Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • help needed for survival analysis / kaplan meier, but with multiple (four) outcomes - alive, revised, died, lost to follow up

    Hello everyone,
    I am new to this game - I'm an orthopaedic surgery resident in the UK , and I'm trying to analyse survival data for a big group of people who had hip replacements. we have follow up data at a mean time of 10 years, (range 1-15 years), and one of four things has happened to the patients.
    option 0: their hip replacement in still in place, working fine.
    option 1: the hip replacement has worn out and it has been revised - that means replaced for a second time
    option 2: the patient died with the original hip replacement still in place, working fine
    option 3: the patient is lost to follow up - we have no idea if they are alive, dead, living abroad, whatever

    so this column heading is called "end point", and either has a value of 0 1 2 or 3
    we have the time of censoring data , in years - this is the time (in years) after the original hip replacement operation was done - and that's the time at which either we saw them in clinic, and everything is fine (end point = 0), or the time after original operation that the hip replacement was replaced again (end point = 1), or the time of their death (end point = 2), or the time point that they were lost to follow up (end point = 3)

    we also have other columns with data such as age at original operation, gender, diagnosis i.e. reason why they needed a hip replacement (e.g. the hip wore out (osteoatrrthritis = OA), or they were born with malformed hips (developmental dysplasia of the hip = DDH), or they had a fracture (trauma) etc,

    can any of you help me work out how to do a kaplan-meier survival curve, with 95% confidence intervals. ??

    I am not sure if i need to tell stata that there are 4 possible outcomes, rather than 2? and also, I am not sure which way stat wants to know the time at which the event occurred - so in my spreadsheet I have time data in 2 formats - first is a single column with the number of years from surgery to either census/revision/death/loss to follow up. the second is 2 columns - one column with date of original surgery, the second is date of census/revision/death/loss to follow up. presumably stata will find it easier with the first way - a single number saying how many years after surgery the event happened

    many thanks, pierre

  • #2
    Hello Pierre,

    Welcome to the Stata Forum / Statalist.

    Let's say you wish to - stset - your data according to events 1, 2 and 3, the time variable being "timevar" and the event variable being "event":

    Code:
    . stset timevar, failure(event = 1,2,3)
    To perform Kaplan-Meier curves, please type:

    Code:
    . help sts graph
    To end, since you said you are "new to this game", please take some time to grasp the core knowledge related to survival analysis, including its principles and assumptions. Not to forget, the Stata Manual is a must read.

    Hopefully that helps.
    Last edited by Marcos Almeida; 05 Apr 2017, 05:43.
    Best regards,

    Marcos

    Comment

    Working...
    X