Antibiotics and Artificial Intelligence: Clinical Considerations on a Rapidly Evolving Landscape

Area of Science:

• Healthcare Artificial Intelligence
• Clinical Decision Support Systems
• Antibiotic Stewardship

Background:

• Growing interest in artificial intelligence (AI) for healthcare decision-making, particularly in antibiotic prescribing.
• Large language models (LLMs) offer potential for processing and generating text-based clinical information.
• Implementing LLM-based support for antibiotic prescribing presents unique complexities.

Purpose of the Study:

• To explore the application of LLMs in antibiotic prescribing.
• To differentiate LLM use in scientific writing versus clinical practice.
• To examine the expertise paradox and error risks associated with LLMs in this domain.

Main Methods:

• Conceptual analysis of LLM applications in healthcare.
• Discussion of commonalities and differences between LLM use in scientific writing and clinical decision support.
• Exploration of the expertise paradox and risk of error in LLM-assisted antibiotic prescribing.

Main Results:

• Identified crucial conceptual differences in LLM application between scientific writing and real-world antibiotic prescribing.
• Discussed the nuances of the expertise paradox when clinicians interact with LLMs.
• Highlighted specific peculiarities and risks of error in LLM-supported complex tasks like antibiotic selection.

Conclusions:

• LLM implementation for antibiotic prescribing requires careful consideration of distinct challenges compared to scientific writing.
• Addressing the expertise paradox and mitigating error risks are critical for safe and effective LLM integration.
• Further research is needed to navigate the complexities of LLM-based clinical decision support in infectious disease management.