Review and discuss editorial advice for application of concept model in creation of observables content
Page tree
Overview
Content Tools
Apps
Copyright © 2025 SNOMED International. All Rights Reserved - Privacy Policy - Cookie Settings
Powered by a free Atlassian Confluence Community License granted to SNOMED International. Evaluate Confluence today.
This Confluence installation runs a Free Gliffy License - Evaluate the Gliffy Confluence Plugin for your Wiki!
- Powered by Atlassian Confluence 7.19.29
- Printed by Atlassian Confluence 7.19.29
- Report a bug
- Atlassian News
22 Comments
James R. Campbell
An unfortunate feature of the concept model is that drugs have conceptual content in both substance and pharmaceutical products which are a single 'thing' from the standpoint of the clinical world of users. In developing some groupers (query applications) for the observables tech preview, I observe that we are placing our users in the difficult position of writing two queries to answer simple questions like: 'Did the patient have a test for opioids?' Reviewing the MRCM for observables, I note that 'Component' and 'Inheres in' are cardinality 0...* AND have valuesets of both Substances and Pharmaceutical products. To favor efficient use of observable content for query and decision support purposes, should we develop guidance that observables assessing 'presence' or 'process' relating to chemical compounds which are also pharmaceutically active should always be defined with BOTH attribute value pairs, assuring that the naive user will get proper performance of database queries like "Was there a test for opioids?'. So, the tech preview model would change for 'Morphine level in blood', to wit:
Morphine Mass/volume in serum or plasma (observable entity)
fully defined as
IS_A Opitaes Mass/volume in serum or plasma
COMPONENT Morphine(substance)
COMPONENT Morphine(Pharmaceutical product)
DIRECT SITE Acellular blood specimen
INHERES IN Plasma (substance)
PROPERTY TYPE Mass concentration
SCALE TYPE Quantitative
TIME ASPECT Point in time
Michael Osborne
Looking at the concept, it appears to be measuring the drug 'Morphine' not the substance....but it sounds like the query you want is SELECT * WHERE COMPONENT HAS_ACTIVE_INGREDIANT Morphine (substance). Would this query capture all of the observables containing Morphine e.g. free Morphine and others? Probably need better guidance, but it comes down to exactly what the observation is measuring, and in this case it appears to be the drug morphine.
Suzanne Santamaria
The product concepts in SNOMED CT are defined by substance concepts via the “Has active ingredient (attribute).” This relationship can be used in queries if needed, as Michael also points out. Yes, it is an additional step, but it keeps the Observables model “cleaner,” more concise and consistent.
Even in the measurement of the drugs (or any product), what is actually being measured is the active ingredient. So while the range of “Component (attribute)” includes both products and substances, we would still go with substances in the vast majority of cases.
If the LOINC Term states “morphine,” we cannot assume or model anything more specific than the LOINC Term (unless RII clarifies that a different substance is actually measured).
We also want to add that the Product hierarchy in SNOMED CT is currently undergoing changes. One change is related to making FSNs more clear, for example, 73572009 | Morphine (product) will change to 73572009 | Product containing morphine (medicinal product) for the July 2017 release. Another change is that all product concepts will be fully defined with an “Has active ingredient (attribute)” relationship to a substance concept.
We realize we need to update the modeling guidelines in the documentation for the LOINC - SNOMED CT Cooperation Project release files in this area.
If more discussion or clarification is needed, we can bring this topic to the next Observables call.
Daniel Karlsson
There is a similar situation in the procedure model where DIRECT SUBSTANCE has both Substance and Product in range. The solution there is to introduce a right identity (aka property chain) stating that every Procedure with a DIRECT SUBSTANCE to a Product also has a the Substance(s) given by HAS ACTIVE INGREDIENT as DIRECT SUSTANCE(s), or in logical terms:DIRECT SUBSTANCE o HAS ACTIVE INGREDIENT → DIRECT SUBSTANCE
A similar logical rule for Observables would solve the problem and avoid having rules which have to be interpreted by users and modelers: COMPONENT o HAS ACTIVE INGREDIENT → COMPONENT. In summary, this would provide the Substance whenever a Product is used with COMPONENT but no the other way around. Jim, is this something that you test in your SNOMED CT edition?
