Condition vs Measurement domain for test results

Dymshyts · February 14, 2019, 4:59pm

Currently we are working on MedDRA clean up and one of the tasks here is proper domain assignment.
So, there are such a concepts like

concept_id	concept_name
42889410	Acquired gene mutation
42889957	Paravaccinia virus test positive
42892934	BK polyomavirus test positive
42889423	Genetic polymorphism

that stand for the test results

We have similar type of concepts in SNOMED that describes the test and its result:
4135410 BRCA1 gene mutation positive
4135411 BRCA2 gene mutation positive
etc
Currently these concepts have “Condition” domain. the logic behind: some patient condition is stated.
and also a lot of test results interfer with syndromes and disorders, like “Red blood cell count decreased” = “Anemia”, “Granulocytes level increased” = “Granulocytosis”, etc etc.
So not to confuse these we decided that all should be conditions.
please look here:

github.com/OHDSI/Vocabulary-v5.0

Changing Domains

opened 03:27AM - 17 Jun 17 UTC

closed 02:24PM - 05 Jun 18 UTC

ericaVoss

Looks like about 1,198 standard concepts changed domains between Vocab 20161231 …to Vocab 20170503. Most look good, however a few look incorrect: ```{TEXT} CONCEPT_ID CONCEPT_NAME CHANGE 4046886 Smoking reduced Observation --> Condition 4308357 Goal not achieved Observation --> Condition 44807379 Complies with dietary regimen Observation --> Condition 44811248 Goal achievement status unknown Observation --> Condition 40486696 Smoked before confirmation of pregnancy Observation --> Condition 4308000 Noncompliance with infection control regimen Observation --> Condition 4052030 Moderate cigarette smoker Observation --> Condition 46273081 No known exposure to tobacco smoke Observation --> Condition 4190830 Poor compliance with inhaler Observation --> Condition 45770364 Inappropriate use of walk-in centre Observation --> Condition 37016593 Non-compliant with peritoneal dialysis prescription Observation --> Condition 4058137 Trying to give up smoking Observation --> Condition 4246988 Compliance issues discussed with patient Observation --> Condition 4222303 Non-smoker Observation --> Condition 4265571 Finding related to ability to comply with treatment Observation --> Condition 4092281 Ex-cigarette smoker Observation --> Condition 4043060 Chews twist tobacco Observation --> Condition 4276526 Cigarette smoker Observation --> Condition 4032524 Noncompliance with therapeutic regimen Observation --> Condition 45773499 Appropriate use of walk-in centre Observation --> Condition 46270993 Complies with therapeutic regimen Observation --> Condition ``` This could be related to #96. Maybe in OHDSI we have decided smoking is a condition? I didn't capture everything, most of them looked related to smoking or treatment compliance. Just a simple query to review ```{SQL} SELECT '9.18' AS TEST_ID, 'Consistency Check - Changing Domains' AS TEST_DESCRIPTION, 'CONCEPT' AS VOCAB_TABLE, CAST(c1.CONCEPT_ID AS VARCHAR(50)) AS ID, 'CONCEPT_ID' AS ID_NAME, 'CONCEPT_NAME <> CURRENT <> PREVIOUS' AS EVIDENCE_COLUMNS, CAST(c1.CONCEPT_NAME AS VARCHAR(50)) + ' <> ' + CAST(c1.DOMAIN_ID AS VARCHAR(50)) + ' <> ' + CAST(c2.DOMAIN_ID AS VARCHAR(50)) AS EVIDENCE FROM CONCEPT c1 JOIN OMOP_VOCABULARY_20161218.dbo.CONCEPT c2 ON c2.CONCEPT_ID = c1.CONCEPT_ID AND c2.DOMAIN_ID != c1.domain_id WHERE c1.STANDARD_CONCEPT = 'S' ORDER BY c1.DOMAIN_ID, c2.CONCEPT_ID ```

But on the other Hand it make sense to put a concepts in a Measurement domain, so,
If I want to find all patients with “BRCA1 gene mutation positive”
I will search for
measurement_concept_id = “BRCA1 gene mutation test” with value_as_concept_id ='Positive"
or
measurement_concept_id = “BRCA1 gene mutation positive”, because both variants can be present in a data.
Thus I need to look onto MEASUREMENT only.
Currently I need to look on
“BRCA1 gene mutation test” with value ='Positive"
in MEASUREMENT
but on
“BRCA1 gene mutation positive” in CONDITION domain.

@Christian_Reich, @aostropolets, @ericaVoss please advice

Alexdavv · February 14, 2019, 8:31pm

@Dymshyts

I think we should go against the tricky hierarchy of SNOMED (and some other vocabularies) here and work a little bit on the domain selection according to the CDM vision.

