Let's say I have some kind of survival data - i.e. I'm giving a drug that may cause mortality. So I have three patients: A, B and C. All are given the drug at Time t1.

Let's say patient A dies at time t2. Patient B dies at time t3. And Patient C dies at time t100.

Clearly, the likelihood that the drug caused the death of patient A and B is higher than the likelihood that the drug cause patient C's death (i.e. patient C likely died from natural causes whereas patients A and B probably died because of the drug).

Are there any techniques in survival analysis/related fields that allow one to quantify the probability/likelihood that the treatment caused the effect. For example, if I could calculate the probability that the treatment caused the death given that patient was treated at time X and died at time Y.



I think you're going to struggle with any sort of definitive treatment effect, because you lack a comparison group of any sort. You need not necessarily have a control group - there's plenty of methods in the case-crossover literature for using cases as their own controls during unexposed time periods, but if everyone got the drug at t=1, then no one has unexposed time to act as a control.


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.