Further, currently and by design the attributes INHERES IN and INHERENT LOCATION are unrelated. That something has the INHERENT LOCATION of X does not imply that it INHERES IN X. However, if there is a need to query for Observables which are "located" (if that's a good term?) somewhere, we might need a more general super attribute to INHERES IN and INHERENT LOCATION (and possibly also DIRECT SITE).
James R. Campbell
At nebraska we have 'conceptualized' and loaded the beta tech preview so that we have ALL tech preview concepts defined and classified as observables. Our common lab set from our research nework was 1372 concepts. The consolidated tech preview was 23287 but only includes 1029 of our common labs. This leaves 343 concepts that are deployed across GPC withsignificant frequency that are not in the tech preview. I have uploaded the set of concepts as an attachment. I also include a screen shot of our developmental ontology for pathology and laboratory medicine below as installed in i2b2
Jim
James R. Campbell
We are publishing our common labs and pathology ontology alpha release this week and have included all beta tech preview concepts of relevance. However the dozen concepts included in the uploaded 'LOINC ONTOLOGY BUILD.xlsx. represent 20M or so lab results in our system that never came out with the beta tech preview. The coagulation tests are incompletely modeled but I do not want to add attributes only to have to back them out of our extension when we get content next January. I would appreciate comments on accuracy of the Observables concept definitions perhaps during a future call
Jim
Farzaneh Ashrafi
Hi Jim,
Suzanne and I reviewed the file you submitted. The following contains the mapping that we recommend for those terms. Please note that most of the mapping described below include the mapping for the LOINC parts (as opposed to the LOINC terms). In general, the guidelines on mapping of LOINC attributes (including component, divisor, property, time, scale, system, and method) and linkage between LOINC terms and SNOMED CT expressions are described in the Beta release documentation file and we welcome your feedback on the documentation.
1) 2340-8 Glucose [Mass/volume] in Blood by Automated test strip
LOINC part (Method): Automated test strip
The issue here is a missing technique (as it was not part of original top 2000 LOINC parts, but it is now considered for addition in January 2018 release):
New | Automated test strip technique (qualifier value) and New | Manual test strip technique (qualifier value) as children of 702660003 | Test strip technique (qualifier value) |
2) 3097-3 Urea nitrogen/Creatinine [Mass Ratio] in Serum or Plasma
LOINC Part (component/divisor): Urea nitrogen/Creatinine
Existing mapping
Component = 95971004 | Urea nitrogen (substance) |
Relative to = 15373003 | Creatinine (substance) |
Note: The mapping for the LOINC parts and the LOINC term have been included in the Beta release and will be part of August 2017 production release. Please check the version of the file.
3) 18182-6 Osmolality of Serum or Plasma by calculation
LOINC part (component): Osmolality
Suggested mapping: This LOINC part will not be mapped as component; it will be mapped as a property (See similar cases for Appearance, Color, etc in the guidelines).
723196003 | Osmolality (property) (qualifier value) |
4) 4544-3 Hematocrit [Volume Fraction] of Blood by Automated count
LOINC part (component): Hematocrit
Hematocrit is the volume percentage (vol%) of red blood cells in blood
Suggested mapping:
Component = 418525009 | Erythrocyte component of blood (substance) |
Relative to = 420135007 | Whole blood (substance)
5) 787-2 Erythrocyte mean corpuscular volume [Entitic volume] by Automated count
LOINC part (component): Erythrocyte mean corpuscular volume
The mean corpuscular volume, or MCV, is a measure of the average red blood cell volume. https://s.details.loinc.org/LOINC/62242-3.html?sections=Comprehensive
Suggested mapping:
Component = 41898006 | Erythrocyte (cell)
Technique = New | Mean corpuscular volume (MCV) calculation [fl = (Hct [in L/L]/RBC [in x10 12/L]) x 1000]
6) 788-0 Erythrocyte distribution width [Ratio] by Automated count
LOINC part (component): Erythrocyte distribution width
