I think you can only stset with one origin.
I wrote a function, survtime, to find a event date and
corresponding indicator for one or more events or
censors.
If you have any interest, I would be happy to share
it.
An example of the syntax is in comments /* ... */
below
generate dolast=dodischg+90
assert dolast~=. /* everyone must have a valid censor
point */
gen temp0=(doreadm<=dolast)
gen byte readmit=temp0
gen readmit_d=doreadm if temp0==1
replace readmit_d=dolast if temp0==0
gen readmit=1 if temp0==1
replace readmit=0 if temp0==0
drop temp0
/* survtime, event(doreadm) censor(dolast) seq
generate(readmit) */
stset readmit_d, failure(readmit==1) origin(doadmit)
id(ptid) scale(30.4) /* e.g. scale _t for months */
-Allen Buxton
--- Alexander Tsai <[email protected]> wrote:
> Dear Statalisters,
>
> I have data on patients, readmissions, and deaths:
>
> . list ptid doadmit doreadm dod
>
>
>
+----------------------------------------------------+
> | ptid doadmit doreadm
> dod |
>
>
|----------------------------------------------------|
> 1. | 00000125 29 Aug 1996 14 Nov 96 20
> Oct 1997 |
> 2. | 00000133 20 Jul 1991 .
> . |
> 3. | 00000491 21 Mar 1997 . 26
> Jun 1999 |
> 4. | 00000865 06 Mar 1996 .
> . |
> 5. | 00000932 13 May 1997 11 Jul 97
> . |
>
>
|----------------------------------------------------|
> 6. | 00001354 04 Jun 1994 .
> . |
> 7. | 00001610 10 May 1995 .
> . |
> 8. | 00001804 24 Sep 1996 . 12
> Jul 1997 |
> 9. | 000020 20 Aug 1992 . 09
> Sep 1993 |
> 10. | 000025 28 Oct 1996 . 24
> Aug 1999 |
>
>
|----------------------------------------------------|
> 11. | 000034 20 Nov 1992 07 Jan 93
> . |
> ...
>
> I want to examine readmissions only up to 90 days
> post-discharge, so I
> create an extra variable, 'dolast':
>
> generate dolast=dodischg+90
> format dolast %dD_m_CY
>
> But all of the observations with doreadm==. are,
> obviously, missing. So
> when I -stset- my data, I obtain the following
> results:
>
> . stset doreadm, id(ptid) time0(doadmit) origin(time
> doadmit)
> failure(event) exit(event==2 time dolast)
>
> id: ptid
> failure event: event != 0 & event < .
> obs. time interval: (doadmit, doreadm]
> exit on or before: event==2 time dolast
> t for analysis: (time-origin)
> origin: time doadmit
>
>
------------------------------------------------------------------------
> ------
> 23324 total obs.
> 18608 event time missing (doreadm>=.)
> PROBABLE
> ERROR
>
------------------------------------------------------------------------
> ------
> 4716 obs. remaining, representing
> 4716 subjects
> 4716 failures in single failure-per-subject
> data
> 184319 total analysis time at risk, at risk from
> t = 0
> earliest observed entry
> t = 0
> last observed exit
> t = 90
>
> What I would like to do is:
>
> a) I would like to remove patients from the study if
> they die within 90
> days post discharge. But I can't have two time
> variables in the exit
> option -- for example, (exit time dod dolast). So
> what I did is set
> event==2 for death (event==1 for readmission).
>
> b) How do I keep the patients with no readmission in
> the study? I
> thought about replacing doreadm=dolast if doreadm==.
> but that doesn't
> seem right.
>
> Cleves/Gould/Gutierrez' book "An Introduction to
> Survival Analysis Using
> Stata" has been exceedingly helpful, but I am stuck
> here. I'd be
> grateful for any help.
>
> Thanks,
> Alex
>
>
>
> *
> * For searches and help try:
> *
> http://www.stata.com/support/faqs/res/findit.html
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
