Harmonizing Mental Health Assessment: Breaking Down Formal Thought Disorder into Manageable Subconstructs

Despite being a common and cross-diagnosed symptom dimension across mental disorders, formal thought disorder (FTD) remains plagued by inconsistent terminology and conceptual standardization. Clinicians use a mix of FTD-specific scales and items from generic scales, which can lead to misunderstandings and misclassifications. Researchers have now developed a novel method using natural language processing to compare definitions, labeling, and overlap of FTD symptoms across psychopathological scales. This method, involving a three-pronged validation approach, aims to identify underlying FTD subconstructs and improve consistency across rating scales.

Key Takeaways:

• Researchers used a three-pronged validation approach to analyze semantic clusters of FTD scale items, involving sentence-BERT, item-to-item similarity matrix, and data-level correspondence.
• The BERT-informed clustering of cross-scale items highlighted four coherent FTD groupings: muddled communication & incomprehension, abrupt topic shifts, inconsistent narrative structure, and restricted speech.
• Expert raters showed moderate-to-high overlap (Fleiss' kappa = .617) with computational clusters.
• The study indicates that FTD rating scales measure overlapping, semantically related constructs that drive item-level correlations.
• Semantic clustering acts as a novel method to harmonize multi-scale data and pinpoint discrepancies between expert and machine classifications.
• The study involved 98 participants (49 healthy controls and 49 schizophrenia/affective psychosis) and used 103 items across seven scales.
• The study's results were validated through correspondence with published factor-analytic divisions.
• The researchers observed a 73% match with prior factor analyses when assigning TALD items to BERT-derived positive or negative FTD groupings.

Statistics:

• 73% match with prior factor analyses when assigning TALD items to BERT-derived positive or negative FTD groupings (p
• 4 coherent FTD groupings identified through BERT-informed clustering: muddled communication & incomprehension, abrupt topic shifts, inconsistent narrative structure, and restricted speech.
• 98 participants (49 healthy controls and 49 schizophrenia/affective psychosis) used in the study.
• 103 items across seven scales used in the study.
• 30 Thought and Language Disorder scale (TALD) items used in the study.
• 49 matched pairs of CLANG-TLC items used to map to BERT-derived clusters.

Sources:

• osf.io/preprints/psyarxiv/9eqtp_v1/ (preprint abstract, not peer-reviewed)