Delusional themes and forensic status in schizophrenia spectrum disorders: a machine learning based discriminative...

Area of Science:

• Psychiatry
• Forensic Psychology
• Machine Learning

Background:

• Schizophrenia spectrum disorders (SSD) are linked to increased risk of offending behavior.
• The role of specific delusional content in this association requires further clarification.
• Understanding phenomenological differences between forensic psychiatric patients (FPP) and general psychiatric patients (GPP) is crucial for forensic assessment.

Purpose of the Study:

• To investigate whether specific psychopathological symptoms can differentiate FPP from GPP with SSD.
• To identify key discriminators of forensic status within the context of SSD.

Main Methods:

• Retrospective analysis of 740 patients with SSD (370 FPP, 370 GPP).
• Selection and analysis of 103 symptom-focused variables.
• Application and comparison of supervised machine learning algorithms using nested cross-validation.
• Performance assessment using AUC, balanced accuracy, sensitivity, and specificity.

Main Results:

• The Naïve Bayes algorithm demonstrated the best performance (AUC=0.81, balanced accuracy=78.2%).
• Bizarre delusions (46% in FPP vs. 2% in GPP) and aggressive delusional content (69% vs. 19%) were key discriminators.
• Global symptom severity measures did not significantly contribute to the model.

Conclusions:

• The qualitative content of delusions, specifically bizarre and aggressive themes, holds significant discriminative value for forensic status in SSD.
• These findings emphasize the relevance of symptom-specific cognitive-affective mechanisms in forensic evaluations.
• Structured assessment of delusional content can enhance existing forensic psychiatric evaluations.