Left truncated survival data?

Question

I am attempting Attrition Analysis in R using the Survival & KMsurv Package. My question is more related to how to use the R package / functionality for my situation.

Let us say the analysis is for Department B.

I have the following dataset:

• All employees who were associated with Department B for a period (say 1 Jan 2013-31 Dec 2015) (So this has some employees who joined way before 2013 )
• Each employee has a start_date and an end_date

While setting up the Survival object, I have done the following:

spell is : end_date - start_date -> for employess who have left: 31-dec-2015 - start_date -> for employees who had not left by 31-dec-2015

event is: 1 - if employee has quit by 31-Dec-2015 0 - if employee has not quit by 31-Dec-2015

I then build the survival object using: Surv(spell, event).

I have some doubts about this:

1. I do not have data on all employees who joined before 1-Jan-2013 (I only have data for employees who remained till after 1-Jan-2013). Will this corrupt the analysis ?
2. Should I consider the employees that joined before 1-Jan-2013 as "left truncated". So that means in the definition of spell for them, start_date is not their respective start_date but 1-Jan-2013.

Answer 1

1. The potential issue I see here is that new vs. long-term employees may have different hazards (e.g., new employees may be more likely to quit than employees who have been around for years). An easy way to deal with this would be to include a covariate in your model for the number of years an individual has been an employee as of 1/1/2013.

2. Actually, if your t = 0 corresponds to the start of observation time (it seems like that is what you are going for), then you do not have left truncation because you aren't considering individuals "at risk" until you start observing. On the other hand, if you want t = 0 to correspond to the start of employment, then you do have left truncation because individuals were "at risk" during the period between start_date and 1/1/2013, but you were not observing them.