Retraction to: S117. DOES ADDING AUTOMATED ANALYSIS OF SPEECH COHESION TO DIMENSIONALIZED R-PAS COGNITIVE CODES IMPROVE OUR ABILITY TO ASSESS DISORGANIZED THINKING?

Abstract

Abstract Background Research is increasingly pointing toward the need for a dimensional, rather than categorical, conceptualization of psychopathology. This is particularly true within the literature on psychosis and related disorders. Taking a dimensional approach to conceptualizing psychosis requires deconstructing its components in assessment. The Rorschach Performance Assessment System (R-PAS) contains meta-analytically supported scales with international norms that are used to categorically rate signs of disordered thinking and perceptions in individuals’ responses to the Rorschach inkblots. Recently, a new dimensionalized set of scales called the Scales of Problematic Communication and Thinking (SPCT) have been developed for R-PAS with three main components: Disturbed and Confusing Communication, Illogical Verbal Justification, and Incongruous Perceptual Combinations. Research also suggests that linguistic measures of speech cohesion can be used to successfully predict conversion to psychosis in clinical high-risk patients and may be more related to stable neurocognitive deficits than clinical measures of disorganized speech. Coh-Metrix, an automated speech analysis software, provides over 100 specific measures of linguistic cohesion, such as connectivity, syntax simplicity, and referential cohesion. The purpose of this presentation is twofold: (a) to report recent findings showing the SPCT provides incremental validity over the traditional R-PAS measure of disordered thinking in predicting clinician ratings of disorganized thinking on the PANSS and (b) to determine if Coh-Metrix measures of linguistic cohesion provide incremental validity to SPCT ratings in predicting these PANSS ratings. Methods R-PAS protocols from a maximum-security inpatient forensic hospital (N = 91) aged 19 to 80 years (M=40) and predominantly male (89%) were coded for thinking and perceptual disturbances using the R-PAS traditional measure of disordered thinking and the 6-point dimensional SPCT measure. The patients’ primary clinician provided PANSS ratings. Protocols will be coded for speech cohesion using 15 indices from the automated speech analysis software Coh-Metrix, which were chosen based on a literature review. Results Interrater reliability was excellent for the SPCT ratings (ICC = 0.97) and good for the PANSS clinician ratings (ICC = 0.71). In a previous study, SPCT ratings showed significant associations with the clinician ratings of disorganized thinking on the PANSS (r = 0.42, p < 0.01, N = 90); hierarchical regression analyses demonstrated incremental validity over the traditional R-PAS measure of disordered thinking (ΔR = 0.28, p < 0.01). Using hierarchical regression analyses, the Coh-Metrix indices of cohesion are predicted to provide incremental validity to the SPCT ratings. Discussion The SPCT shows promise as a reliable and valid dimensionalized measure for assessing the continuum of clear thinking to psychosis-level disturbances. The measures of linguistic cohesion provided by Coh-Metrix have the potential to offer clinicians a quick, efficient, and objective method for assessing disorganized thinking. The R-PAS international norms are currently being coded for SPCT and, if Coh-Metrix measures provide incremental validity, will be coded for these measures as well. Using these measures in combination with SPCT ratings can provide clinicians with a clearer understanding of this significant component of psychosis.