Add support for ICD-9

[salmasian]

It would be extremely helpful if ICD-9 support is also added (based on the Deyo et al. paper for Charlson, and a similar source for Elixhauser).

[ellessenne]

Hi! I am not really familiar with ICD-9 coding, what is the difference between the Deyo et al. paper and the Quan et al. paper? Is it ICD-9 vs ICD-9-CM?

[aalexandersson]

Hojjat: I agree. However, the R package icd already does this though its AHRQ Elixhauser implementation is dubious.

Alessandro: No, both Deyo et al. (1992) and Quan et al. (2005) use ICD-9-CM codes as starting point. Deyo et al. coded only Charlson (1987) comorbidities whereas Quan et al. also code Elixhauser (1998), and for both ICD-9-CM and ICD-10.

On Wed, Feb 7, 2018 at 11:24 AM, Alessandro Gasparini < notifications@github.com> wrote:

Hi! I am not really familiar with ICD-9 coding, what is the difference between the Deyo et al. paper and the Quan et al. https://www.ncbi.nlm.nih.gov/pubmed/16224307 paper? Is it ICD-9 vs ICD-9-CM?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/ellessenne/comorbidity/issues/1#issuecomment-363824412, or mute the thread https://github.com/notifications/unsubscribe-auth/AThPml-c5TQqqTciLX8ZtsKa1NzaDrMQks5tSc46gaJpZM4R8z-U .

[salmasian]

The icd package is not well maintained, and cannot handle even modest-size data. I have found your package extremely useful, and I thin it would be great if you would extend it to support ICD-9 too!

[ellessenne]

@aalexandersson, Thanks for the clarification! I will refer to the Quan et al. paper then.

@salmasian, I will definitely look into adding support for ICD-9-CM coding, please keep an eye on this issue as I start working on it. Do you know (by any chance) if it is safe to just disregard any punctuation (e.g. the dots) when identifying/recognising ICD-9 codes? With ICD-10 that works well!

[salmasian]

@ellessenne the dot should be remove from the ICD-9 input, before checking it against the Deyo/Quan lists.

PS: I might submit a pull request myself, so just give it a few days, and if you don't hear from me, start working on it.

[ellessenne]

@salmasian Ok thanks! It should be enough to add a charlson_icd9 internal function that gets called from the main comorbidity function!

[aalexandersson]

@ellessenne Please also add an elixhauser_icd9 function (ideally AHRQ version) because Elixhauser is often regarded as an improvement over Charlson-Deyo. I am not aware of anything better than icd for ICD-9-CM in R. There are some more options in Stata and SAS.

@salmasian How bad is the performance in icd? Does it stop working at thousands of records or at millions or what is modest-size data here to you? I recommended icd because the documentation looks good and it works on the small Vermont sample data it comes with. But I have not tried the package on realistic datasets. I need to work with complicated cancer datasets that have about 4 million observations. Your comment makes me very concerned about the usefulness of the package icd.

On Wed, Feb 7, 2018 at 4:56 PM, Alessandro Gasparini < notifications@github.com> wrote:

@salmasian https://github.com/salmasian Ok thanks! It should be enough to add a charlson_icd9 internal function that gets called from the main comorbidity function!

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ellessenne/comorbidity/issues/1#issuecomment-363924213, or mute the thread https://github.com/notifications/unsubscribe-auth/AThPmkRnQ9Pu9CxuuOUetHljgae70aurks5tShwEgaJpZM4R8z-U .