[R-pkgs] icd9 - a new R package

[Jack Wasey]

Dear R people,

The new packge 'icd9' provides a range of tools for working with ICD-9-CM codes.

http://cran.r-project.org/web/packages/icd9/index.html
https://github.com/jackwasey/icd9

ICD-9 (clinical modification) is primarily used for categorizing
diseases in the USA for hospital administration, whereas ICD-10 is
used by the rest of the world for disease surveillance. This package
is currently restricted to ICD-9-CM codes.

I've seen other R code which manipulates ICD-9 codes, but the mistake
is often made of thinking they are numeric. This is not the case, e.g.
100.0 is different from 100 and 100.00 . This package takes care of
validating these codes, explaining them (converting code to plain
English), comparing them, and attributing codes to groups of codes to
assign co-morbidities to patients. ICD-9 codes are often provided in a
shortened format without a decimal place, and these have distinct
validation rules. Functions to convert between decimal and short forms
are provided. All key parts use vectorized code, and comorbidities for
a million patient visits can be assigned in a few seconds on a modest
workstation.

SAS code is published by AHRQ to allow assignment of ICD-9 codes to
comorbidities. This package contains some SAS source to R code
translation, so that the canonical ICD-9-CM to comorbidity mapping
provided by AHRQ can be derived directly without the cumbersome and
error-prone manual task of re-encoding the relationships in R. I
believe a SAS to R converter was an April Fools' joke some time ago,
but this is indeed a very limited answer to that problem.
http://www.biostatistics.dk/sas2r/index.html

A short vignette covers the major use-cases.
http://cran.r-project.org/web/packages/icd9/vignettes/icd9.pdf

The code is supported by a fairly thorough test suite, and is well
documented in the hope that it will be easier for users of the package
to understand it, and to get involved. I chose only to export key
functions where I had thought carefully about the external API, but
all internal functions are documented and contain potentially useful
nuggets for power users.

Comments and contributions are most welcome. In particular, I'd love
to see unit tests corresponding to any failures you may encounter
working with your own ICD-9 data.

Hope you find this useful.
Jack

--
Jack Wasey
Resident Physician, Anesthesiology and Critical Care Medicine
Johns Hopkins Hospital
Baltimore, MD, USA