OHDSI Home | Forums | Wiki | Github

How to interpret Diabetic complications not included in T1 and T2 disorders

Implementers
, , , ,
#1

Hello Everyone,

I was looking at the concept hierarchy in Athena for Diabetic complication Click here for hierarchy

I see that under “Diabetic complication”, we have 37 terms. 2 of which are “Disorders due to Type 1 Diabetes mellitus” and “Disorders due to Type 2 Diabetes mellitus”.

But we still have remaining 35 terms under Diabetic complication which isn’t under any of the above two categories (Type 1 or Type 2)

May I know what does that mean?

Being new to healthcare, I wa/s of the assumption that Diabetes can either be Type 1 or Type 2. And In extreme cases, Type 2 can become Type 1.

But the complications are usually grouped under Type 1 or Type 2.

Since these 35 terms don’t fall under Type 1 or Type 2 disorders, what do they mean or how should I interpret them? Whether it was caused due to Type 1 or Type 2? Or it’s like these complications can occur for both Type 1 and Type 2. Overlap is possible?

Can anyone help me with this?

#2

Hello!

OMOP CDM primarily uses SNOMED CT hierarchy for disorders. It is provided by organization called IHTDSO and used in standardized vocabularies as is.

It is important to note that SNOMED tries to keep concept names unambigious and otherwise follows “open world model”. It means that if a concept name is stated like Diabetic foot ulcer, it means “Any kind of ulcer of any degree anywhere on foot caused by any kind of diabetes”. All concepts with narrower meaning will be descendants of this concept.

SNOMED allows for for multiple axes of hierarchy, allowing to classify each concept as a subtype of more than one concept. It means that concepts like Disorder of eye with type 2 diabetes mellitus will be a direct descendant of both Disorder of eye due to diabetes mellitus and Disorder due to type 2 diabetes mellitus; both of which in turn are descendants of Diabetic complication on the same hierarchical level.

Depending of what the nature of your study is (e.g. comparing T2DM vs. T1DM complications or looking for specific complication regardless of type of underlying diabetes), you can use different ancestors for concept set definition.

To explore and understand SNOMED’s internal hierarchy, I can recommend using the SHRIMP browser. To me Athena’s hierarchy browser feels cumbersome.

1 Like
#3

Hi @Eduard_Korchmar - Thanks for your response. Yes, I understand now.

I am trying to segregate patients as T1 and T2DM based on complications.

So in this case, disorders due to Type 1 and Type 2 would be the right thing to do? Am I right?

But for example, in our data of T2DM patients, one patient has conditions like below during his hospital journey (multiple visits)

"Gastroesophageal reflux disease"
"Tinnitus"
"Blood chemistry abnormal"
"Gastroesophageal reflux disease without esophagitis"
"Manifest vertical squint"
"Nonulcer dyspepsia"
"Chronic gastritis"

and other patient has condition like

"sleep apnea"
"obesity"

In a case like above, he/she may not be tagged as T2DM because the above conditions are not present in the disorder hierarchies. But I know for sure that he/she is a T2DM patient.

So do I have to manually add them in my concept set to be tagged as T2DM? Or it’s mandatory that person who has been diagnosed with T2DM with have atleast one of the disorders from the hierarchy? Or it’s possible that our data source doesn’t have info about his disorders but only the above condition info for the patient discussed. Might be his disorder info is recorded in some other hospital?

How should the above two patients be considered when they don’t have any disorders (present in hierarchy) of T2DM? But I know for sure that they are T2DM

#4

If you know that the patient is diagnosed with T2DM, you could add a corresponding entry into CONDITION_OCCURRENCE table. You could use Observation recorded from EHR as condition_status_concept_id if it is unclear how exactly were the patients diagnosed and the date of earliest known encounter as condition_start_datetime .

If you need to transform a disorder to a T2DM complication, you could look for the nearest common descendant of Disorder due to T2DM and the condition in question. If needed combination is missing, you could create local 2-bilion Standard concepts which could be made direct descendant of both.

#5

Hi @Eduard_Korchmar - Thanks. And yes, I am aware of the 2 billion concept id etc. Actually I was trying to understand the clinical aspects of it

t