Journal Article

DZNE-2026-00872

http://join2-wiki.gsi.de/foswiki/pub/Main/Artwork/join2_logo100x88.png Delusional themes and forensic status in schizophrenia spectrum disorders: a machine learning based discriminative study.

Eleryan, A. A. N. ; Boudriot, F. ; Dörner, M.DZNE* ; Hofmann, A. B. ; Machetanz, L. ; Kirchebner, J.

2026
Frontiers Research Foundation Lausanne

This record in other databases:    

Please use a persistent id in citations: doi:10.3389/fpsyg.2026.1840906

Abstract: Schizophrenia spectrum disorders (SSD) are associated with an elevated risk of violent and offending behavior; however, the specific contribution of delusional content remains insufficiently understood. Prior research has yielded inconsistent findings regarding the relationship between particular psychopathological symptoms and delusional themes and criminal behavior. A better understanding of symptom-specific phenomenological differences between forensic psychiatric (FPP) and general psychiatric patients (GPP) may inform clinical assessment in forensic settings. This study therefore aimed to examine whether specific psychopathological symptoms discriminate between FPP and GPP with SSD.In this retrospective study, data from 740 patients with SSD (370 FPP, 370 GPP) treated at a Swiss university hospital were analyzed. From an initial dataset comprising over 500 variables, 103 symptom-focused variables were selected to isolate psychopathological characteristics. Following preprocessing, the dataset was split into training and validation subsets. After the application of dimension reduction techniques, multiple supervised machine learning algorithms were applied and compared using nested cross-validation. Model performance was assessed using area under the curve (AUC), balanced accuracy, sensitivity, and specificity. A post hoc analysis further examined the impact of comorbid substance use and personality disorders.Naïve Bayes achieved the best performance (AUC = 0.81, balanced accuracy = 78.2%), with stable results in the validation sample. Two variables emerged as key discriminators: bizarre delusions (46% in FPP vs. 2% in GPP) and aggressive delusional content (69% vs. 19%). Global symptom severity measures, including PANSS scores, did not contribute to the model. The post hoc analysis resulted in only minimal changes, with estimates remaining within the original confidence intervals, indicating robust model stability.The qualitative content of delusions - particularly bizarre and aggressive themes - shows substantial discriminative value for forensic status in SSD, beyond overall symptom severity and established comorbid risk factors. These findings highlight the relevance of symptom-specific, cognitive-affective mechanisms in forensic evaluations. Incorporating structured assessment of delusional content may complement existing psychometric approaches in forensic psychiatric evaluations.

Keyword(s): criminal behavior ; delusions ; machine learning ; psychopathology ; schizophrenia spectrum disorders

---

Contributing Institute(s):

1. Mixed Cerebral Pathologies and Cognitive Aging (AG Schreiber)

Research Program(s):

1. 353 - Clinical and Health Care Research (POF4-353) (POF4-353)

---

Database coverage:
; ; ; Article Processing Charges ; Clarivate Analytics Master Journal List ; Current Contents - Clinical Medicine ; Current Contents - Social and Behavioral Sciences ; DOAJ Seal ; Essential Science Indicators ; Fees ; IF < 5 ; JCR ; PubMed Central ; SCOPUS ; Social Sciences Citation Index

---