Andrea Polari1,2,3, Suzie Lavoie2,3, Hok Pan Yuen2,3, Paul Amminger2,3, Gregor Berger4, Eric Chen5, Lieuwe deHaan6, Jessica Hartmann2,3, Connie Markulev2,3, Dorien Nieman7, Merete Nordentoft8; 1Orygen Youth Health, 2Orygen, The National Centre of Excellence in Youth Mental Health, Melbourne, Australia, 3Centre for Youth Mental Health, The University of Melbourne, Melbourne, Australia, 4Department of Child and Adolescent Psychiatry and Psychotherapy, University Hospital of Psychiatry Zurich, University of Zurich, Switzerland, 5Department of Psychiatry, the University of Hong Kong, Hong Kong, China, 6Academic Medical Centre, University of Amsterdam and Arkin Institute for Mental Health, Amsterdam, The Netherlands, 7Academic Medical Centre, Department of Psychiatry, Amsterdam, The Netherlands, 8Psykiatrisk Center København, Forskningsenheden, København, Denmark
Introduction Several prediction models have been introduced to identify young people at greatest risk of transitioning to psychosis but to date none have examined prediction of outcomes other than transition. Aims To examine associations between baseline clinical predictors and outcomes besides transition to psychosis and to develop a prediction model for those outcomes. Method Several evidence-based variables previously associated with transition to psychosis and some important clinical comorbidities experienced by Ultra-High Risk (UHR) individuals were identified in 202 UHR individuals. Analyses were conducted to investigate the associations between these variables and favourable (remission and recovery) or unfavourable (transition to psychosis, no remission, any recurrence and relapse) clinical outcomes, which were defined using the individuals' responses in terms of positive symptoms and functioning at months 3, 6, 9 and 12. Logistic regression, best subset selection, Akaike Information Criterion and Receiver Operating Characteristic curves were used to seek the best prediction model for clinical outcomes from all combinations of possible predictors. The baseline variables considered included MADRS, SANS, BPRS, Attenuated Psychotic Symptoms, SOFAS, Substance Use Disorders, Duration of Untreated Symptoms (DUS) and Borderline Personality Disorder. Results When considered individually, only higher Total BPRS (p=0.023) and increased DUS (p=0.042) were associated with unfavourable outcomes. Two best prediction models of clinical outcomes were identified: 1. Total BPRS and SOFAS and 2. Total BPRS, SOFAS and DUS. Discussion Although Total BPRS, SOFAS and DUS appear to be predictive of unfavourable outcomes, the predictive performance of the resulting model requires improvement and further research is needed.
