Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
| From | Ronán Conroy <rconroy@rcsi.ie> |
| To | "statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu> |
| Subject | Re: st: Cox multilevel |
| Date | Wed, 17 Nov 2010 13:31:08 +0000 |
On 16 Nov 2010, at 15:34, Celine Genty wrote:
I am trying to make a multilevel model of survival data, so with the -shared()- option. I have about 5000 patients included by 300 center. But 45 center have only 1 patient (with a median of 7 patients per center IQR 3-21).Is multilevel model appropriate in this case ? Is there too much center? Is there too much center with only 1 patient ?
If you really want to examine the effects of context (centre, in your case) then clearly you need an informative sample of patients from each context.
It seems to me that you might begin with the obvious: the number of patients per centre is probably a good index of turnover, and you might wonder if patients at centres with high turnover had a different disease profile and a different prognosis. Different disease profile because the more difficult patients tend to get referred onwards to the centres of excellence, and different prognosis because the busy hospitals are supposed to be the centres of excellence.
Ronán Conroy rconroy@rcsi.ie Associate Professor Division of Population Health Sciences Royal College of Surgeons in Ireland * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/