B1306
Title: Estimation of marginal hazard ratios from observational data under noninformative censoring
Authors: Guilherme Wang de Faria Barros - Umea University (Sweden) [presenting]
Jenny Haggstrom - Umea University (Sweden)
Abstract: When using observational data to study causal relationships, the process of balancing covariates is a critical part of the process to estimate causal effects. Two main distinct approaches are popular when balancing covariates: weighting and matching. In medical and epidemiological studies, one of the most common settings is the time-to-event setting, also known as survival, in which a common causal effect to be estimated is the marginal hazard ratio (MHR). This setting has the particularity of the existence of censoring of the time-to-event, which has been shown to cause bias when estimating causal effects, especially when the censoring mechanism is informative. The goal is to study the properties of weighting and matching for the estimation of the MHR in time-to-event settings under high censoring proportions with noninformative censoring, compare the results of different approaches and suggest possible solutions.