In my opinion, all the syndroms (anemia, granulocytosis, …cytosis, …penia, NOS genetic polymorphism, NOS mutations, etc.) should live in the Condition domain. I believe they don’t always reflect just the results of the measurement performed, or even may represent the results of the series of tests.

While the concepts of single test, especially having the concrete result (low, high, increased, decreased, POS/NEG) may be moved to Measurement domain.

Apparently, we’ll find some complicated cases here, e.g. increased function level, abnormal result, etc.
But let’s hear what people say first.

Christian_Reich · February 15, 2019, 3:58am

A lab test doesn’t make a condition, positive or not. A condition is defined as per OMOP CDM: “Conditions are records of a Person suggesting the presence of a disease or medical condition stated as a diagnosis, a sign, or a symptom, which is either observed by a Provider or reported by the patient.”

So: Measurements. Looks like you need to re-assign the domain and map them.

Dymshyts · February 18, 2019, 10:13am

Hm, let’s do this way:
when it’s definitely a Measurement with result then it’s Measurement, for example MedDRA has these packs of concepts:

10007977	Central venous pressure
10007978	Central venous pressure abnormal
10007979	Central venous pressure decreased
10007980	Central venous pressure increased
10081773	Modified Rodnan skin score normal
—	—
10081776	Modified Rodnan skin score abnormal
10081780	Modified Rodnan skin score
10007981	Central venous pressure normal

or SNOMED

412736006	BRCA1 gene mutation negative
412739004	BRCA2 gene mutation negative

When it’s a concept more or less defining a condition, then it’s Condition, for example

165509000	White blood cell count abnormal
762656009	Abnormal blood cell count
165397008	Hemoglobin low

all these concepts are ancesors of different types of Blood Disorders, so they are conditions.

Dymshyts · February 21, 2019, 3:12pm

@Christian_Reich,
and what about these:

10002440	Angiogram pulmonary
10002441	Angiogram pulmonary abnormal
10002442	Angiogram pulmonary normal
10006465	Bronchogram
10006466	Bronchogram abnormal
10006467	Bronchogram normal
10006479	Bronchoscopy
10006480	Bronchoscopy abnormal
10006481	Bronchoscopy normal

Procedure without result is a Procedure,
with result is an Observation?

Christian_Reich · February 22, 2019, 11:40am

Nope. It’s just a Measurement.

Can’t say I agree. It’s not even clear whether it is high or low, at which point you might call them leukocytosis or leukopenia, but even those are not Conditions.

Procedure

These are tricky, because the Procedure “Angiography” has a result. But the result is most likely a Condition, not a Measurement. This is a tricky case. You don’t know what that “abnormal” thing is. A pulmonary embolism?

So, I’d call them Procedures.

Dymshyts · February 22, 2019, 4:29pm

Those are Conditions as they are parents of “Anemia”.
It’s like 118939000 “Disorder of Neck”, It’s not even clear which organ is affected and how. But it’s a parent of 34486009 “Hyperthyroidism” (exact disorder).
So, you never see something like |165509000|White blood cell count abnormal in a real patient data. But we need to deal with its children domains. So it’s a Condition.

I agree.
We have the same logic for this one as for “Angiogram pulmonary”. We expect to have explicit result in NOTE table connected to PROCEDURE table.

Christian_Reich · February 22, 2019, 10:14pm

[quote=“Dymshyts, post:7, topic:6128”]th
Those are Conditions as they are parents of “Anemia”.
[/quote]

Need to complain to SNOMED. Same issue like with Polyuria, remember? They fixed it.

There is one thing to have peace with the SNOMED hierarchy decision, the other one is to assign a domain to an OMOP concept, so that the analytical use cases are met. And quantitatively measuring something in the blood is by our own definitions a Measurement.

Chris_Knoll · February 23, 2019, 6:11am

I really this all was as simple as:

It’s a measurement if it is a value (possibly with units).

It’s a condition if there is a judgement about a person to assign it a diagnosis.

Measurement: Patient weights 250 Kilograms
Condition: Patient is Obese. Patient is Abnormally Heavy, etc

It’s a grey area that certain observations are measurement because they say ‘positive’: you could call everything found in Condition Occurrence as ‘Positive finding of X’.

Someone could take a measurement where the finding was negative, but that’s implying that there was no measured value of what they were looking for…so…measurement. Likewise, a positve test means they were looking for something and they found it. But these cases of ‘abnormal findings’…i think they are describing the person is in the state of disorder (it’s not normal) and not necessarily reporting a measurement was made.

