Survival Function List - Cox Regression

[katie moll]

Hello,

I am able to generate the survivor function overtime with the following code and associated output:
sts list if fullcohort_rosey==1 & _weight==1, at(0 1 2 3 4 5) by(drug)

failure _d: readmitdead_90 == 1
analysis time _t: readmitdeaddays_90

Beg. Survivor Std.
Time Total Fail Function Error [95% Conf. Int.]
-------------------------------------------------------------------------------
drug1=0
0 0 0 1.0000 . . .
1 2692 23 0.9915 0.0018 0.9872 0.9943
2 2666 17 0.9851 0.0023 0.9798 0.9891
3 2647 25 0.9758 0.0030 0.9693 0.9810
4 2621 10 0.9721 0.0032 0.9651 0.9777
5 2610 15 0.9665 0.0035 0.9590 0.9727
drug1=1
0 0 0 1.0000 . . .
1 2692 16 0.9941 0.0015 0.9903 0.9964
2 2674 19 0.9870 0.0022 0.9819 0.9906
3 2655 14 0.9818 0.0026 0.9760 0.9862
4 2641 16 0.9758 0.0030 0.9693 0.9810
5 2625 16 0.9699 0.0033 0.9627 0.9757
-------------------------------------------------------------------------------


However, I want to generate the same output/list in order to get the survival function overtime of the cox-hazard regression output. So I would like to generate a list of adjusted survival functions. I cannot seem to find a way to do this within the stcurve or stslist code. Does anyone have any suggestions?

Thank you!!~

Katie

[Clyde Schechter]

-stcurve- has an -outfile()- option which will let you save the survival function to a new Stata data set. You can then -use- and -list- that file.

[katie moll]

Awesome!! got it!! Thank you!

[katie moll]

Clyde, I can see the survival function with outfile option. However, the data does not include the 95% confidence intervals or standard deviations. Do you know of any way to get (or calculate) that data?
Thank you,
Katie

[Clyde Schechter]

I'm afraid I don't see that as a saved output to any of the survival analysis commands I am familiar with. When I first learned survival analysis, I did learn how to calculate the standard errors of the survival function, but that is a long time ago and I have neither a memory of how to do it nor the textbook I used back then. I hope somebody else following along either knows how to calculate it or knows of a command that will give it to you.

[Paul Dickman]

I don't know how to get what you want following a Cox model (although I imagine it should be possible). You can definitely get what you want after fitting a flexible parametric model using -stpm2- (available at SSC). See this tutorial for details.

That tutorial is more advanced than what you actually need.

If your cox model is

stcox x1 x2 x3

then the corresponding code is something like

stpm2 x1 x2, scale(hazard) df(4) eform
predict s, survival ci

For each observation, the predicted survival will be in s with 95% confidence limits in s_lci and s_uci.

Predictions will be very similar to the Cox model.