proc lifetest

[lone0708]

Hello all, 

This is probably very simple for someone used to use SAS. But i am interested in making a simple "survival curve" on my data. 

My dataset is containing information about prescriptions filled in a 6 months period. I have two groups i want to illustrate in the same plot. The first group in persons who did not fill a prescription in the first 10 weeks and the other group in persons who did fill a prescription.

data have;
input person_id week_of_use group
datalines;
1   3    1
1   3     1
1    7     1
2   11     2
3    8     1
3    12   1
4   23     2
5  2       1
6   13    2
6    14   2
6    20    2
6    23    2
;

I hope, you can help me

[lone0708]

[WarrenKuhfeld]

https://support.sas.com/documentation/onlinedoc/stat/151/kaplan.pdf

Examples in this chapter start with the simple and continue to the complex.

[lone0708]

@WarrenKuhfeld,
Thanks for your reply.
The problem for me is, that i do not understand the coding of status in my particular case, as all participants are contributing in all of the study period

[Reeza]

For survival analysis you need your data in the form of:

ID Duration Censor

How does that work for your data if all participants are available throughout the study? What is the time component that causes people to drop off?

Please explain your data and the graph you're trying to create. 

---

@lone0708 wrote:
@WarrenKuhfeld,
Thanks for your reply.
The problem for me is, that i do not understand the coding of status in my particular case, as all participants are contributing in all of the study period

---

[lone0708]

Hi Reeza,
Thanks for your reply!
I can create a variable with duration of use, and I also have entering and end date of study period for all participants (not shown in the sample data). All participants have the same length of follow-up, but the entering and end dates are different for each person.
What should the censor variable contain?

I am sorry for my clumsiness. I am far from being a statistician……

[Reeza]

Why does one group start at a fixed number and drop off and one starts at nothing and grows?

[lone0708]

It’s just how I expect the curve to look like.
After a certain event (entering date) some will not use any medication, but at some point (10 weeks after event) they will be more likely have a need for medication.
The other way around with the other group. They have an immediate need for medication right after the event, but will not need it with time.
I hope it makes sense

[Reeza]

Your censor variable tells you of the event of interest and if it's met. So what are you looking for the graph to tell you? How many patients were in the study and when they needed the medication? If so, the timepoint of when they started the medication and if they took the medication is your censor variable.

[lone0708]

The purpose is to illustrate how many persons took/got a preacribtion x weeks from the entering date (which is different for every person).
Should the censoring variable then contain the date of end of follow-up?

[Reeza]

So duration is the time to start taking the drug or end of follow up from when they started.
Censor is if they took the drug, a 0/1.