Views of People With Psychosis About Algorithm-Based Relapse Prediction and Data Sharing: Qualitative Study

Area of Science:

• Digital mental health
• Psychosis relapse prevention
• Algorithmic prediction

Background:

• Preventing psychosis relapses is challenging, with digital remote monitoring (DRM) systems and algorithmic relapse prediction emerging as key tools.
• Limited research exists on the perspectives of individuals with psychosis regarding these algorithmic prediction systems.

Purpose of the Study:

• To conduct an in-depth examination of the views of people with psychosis on algorithmic relapse prediction within DRM systems.
• To explore perspectives on systems integrating active symptom monitoring and passive sensing data.

Main Methods:

• Qualitative interviews were conducted with 58 individuals with psychosis across 6 UK regions.
• Reflexive thematic analysis was used to analyze interview transcripts.
• Individuals with lived experience were integral to study design, analysis, and reporting.

Main Results:

• Participants stressed the importance of algorithm accuracy (sensitivity/specificity) and transparency, highlighting risks of false positives/negatives.
• A human-in-the-loop approach, blending digital monitoring with human oversight and feedback, was favored to mitigate errors.
• Key themes included trust in the system and clinical team, fears of over/under-reaction, the necessity of user choice, and perceived benefits like early intervention and reduced bias.

Conclusions:

• Individuals with psychosis see potential benefits in algorithm-assisted relapse prediction for timely and efficient care.
• Effective implementation requires accurate algorithms, interpretation by informed human oversight, valid consent, and respect for user autonomy.
• The use of these systems should avoid increasing restrictions and prioritize user choice and voice.