Role of artificial intelligence in ICU therapeutic decision-making for severe infections.

Abstract

Purpose of review: To discuss current and future role of artificial intelligence in predicting severe infections and supporting decisions on antibiotic treatment in critically ill patients in intensive care units (ICU), focusing in particular on some relevant conceptual changes compared to classical clinical reasoning.

Recent findings: Several studies have evaluated the ability of machine learning techniques for severe infection prediction, while other studies have explored the potential of large language models (LLM)-based tools to assist clinicians in deciding which antimicrobial agent(s) to prescribe to patients with severe infections.

Summary: The support of artificial intelligence for infection prediction and antimicrobial prescribing has shown the potential to improve the treatment of severe infections in ICU. However, the limited number of studies focused on ICU should be highlighted, along with the need to thoroughly address the issue of patients' privacy and to improve the ethical and legal frameworks for decision accountability, as well as the transparency and quality of training data. A standardized approach to the accuracy-interpretability trade-off would also be essential to outline a correct and shared approach both for the future conduct of studies and for the interpretation of their evidence for clinical practice.