Attenuated symptoms are associated with connectedness and emotional expression in narratives based on emotional pictures in a Brazilian clinical high-risk cohort

Abstract

Natural language processing tools have been applied to identify symptoms through the analysis of narratives. Here we use this strategy to map attenuated symptoms when applied to prodromal stages of psychosis. We assessed narratives in a population-based prodromal screen-positive Brazilian cohort of individuals confirmed through structured interview as at clinical high-risk for psychosis (CHR) (N = 42) or as controls (N = 29) to identify language markers predictive of transition to diagnosis or remission. Participants were identified by screening >4500 individuals with the Prodromal Questionnaire–Brief and the Perceptual and Cognitive aberrations scale, followed by the Structured Interview for Psychosis-Risk Syndromes. After 2.5 years of follow-up, 23 CHR individuals were found to have transitioned to mood and anxiety disorders and 4 to psychosis. We assessed word-recurrence graph connectedness and emotional content in baseline narratives elicited with affective pictures, finding CHR markers and associations with mood symptoms. Specifically, greater recurrence (vs. randomness) in reciprocal connectedness was associated with dysphoric mood (ρ=0.64), social anhedonia (ρ=0.56), and perseveration (ρ=0.53), suggesting these may be related to rumination related to depression in individuals with psychosis risk. Language markers combined with years of education explained 58 %, 57 %, and 44 % of dysphoric mood, social anhedonia, and perseveration severity variance, respectively.