July 25, 2024

Large language models (LLMs) offer immense promise for neurology-related information tasks. But the performance of these advanced artificial intelligence systems in daily clinical practice remains uncertain.

Mayo Clinic neurologists are among the researchers who outlined the challenges of using LLMs and the importance of harnessing their potential in neurological care. The researchers' review was published in a June 2024 issue of Neurology.

LLMs' constraints include:

• Limited clinical reasoning.
• Variable reliability and accuracy.
• Reproducibility, self-serving and sponsorship biases.
• Potential for exacerbating healthcare disparities.

"These challenges are further compounded by practical business considerations and infrastructure requirements, including associated costs," says Mayo Clinic neurologist Lyell K. Jones, Jr., M.D.

Overcoming these hurdles requires healthcare organizations to:

• Cultivate a culture that welcomes AI solutions and aligns them with healthcare operations.
• Create clear objectives and business plans to guide the selection of AI solutions.
• Engage with clinical and nonclinical stakeholders to secure necessary resources and foster trust.
• Provide testing, validation, training and ongoing monitoring to ensure successful integration.

The researchers note that LLMs have potential to reshape the landscape of neurological care. "Awareness of the associated challenges is vital for harnessing AI's potential," Dr. Jones says. "When harnessed in accordance with best practices, LLMs are poised to enhance the field of neurology, ultimately improving outcomes and efficiency in patient care."

For more information

Moura L, et al. Implications of large language models for quality and efficiency of neurological care. Neurology. 2024;102:e209497.

Refer a patient to Mayo Clinic.