Large Language Models for Mental Health Applications: Systematic Review

Area of Science:

• Artificial Intelligence in Medicine
• Digital Mental Health
• Clinical Informatics

Background:

• Large language models (LLMs) are advanced AI with potential in digital health.
• Their application in mental health clinical settings is debated.
• This review critically assesses LLM use in mental health.

Purpose of the Study:

• To systematically review the applicability and efficacy of LLMs in mental health.
• Focus on early screening, digital interventions, and clinical settings.
• Analyze models, methodologies, data, and outcomes to highlight potential and challenges.

Main Methods:

• Systematic review following PRISMA guidelines.
• Searched MEDLINE, IEEE Xplore, Scopus, JMIR, and ACM Digital Library.
• Included English articles from Jan 2017-Apr 2024 using keywords (mental health OR mental illness OR mental disorder OR psychiatry) AND (large language models).

Main Results:

• Evaluated 40 articles: 38% on condition detection, 18% as conversational agents, 45% on other applications.
• LLMs demonstrate effectiveness in detecting mental health issues and providing accessible eHealth services.
• Current clinical risks (inconsistencies, hallucinations, lack of ethical framework) may exceed benefits.

Conclusions:

• LLMs show potential as clinical aids in mental health but are not substitutes for professional services.
• Identified issues include lack of multilingual expert-annotated datasets, accuracy concerns, interpretability challenges, and ethical dilemmas.
• Continued research is crucial for developing LLMs responsibly for mental health applications.