OHDSI Home | Forums | Wiki | Github

ICD9CM vs ICDProc

Vocabulary Users
#1

@Dymshyts, @aostropolets, or @Christian_Reich,

Could I ask you what makes up the ICD9Proc in the OMOP Vocabulary?

@TBanokina asked what is the difference between ICD9Proc and the procedure codes in ICD9CM. She did an experiment of mapping to Truven and found the ICD9Proc does better than ICD9CM (~100% vs ~60% respectively). The Truven documentation says that we should use ICD9CM to map to procedure codes but we seem to get a better mapping with ICD9Proc.

Is the ICD9Proc in the Vocab directly from WHO? Does anyone know if in the US people primarily use ICD9Proc directly from WHO instead of ICD9CM?

This is related to this thread:

#2

Hi @ericaVoss,

Originally ICD9CM contains 2 code systems -for diagnoses and procedures.
These are divided on to 2 different vocabularies in OMOP CDM : ICD9CM (for diagnoses) and ICD9Proc (for procedures).

Experiment of mapping is good showcase of codes intersections, so it should be specified in source data whether these codes are for diagnoses or procedures to choose proper OMOP CDM vocabulary.

1 Like
#3

@Eldar thank you for the reply.

Vendor documentation was suggesting that the procedure codes were ICD9CM however based on how it shows in the OMOP Vocabulary, it looks like we should use ICD9Proc and ignore the vendor comment here.

#4

60%? Impossible. ICD9Proc is indeed 100%, because all Concepts are Standard, but ICD9CM is in the 99% mapping range. And the stuff that doesn’t map is mostly unmappable junk.

#5

@TBanokina could you provide more detail on what you saw?

#6

Actually you shouldn’t ignore but follow this comment - it contains word ‘procedure’ :smile:

@Christian_Reich
I think they don’t mean mapping rate of ICD9CM to SNOMED as standard one, but rate of source codes mapping.
Seems like assuming ICD9Proc to be source vocabulary gives 100% source codes detected while ICD9CM doesn’t contain 40% of them.

#7

@Eldar:

Not sure what you mean. Are you saying there is a combined ICD9CM and ICD9Proc field, and therefore you get 60% mapping against ICD9CM only? Then ICD9Proc should get only 40%.

Anyway. Let’s hear from the horse’s mouth.

#8

Hi @Christian_Reich, @eldar, @ericavoss,

I meant mapping of the source codes to the source vocabulary. In TRUVEN all source codes don’t contain a decimal point and therefore the same codes can be mapped to different vocabularies at the same time. From the analysis I saw that 60% of the codes may belong to ICD9CM and if the same codes to map to ICD9Proc – then, 99% of the matches.

The questions that I was having is: why in the documentation it is written that procedure fields coded in ICD9CM while on the data I see a great mapping to ICD9Proc.

As far as I understood: ICD9Proc vocabulary exists only in OMOP CDM and it is part of the ICD-9-CM classification system. So, when in the documentation it is written “ICD9CM procedure codes” and we have a good mapping rate to ICD9Proc– it means that we need to map to ICD9Proc vocabulary. Is it correct?

#9

@TBanokina:

Well, I can’t help with Truven. That’s the database we are not selling. :smile:

Regarding ICD-9-CM and -Proc: ICD-9-CM was released by the CMS, and it contained three Volumes:

  • Volume 1: The numeric listing of diseases, classified by etiology and anatomical system, along with as a classification of other reasons for encounters and causes of injury. This is called the tabular section of ICD-9-CM. Volume 1 is used by all health care providers and facilities.
  • Volume 2: The alphabetic index used to locate the codes in Volume 1. Volume 2 is used by all healthcare providers and facilities.
  • Volume 3: A procedural classification with a tabular section and an index. This set of procedure codes is used only by hospitals to report services.

So, what folks generally call ICD-9, and we call ICD9CM, are Volume 1 and 2 diagnostic codes. What folks generally call ICD-9 Procedure Codes or ICD-9-CM Procedure Codes, and we call ICD9Proc, are Volume 3 procedure codes.

Makes sense?

What Truven does, and how to distinguish dotless ICD9CM from ICD9Proc - beates me.

t