Observables and Evaluation procedures discussion

Slides: https://docs.google.com/presentation/d/1l9qttsqa_MKjFQs5TB-3_iofIraiGzP7hjUPPpltUk4/edit?usp=sharing

Summary of discussion so far

Observables and Evaluation procedures came from two different strands of work; Observables originally from NHS Clinical Terms and Evaluation procedures from SNOMED RT. In an effort to bond LOINC and SNOMED RT in 1996-97 created much of the structure for Evaluation procedures and the specific attributes used for Evaluation procedures cam from this LOINC integration. Observables did not have a concept model until the first draft in 2009 and implemented in SNOMED CT not until January 2017. Based on this historical analysis between the original source of Evaluation procedures and the current Observables concept model, there is no difference in the use cases accounted for.

The Observables concept model is designed to support the representation of observations which have one and only one property (potentially with a result as a list of entities of the same kind). This overlaps with some Evaluations procedures, particularly those resulting from the LOINC import.

Evaluation procedures include content of different categories. Some are "pure" methods, e.g. 9718006 | Polymerase chain reaction analysis (procedure) | or 76978006 | Enzyme-linked immunosorbent assay (procedure) | which might better be represented using the 272394005 | Technique (qualifier value) | hierarchy. Imaging procedures are Evaluation procedures but do not use the Evaluation procedure concept model. These procedures are generally neither pure methods nor do they distinguish one particular property. Some of the Evaluation procedures are batteries/panels and have previously been considered out of scope for Observables, but this can be revisited.

From previous discussions there are some requirements for having different codes for requesting observations to be done and reporting the results of those observations. Currently, by some SNOMED CT users, it has been interpreted that Evaluation procedures are for requesting and Observables are for reporting. The use case for having two codes for reportable and requestable should be investigated and pros and cons of keeping two codes vs. one should be listed.

Decisions to be made

• What is the scope of Observable/Evaluation procedure? What use cases should be supported?
  • There is a difference between representing (1) a request or a report from (2) that what is requested or reported on
• To keep separate representations for supporting request and report use cases respectively or not?
• If the decision is to keep two separate representations:
  • What are the requirements to keep the two in sync?
  • How should the two be kept in sync?
  • What guidance is needed to avoid user confusion?
• If the decision is not to keep separate representations
  • What hierarchy is kept, Observables or Evaluations procedures?
    • Hierarchy and concept model decisions are separate!
  • How is the content of the deprecated hierarchy moved into the other?
    • How are concept models 
    • Can concept id:s be kept?
    • What is the classification impact of the changes?
• How are input from the community collected?
• How are any decisions communicated to the community?

Proposals

Keep both hierarchies

• Synchronize content
• Update documentation

Move relevant Evaluation procedures to Observables

See this document from James R. Campbell

Move Observables under Evaluation procedures