Observables technology preview analysis

The tech preview for Observables contains 13621 LOINC codes.

 The results database at University of Nebraska contains results for 1352 distinct LOINC codes.

 593 LOINC codes have been reported at Nebraska but are not in the tech preview, constituting about 36% of results by volume. 

251 LOINC coded results reported at Nebraska have 1000 or more items in our database (0.0005% of total).

16 LOINC code results reported have 100,000 or more items in our database(0.05% of total)

Sp[readsheet attached to this discussion with code frequencies

Jim

Hi Jim,

Thank you for sending the file. At the moment Farzaneh and I are preparing the Beta release files for the LOINC - SNOMED CT Cooperation Project, to be released on 31 March, 2017. This upcoming release will have almost 22000 LOINC Terms. This release includes a majority of the Top 2000 LOINC Lab Observations. Some content has been excluded from the Beta Release: 

• We have not yet mapped all the LOINC Parts associated with the Top 2000 LOINC Terms to SNOMED CT concepts (or have not created the SNOMED CT concepts). The main reason for this exclusion is the requirement for further clarification with the Substance Project, RII, and Observable Model Redesign Project.

• There are Terms being discussed internally or with Regenstrief for modelling reconsiderations and possible modification:

• Susceptibility testing

• Genetic testing

• Terms defined by LP Components containing “tested for”

• Terms containing disjunction

• Physiological measurements and Vital signs Terms are not included because their scope is still being discussed with RII. Furthermore IHTSDO does not yet have the list of LOINC Parts from RII used to define these Terms.

• Veterinary specific terms are not considered in the scope of SNOMED CT and this project

• Although panel names are in the scope of the Cooperative Agreement, modeling of panel Terms is not. However, SNOMED International decided not to include the LOINC Term with panel names.

We appreciate having this list with frequency stats and will consider this in the future when we are prioritizing the next phase of mapping.

Thanks,

Suzanne

Speaking as an IHTSDO member, I wonder why the observables content currently in our EHR cannot be moved to the head of the queue?  We are not talking about a lot of concepts, and the ability to deploy a functional product fully encoded with SNOMED CT is an IHTSDO strategic imperative!

Jim

Having loaded the tech preview as an observables ontology, some questions arise in my mind concerning the modeling:

1) Chemistry tests on 24 hour urine are modeled as direct site spot urine specimen (not my choice) and do not inhere in Urine(substance) (Inhere in attribute is not employed in the concept definition). Virtually all other clinical chemistries including those on random urine 'Inhere in' the substance they are analyzing.  I think this is logically inconsistent and incomplete.

2) Chemistries on red blood cells (Bicarbonate in red blood cells and others) Inhere in "Population of all erythrocytes in portion of fluid".  This might be ontologically sound  and would be appropriate for eg: count the number of RBCs in a volume - but it appears illogical to me when assessing a feature of the red blood cell such as the concentration of bicarbonate it contains.

Jim

James R. Campbell

Hi Jim,

Thanks for the feedback. Suzanne and I looked and we can't locate the problem areas you noted. Can you please send us an example or two for each of the items noted above?

Thanks,

Farzaneh

CC: Suzanne Santamaria

Examples that I pulled at random....we use the long name to form an FSN:

1-Methylhistidine [Moles/time] in 24 hour Urine (observable entity)

11-Dehydro thromboxane beta 2 [Mass/time] in 24 hour Urine (observable entity)

11-Deoxycorticosterone [Mass/time] in 24 hour Urine (observable entity)

11-Deoxycortisol [Moles/time] in 24 hour Urine (observable entity)

11-Hydroxyandrosterone [Mass/time] in 24 hour Urine (observable entity)

Aldosterone [Moles/time] in 24 hour Urine (observable entity)

Estriol (E3) [Moles/time] in 24 hour Urine (observable entity)

Bicarbonate [Moles/volume] in Red Blood Cells (observable entity)

2,3-Diphosphoglycerate [Mass/volume] in Red Blood Cells (observable entity)

Adenylate kinase [Enzymatic activity/volume] in Red Blood Cells (observable entity)

Calcium [Moles/volume] in Red Blood Cells (observable entity)

Glutathione peroxidase [Enzymatic activity/mass] in Red Blood Cells (observable entity)

Uroporphyrin [Mass/volume] in Red Blood Cells (observable entity)

Hi James R. Campbell,

Thank you for providing the examples. 

Concerning the first group of examples, our current guidance on this type of modeling, to be included in the upcoming LOINC - SNOMED CT Cooperation Project Beta Release documentation, was recently shared in a different post on the Observables Confluence site and is being discussed with Daniel now. See my lengthy comment with the text from the documentation on 20 February here. We need to add your question to the discussion, namely whether Inheres in should also be used when modeling these LOINC Terms.

Your second group of examples relates to mapping of LP RBC, which is an issue that has been identified previously in the LOINC Project in 2016. Since it has not been resolved yet, related content was excluded from the Beta Release. We will be discussing this with the Observables Project Group soon. 

Suzanne

cc: Farzaneh Ashrafi

MULTI-INGREDIENT OBSERVABLES

My primary use case for laboratory tests is to be able to retrieve data across multiple testing methods, process durations and specimens of my choice for specific analytes of interest.  For example, all hemoglobin A1C tests  from blood, all influenza antigens from any biological specimen, etc.  If an influenza antigen happens to be one member of a panel test reported only with summary results for any component, identifying that as an influenza test is valuable althought it will NOT answer the question: was influenza antigen found.

With the load of tech preview, I took one multi-component test ( 122441001000004107 |Tiglylcarnitine+methylcrotonylcarnitine (C5:1) [Moles/volume] in Serum or Plasma (observable entity)|

And defined it fully with the two components:

I then defined grouper concepts for tests of the two components:

When I classified SNOMED CT, each grouper concept subsumed the multi-ingredient test

And the classified multi-component observable concept was a subtype of the two types of tests.  So, to serve the query use case “was any other type of test done for this component”, and NOT for the case “was this analyte confirmed present”,  I think that fully defining the multi-component tests supports an ontology with useful features.  This will have to be clearly documented and the user community trained in its importance.  It would be a deployment with opportunity for misuse.

Jim

(I attached Word file of this message and graphics)

While building an Observables ontology for common labv, I noticed that Coagulation tests seemed to be missing.  In particular, Prothrombin time(6302-4), INRs(6301-6,34714-6,52129-4), Partial thromboplastin time and Bleeding times (3178-1,3179-9) are all missing while the much less commonly performed coagulation factor assays and autoimmune anticoagulants appear to be abundant.  I have to be sure that those are added to the development list.  Those tests are done a lot in the Coagulation lab.

I could use some guidance on applying the concept model to these as most are assessing process times for coagulation under various controlled conditions.

Jim

I have been defining the groupers for our common lab ontology and woirking on cell surface markers.  I noted that the definition of

"CD19 (Lymphs) cells [#/volume] in Blood (observable entity)" has component Lymphocyte antigen CD19 but Inheres in blood whereas I thought it should be lymphocyte.  When I compared this to the definition of many other cell markers such as "CD10 cells [#/volume] in Blood (observable entity)" I noted that the others all had component "Cell positive for CDXX antigen" which could reasonably Inhere in blood.  I think this represents a minor inconsistency in the tech preview but arguably they should all be corrected since the observables relate results for staining of celkl surface antigens in families of white blood cells, more specifically lymphocytes.

Jim Campbell