The value of machine and deep learning in management of critically ill patients: An umbrella review.

Abstract

The integration of Artificial Intelligence and Machine Learning methods in healthcare, particularly in Intensive Care Units (ICUs), has great potential to transform medical care. This umbrella systematic review explores the possibility of using machine and deep learning (ML and DL) in the management of critically ill patients. The umbrella review was conducted according to PRISMA guidelines. Systematic reviews and meta-analyses were searched in three databases from the beginning until March 1st, 2025. The articles that included machine or deep learning in the management of critically ill patients. The information about the author's results, outcomes, goals, strengths, limitations, and ML or DL model types were extracted. After a careful search of electronic databases, 2148 records were found, of which 42 studies met the criteria for analysis. Key findings highlight the potential of artificial intelligence, especially deep learning techniques, to revolutionize clinical decision-making and patient outcomes in the ICU for sepsis, respiratory, cardiovascular, renal, and neurological diseases. They manifested as early detection of complications, individualization of treatment approach and accurate prediction of clinical outcomes. ML and DL models offer promising opportunities for treating patients in ICUs, but their clinical application remains limited due to insufficient external validation, methodological inconsistencies, and unresolved ethical issues. To narrow the gap between research and practice, future work should focus on developing clinically interpretable models, standardization of assessment protocols, and addressing ethical implications through proactive governance frameworks.