Understanding Physician's Perspectives on AI in Health Care: Protocol for a Sequential Multiple Assignment Randomized...

Area of Science:

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

Background:

• Artificial intelligence (AI)-based clinical decision support (CDS) systems are increasingly integrated into patient care.
• Physicians need to understand how factors like clinical risk and AI output influence their use of these tools.
• Hypothesis-driven research is crucial for anticipating AI's impact on frontline healthcare providers.

Purpose of the Study:

• To design and implement a web-based vignette survey exploring AI-based CDS system applications.
• To systematically vary factors such as clinical risk, AI information, and AI results within hypothetical scenarios.
• To determine the causal effects of these factors on physicians' judgments and decision-making.

Main Methods:

• Recruitment of US-based family or internal medicine physicians (target n=420).
• Utilizing a web-based survey with a novel "sequential multiple assignment randomization trial (SMART) vignette" design.
• Randomizing participants to vignette variations to test the impact of clinical risk, AI information, and AI output certainty.

Main Results:

• The study is currently in progress with data collection expected to conclude in 2024.
• Results will quantify the influence of specific contextual factors on physician decision-making when using AI-CDS.
• The "SMART vignette" design is anticipated to yield increased analytical validity and balanced group sample sizes.

Conclusions:

• This web-based vignette study will elucidate how clinical risk, AI information, and AI results impact physician responses to AI in healthcare.
• The novel "SMART vignette" design offers advantages over traditional methods for analyzing AI's influence on clinical decisions.
• Findings will enhance understanding of physician decision-making processes involving AI-based clinical decision support tools.