Pseudo-observations and super learner for the estimation of the restricted mean survival time

This article delves into the estimation of the restricted mean survival time (RMST) using pseudo-observations and super learner algorithms, addressing the challenges posed by right-censored data. The study introduces split pseudo-observations, a novel type of pseudo-observations that facilitate theoretical analysis and practical implementation. The super learner algorithm, known for its ability to combine multiple prediction models, is adapted for RMST estimation, demonstrating superior performance compared to individual models. The article presents extensive simulations and real-world applications, showcasing the method's accuracy and robustness. Theoretical results are derived to support the efficacy of the proposed approach, and comparisons with existing methods such as the Cox model and random survival forests are conducted. The findings highlight the potential of this novel method to enhance survival time predictions in various applications, offering a powerful tool for researchers and practitioners in the field.