OHDSI Home | Forums | Wiki | Github

Multiple Modifier with Single Procedure

cdm

(Tarun Shah) #1

There are multiple modifiers mapped with single procedure. Do we have to insert row for each modifier in PROCEDURE_OCCURRENCE?


(Don Torok) #2

An example of what you are experiencing will help in answering your question.


(Melanie Philofsky) #3

You don’t have to and I would definitely assess the time cost versus the return before adding multiple modifiers to the CDM.
Many of the CPT4 modifiers are not related to the clinical aspect of a Procedure and are instead supplemental facts for billing purposes and/or additional information needed for healthcare workflow. See the list of modifiers. I would critically analyze the data and then make a decision on whether all the modifiers are necessary for your use cases.


(Tarun Shah) #4

Hi, @MPhilofsky
How can we differentiate Modifiers in clinical, supplemental and as additional information. Also what if there are two clinical modifiers are attached?


(Melanie Philofsky) #5

Hi @TMS,

Have a clinical informatician assess the data. I suggest writing some SQL to pull the data into a csv:

SELECT
Modifier1,
Modifier2,
Modifier3,
count (*)
FROM Your data source
GROUP BY
1,2,3
ORDER BY
4 desc

I think this will be a rare occurrence. But you can add another row in your Procedure table with procedure_concept_id = 0 (you don’t want to repeat the procedure_concept_id because your Person only got one Procedure) and then the modifier in the modifier_concept_id and link them using the Fact Relationship table. O


(Ajit Londhe) #6

Hi all,

Revisiting this issue. So the options are:

  1. Select the most clinically important modifier for each patient’s procedures
  2. If we want to keep all of them, then insert additional records into procedure occurrence and use fact relationship to tie them to the event

Are there any other methods people are using?

Thanks,
Ajit


(Melanie Philofsky) #7

Colorado puts the first modifier in the CDM and leaves the rest in the source. When someone needs more than the 1st modifier, then we go back to the source to grab the data. We haven’t had an Atlas use case for this data, yet.

Option #2 is a bunch of extra work considering Atlas doesn’t use the Fact Relationship table. #1 sounds somewhat feasible, but what if you have two clinically important modifiers? And ‘important’ probably depends on the point of view. Clinician researchers are probably more interested in clinical modifiers while those studying health economics are more concerned how many Assistant Surgeons were present.


(Robert Miller) #8

@Ajit_Londhe Another approach would be to use the expanded measurement table from the oncology extension, which allows you to store unlimited modifiers for any domain.


(Qi Yang) #9

I believe what @rtmill said is the reason that @Christian_Reich is considering to remove the modifier_concept_id from procedure_occurrence table.

The 2 extra columns in measurement table, namely modifier_of_event_id and modifier_of_field_concept_id can be used as modifier of any clinical event tables, such as condition_occurrence, procedure_occurrence etc. And it does not have to be restricted to the oncology space.


(ella) #10

That’s very interesting… to date, we have transformed ‘bilateral’ procedures in our source to two procedure records in OMOP - one with a modifier ‘left’ and one with a modifier ‘right’. We repeat the procedure_concept_id (don’t use 0 for the second record). Since we use i2b2 cohort browser to represent the data, there are no risks of duplications. I’m wondering if anyone else does this or are there other/better ways ? Thanks !


(Qi Yang) #11

@ellayoung

I would just use Bilateral procedure as the modifier. Save you a record.


(ella) #12

I guess that’s an option, but then wouldn’t we lose the right/left part? that’s very important for our users to know…


t