I want to assess the association of a treatment with survival time in patients. The background information of patients like age and severity differ between the treatment group and non-treatment group. So I want to do propensity score matching to pick up patients with the treatment and background-matched patients without the treatment. After the matching, I want to assess the association using cox regressoin analysis in the matched data.

The treatment is a time-dependent covariate. Thus, the start of follow-up is not same as the treatment time. The background information of patients are not time-dependent covariates. The information was obtained at the start of follow-up.

In general, the propensity score was calculated by the logistic regression analysis. (dependent variable: treatment or non-treatment, independent variables: background information) In this case, is it possible to calculated the propensity score by the cos regression analysis? (dependent variable: tmee-to event data (event = treatment), independent variables: background information)

Could you give me any advice?

  • $\begingroup$ Have you looked into structural equation models? I believe they would be appropriate for this type of situation. $\endgroup$ – Björn Dec 18 '18 at 6:53
  • $\begingroup$ Thank you for your comment. Thank you for your comment. I don't known about it very well. Can the time-to-event data be applied to the model? $\endgroup$ – Totti Dec 18 '18 at 6:54
  • $\begingroup$ I could perform the calculation by using the marginal structural model! $\endgroup$ – Totti Feb 6 '19 at 14:53

Your Answer

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

Browse other questions tagged or ask your own question.