Re: st: SVY medians and Elixhauser

[Richard Goldstein <richgold@ix.netcom.com>]

Mike,

based on my reading of the literature, the use of up to 29 indicator
(dummy) variables is what Elixhauser calls for

I have never used the -elixhaus- command or dialog box; however, the
diagnosis code being asked for are comorbidity diagnoses in your data;
e.g., I am currently using a data base that has 11 ICD-9 variables:
primary diagnosis and up to 10 comorbidity diagnoses - so, I use my code
on the 10 comorbidity ICD-9 variables -- this is not to be used on your
outcome variable

sorry, not familiar with "diagnostic related groups"; but I understand
that there is lots of literature on this (search DRG has well as the
whole words)

I assume that "input and output files" mean exactly what you guess in
your parenthetical to your point "3"

Rich

On 4/14/13 2:56 PM, Mike Butterfield wrote:
> Hi Rich, I appreciate your help on this.
> 
> My general question has been--how do you use this elixhauser command?
> 
> 1. So there are 29 Elixhauser comorbidity variables that should be
> created (somehow).  Does this mean that when I run an analysis, I will
> have to pick and choose which ones to include as covariates?  For
> example, in predicting death from an infection:
> 
> logistic died infection elix1 elix12 elix17?
> 
> I would think there'd be a more elegant way to do this.
> 
> 2. I don't understand db elixhaus command itself.  My understanding is
> that it's trying to create elixhauser comorbidities for my database,
> creating a copy of the database in the process.   When I type "db
> elixhaus," I get a pop-up display that asks me to:
> 1.  Enter a diagnosis code prefix (Is this for my outcome variable?)
> 2. Enter a diagnostic related group variable name (?)
> 3. select input and output files (? Does this mean my database and the
> new name/location for the copy of the database with these new
> elixhauser variables added?)
> 
> The "help" information for elixhaus doesn't talk about any of these.
> 
> -I supposed that by "diagnosis code prefix" they mean ICD9 for my
> outcome variable, but the program wouldn't accept any version that I
> offered ("no such diagnosis code prefix")
> -Supposing that they were asking for comorbidities to include in the
> Elixhauser index to be applied in this particular case, I used the
> example of hepatocelluar carcinoma (ICD9=155) and corresponding DRG
> code but got the same result.
> 
> Best,
> -Mike
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/