1.which domain should I choose,observation、procedure or measurement?
2.Whether each scale has a total score ID,for example “PHQ-9 - Patient health questionnaire 9” (concept id 44804610)/“Patient Health Questionnaire 9 item (PHQ-9) total score [Reported]”(concept id 3042932)? If I can not find corresponding total score ID,how can I deal with it?
3.whether it is correct or not if i store every question of the scale as the same way like the total score?
4.what does “class” mean in the athena?For example,a concept name has two kinds of class,but they have same domain.How should i choose?
1.which domain should I choose,observation、procedure or measurement?
This allows you to connect questions to a questionnaire using SURVEY_CONDUCT table.
For the PHQ-9 case you can use Patient Health Questionnaire 9 item (PHQ-9) total score [Reported] to represent the total score (put in Observation table),
and you put in the OBSERVATION table related questions and answers (if you have this information):
Feeling down, depressed, or hopeless in last 2 weeks [Reported.PHQ]
Feeling tired or having little energy in last 2 weeks [Reported.PHQ]
If you can’t find some scales among standard OHDSI concepts, please report, and we consider adding them.
Concept_class identifies distinct semantic group of concepts. Nothing to do with CDM building, so disregard it.
Thanks a lot!You are a star!
1.How can i find a SURVEY_CONDUCT_ID?Should i use SURVEY_CONCEPT_ID,or use a Procedure_ID like “Assessment using Young mania rating scale36684932”?
2.As we have a question / answer record in the observation table,why should i link it to the patient questionnaire?What does the patient questionnaire mean?Which table should i put it in?
- SURVEY_CONDUCT_ID is just a number of a row in the table.
There’s no domain specified for SURVEY_CONCEPT_ID, so you can use Procedure concept (“Assessment using Young mania rating scale36684932” or Measurement concept_id 458501000124102 Young mania rating scale
- “Young mania rating scale” is a given questionnaire.
I suppose the idea of SURVEY_CONDUCT is to help to track which questions belong to which Scale (questionnaire).
So, if you connect OBSERVATION entries
Feeling down, depressed…
Feeling tired or having…
to the PHQ9 survey_concept_id you’ll know which questions make PHQ9.
(Not concepts but event_ids as said in the documentation).
And SURVEY_CONDUCT allows you to add additional information such as RESPONDENT_TYPE_CONCEPT_ID, so you can indicate whether it was done by patient or by caregiver, doctor, or someone else.
TIMING_CONCEPT_ID - so you indicate the follow-up period.
As far as I know there are not so many use cases of SURVEY_CONDUCT usage yet, so you’re welcome to share your thougths.
Anyway, I suppose, the main focus during studies would be on total scores, so people will just query Observation table for this purpose.
We have several scales:“Young mania rating scale”,“Quick inventory of depressive symptomatology”,and “Sheehan disability scale”.They have not got a total score id like " Patient Health Questionnaire 9 item (PHQ-9) total score [Reported]".How do I store their scores into OMOP?
You can use “Young mania rating scale” as MEASUREMENT.measurement_concept_id with total score in value_as_number.
We can make a convention: The scale stays in MEASUREMENT and even its description doesn’t have “total score” we keep total score in mind.
Let it be temporal solution until we collect more information about Psychiatric scales to make more mature decision
There’s a Psychiatry Working Group
You can make a list of questions and possible solutions (if present) and present in the one of the meetings.
There are a (growing?) number of standardized assessment and diagnostic survey instruments encoded in LOINC. There are LOINC Panels for the entire survey, LOINC codes for each question within the survey, and enumerated value choices (answer options) for those questions. For some of the answers, there are mappings to SNOMED to represent the clinical meaning.
Should surveys encoded in LOINC be stored in the MEASUREMENT table? If you have surveys with results stored across the MEASUREMENT and OBSERVATION tables, does that complicate querying for the presence of a survey administration and/or the scored result of survey scales?
Thanks a lot!
Yea,you’re right.It will make it complicate for a survey administration,because the store logic is complicate,and we still havn’t got enough survey data in the OMOP CDM.That results we won’t be able to do researches using the scales that they don’t have by Atlas.
We had a small discussion with @Dymshyts regarding the problems of representation of psychiatric scales in OMOP CDM. The sad truth is our usual favourite vocabularies that we use for standardization are poorly adapted to represent Psychiatric data.
- Some Scales don’t have any standard representation whatsoever, even if they are known and popular and not burdened by licenses (like HAM-A).