RDW-SD) is a measure of the range of variation of red blood cell (RBC) volume (https://s.details.loinc.org/LOINC/30384-2.html?sections=Comprehensive)
Suggested mapping:
Component = 41898006 | Erythrocyte (cell)
Technique = New | Erythrocyte distribution width calculation [Standard deviation of MCV ÷ mean MCV) × 100]
7) 786-4 Erythrocyte mean corpuscular hemoglobin concentration [Mass/volume] by Automated count
LOINC part (component): Erythrocyte mean corpuscular hemoglobin concentration
The mean corpuscular hemoglobin concentration, or MCHC, is a measure of the concentration of hemoglobin in a given volume of packed red blood cell. It is reported as part of a standard complete blood count. It is calculated by dividing the hemoglobin by the hematocrit. https://s.details.loinc.org/LOINC/62246-4.html?sections=Comprehensive
Suggested mapping:
Component = 38082009 | Hemoglobin (substance)
Technique = New | Erythrocyte mean corpuscular hemoglobin concentration (MCHC) calculation [Hb / (Hct/100)]
8) 48642-3 Glomerular filtration rate/1.73 sq M predicted among non-blacks [Volume Rate/Area] in Serum or Plasma by Creatinine-based formula (MDRD)
9) 48643-1 Glomerular filtration rate/1.73 sq M predicted among blacks [Volume Rate/Area] in Serum or Plasma by Creatinine-based formula (MDRD)
LOINC parts (component/divisor): Glomerular filtration rate (as nominator) and /1.73 sq M (as denominator):
Creatinine clearance can be precisely determined by measuring the amount of creatinine present in a sample of urine collected over 24 hours. This method requires a person to urinate exclusively in a plastic jug for one day, then bring it in for testing. Although the urine creatinine measurement method is inconvenient, it may be necessary to diagnose some kidney conditions.
GFR can be estimated using a single blood level of creatinine, which your doctor enters into a formula. Different formulas are available, which take into account age, sex, and sometimes weight and ethnicity. The higher the blood creatinine level, the lower the estimated GFR and creatinine clearance. See details here: http://www.webmd.com/a-to-z-guides/creatinine-and-creatinine-clearance-blood-tests#1
and https://www.niddk.nih.gov/health-information/health-communication-programs/nkdep/lab-evaluation/gfr-calculators/Pages/gfr-calculators.aspx
Suggested mapping:
CHARACTERIZES|= New | Filtration process (qualifier value), Parent: 716234001 | Physiological process (qualifier value)
PROCESS OUTPUT = Creatinine
PROCESS AGENT = Parenchyma of kidney
TECHNIQUE = The “/1.73 sq..” divisors are all techniques (a formula that revises the measurement). Map divisor LPs to SNOMED CT techniques as follows:
LP20516-8 1.73 sq M 702667000 Clearance calculation formula (qualifier value)
LP69197-9 1.73 sq M.predicted 703505007 Clearance calculation relative to 1.73 square meters body surface area (qualifier value)
LP69200-1 1.73 sq M.predicted.black 703506008 Clearance calculation relative to 1.73 square meters body surface area and adjusted for African race (qualifier value)
LP69848-7 1.73 sq M.predicted.non black 703507004 Clearance calculation relative to 1.73 square meters body surface area and adjusted for non-African race (qualifier value)
10) 2708-6 Oxygen saturation in Arterial blood
LOINC component: Oxygen saturation
Oxygen saturation (sO2) is defined as the ratio of oxyhemoglobin to "active" hemoglobin (oxyhemoglobin + deoxyhemoglobin). Some say the term oxygen saturation has been erroneously redefined with a denominator of "total" instead of "active" hemoglobin, an ambiguity that can sometimes create confusion when viewing reported values from different instruments (Anesth Analg. 2007 Dec;105(6 Suppl):S5-9.) Oxygen saturation is commonly measured by blood gas analyzers and pulse oximetry.
Suggested mapping:
Component: 80873004 | Oxyhemoglobin (substance)
Technique: 703463001 | Calculated from oxygen partial pressure (qualifier value)
Relative to: new |oxyhemoglobin and deoxyhemoglobin (substance) --> Asked substance team if such concept can be added. The answer was: ““The editorial guidelines we agreed to for combined substances is listed below. Unless I'm missing something, I don't think this new concept meets the criteria. It seems more similar to the ones we recently inactivated.
