Assessing Barriers to Implementation of Machine Learning and Artificial Intelligence-Based Tools in Critical Care:...

Area of Science:

• Critical care medicine
• Health informatics
• Machine learning applications

Background:

• Limited implementation of machine learning (ML) and artificial intelligence (AI) in critical care practice despite high interest.
• Need to understand physician and healthcare provider perspectives on ML tools.

Purpose of the Study:

• Assess physician views on a novel intubation prediction tool.
• Understand provider and non-provider perspectives on ML in healthcare.
• Identify implementation barriers and determinants for ML/AI tools in critical care.

Main Methods:

• Two anonymous surveys (single-center for physicians, social media for broader audience) were conducted.
• Surveys included categorical, Likert scale, and free-text questions.
• Statistical analysis (t-tests) and content analysis of qualitative responses were performed.

Main Results:

• Physician knowledge of ML was low (mean 2.4/5), with moderate willingness to use ML tools (mean 3.32/5).
• Non-providers showed significantly lower perceived knowledge and comfort with ML compared to providers.
• Common concerns included accuracy, data bias, patient safety, and privacy.

Conclusions:

• Both providers and non-providers generally have positive perceptions of ML tools.
• An intubation prediction tool is of interest to critical care providers.
• Addressing shared concerns is crucial for successful ML/AI implementation in healthcare.