Inference via Wild Bootstrap and Multiple Imputation under Fine-Gray Models with Incomplete Data

Abstract: Fine-Gray models specify the subdistribution hazards for one out of multiple competing risks to be proportional. The estimators of parameters and cumulative incidence functions under Fine-Gray models have a simpler structure when data are censoring-complete than when they are more generally incomplete. This paper considers the case of incomplete data but it exploits the above-mentioned simpler estimator structure for which there exists a wild bootstrap approach for inferential purposes. The present idea is to link the methodology under censoring-completeness with the more general right-censoring regime with the help of multiple imputation. In a simulation study, this approach is compared to the estimation procedure proposed in the original paper by Fine and Gray when it is combined with a bootstrap approach. An application to a data set about hospital-acquired infections illustrates the method.