2022-04-06 - OBSERVABLE Face-to-face Meeting

Remember recording!

Daniel Karlsson

Daniel Karlsson 

James R. Campbell

Sarah Harry

The E2O project has produced a larger set of commonly used Evaluation procedure concepts as new concepts modeled using the Observables model (n=800+). The set was choosen from the 1000 most commonly used British primary care lab tests, excluding pure orderables.

There is evidence that re-modelling the content can bring significant improvements. There is also evidence that current content is not consistently represented and is in need of work, particularly if some content will stay in the procedure hierarchy.

There is still some disagreement about the requirement to represent orderables and reportables as distinct concepts, and if so, how these may be coordinated. There is still a requirement to represent "pure" orderables (e.g. panels) for which there is no current observables pattern.

Two outcomes of the discussions could be:

1. We decide to put the work done by Sarah Harry and James R. Campbell in a (separate) community edition for review by the community of practice accompanied by relevant communication.
2. We decide that we need to look further into how Procedures and Observables may be coordinated and/or representation of orderables as observables.

2022-04-06:

A discussion of options for Observables and Evaluation procedures followed after this summary.

There is an appetite, particularly from UK users (Jeremy Rogers, Jay Kola) to be able to use both Evaluation procedures (for ordering) and Observables (for reporting).

There are still issues to be addressed and resolved for Evaluation procedures which (truly) have multiple components, a.k.a. panels. Currently all Observables attributes are self grouped, with the intent to represent one and only one observation (which can have one and only one result). This means that the Observables model is not safe for (simple) combination representation of panels in terms of their constituents. So, currently, not all Evaluation procedures can be represented as Observables.

The work done to represent the 800+ observables could be used as a basis for discussion among stakeholders. Paul Amosasked if (1) any meaning was added to Evaluation procedures and (2) if there is any expected impact on modeling using the Evaluation procedures as value of 363714003 |Interprets (attribute)|. 

1. No meaning was added compared to the Fully specified name of the Evaluation procedure. Additional modeling was done for addressing modeling issues and (conservatively) add meaning of the FSN not modeled in the Evaluation procedure. E.g. "level" in the FSN was interpreted as 370130000 |Property (attribute)| = 118594004 |Quantity concentration (property) (qualifier value)|, "count" in FSN was interpreted as 370130000 |Property (attribute)| = 118550005 |Number concentration (property) (qualifier value)|.
2. The impact will preferably be that quality of content increases, with cleaner, more maintainable hierarchies and thus would have a positive impact also on any concepts using the current Evaluation procedure concepts as values (e.g. findings).

As a next step, a Community content area could be set up where the existing work could be made available. The 800+ concepts should be added as a start and continued development of E2O should make use of this Community content area.

In addition, any publication in the Community content area should be followed by communication to stakeholders allowing them to make an informed decision about moving Evalutation procedures to Observables.

James R. Campbell

2022-02-21:

Presentation of Race and Gender assignment observables.

Alejandro Lopez Osornio in latin america race typically is not recorded, with few exceptions. 

Nashar Karim patient-reported, self-assigned, where is that seen in the modeling? It is represented through Interview technique.

Daniel Karlsson recording race is illegal in many European countries.

2022-03-21:

Presentation of proposals for modeling of race, gender, and sex observables.

For all these observables, a general pattern of role assignment was used, i.e. race is assigned for a person by an agent, e.g. the community. There was agreement in the group that this pattern was useful.

For gender and sex observables, there was discussion about the relation between the two. Most agreed that sex and gender are distinct entities, where one is not a kind of the other. There are also many variants of both gender, e.g.  and biological sex, e.g. fenotypic (external genitalia, hormone levels) or genotypic.

Further, other standards organizaiton is doing work, paritcularly in the sex and gender area, indluing the HL7 Gender Harmony project https://confluence.hl7.org/display/VOC/The+Gender+Harmony+Project and the ISO TC 215 work on "Sex and Gender in Electronic Health Records" https://www.iso.org/standard/83431.html

2022-04-06:

Postponed

Daniel Karlsson

In relation to the change in modeling of primary and secondary/metastatic neoplasms, there has been a comment from Ed Cheetham about its implication for Observables concepts.

I can see how this approach will model existing combined primary and metastatic disorders (as now, through role grouping but including the new pathological process role), but is it suitable for complex observables such as 2960001000004100 |Presence of direct invasion by metastatic carcinoma of lymph node in excised specimen to perilymphatic tissue (observable entity)|? This too makes reference to both metastasis and local invasion, but its definition is made up of multiple self-grouped attributes. If the current 'inheres in=carcinoma, metastatic' becomes 'inheres in=carcinoma' and 'characterizes=discontinuous metastatic spread", how will this 'work' with the existing 'characterizes=direct local invasion' pair and current cardinality rules?

 Maybe this is covered by the line in the EAG paper that states "...The proposal involves creating new more specific subtypes of 1204295007 |Malignant proliferation of neoplasm (qualifier value)| as required e.g. Direct local invasion of primary malignant neoplastic proliferation (qualifier value) to model content..."? If so, this further complicates the role of 'characterizes' - since it now has a bearing on the interpretation of the 'inheres in' attribute (which is not included in its text definition – “This attribute specifies the process which the property describes, and on which the property (of this observable) depends.”).

2022-04-06:

A presentation of the issue was made.

The meaning of "Presence of direct invasion by metastatic carcinoma of lymph node in excised specimen to perilymphatic tissue" represents a nested meaning (the tumor has (1) spread to a lymph node by metastasis and has then (2) directly invaded the surrounding tissue). SNOMED CT cannot represent nesting without creating intermediary named/precoordinated concepts for the different levels of nested meaning. In this particular case, the MRCM for morphologies does not support this modeling.

Only way of addressing this issue (without changing the underpinnings of SNOMED CT) would be to develop editorial guidance which allows the identification of the focal aspect which is to be used for modeling. In this example, determining if the value of 704321009 |Characterizes (attribute)| is 1505281000004101 |Direct local invasion (qualifier value)| or 1505291000004103 |Discontinuous metastatic spread (qualifier value)|. Such concepts would also have to be primitive.

This particular problem arises as a side-effect of the decision to remove the proliferation process aspect from morphologic abnormalities.

Sarah Harry: other parts of the concept model have attributes like "direct site" and "indirect site" to represent similar kinds of nesting.

Piper Allyn Ranallo: nested processes can be created like "local invasion by metastatic neoplasm.

Using new attributes, we would still have to relate to the different types of neoplasm to the different processes, and the change was introduced to decouple the morphologies from the processes. The use of new attributes have been applied in the model before (e.g. 704319004 |Inheres in (attribute)| and 718497002 |Inherent location (attribute)| is a kind of hard-coded nesting.). There are limits to what can be expressed with the (current) Observables model and this could be one such case. There are currently only two concepts of this particular kind.

So, editorial guidance including a repeatable pattern would have to be developed, e.g. in case of nesting where the attributes could refer to any of the nested entities, the attributes should refer to the outermost nested entity and to that entity only if the attributes could also apply to multiple nesting levels, the concept must be marked as primitive.

Daniel Karlsson

Next meeting is April 25 2022 20.00 UTC.

Agenda:

• Quality and Process observables - clarification and potential change of guidance
• Impulsivity observables