- SNOMED CT often has multiple standard concepts representing the same thing:
36674488 774093002 Global Deterioration Scale (assessment scale)
36676826 774094008 Global Deterioration Scale level (observable entity)
36676827 774095009 Assessment using Global Deterioration Scale (procedure)
4250116 407628009 Global deterioration scale finding for assessment of primary degenerative dementia (finding)
First concept represents the scale itself, second represents result, third represents the process of assessment and fourth the condition of the patient as having a certain score. Semantically they may be different, but clinically they mean the same thing. They should not all exist as standard concepts.
There should be a single standard concept representing the scale, belonging to Measurement domain. This domain should be chosen, because such scales are standardized tools that provide consistent standardized numeric (or qualitative in case of ranges) results.
3. LOINC is widely inconsistent in representing scales and scores (like PHQ-9)
This concept seems to be the root for all PHQ-9 questions (in two variations – as presence of symptoms and frequency of symptoms), but it includes total score on different levels to particular questions themselves.
From this, it seems that we need a new standard vocabulary that would include psychiatry Scales as standard Measurement concepts. This vocabulary could even include content of questionnaires, where the license of the Scale allows to share its contents.
The problem is that seemingly no such vocabulary exists. We could create one and support it or cooperate with a professional entity such as APA to create a maintained list. But it in any case is a longterm solution that would need ample time to implement.
So for the time being, the best advice is to use standard concept from Measurement domain and fill both measurement.value_as_number and measurement.value_as_concept_id (e.g. 10 in value_as_number and 4116992 (Mild) as value_as_concept_id).
SNOMED concepts representing scales and scores will be made Measurements next release. But we do need to find a new vocabulary to better represent psychiatric scales.
let me double-check this issue with you.
my understanding from conversation with Dmytry from last psychiatry WG is that
total score goes to Measurement table, but subscales/subquestions go to Observation table.
is it correct? or all go to Measurement.
Yes, the subquestions like
“Do you feel full of energy [GDS]”
“Do you feel that your situation is hopeless [GDS]”
“Do you think that most people are better off than you are [GDS]”
should be placed into Observation table.
Because is something reported by the patient.
And the total score should go to Measurement, because it’s something calculated by doctor.
Subscales should belong to Measurement domain as well,
MMPI consists from 10 subscales:
Scale 1 (AKA the Hypochondriasis Scale)
Scale 2 (AKA the Depression Scale)
Scale 3 (AKA the Hysteria Scale)
Scale 4 (AKA the Psychopathic Deviate Scale)
Scale 5 (AKA the Femininity/Masculinity Scale)
Scale 6 (AKA the Paranoia Scale)
Scale 7 (AKA the Psychasthenia Scale)
Scale 8 (AKA the Schizophrenia Scale)
Scale 9 (AKA the Mania Scale)
Scale 0 (AKA the Social Introversion Scale)
Those are verifyed calculated scores.
And I can’t find them in the OMOP vocabulary, so another need for a new vocabulary creation.
Now we are doing the following excercise:
We take all the scales met in source data and map them to all possible concepts regardless of domain.
This will give us the pattern and help to make weighted decision.
We’ll share the results.
Here’s the result of mapping to any possible concepts:Psychiatric scales mapped.xlsx (33.4 KB)
Here we can see two kinds of problems:
- No mapping - no concept in standard terminologies;
- Multiple mapping concepts possible, within different domains and vocabularies.
Among multiple mapping options semantically closest are
‘[Scale name] score’ concepts, for example: Beck Anxiety Inventory score,
but for some scales this particular type of concept is missing, see Young mania rating scale, for example.
So the possible solution is the following:
- We change the domain for all ‘[Scale name] score’ like concepts to measurement.
- For the cases when such concepts are missing we create OMOP Extension concepts.
- For such concepts as ‘[Scale name]’, describing scale as a tool, not as a result, for example Young mania rating scale, we change domain to Observation.
This way users will have the list of concepts describing various scales and subscales met in a patient data.
Such OMOP Extension vocabulary will be supported and will include new scales met in a real world data.
@Christian_Reich @aostropolets @Eduard_Korchmar @Alexdavv
What do you think?
And how are we gonna use these concepts? As attributes for something? I’d leave them in the Measurement Domain and use them in the same way as concepts expecting the score.
It’s just a mixture of concepts from 1 and 3. They are created by the same rules, have no attributes, not used as attributes. The only problem - SNOMED doesn’t follow the naming convention. We can ignore this.
We had a call with SNOMED and they might add the missing scale concepts.
We’ll work on new concept submission to SNOMED.
Dima, This is great news!
What about the issue with multiple concepts within different domains and vocabularies? Was there any recommendation from SNOMED?