0

I'm working with the rms package to derive a score for predicting response to a drug. I actually derived two scores: one including only pre-treatment variables (x and y), and a second including also variable that can be measured only after 6 months of therapy (z).

So the models are:

score_baseline <– cph(Surv(t, response)~ x + y, data = df)
score_6months <– cph(Surv(t, response)~ x + y + z, data = df)

By doing that, the two scores have different scales of the linear predictor and different baseline survival values (S0(t)). Is there a way to rescale the coefficients of one of the two models (e.g. score_6months) and have the two linear predictors on the same scale and also same baseline survival estimates?

Thanks for help.

Devanto
  • 33
  • 5

1 Answers1

1

Not only do I not think there is a way to do that, I don't think it's a good idea to try to model the two outcomes using the same time zero. You are treating z as if it's measured at time zero but it's not, and patients must survive 6 months to have z apply. So it seems that the second model needs to be a conditional one, that is, conditional on (subsetting on) patients surviving at least 6m.

Frank Harrell
  • 1,954
  • 2
  • 18
  • 36