生存包R：survSplit“开始时间<结束时间”

Question

I am trying to create a dataframe with cutpoints for piecewise hazard modelling using the survSplit function from survival. The data is right-censored, but not left-truncated/censored. The data looks like this

stay event sex marstat agegr health ID 0.101 1 f 0 (75,90] 1 1 0.167 1 f 0 (75,90] 2 2 2.968 0 f 0 [65,75] 2 3

With stay being the time variable, and event being the event indicator

Based on previous experience with the survsplit I started with the following code:

             datasplit <- survSplit(data= data, cut = c(0.25, 0.5, 1, 2,3), 
                          end = "stay", start = "start", episode= "interval")

This did not work because it did not have a formula (but it apparently worked for me back in november, with what I guess was an older version of the survival package?) After a bit of googling of what the formula should be I entered the following

           datasplit <- survSplit(Surv(stay, event)~., 
                        data= data, cut = c(0.25, 0.5, 1, 2,3), 
                        end = "stay", start = "start", episode= "interval")

I get the error message: " start time < end time", which is odd seeing as there is no start time. Based on a previous answer I saw on Stack overflow, I retried without end = and event = , but the error message did not go away. I am not sure what I can do to remedy it at this point.

Answer 1

I see that you have already provided a response , but I wanted to discuss a little bit more the topic.

There are two things to take into account here, sometimes in survival analysis it makes no sense that time happens at 0, so probably you are right that the events at time 0 should be added error to make sense of them.

  longdata <- survival::survSplit(Surv(time2event, os_event ) ~.,
                        data = d %>%
                        dplyr::mutate(os_event = ( status == event_type),
                                      time2event = time) ,
                        cut = tau,
                        subset = (d %>% dplyr::select(time) %>%
                          unlist) != 0,
                        start = "start",
                        end = "event",
                        episode = "interval")

I found that the function has also a option called subset that could work to drop the observation at time 0 otherwise, I have read sometimes that observations at time 0 should be dropped, however I make more sense of your solution of adding a small error.