• Recording of meeting approved by participants.

Conflicts of interest and agenda review

UK has high usage of a few of these:

CONCEPTID             FSN                                                                                                                         USAGE 2011-2022

394617004              Result (navigational concept)                                                                                 48,227,610

160237006              History/symptoms (navigational concept)                                                             10,146,392

309157004              Normal laboratory finding (navigational concept)                                                 581,209

267368005              Endocrine, nutritional, metabolic and immunity disorders                                   160,475

243800003              Test categorized by action status (navigational concept)                                      108,176

250541005              Biochemical finding (navigational concept)                                                           100,632

Currently, most "417662000 |History of clinical finding in subject (situation)|" concepts (~1000) are modeled with a TEMPORAL CONTEXT of 410513005 |In the past (qualifier value)|.  The exceptions are "Family history of X" concepts.  While this makes sense in the context of procedures, it is not so clear with clinical findings.  In many cases clinical history involves conditions that are both current and in the past (e.g. history of cough for last 10 days).  While we have a few concepts that specifically state "Past history", for the most part History of clinical findings represent conditions that were in the past and do not represent ongoing clinical conditions.  

It is recommended that these concepts be remodeled with a new TEMPORAL CONTEXT, either the existing 410511007 |Current or past (actual) (qualifier value)|, or a new "Current and/or past (qualifier value).  As this change is more general than the current modeling, it should have no impact on the taxonomic structure of the hierarchy.  

Discussion:

In the past was modeled a long time ago and there was an argument that this was sufficient to model these concepts.  There are some conditions that need to be evaluated as to whether they only occur in the past.  Might need to look at Family history as well.  Loosening this context up would be an improvement.  Suggested to ask Ed Cheetham for background.  There may also be some duplication.  There is still a valid use case for "in the past" and "Current and/or past".  The problem list use case requires that the context be explicitly represented.

Do we need additional contexts that refer to resolved? The Situation model was intended to be a "mini-information model".

Decision:

No opposition to starting with this proposal

We have examples of clinical findings and disorders that have identical modeling, and in some cases have identical descriptions.  An example brought to our attention is that of "Inflamed joint" vs. "Arthritis".  These are modeled identically and 3723001 |Arthritis (disorder)| has the additional description of "Joint inflammation".  During internal discussions it was commented that there is a need to distinguish between observations made during clinical exam and those that represent a "diagnosis".  Other examples include 281795003 |Inflamed tonsils (finding)| vs. 90176007 |Tonsillitis (disorder)|; 298170003 |Knee joint inflamed (finding)| vs. 371081002 |Arthritis of knee (disorder)|, etc.

Some proposed use cases were nursing findings, and the "need" to differentiate a finding on observation vs. diagnosis.  We have recently completed inactivation of a large number of O/E (on examination) findings.  However, we do not have a comprehensive representation of this pairing of findings and disorders in SNOMED CT.  This impacts where specific descriptions may be applied; for example, we would not put the description "Inflammation of knee joint" on the concept 371081002 |Arthritis of knee (disorder)| (which is currently the case), since it would be more appropriate for 298170003 |Knee joint inflamed (finding)|.  There is considerable inconsistency in the current content, for example there is 298171004 |Ankle joint inflamed (finding)|, but no "Arthritis of ankle". 

Q:  Is there a need to maintain this distinction and, if so, where should it be applied?

Discussion:

This is a discussion that has gone on for over 20 years, the finding/disorder dichotomy.  The nursing use case (i.e. clinical observation) makes these findings vs. disorders which are diagnoses.  The distinction is vague and inconsistent.  We do not have a comprehensive representation of finding and disorder for many of these.  

The distinction of who can make what type of finding/disorder should not be the responsibility of SNOMED.  How far do we need to go to maintain the distinction?  Arthritis seems like more of a definitive diagnosis as opposed to just a clinical finding.  This is a problem with the use of and interpretation of the language.  Because the terms are defined the same, what are the reasons for maintaining the difference?  It is the context in which these concepts are used that makes the difference.

