在r中拟合时变cox模型

Question

Please, I have been having some challenges trying to fit a time varying cox model. Precisely I find it hard getting the same result after following examples I read about in which the survSplit command was used. I end up with one error or the other, here is an example and the error message.

hip2=survSplit(Surv(Time,status)~Operation.Performed,
           data=hip,cut=c(200,500,1000,1600),
           end='Time',event='status',start="start")

Error in survSplit(Surv(Time, status) ~ Operation.Performed, data = hip,  : 
start time must be < stop time

Note: My data set initially has no column labelled start. But I read it will be created by the command, I do not know if this is right?

Please, is there anything am not doing correctly to get the right result??? or better still is there a better way to implement the survSplit command?

Answer 1

It's hard to say without seeing your data head(hip) , but the error message states that you have the start and end times confused - you can see the message in the code in github here . survSplit Indeed creates a start time if you don't have one and calls it "tstart" by default. Unless you use an old version of the package, don't use the event And start Arguments as they are deprecated and for legacy use alone. Give a name to the episode for use in the co ph model as an interaction, and you should be OK.

hip2 <- survSplit(Surv(Time, status) ~., data=hip, cut=c(200,500,1000,1600),episode="tgroup")

model.coxph <- coxph(Surv(tstart, Time, status) ~ Operation.Performed:strata(tgroup), data=hip2)

And that should do it.