But other cases where the person is described as having low blood sugar and so we feel we put that in measurement because it sounds like something was measured to be ‘low’, I think that’s more of a type of condition diagnosis of the state of the person.

Is ‘HIV Positive’ a condition or a measurement? Seems like you get diagnosed as HIV Positive, even tho somewhere a test/measure was made…

Chris_Knoll · February 23, 2019, 6:13am

What was the fix here? Dr. Wikibedia describes it as a Genitourinary Disorder…did SNOMED make it a measurement or a condition?

Christian_Reich · February 23, 2019, 12:35pm

@Chris_Knoll: You learn that in Med School like a gazillion times: Test results are not Conditions. Conditions are conclusions, that take into account a diagnostic process with many inputs. Even HIV positive - you can be totally healthy with that (either before the manifestation or later under drugs). The Condition is called AIDS.

Plus: We have to be clear on this, otherwise we perpetuate the mess that those MedDRAs, ICDs, HCPCS, CPT4s etc. put us in: Where everything is mixed together.

You are right with the grey area of “Findings”. If the finding is pronounced by the patient or doctor (usually called symptom or sign) we still consider it a Condition. Mostly because there is a continuum between them and we cannot distinguish. But if the finding is found by running a test it is a Measurement. Doesn’t matter if the Measurement comes to us through a LOINC code in a lab test table, or these ugly diagnostic codes.

They had it as a Condition, and we suggested making two out of them, one a Condition and one a Measurement. And they did, but now I found out they deduped them and it is back a Condition. Urgh. Need to complain again.

Chris_Knoll · February 23, 2019, 3:48pm

You didn’t really address this example where there’s a measurement (I think we agree patient weight is a measurement) and a condition (something determined as a result of a test: the measure of weight)

I guess we’ll agree to disagree here. Taking some of your examples that you call measurements:

leukocytosis : Dr Wikipedia says: “it is white cells (the leukocyte count) above the normal range in the blood. It is frequently a sign of an inflammatory response,[3] most commonly the result of infection, but may also occur following certain parasitic infections or bone tumors as well as leukemia”.

So, in order to determine that the patient has a disorder that their white blood count is above normal they measured leukocyte count, got a value (that’s the measure) and due to the measure they diagnosed a disorder (leukocytosis). Same thing with leukopenia, just they did the same measurement but came to a different diagnosis. Like you said, the measurement is not the diagnosis, but in the same statement you’re saying the diagnosis is the measurement.

How about this: overactive bladder. The way you know you have an overactive bladder is a certain frequency of urination in a time period. The measurement would be count (the value) per hour (the unit). From that frequency, someone determined you have ‘overactive bladder’ (the condition).

And to be clear, i’m not trying to joust with you about medical expertise and knowledge, I have none of that. I"m trying to set up some simple rules that the tools can adhere to and our users can follow. I looked at some of the conditions you are calling ‘measurements’ and in all those cases they describe it as ‘caused by something’. Measurements aren’t ‘caused by’ anything. Measurements just are, and they don’t make any judgement about the condition of the patient. That’s why a ‘score’ is a measure but saying it’s ‘abnormal’ should be the condition.

Christian_Reich · February 23, 2019, 4:06pm

@Chris_Knoll:

Not sure where the disagreement is supposed to be, really.

Because yes. The weight of 250 kg is a Measurement, the Obesity is a Condition. Note: Obesity is a lot more than the weight. It’s an accumulation of fat in various places, and the consequences of this being a chronic state with consequences on the metabolism and musculoskeletal system. The max speed of 200 mph in a sports car is a Measurement, the Porsche is a sports car.

Same thing. Leukocytosis is the Measurement, inflammation, infection, bone tumor and leukemia are the conditions.

Your own quotation above says it’s a sign, and it is white blood count above normal. How did you now come to the conclusion somebody diagnosed it is a disorder?

No difference. The frequency of urination is a Measurement, the Overactive Bladder a Condition. But you can pee too frequently by just drinking too much beer.

Sometimes, these Measurements only leave one conclusion, and only one Condition can cause it. But that is not typical, and the process of diagnosing will reveal the Condition.

Simple rule: Measurements are “records of structured values (numerical or categorical) obtained through systematic and standardized examination or testing of a Person or Person’s sample.” Often they come with thresholds that tell us whether or not they are physiological (normal) or not. But these are not perfect. They depend on circumstances: If you e.g. lived on the Andes your hemoglobin would be certainly above normal, yet you are just fine. Conditions are "records of a Person suggesting the presence of a disease or medical condition stated as a diagnosis, a sign, or a symptom, which is either observed by a Provider or reported by the patient.”

Don’t think we need to change anything.