Announcement

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

  • Lincom and competing risk regression

    Hi All,

    I'm running a competing risk regression model with subhazard ratios and I became stuck at the last step of the analysis. In essence, I am comparing survival for different racial/ethnic groups of transplant surgery candidates before and after a new surgical technique was implemented. Using -lincom-, I already figured out how to compare survival within an era of the surgical technique but now I would like to compare across surgical technique periods (pre. vs- post-implementation of the surgical technique). For the race variable, 1=white (reference group), 2=black, 3-Asian, 4=Hispanic. For event variable, 1==successful transplant and 2==death

    stset survival_time, failure(event==1)
    stcrreg i.race, compete(event==2)
    stcrreg i.surgical_technique_era##i.race, compete(event==2)

    Compare all races to white in pre-surgical technique implementation
    lincom 2.race + 2.race#0.surgical_technique_era, shr


    Now, how do I set up the lincom equation to compare survival between race groups across eras of surgical technique implementation? For instance, how do I compare survival for Asian patients pre- vs. post rather than just survival of Asian patients to white patients in one given era? Also, how would I compare survival for white patients in pre-era to survival of black patients in post-era?

    Happy to clarify further if I made this unclear!

    Thank you in advance!
    Emile

  • #2
    I would do this with -margins- rather than -lincom-.

    Code:
    margins race, dydx(surgical_technique_era) // EFFECT OF ERA, BY RACE
    margins race#surgical_technique_era, pwcompare // COMPARE ALL RACES & BOTH ERAS

    Comment

    Working...
    X