Intractable problem?  The clinical interpretation is an important aspect that is not captured by the current concept model.  There is an implicit notion of temporality.  Findings are made at a point in time, whereas a "disorder/disease" requires more data.  How do we keep the two "meanings" different?

Decision:

Need to come back with a proposal to determine how many of these would be needed if the distinction is maintained.  Also, for arthritis, what are the necessary conditions to make something a disorder?  Can this be done? 

A recent query from members brought to light the fact that SNOMED CT represents "Clinical finding present" concepts in two ways.  The majority of the concepts (>550) are located in the Clinical finding hierarchy (e.g. 297990009 |Hair present (finding)|), with the majority of those concepts having a definition status of primitive.  A small number (~55) are located in the Finding with explicit context hierarchy and are sufficiently defined, having been moved from the Clinical finding hierarchy, first to the "Context dependent category" (precursor to the Situation hierarchy) in 2004 and then to the Situation with explicit context hierarchy in 2006.  

It may seem clear that the remaining "X present (finding)" concepts would logically belong in the Situation hierarchy; however, large changes in hierarchy have historically caused issues with implementations.  Advice is sought concerning the potential move of these concepts.

1. Clinical findings have an implied context that is dependent on where in the medical record they are used.  The default context is "known present" and "current".
2. Clinical findings with the finding context explicitly stated do not have alternative interpretations. 
3. Clinical findings with stated context have historically been moved to the Situation hierarchy: however, this has nearly always caused issues with implementations
4. Movement of context-specified content allows them to be sufficiently defined, whereas they are usually primitive in the Clinical finding hierarchy. 
5. Historically, movement of large numbers of concepts to the Situation hierarchy retained the original concept ID as there was no meaningful change in meaning of the term, just a change in hierarchical location.

Ancillary to this is the use or the TEMPORAL CONTEXT value 410512000 |Current or specified time (qualifier value)|.  The addition of "specified time" is assumed to have been used to state that the referenced condition was known to be present based on a timestamp that is specified in the EHR, but has caused confusion for editors regarding the appropriate selection of "current" vs "current or specified time", leading to inconsistency in modeling of TEMPORAL CONTEXT.. 

An underlying assumption is made that the recording of the finding is contemporaneous with the the observation and that all modern EHR systems provide a timestamp for each recorded instance.  If the observation is recorded at a future date, then a determination of whether the condition is still present (i.e. 15240007 |Current (qualifier value)|), or has resolved (i.e. 410513005 |In the past (qualifier value)|) must be made. This makes questionable the use of any temporal context value that refers to "specified time".  The use of these temporal context values has been extremely limited, although subtypes (e.g. 410589000 |All times past (qualifier value)|) have been used for a relatively small number of concepts (<100).  

Q: Given their limited use an questionable value, should Temporal context values that use "specified time" be inactivated and affected concepts be remodeled with the appropriate parent value?

Discussion:

Many of these cannot be expressed within the situation hierarchy because there is no way to represent the ASSOCIATED FINDING (e.g. testes present).  Other terms that represent disorders can be moved over.  Many of the existing finding concepts fall into the first category.

Should remove the description "Disease present" should be removed from "Disease".  Determine whether there are approaches to make these sufficiently defined within the findings hierarchy.

Which is a better representation of currency, "Specified time" or "Current"?  The use of Specified time explicitly states that the time associated with the finding is recorded elsewhere in the record.  Issue with how to represent a specified time in the past.  

For the issue around "Finding present" there may not be a need for temporal context at all. 

Decision:

SI will test some potential modeling patterns in the findings hierarchy to determine if we can sufficiently defined.  

Will test whether leaving off temporal context when not stated has an impact on classification. 

Put out a doodle poll to determine a new time for the conference calls.