Using AI to Train Future Clinicians in Depression Assessment: Feasibility Study.

Area of Science:

• Medical Education
• Artificial Intelligence in Healthcare
• Psychiatry

Background:

• Depression poses a significant global health challenge, necessitating accurate and timely diagnosis.
• Generative pretrained transformer (GPT)-powered chatbots can simulate patient interactions for medical training.
• Developing diagnostic skills for depression and suicidality is crucial for future clinicians.

Purpose of the Study:

• To evaluate a GPT-4 powered chatbot's ability to simulate psychosomatic cases, specifically depression and suicidality.
• To assess the chatbot's accuracy in role-playing, authenticity, and suitability for practicing diagnostic skills.
• To determine the chatbot's effectiveness in training students to diagnose depressive disorders and assess suicidality stages.

Main Methods:

• Developed 3 virtual patient role scripts for depression and suicidality, aligned with expert input and clinical guidelines.
• 148 medical students interacted with the GPT-4 powered chatbot, followed by a user experience questionnaire.
• Chat logs were analyzed for diagnostic accuracy, suicidality assessment, role adherence, and AI authenticity.

Main Results:

• The chatbot maintained its assigned role in over 90% of interactions.
• Students achieved 60% accuracy in diagnosing depression severity and 67% in identifying suicidality phases.
• A significant portion of students inadequately assessed suicidality, despite prior instructions.

Conclusions:

• GPT-powered chatbots can effectively simulate patients with depression, offering a valuable training tool.
• Over two-thirds of students perceived the AI patients as authentic, with nearly 80% desiring further practice.
• AI-driven virtual patients provide a flexible, standardized, and accessible method for clinical skills development.