Cox PH time-dependent covariate formatting in R

Question

I am trying to perform a Cox proportional hazards analysis with a time-dependent covariate in R. I have read the relevant documentation (Therneau et al.) and several tutorials but am struggling to format my data properly given how the covariate is structured in the data. It is a segmented time-dependent/time-varying covariate.

The outcome of interest is "death", the time variable is "fu", and the time-dependent covariate is included in the "asp" AKA aspirin columns. Eg for the first case, the patient was not taking aspirin at baseline, 30 days, and 1 year but was taking it at the time of last follow-up, in this case 2275 days. Periods represent missing data.

| ID    | aspb | asp30 | asp1y | aspfu | death | fu        |
|-------|------|-------|-------|-------|-------|-----------|
| 1479  | 0    | 0     | 0     | 1     | 0     | 2275      |
| 10523 | 1    | 1     | .     | .     | 1     | 41        |
| 25436 | 0    | 0     | 1     | 1     | 0     | 1773      |

I cannot figure out how to achieve the (start, stop] format of the table en masse, as there are 1000+ cases. It's also complicated by the fact that the time intervals are not necessarily consistent between cases, ie the time corresponding to "aspfu" can fall anywhere between 0 days, 30 days, 1 year, ...

I did accomplish this analysis in SPSS using the following notation for the time-dependent covariate: (T_ < 1)*aspb + (T_ >=1 & T_ < 31)*asp30 + (T_ >= 31 & T_ < 366)*asp1y + (T_ >= 366)*aspfu. But I am struggling to translate this to R.

Any guidance would be appreciated! Thank you very much!

Answer 1

Try this:

## start time
df <- df %>%
  group_by(ID) %>%
  mutate(start_time = 0:(n() - 1))
df <- ungroup(df)

## end time
df <- df %>%
  group_by(ID) %>%
  mutate(end_time = 1:(n()))
df <- ungroup(df)
```