1
$\begingroup$

In the example mentioned in the this doc for Survival Analysis on how to define time intervals, it says that the logic for creating the time variables is that, if a subject has no events, there is a single interval from 0 to last follow-up, with a status of 0. If there are events, then there is an interval for each event (0, event1], (event1, event2], etc. Each with status =1.

But what if the subject had different events in same time period? Say, Part 1 and part2 both failed on an equipment id 1 during the interval 219-373? How would one represent that case? Also, if subject ID 1 was observed from time 0-414, but subject ID 2 was observed from 8-241, should I worry about the missing time interval 0-8 for subject 2?

id       tstart   tstop   status enum  treat

 1         0      219      0       1  rIFN-g
 1        219     373      1       2  rIFN-g
 1        373     414      0       3  rIFN-g
 2        0       8        1       1  placebo
 2        8       26       1       2  placebo
 2        26      152      1       3  placebo
 2       152      241      1       4  placebo

cfit <- coxph(Surv(tstart, tstop, status) ~ treat + sex + age + inherit + cluster(id), data=cgd)

$\endgroup$
  • $\begingroup$ If you format your survival data in that way, you would need multiple event indicators. You would censor the patient for event 1 at the time event 2 happened, coding event 1 as 0 and event 2 as 1. Then "re-enter" that participant into the cohort at that event 2 time, and follow-them until their next event. If it is 1 then Event 1 takes value 1 and Event 2 takes value 0. The survival analysis will be no different. That is because you do not model multiple events in a Cox model, only single or composite events. $\endgroup$ – AdamO Feb 5 '18 at 17:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.