Paper accepted in NEJM AI
We are pleased to announce that our follow-up study on real-world validation of our DCI risk model has been published in NEJM AI. Building on our previously developed longitudinal model for predicting delayed cerebral ischemia (DCI) after subarachnoid hemorrhage, this study evaluates the model in a silent, real-time clinical deployment. The algorithm ran prospectively in the background without influencing medical decisions, allowing us to assess its performance under routine ICU conditions. Using time-dependent performance analyses, we confirmed that the combined static–dynamic approach provides stable and clinically meaningful risk trajectories over the course of hospitalization. As in the original study, integrating admission characteristics with evolving laboratory data outperformed models relying on only one component. This work represents a key translational step—from retrospective model development to prospective validation in real clinical workflows—and brings us closer to real-time risk monitoring for DCI in neurocritical care.
Read the full article here: https://ai.nejm.org/doi/full/10.1056/AIoa2500749