0
$\begingroup$

I currently read a publication about a prediction model to estimate heart failure specific events, providing hazard ratios for multiple risk factors (see below only the HRs of one state of 3 that are prediced):

enter image description here

If I have new patients, where I have distinct values of age, gender and all the risk factors given, will I be able to derive a model from the hazard ratios and the confidence intervals, and calculate their 1 year probability of having the event (here HF hospitalization)?

If, yes, how? And is there a way to do this in R?

Edit: Added two example patients, which is also provided in the paper, does this help?:

enter image description here

Improve this question
$\endgroup$
2
  • 1
    $\begingroup$ To compute probabilities of survival, the baseline survival (or hazard or cumulative hazard) is also necessary - was that provided? $\endgroup$
    – Cam.Davidson.Pilon
    Sep 14, 2020 at 13:39
  • $\begingroup$ I don't see that anywhere in the paper, however they provided some patients with example risk values and a one year probability of hospitalization. I will edit the question and add an example. Is it possible to use this information to deduct the model? $\endgroup$
    – aldorado
    Sep 15, 2020 at 8:49

1 Answer 1

Reset to default
1
$\begingroup$

The semi-parametric nature of Cox models is both a strength and a weakness.

The strength is that you don't need to specify a model for the overall shape of the survival curve. The baseline survival curve is estimated from the data themselves.

The weakness is that unless you have access to the baseline survival curve upon which the model was built, you can't estimate absolute probabilities of survival versus time. All you can do is estimate relative hazards among conditions.

Sometimes, but not frequently, a nomogram is provided from which survival probabilities can be estimated. Otherwise, if there is a graph in the paper of the survival curve for some specified set of covariate values, you could digitize that and then use the differences of covariate values between your new cases and that specified set to adjust the curve accordingly. Or you could ask the authors for the underlying raw data or at least a graph of a baseline survival curve. You would in any event need to assume that the baseline survival in the report fairly represents the population from which you are drawing the new cases.

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.