A1288
Title: Inference via wild bootstrap and multiple imputation under fine-gray models with incomplete data
Authors: Marina Dietrich - University Augsburg, MNTF, Prof. Friedrich (Germany) [presenting]
Dennis Dobler - TU Dortmund University (Germany)
Mathisca de Gunst - Vrije Universiteit Amsterdam (Netherlands)
Abstract: The wild bootstrap is a popular resampling method in the context of time-to-event data analysis. It can be used to justify the accuracy of inference procedures such as hypothesis tests or time-simultaneous confidence bands. In previous works, wild bootstrap confidence bands have been established for inference on cumulative incidence functions under the Fine-Gray proportional sub-hazards model if the data are censoring-complete. However, it is rather unusual that data are censoring-complete, and hence, this assumption is restrictive. In order to overcome this limitation, a novel wild bootstrap and multiple imputation-based (WB-MI) confidence band is proposed for the cumulative incidence function under the fine-gray model with incomplete data. Furthermore, the asymptotic validity of the proposed WB-MI confidence band is justified, and its reliability is numerically assessed and compared with already existing methods. The approach is illustrated by investigating the impact of pneumonia for intensive care unit patients on the probabilities of hospital death competing with live discharge.
