesia (IMO), Anaesthesia is sibling with sedation with children intravenous analgosedation, intravenous sedation, inhalational sedation, inhalational analgosedation. Ether analgesia would put as a child of analgo sedation.
Q4: 373266007 |Anesthetic (substance)| is outdated. It needs new term in there General Inhalational Anesthetic (Substance) as a child under General anesthetic and as a sibling of intravenous anesthetic. I would put Ketamine under intravenous anesthetic.
Under General Intravenous anesthetics I would remove Alfentanil, Fentanyl, Sufentanil and Remifentanil and their derivatives as these are opiates and not Anesthetics. Interestingly Opiate (substance) doesn't exist in SNOMED, but 373529000 | Substance with opioid receptor agonist mechanism of action (substance) (but what is a Morphinan - Wikipedia?) a pharmacologist needs to look through this.! I have never heard of this term
Andrew Marchant
Similarly to Steven:
Q1. Expressing LA and RA as siblings is reasonable.
Q2. Depends what is meant by "Local Anesthesia". I agree with Steven because I think it means "anesthesia which is local to the surgical site". For RA to be a subset of LA, LA would have to be defined as "anesthetic procedure using (principally) a local anesthetic drug", which is another way the term is used in everyday speech (often the distinction doesn't matter).
Q3. There is already a higher-order term "410011004 | Administration of anesthesia AND/OR sedation (procedure)", which has sedative procedures as children.
Q4. 373266007 |Anesthetic (substance) seems to encompass anything with qualities which are somehow described as anaesthetic (encompassing tribromoethanol, sevoflurane and bupivacaine). 372752008 | Central nervous system agent (substance) and 373266007 |Anesthetic (substance) are at the same level. 372614000 |Sedative is further down the tree, below 372752008 | Central nervous system agent (substance). Perhaps, then, General Anesthetics (substances rather than procedures) would be a subgroup under 9680003 | Central depressant (substance), as 372614000 |Sedative already is? (using the UK release for convenience).
Steven's point about Alfentanil, etc., is well made although common usage outside Anaesthesia does not entirely make the distinction clear.
Victor Medina
So far, agreement about hierarchization of Local and Regional anesthesia: These two concepts should be siblings.
About 410011004 | Administration of anesthesia AND/OR sedation (procedure). Being a Classification derived component, will be deactivated and split into Administration of anesthesia and Administration of sedation. After responses, I understand that these two concepts should be siblings.
Substances need review for classification and standardization.
Ed Cheetham
If I might offer an alternative view here, since there are always trade-offs...
The Local/Regional question:
This is a good example of what Rector characterised (1999 - section 3) as the tension between linguistics, 'clinical pragmatics' and logical concept representation in terminology management. Andrew offers two possible meanings for "Local Anesthesia":
386761002 has been modelled consistent with definition 2 for more than 20 years (grouping and method have varied a bit, but direct_substance=local_anaesthetic has been a steady feature). So long as it is modelled this way (and is regarded as sufficiently defined), and so long as 27372005 |Regional anesthesia (procedure)| also includes direct_substance=local_anaesthetic in its formal definition, in SNOMED CT's opinion the latter will (AFAIK) always classify as a subtype of the former.
As such the formal definition (Rectors 'logical concept representation') is at odds with what appears to be the preferred 'clinical pragmatic' definition (a disjunction of 'topical' and 'infiltration'). If there is truly a desire to organise SNOMED CT according to the pragmatic preference then it is worth considering the costs:
If 386761002 |Local anesthesia (procedure)| survives the changes without inactivation:
If 386761002 |Local anesthesia (procedure)| is inactivated as a result of the proposed changes:
Regarding 410011004 | Administration of anesthesia AND/OR sedation (procedure):
Why inactivate it? Time and effort was spent 12 months ago applying a number of GCI axioms to its stated definition making it the formal disjunction of the proposed replacements - which are already siblings. How would the removal of 410011004 improve the data?
Regarding the original Q4, is it possible to explain how any proposed changes will be reconciled with the approaches taken in the substance redesign project (notably the use of the 'has_disposition' attribute) and that taken in the product redesign work (notably the 'plays role' attribute)? Whilst only the former is likely to affect the substance and procedure content directly it would be helpful to know what design philosophy is being used.
