Sign in Subscribe
Natural language processing

Early Relapse Prediction in Psychosis Using Speech-Based Markers

Strategic Insights

— 1 min read

A recent study from the Douglas Mental Health University Institute has made significant progress in predicting psychotic relapse using speech-derived markers. The research aimed to identify likely relapse in individuals with schizophrenia by leveraging Natural Language Processing (NLP). The study analyzed speech samples of 68 individuals with acute psychosis and found three lexical, syntactic, and narrative markers that strongly predicted relapse status. These markers included semantic similarity, clause complexity, and analytic thinking index. The findings suggest that a Bayesian approach can be used to select psychopathology-guided variables for speech-based relapse prediction, complementing clinical intuition in practice.

Key Takeaways:

  • The study proposes a psychopathology-based systematic approach to identify likely relapse in individuals with schizophrenia, focusing on the presence of schizophrenia in its untreated early stages and tracking disorganization in psychosis.
  • The research used NLP to derive three speech-based markers: semantic similarity, clause complexity, and analytic thinking index, from speech samples of 68 individuals with acute psychosis.
  • The speech-based model predicted relapse status with strong evidence (Bayes Factor BF = 79.5) against the clinical intuition model.
  • The study suggests that a Bayesian approach can be used to select psychopathology-guided variables for speech-based relapse prediction, complementing clinical intuition in practice.
  • The research was conducted by Min Tae M. Park and his team at the Douglas Mental Health University Institute, Dept. of Psychiatry, McGill University, Montreal, QC, Canada.
  • The study's findings have implications for early intervention and treatment in psychosis.

Statistics:

  • 68 individuals with acute psychosis were included in the study.
  • 12 out of 68 individuals experienced subsequent relapses over a year.
  • The Bayes Factor BF for the speech-based model was 79.5, indicating strong evidence for relapse prediction.
  • The study represents a preliminary step towards developing a psychopathology-guided speech-based early relapse prediction model.

Sources:

  • Park, M. T. M., Dalal, T. C., Silva, A. M., Iskhakova, S., Voppel, A., Brierley, N. J., ... & Palaniyappan, L. (2025). Clinical psychopathology-based early relapse prediction model using speech and language in psychosis. Schizophrenia Research Cognition, 43, 100392.
  • Park, M. T. M. et al. (2025). Report Summarizes Psychosis Study Findings from Douglas Mental Health University Institute (Clinical psychopathology-based early relapse prediction model using speech and language in psychosis). Mental Health Weekly Digest, October 20, 2025; p 723.

Read more