cause specific survival regression

Discuss statistics related things

by pieter zanen » Wed Nov 04, 2020 8:09 am

Dear all

I started using jamovi survival to analyse the ‘risk’ on receiving a lung transplant. Patients are not only removed from the waiting list because they are transplanted (= the event), but also due to death, personal reasons or doctor decisions. In other words, competing risks and a cause specific cox regression or Fine/Gray analysis.

Text books, like the one by Singer/Willet, indicate that one needs to carry out several analyses, each for an event. So basically standard Cox regressions suffice.
In jamovi I am not sure how to approach the issue. I see a section on multiple events, but I lack the guidance. The underlying R package contains a Fine/Gray approach: is this what the section is tempting to do?

Can any one direct me to a manual or reveal details?

Kind regards

Pieter Zanen
pieter zanen
 
Posts: 2
Joined: Wed Nov 04, 2020 7:53 am

by sbalci » Wed Nov 04, 2020 8:27 am

Dear Pieter Zanen,

You probably need a competing risk analysis. I have started working on implementing competing risks to jsurvival module, but I could not finish it and it will take some time.

If you use R, I may recommend you finalfit package (which jsurvival uses). It works very nice:
How to define outcome status: https://finalfit.org/articles/survival.html#death-status
How to use cox regression: https://finalfit.org/articles/survival.html#competing-risks-regression
sbalci
 
Posts: 48
Joined: Sat Jan 06, 2018 10:25 pm

by pieter zanen » Wed Nov 04, 2020 10:40 am

Hi

thanks for the prompt reply: appreciated!

The link to finalfit.org website solved the issue. It lists the text copied below
"Overall survival: considering all-cause mortality, comparing 2 (alive) with 1 (died melanoma)/3 (died other);
Cause-specific survival: considering disease-specific mortality comparing 2 (alive)/3 (died other) with 1 (died melanoma);"
This text explains the approach taken by the multiple events option in Jamovi

By the way: SPSS has that nice R plug-in option and through that option I can use the cmprsk module of Bob Gary, so no need to turn to R

pieter
pieter zanen
 
Posts: 2
Joined: Wed Nov 04, 2020 7:53 am


Return to Statistics