In an observational study, the estimation of causal treatment effect is prone to selection bias. Such bias may result from the presence of confounders. Another source of selection bias is censored outcome due to loss to follow-up, which may yield invalid causal inference when the censoring is not random. Methods based on propensity score (PS) are widely used to adjust for confounding effect in the binary treatment case. However, the applications of PS methods in studies with multiple treatments remain limited, and the performance of these methods remain unclear for comparing multiple treatments. To correct the bias due to censoring, we combine the existing PS methods with weighting by inverse probability of not being censored (IPCW) given baseline covariates for each treatment. We conduct simulation studies to compare different PS methods combined with IPCW. We apply these methods to estimate the effect of four common treatments for metastatic castration-resistant prostate cancer, using claims data from a large national private health insurance network with the outcome being admission to the emergency room within a short time window of treatment initiation.