Date:         Wed, 27 Aug 2003 19:12:25 -0400
Reply-To:     Melissa McClung <melissa.mcclung@DHHA.ORG>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Melissa McClung <melissa.mcclung@DHHA.ORG>
Subject:      PROC GENMOD w/GEE

Fellow Users - I have a question regarding the applicability of GENMOD
w/GEE to perform logistic regression.  I have a dataset of visits to a
health clinic where 75,000 patients have made 125,000 visits over the
course of 12 years.  70% of patients have only made 1 visit to the clinic
during the study period, 15% have made 2 visits and the remaining 15% have
made more than 2 visits, ranging up to 34 visits/patient.  Data are
analyzed at the visit level and a specific diagnosis at the time of the
visit is the dependent variable.  Is it valid to use GENMOD w/GEE to
perform logistic regression in order to account for multiple visits per
patient?  I obviously have a lot of missing data from SAS' standpoint.  Any
thoughts or suggestions? What should I use for the structure of the
correlation matrix?

Thanks for your help.
Melissa