Artificial intelligence-based models for clinical prognosis at the end of life in the ICU

Introduction: Artificial intelligence (AI) has emerged as a promising tool to enhance clinical prediction in critically ill patients near the end of life in the intensive care unit (ICU), overcoming the limitations of traditional prognostic scales. Aim: To identify and describe artificial intelligence models implemented to support prognostic decisions at the end of life in the intensive care unit (ICU). Methods: A scoping review was conducted across three databases (PubMed, Scopus, and Web of Science), encompassing studies published between 2015 and 2025 that utilised AI algorithms to predict relevant clinical outcomes in critically ill patients near the end of life. Results: Twenty-eight studies were included. The algorithms applied in the ICU showed good performance in terms of anticipating outcomes such as mortality, withdrawal of life support, or clinical deterioration. There was variability in the type of algorithms used, data sources, and metrics reported, which limited comparisons between studies. Conclusions: Artificial intelligence represents a promising tool for improving the clinical prognosis of critically ill patients at the end of life. However, greater methodological robustness and validation in real contexts are needed to strengthen its usefulness in clinical decision-making.