Cox-regression with a time-varying covariate registered only at 1 year post surgery - which command to choose?

[Vilma Antonov]

Hi!

Basic question, I fear, but I'm having a bit of trouble navigating the guides.

I am studying whether risk of dying post surgery for patients is increased if the patient has been hospitalized for an infection at 1-year post surgery (no exact dates provided, only a categorical variable with y/n). I would like to fit a Cox regression with the time-varying covariate hosp_1yr included as well as other covariates such as age, sex, hypertension etc.. If possible, I would thereafter like to illustrate the effect of having been hospitalized for infection 1yr post surgery on survival time in a graph, preferably divided by sex. I can't figure out whether I should use the tvc option, stsplit at failures, or the stcoxint. Any ideas would be greatly appreciated. Thanks for helping a newbie out!

[Vilma Antonov]

Still struggling with this, if anyone has a good idea please let me know!

[Clyde Schechter]

I think the reason your question went unanswered for so long is that people are having a hard time understanding what you want to do. I passed it by when I first saw it for that reason. On re-reading, to the extent that I do understand what you are asking, it doesn't make any sense to me. If all you have for hospitalization for infection 1 year post-op is a yes-no variable with no dates, how can it possibly be a variable that varies with time, let alone a "time varying covariate" (which, in a survival analysis, is not the same thing as a variable that changes over time)? I think a clearer explanation of what's you have and what you are trying to do, accompanied by example data (posted using the -dataex- command) would increase your chances of getting a helpful response.