Distributionally Robust Learning in Survival Analysis
This work addresses the challenge of model sensitivity and data perturbations in survival analysis, which is crucial for fields like healthcare and biostatistics, though it appears incremental as it builds upon existing Cox regression methods.
The authors tackled the problem of improving robustness and accuracy in survival predictions by incorporating a Distributionally Robust Learning (DRL) framework into Cox regression, resulting in a model that demonstrated superior performance in simulations and real-world studies.
We introduce an innovative approach that incorporates a Distributionally Robust Learning (DRL) approach into Cox regression to enhance the robustness and accuracy of survival predictions. By formulating a DRL framework with a Wasserstein distance-based ambiguity set, we develop a variant Cox model that is less sensitive to assumptions about the underlying data distribution and more resilient to model misspecification and data perturbations. By leveraging Wasserstein duality, we reformulate the original min-max DRL problem into a tractable regularized empirical risk minimization problem, which can be computed by exponential conic programming. We provide guarantees on the finite sample behavior of our DRL-Cox model. Moreover, through extensive simulations and real world case studies, we demonstrate that our regression model achieves superior performance in terms of prediction accuracy and robustness compared with traditional methods.