Fast information and slow evidence in the large language models era.

Area of Science:

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

Background:

• Large language models (LLMs) offer rapid synthesis of medical information.
• Over-reliance on LLM outputs without critical appraisal poses risks.
• Distinguishing information from evidence is crucial in healthcare.

Purpose of the Study:

• To evaluate the role of LLMs in medical information processing.
• To define the necessary steps for transforming LLM-generated information into clinical evidence.
• To explore how LLMs can support and enhance evidence-based medicine.

Main Methods:

• Analysis of the data-information-evidence-practice hierarchy.
• Conceptual framework for appraising LLM outputs in clinical settings.
• Review of LLM capabilities in supporting evidence infrastructures.

Main Results:

• LLM outputs function as preliminary information, not direct evidence.
• Appraisal, validation, and contextual judgment are essential for evidence generation.
• LLMs can augment, but not replace, clinical expertise and evidence-based practices.

Conclusions:

• LLMs accelerate medical information work but necessitate rigorous validation.
• Integrating LLMs requires clarifying evidence boundaries and strengthening appraisal processes.
• LLMs should be utilized to support evidence-based medicine, enhancing rather than supplanting expert judgment.