Combined substances:
Combined substances added to the Substance hierarchy must meet the following criteria:
The physiologic or biologic action of the combination must be enhanced or synergistic
Combinations that do not have an enhanced or synergistic effect (e.g. combinations created for convenience) are out of scope.”
Question for OIMP: How to handle such cases?
11) 6301-6 INR in Platelet poor plasma by Coagulation assay
12) 3184-9 Activated clotting time (ACT) of Blood by Coagulation assay
Coagulation assays are being discussed here.
CC: James R. Campbell Suzanne Santamaria Daniel Karlsson
Daniel Karlsson
Hi, some reflections/suggestions:
Use of IsAs seem like prototypical use case for refsets
When is | Direct site | populated?
When is | Population of erythrocytes in fluid | used vs. | Erythrocyte | (4544-3 vs. 787-2)
Property for 788-0? "ratio" is "Quotient of quantities of the same kind for different components within the same system." according to IUPAC Gold book. This quotient is SD / mean.
/Daniel
James R. Campbell
I was reviewing cardiac Troponin-T tests requested by one of our sites for our common lab/path ontology build. I noticed that Troponin T(T-T) in venous blood classified as subtype of T-T in blood, but that T-T in serum/plasma did not even though ser/plasma specimens and venous blood specimens are both subtypes of blood specimen(the direct site). This is apparently because the Observables are defined as Inheres_in (Blood(substance) and Plasma(substance)) but Plasma is not a subtype of blood.
A common use case in clinical lab decision support is "Did the patient have a blood test for XXX"? which usually means a mass concentration check for component XXX in blood/serum/plasma. The redundancy and ontologic inconsistency between specimens and substances leads us to another problem similar to past discussions regarding medicinal products and substances.
IS THERE AN ONTOLOGIST IN THE HOUSE?
PS: This is an example of the use case that is probably one driver to the creation of LOINC groups.
Daniel Karlsson
Hi Jim,
keep calm and look at <<256906008 | Blood material (substance) |
. Alternatively, the query your looking for is "any test performed on a blood sample" which would also work. The distinction between what is observed and how that is observed is one of the features of the Observables model that allows us to aggregate tests for the same "thing" (e.g. quantity) without having to consider how that was measured.
PS. I would really like to try to reproduce the groups using logic to see what's possible (and what's not as easy). Any one up for it? DS.
/Daniel
James R. Campbell
We have been using our laboratory/pathology Observables ontology build to create metadata for i2b2 interoperation. A collaborating site had large volumes of research data stored with 25 LOINC codes not in the technology preview. I have uploaded a spreadsheet with the 25 additional LOINC codes I modelled and classified with our ontology. I can spin off the expressions for these 25 concepts if we want them added to the tech preview
Jim
James R. Campbell
I have just completed a review and edit of the technology preview which we have converted to SNOMED CT concept definitions and classified with the rest of our observables work. I offer the following nsuggestion for revision of the tech preview expression set:
1) Neisseria meningitidis Z' Ag [Presence] in Isolate by Agglutination (observable entity) should not point to same component as ‘Z’. It looks to me like the 'Z' and ‘Z’’ antigens are probably different
2) Ionized calcium (and magnesium) observables are modelled as 'calcium electrolyte which has face validity, however the observables of total calcium and magnesium are moelled with the same component. I suggest that we add 'unbound calcium electrolyte' as a substance and use that to fully define all the measurements of ionized calcium and magnesium. I tested a couple concepts and it gives the desired behavior.
3) There are lots of observables for 'Leukocytes other' and other cell types. These are meant to be 'NEC' and are just incorrect in the modelling. I suggest discussion and revision
4) I don't think HTLV-1 g46 antigen is gp46 Ag which is how it has been deployed in concept definition
5) I added 3-Hydroxymyristoleylcarnitine (substance) so that I can properly define several concepts that were incorrectly modelled with another moleculae
6) I added Actin smooth muscle antibody (substance) since it is not ASMA so that I could properly define a concept
7) I added 11-oxo-Androsterone substance in order to correct modelling of a couple observables.
Jim
Suzanne Santamaria
Hi Jim,
Farzaneh Ashrafi and I reviewed your latest comments and have these responses:
1) We agree. We have created a new concept in SNOMED CT for the January 2019 release: Antigen of Neisseria meningitidis E (substance) with synonyms Antigen of Neisseria meningitidis 29E and Antigen of Neisseria meningitidis Z'.
