Exploring Stakeholder Perceptions about Using Artificial Intelligence for the Diagnosis of Rare and Atypical...

Area of Science:

• Critical care medicine
• Infectious diseases
• Medical informatics

Background:

• Delays in diagnosing rare and atypical infections are a significant challenge in critical care.
• Current diagnostic practices may be limited by insufficient assessment and consideration of specific patient factors.

Purpose of the Study:

• To evaluate critical care provider perspectives on diagnosing rare and atypical infections.
• To assess the potential of artificial intelligence (AI) as a decision support system (DSS) for improving these diagnoses.

Main Methods:

• An anonymous web-based survey was conducted among critical care providers at Mayo Clinic Rochester.
• The survey evaluated experiences with rare infection diagnostics and AI-based DSS, including perceived usefulness, impact, risks, and benefits.

Main Results:

• Most providers (38/47) agreed on diagnostic delays, citing limited patient factor assessment.
• While familiarity with AI DSS is high, only 18/38 found them consistently valuable for general patient care.
• However, 34/47 believed AI DSS could improve rare infection diagnosis, emphasizing reliability, interpretability, and workflow integration as crucial.

Conclusions:

• Critical care providers perceive diagnostic delays for rare infections, primarily due to inadequate patient factor assessment.
• AI-based DSS show potential for improving rare infection diagnosis, but usability hinges on reliability, interpretability, and workflow integration.
• Addressing alert fatigue is essential for successful AI DSS implementation in critical care settings.