Prediction of the functional outcome in Children and Adolescents with and without clinical high risk- a Machine Learning Approach

Aims: Studies on the early detection of psychosis have shown that numerous patients at clinical high risk (CHR) do not develop a manifest psychosis but, nevertheless, do present poor clinical outcomes. Therefore, psychosocial functioning has become a valuable parameter to quantify disease state in CHR patients. Methods: Within a Swiss-German naturalistic longitudinal study, we examined psychosocial functioning in 8- to 17-year-olds using the Social and Occupational Functioning Scale (sofas) within 2 years. The sample included 69 clinical high-risk patients, 147 inpatients not suspected to develop psychosis and 110 community subjects. Participants were examined with the Schizophrenia Proneness Instrument-Child & Youth version (SPI-CY), the Structured Interview for Psychosis-Risk Syndromes (SIPS), the Structured Clinical Interview for DSM-IV and a neurocognitive test battery. We focussed on participants with poor psychosocial functioning (sofas < 70) at T0 and differentiated between 'remaining impaired' versus 'recovered' at either T1 or T2. An external validation was performed with a 13- to 17-year-old CHR sample of an independent prospective naturalistic Swiss study (N=41, 19 recovered). Results: One hundred and ninety participants presented a recovered psychosocial functioning in the model developing sample. Twenty-three significant predictors were identified. Using these predictors only, BAC increased to 75.9% (sensitivity: 67.9%, specificity: 83.8%). External validation revealed a sufficient generalizability of BAC = 64.7% (sensitivity: 61.5%, specificity: 67.9%). Conclusions: The higher number of predictors, compared to similar analyses in adults, indicate that predictions in children and adolescents might be more complex than in adults, due to ongoing developmental processes