Thanks - Ed
Steven Dain
I'm a pragmatist, greatly adhering to architect Louis Sullivan's axiom "Form follows function" design philosophy and the KISS principle–keep it simple and safe. A controlled vocabulary has to be usable, non-ambiguous and interoperable. I also have a dislike of pre-coordinated terms, which can make analysis challenging. Want to look at data about hips, have to search hip, right hip and left hip, rather than just hip alone if it is postcoordinated with right and left qualifiers!
When using a local anesthetic (substance), I can use it in as "44596000 l Local anesthesia, by infiltration" (procedure), "27372005 regional anesthesia (procedure)" or "56333001 Nerve Block" (procedure) Perhaps we need to deprecate "386761002 local anesthesia (procedure)" as I believe that is at the root of the problem. Its 28 current children can be better classified into one of the 3 above.
With regard to 410011004 | Administration of anesthesia AND/OR sedation (procedure) It may be of use for organization of an ontology, but it is of little or no use in clinical practice where one has to specify a procedure as either an anesthetic or sedation for medical records and billing purposes. It is a code I would never use in a EHR, or for administrative or QA purposes
If EHRs are designed properly, one doesn't need expensive and inaccurate NLP, and as NLP gets better with generative AI, hopefully it will figure out the nuances. Local anesthesia procedure is a layman's term and in my opinion, has no place in SNOMED, as it is not sufficient.
Where is come to medications, it is particularly challenging. Ozempic is a great example. Its intended use was for the treatment of type 2 Diabetes, but now is primarily used for the treatment of Obesity! Propofol, midazolam, ketamine are sedatives in low doses and general anesthetics in higher doses. The opiates fentanyl and its analogues are not anesthetics at all. There are a lots of cases of recall with the older high dose opiate techniques used in conjunction with accidental inadequate amounts of anesthetics. Patients were pain-free during surgery, but they could recite the conversation that the surgeon and nurses were having.
General anesthesia (procedure) means causing loss of pain and loss of consciousness, obtundation of reflexes. General anesthetic substances are drugs that cause loss of pain and loss of consciousness, obtundation of reflexes.
Ed Cheetham
Thanks Steve
It's probably worth pointing out that Rector's notion of 'clinical pragmatics' [1] is not the same as a more general notion of 'being pragmatic' - nevertheless your point (in particular a preference for 'simple and safe') is significant in the context of terminology maintenance.
DL-based classification (the dominant organising paradigm for much of SNOMED CT) done 'simply' results in a structure which - much of the time - corresponds to clinical expectation (one branch of 'pragmatics'), but not always. This is why, for example, we see 72577009 |Excision of nipple| classified as a kind of 1231734007 |Mastectomy| and 274025005 |Colonic polypectomy| classified as a kind of 23968004 |Colectomy| - both of these assertions might well also conflict with expert clinical expectation. Extended DL dialects do exist that could be employed to try and replicate clinical expectation at this level of the product, but this often comes at the expense of 'simplicity' (and historically, computational tractability) with knock-on consequences to maintainability and performance [2].
When introducing his 'Hypothesis of separability' (between clinical linguistics, 'clinical pragmatics' and logical concept representation) Rector says "... It is implausible that a single integrated system could meet them all simultaneously...". An alternative is therefore to note (and highlight) the limitations of SNOMED CT's 'simple' (but achievable) organising principles, but consider producing complementary navigational products to replicate 'practical' views of the same data which corresponds better to clinical expectation.
With regard to 'perhaps we need to deprecate 386761002 |Local anaesthesia|' and its characterisation as a layman's term: Whilst a specialist anaesthetist will want to record more specific activities, there is a risk in ejecting more abstract/imprecise classes which may still be deemed suitable in alternative contexts. looking at UK usage data in primary care, 386761002 |Local anaesthesia| has been one of the top 3 codes for recording local anaesthesia activity for over a decade. This is why I asked before about the precise nature of such a change, and for a consideration of the impact of its removal.
If SNOMED CT is to achieve its goal of being a common terminology for all user groups and suitable for all recording contexts and styles, it should include content that supports these various (and varied) perspectives.
[1] https:/www.cs.man.ac.uk/~rector/papers/Why-is-terminology-hard-single-r2.pdf
[2] see, for example, how the recent introduction of a GCI to define 264274002 |Endoscopic operation| means it is now no longer a descendant of 363687006 |Endoscopic procedure| - thus making the task of identifying all 'endoscopic procedures' harder.