Using AI to Detect Psychosis Relapse: Scoping Review

Area of Science:

• Digital Health
• Artificial Intelligence in Psychiatry
• Computational Psychiatry

Background:

• Psychotic disorders are a major cause of global disability, with frequent relapse.
• Artificial intelligence (AI) offers potential for enhanced clinical monitoring in psychosis.
• Early detection of relapse is crucial for managing psychotic disorders.

Purpose of the Study:

• To systematically review and map the literature on AI-based methods for detecting relapse in psychotic disorders.
• To identify and analyze AI approaches including machine learning, deep learning, and natural language processing.
• To assess the current state and limitations of AI in psychosis relapse detection.

Main Methods:

• A systematic search of PubMed, PsycINFO, and Embase databases was conducted.
• Included studies were observational, RCTs, and quasi-experimental using AI for psychosis relapse detection.
• Data extraction and narrative synthesis were performed by independent reviewers.

Main Results:

• Ten studies utilized digital tools like smartphones, smartwatches, and social media for data collection.
• Digital phenotyping via smartphones and wearables was the most common data collection method.
• AI model efficacy varied significantly, with sensitivity from 0.25-0.77 and specificity from 0.06-0.88.

Conclusions:

• AI, particularly passive digital phenotyping, shows potential for psychosis relapse detection.
• Personalized, individual-level AI modeling demonstrates promise but requires validation.
• Future research needs larger cohorts, advanced AI methods (e.g., large language models), and collaborative efforts for clinical implementation.