Artificial intelligence-driven decision support for patients with acute respiratory failure: a scoping review

Area of Science:

• Medical Informatics
• Clinical Decision Support
• Artificial Intelligence in Healthcare

Background:

• Artificial intelligence (AI) offers potential for managing acute respiratory failure.
• The real-world clinical impact and validation of AI tools in this area require further investigation.
• This review focuses on clinically validated AI tools and their evaluation quality measures for deployment.

Purpose of the Study:

• To identify and evaluate AI-driven decision support tools for acute respiratory failure.
• To assess the quality of reporting for key evaluation measures in clinical studies of AI tools.
• To understand the prerequisites for broader deployment of AI in managing acute respiratory failure.

Main Methods:

• Scoping review of studies comparing AI interventions to control groups in adult patients with acute respiratory failure.
• Systematic literature search in PubMed, CINAHL, and EmBase until January 2025.
• Data extraction and quality assessment using the DECIDE-AI framework for early clinical evaluation.

Main Results:

• Six studies met eligibility criteria, focusing primarily on predicting weaning from mechanical ventilation.
• Only 50% of studies demonstrated statistically significant and clinically meaningful outcomes.
• Studies met a median of 3.5 out of 17 DECIDE-AI criteria, with critical aspects like error reporting and fairness largely unaddressed.

Conclusions:

• AI tools for acute respiratory failure show promise, particularly in predicting mechanical ventilation weaning.
• Methodological rigor in early clinical evaluation of these AI tools is inconsistent.
• Further high-quality assessments of reliability, usability, and real-world implementation are needed for AI to transform patient care.