2) We have submitted this question to RII for clarification and will discuss with the OIMP after we receive their feedback.
3) This has been discussed extensively in the past and it was decided that “Leukocytes other” could not be consistently defined, therefore we map it to the generic Leukocyte concept and mark expressions with these as primitive and LOINC narrower. Can you tell us more about how these should be revised?
4-7) We have submitted this to RII for clarification
Daniel Karlsson
Regarding 2) I tried to identify ions in SNOMED CT and could identify only 14(+5) (~specific substances) which had been named using three different naming patterns for ions ("X ion", "X electrolyte", "ionized X") plus some anions.
James R. Campbell
At this time our load of tech preview lab expressions into Nebraska Lexicon for our build of LOINC-on-OWL has ~170 validation errors, mostly due to defining relationships that point to inactive concepts in the substance hierarchy - presumably from all the model change work that Toni and the staff have been doing. Was any work done on maintaining or correcting those expressions to stay current with SNOMED CT?
Jim
Farzaneh Ashrafi
Hi James R. Campbell
Suzanne Santamaria and I have been working on maintaining the expressions up-to-date with SNOMED CT changes in different hierarchies (including the changes in the Substance hierarchy) since the release of LOINC expression production release in August 2017 (as a derivative to July 2017 SNOMED CT international release).
Thanks,
Farzaneh
Toni Morrison
James R. Campbell
Farzaneh
Great! Where can I access the updated refset?
Jim
Farzaneh Ashrafi
Hi James R. Campbell ,
At this time, only the activities related to maintaining the LOINC-SNOMED post-coordinated expressions are progressing. However, there is no scheduled release date for the updated RefSet since the project is on-hold.
Thanks,
Farzaneh
CC: Suzanne Santamaria
James R. Campbell
We are getting our microbiology in shape and of course we have started with the observables coding from the Technology Preview to jump start our LOINC terminology definitions. I am extending the material to include all culture and susceptibility reports from our micro lab. The new DNA and RNA detection methods have changed the clinical paradigm for early diagnosis of infectious disease and I have been starting to deal with those in our database.I enclose the Tech Preview definition of
40975-5|Adenovirus DNA Identifier in unspecified specimen by restriction fragment length polymorphism (observable entity)| from our SNOW OWL platform.
INHERES IN Adenovirus DNA
DIRECT SITE Specimen
PROPERTY Presence or identity
SCALE TYPE Nominal
TECHNIQUE Restriction fragment length polymorphism technique
I note that the test result is defined as the Presence or identification of Adenovirus DNA in the specimen. This is technically correct but the test is really about proving the presence of the VIRUS in the specimen. I therefore ask why it INHERES IN the DNA and not the virus with the DIRECT SUBSTANCE = Adenovirus DNA?
Jim
James R. Campbell
I have been validating the concept model definition of the most common clinical and lab observables deployed in our electronic record and started to think about
LOINC defines it as a mass fraction of oxygen in a blood arterial sample but this is not scientifically accurate. It is actually the fraction of oxyhemoglobin relative to total hemoglobin in the arterial blood sample and so I think the correct application of the concept model would be:
Oxygen saturation in blood by pulse oximetry
IS_A Clinical chemistry observable
COMPONENT Oxyhemoglobin
RELATIVE TO Hemoglobin
INHERES IN Blood substance
DIRECT SITE arterial blood specimen
SCALE TYPE quantitative
TECHNIQUE pulse oximetry
TIME ASPECT single point in time
Sarah Harry
I have blood gasses on my backlog for UK users. From the implementation end, I have to have something that can be taken from the concept to assert that this is displayed as a percentage (to avoid external references). 'Relative to' only gives a clue and I'd be keen to see the use of a sound Property that could capture this. Mass fraction etc doesn't do it. We could use 'Percent (property)' if there is consensus on that but strictly it's 'the fraction of oxyhemoglobin relative to total hemoglobin expressed as a percentage'. I also note there is no means of expressing 'per mille' (per 1000).
Daniel Karlsson
There is some guidance here: IHTSDO-979 Vital Signs Observables (search for "saturation")