The field of electronic health record (EHR) analysis and survival prediction is rapidly advancing, with a focus on developing innovative methods to handle irregularly sampled time series data and missing values. Researchers are exploring the use of large language models, diffusion models, and neural networks to improve the accuracy and interpretability of survival predictions. Notable trends include the integration of functional covariates, competing risk modeling, and physically interpretable survival prediction. These advancements have the potential to significantly improve prognostic modeling in critical care and enhance our understanding of disease progression patterns. Noteworthy papers include: Mind the Missing, which proposes a variable-aware representation learning framework for irregular EHR time series using large language models. SurvDiff, which introduces a diffusion model for generating synthetic data in survival analysis, outperforming state-of-the-art generative baselines in distributional fidelity and downstream evaluation metrics.