Objective:

To evaluate the association between short, naturalistic speech samples and concurrent variation in psychotic symptoms using standardized scales, specifically the Positive and Negative Syndrome Scale (PANSS) and the Brief Psychiatric Rating Scale (BPRS).

Approach:

• Cohort Study Design: Conducted a longitudinal study with Dutch and US cohorts, assessing speech samples and symptom severity using PANSS and BPRS, involving participants diagnosed with schizophrenia spectrum or bipolar I disorder.
• Speech Sample Collection: Collected spontaneous speech samples through open-ended questions, with Dutch interviews conducted by trained researchers and US samples collected via a touch-screen tablet application.
• Feature Extraction: Processed speech recordings to extract a total of 160 features for the Dutch cohort and 138 for the US cohort, including acoustic, syntactic, semantic, and sentiment features.
• Data Analysis: Performed principal component analyses to reduce dimensionality and used linear mixed-effects models to evaluate associations with symptom severity, ensuring robust statistical methods were applied.

Key Findings:

• Speech-based models can detect symptom severity with accuracy, although specific metrics are not detailed.
• Models can identify specific symptoms such as hallucinations and blunted affect.
• Speech features align with clinician observations when rating symptoms.
• Models capture psychopathology beyond demographic characteristics.

Interpretation:

The study explores the use of speech analysis as a potential method for monitoring psychotic symptoms, though further validation is needed.

Limitations:

• The study's findings may not be generalizable beyond the specific cohorts studied.
• Variability in speech due to background noise and recording artifacts may affect results, which should be considered in future research.

Conclusion:

The findings suggest the potential for using speech analysis in monitoring psychotic symptoms, but further research is necessary to establish its efficacy.