Is there a way to run survival analysis in R when you have more than one death per event?

Question

For example, I have a data frame consisting of columns ID, Genotype, Sex, time (days), event (deaths), for a number of Drosophila strains of both sexes, where deaths can be anywhere from 0 to 2, 3, etc. Apologies as this is probably a very simple solution, but most tutorials about survival analysis don't answer it and only use datasets where the event is either 0 or 1. Have tried typical loading of data frame and packages (this part seems fine) and then assign survival object and plot with ggsurvplot.

Answer 1

Expanding the data set into one row per individual, as suggested in a comment and in another answer, is a very good choice in this situation and should be both reliable and extensible to more complicated situations.

For reference, it's also possible to use the weights argument to the R coxph() function to accomplish this. According to the help page:

Case weights are treated as replication weights, i.e., a case weight of 2 is equivalent to having 2 copies of that subject's observation.

In your situation, you would have an event marker of 1 for death at a particular time and set of covariate values in a data row, but specify the number of individuals having the event and that set of covariate values as the weight value in that row. You could similarly use the weights argument, with an event marker of 0, for the number of flies that survived to the final observation time, for each set of covariate values.

Answer 2

Assuming your survival data looks like this:

ID	Genotype	Sex	time (days)	event (deaths)
Control_Vial1	Control	F	10	1
Control_Vial2	Control	M	10	1

Option 1: Use the ggbulksurv package at https://github.com/qhuitan/ggbulksurv - I wrote it specifically for this scenario. ggbulksurv::get_indiv_surv() would do the trick.

In order to make your data work with this package, you have to add a column called condition which would reflect the sex. Your input data should look like this:

condition	ID	Genotype	Sex	day	dead
control_F	Control_Vial1	Control	F	10	3
control_M	Control_Vial2	Control	M	10	2

Option 2: Manually code this using tidyr::uncount()

I'm assuming you have no censored data (that is, no flies have escaped/ gone missing/ gotten stuck in food).

Renaming some columns for easy reference:

ID	genotype	sex	day	dead
Control_Vial1	Control	F	10	3
Control_Vial2	Control	M	10	2

sample_data <- read.csv("path-to-file.csv")

# Add a condition column
sample_data <- sample_data %>% 
                tidyr::unite("condition", genotype:sex, remove = FALSE)


## -- Calculate dead -- ##
  df_dead <- sample_data %>%
    # Select only dead column
    dplyr::group_by(condition) %>%
    # Uncount so that each row becomes an individual
    tidyr::uncount(dead) %>%
    # Add status = 1 (dead)
    dplyr::mutate(status = 1)

If you have censored data, use this code instead:

ID	genotype	sex	day	dead	censored
Control_Vial1	Control	F	10	3	2
Control_Vial2	Control	M	10	2	5

 ## -- Calculate dead -- ##
  df_dead <- sample_data %>%
    # Select only dead column
    dplyr::group_by(condition) %>%
    dplyr::select(-c(censored)) %>%
    # Uncount so that each row becomes an individual
    tidyr::uncount(dead) %>%
    # Add status = 1 (dead)
    dplyr::mutate(status = 1)

  ## -- Calculate censored -- ##
  df_censored <- sample_data %>%
    # Select only censored column
    dplyr::group_by(condition) %>%
    dplyr::select(-c(dead)) %>%
    # Uncount so that each row becomes an individual
    tidyr::uncount(censored) %>%
    # Add status = 0 (censored)
    dplyr::mutate(status = 0)

  ## -- Combine -- ##
  df_combined <- dplyr::full_join(df_censored, df_dead,
                           join_by("condition", "day", "status")) %>%
    dplyr::arrange(day) %>%
    dplyr::ungroup()