I am working on a survival analysis using PROC PHREG (SAS EG 17.1). I have recurrent events in my dataset. I am not currently interested in doing a recurrent event analysis (that will be separate) but am using PROC PHREG to analyze time to 1) first any event, 2) death event.
The outcome events are: 2 = died, 1 = hospitalized, 0=censor.
The exposure of interest is Treat: 1=treated, 0=not
I have data stored in counting process as the treatment is allowed to change during the period of study.
The PeriodStart and PeriodEnd are Dates (date9.).
Data Structure:
This is an example. Each PeriodStart is 1 day after the prior PeriodEnd.
Obs | ID | PeriodStart | PeriodEnd | Treat | Event |
1 | 1 | | | 0 | 0 |
2 | 1 | | | 1 | 1 |
3 | 1 | | | 1 | 0 |
4 | 1 | | | 0 | 2 |
5 | 2 | | | 0 | 0 |
6 | 2 | | | 1 | 0 |
7 | 2 | | | 1 | 2 |
... | ... | | | | |
SAS Procedure:
PROC PHREG DATA=dataset COVS(Aggregate);
MODEL (PeriodStart, PeriodEnd) * Event (0,1) = Treat;
ID unique_id;
RUN;
In this model, I have indicated the usual censor value (event=0) and the hospitalization "event" value (event=1) as censored. So, it gives time to death. This makes sense. HR of death.
PROC PHREG DATA=dataset COVS(Aggregate);
MODEL (PeriodStart, PeriodEnd) * Event (0) = Treat;
ID unique_id;
RUN;
In this model, I am allowing either event (1=hosp, 2=death) as the hazard and indicating the 0=censor. I do not know (and can't find) how SAS handles SUBSEQUENT events after the 1st event. Are they dropped from analysis (which would make my life easier)? Or is each observation in the table allowed to continue to contribute to the hazard?
Thanks all!
Aaron