Advancements in Mortality Prediction Models for Intensive Care Units: A Review of Various Learning Approaches and Clinical Applications

Mortality prediction is a critical task in healthcare, particularly in intensive care units (ICUs), where timely and accurate prognostication can significantly impact patient care and resource allocation. Recent progressions in mortality prediction methodologies predominantly leverage the Medical Information Mart for Intensive Care (MIMIC) III database, incorporating a diverse array of machine learning techniques and clinical datasets. A variety of methods are covered by the studies, including deep learning models, ensemble methods, feature selection algorithms, and hybrid predictive models. Key findings include advancements in prediction accuracy, addressing data challenges such as imbalanced datasets and noisy clinical text, ensuring clinical interpretability, and validating model performance across different ICU settings. Overall, these studies contribute to the ongoing effort to enhance mortality prediction in ICUs, highlighting the importance of continued research and innovation in critical care settings.