Artificial Intelligence for Management of Major Depression: Initial Design, Progress, and Research Plans

Area of Science:

• Artificial Intelligence in Medicine
• Digital Health
• Mental Health Technology

Background:

• Development of an autonomous artificial intelligence (AI) system for managing major depressive disorder (MDD).
• The system is a web-based, patient-facing conversational AI designed for comprehensive MDD management.
• It collects medical history, provides diagnoses, recommends treatments, and coordinates care.

Purpose of the Study:

• To report on the initial development, progress, and future plans of an AI system for MDD management.
• To evaluate the system's components, including its knowledge base, bias correction, conversational capabilities, and treatment recommendation module.
• To assess the potential of AI in improving patient outcomes and clinical workflow for MDD.

Main Methods:

• Constructed an AI knowledge base using LASSO logistic regression on patient data to identify factors influencing antidepressant response.
• Developed a conversational Large Language Model (LLM) for efficient medical history collection and a dialogue management system to maintain conversation coherence.
• Incorporated bias correction for patient subgroups and a module for providing evidence-based treatment advice.

Main Results:

• The AI knowledge base predicted 69.2% to 78.5% of antidepressant response variation.
• AI-assisted clinicians showed a 17.5% higher likelihood of patient benefit.
• The system demonstrated efficient data collection and effective dialogue management, with clinicians receptive to referrals when the system advised patients at home.

Conclusions:

• The AI system shows significant potential in predicting treatment response and aiding MDD management.
• Adjustments were necessary to improve accuracy for specific patient subgroups, highlighting the importance of bias mitigation.
• Future research, including a prospective randomized trial, is planned to further evaluate the system's impact on clinical practice and patient outcomes.