24.1 VERBAL COMMUNICATION DISORDERS IN SCHIZOPHRENIA SPECTRUM PATIENTS: SYMPTOM AND SIDE EFFECT

Background: Verbal communication disturbances are a key diagnostic feature of schizophrenia. These disturbances present in different aspects, which can be assessed by looking at form and meaning. However, research on this topic is often confounded by the effects of antipsychotic medication. It therefore remains unclear which aspects of language production are influenced by antipsychotics, and which disturbances can be viewed as true psychotic symptoms. Automated language analysis has recently shown to be a useful tool to characterize verbal communication disturbances in schizophrenia. Methods: The spoken language of 42 healthy controls and 48 patients with a schizophrenia spectrum disorder was recorded using a semi-structured interview designed to elicit spontaneous speech in a natural setting. The audio was analyzed for measures of speed and quantity. For a subset of participants, the transcribed interview was analyzed using a novel Natural Language Processing (NLP) word2vec model to quantify incoherence. For patients, dopamine receptor affinity of their antipsychotic drug was estimated. Symptom severity was assessed by means of the Positive and Negative Syndrome Scale (PANSS). Results: Overall, schizophrenia spectrum patients spoke slower and produced fewer words than the healthy controls. Language measures revealed medium to strong correlations with PANSS negative and general scores. Usage of antipsychotics with strong D2 receptor affinity was found to have the strongest effect on speech. Word2vec trained models were able to differentiate between patients and controls. Conclusions: Automated assessments of aspects of verbal communication show promise in elucidating and quantifying various symptoms. Our results indicate that usage of antipsychotic medication has a marked effect on verbal communication in addition to disturbances that are better interpreted as part of the illness. These medication-effects should be taken into account when analyzing language disturbances in schizophrenia. Word2vec proved to be a useful tool to differentiate between subjects and controls, reflecting disturbances in both meaning and structure in verbal communication. This research illustrates the possibilities in automated assessment of language which can serve as measures of symptom severity, medication effects and open the door to diagnosis and prognosis.