Abstract
In schizophrenia, patient-reported rehabilitation outcomes provide a unique perspective that goes beyond standard clinical practice, which has traditionally relied on socio-demographic and basic clinical data for diagnosis and treatment decisions. However, there is ongoing debate about whether these routinely collected data meaningfully align with patients' rehabilitation goals. In the current study, we used data from the French network of rehabilitation centers REHABase to predict 19 patient-reported rehabilitation outcomes related to quality of life, self-stigma, self-esteem, insight, and medication adherence in schizophrenia. We employed a SuperLearner ensemble combining a variety of base learners. Predictions were based on 21 socio-demographic and basic clinical predictors. The number of observations ranged from 1,332 (medication adherence) to 1,563 (insight). Approximately three-quarters of the participants were male, with a mean age of 33 years. The SuperLearner ensemble achieved an overall R² of 0.048 (range: 0.01-0.08) on the hold-out testing sets. Most base learners, including general linear models, exhibited a performance gap between the training and testing sets. Overall, socio-demographic and basic clinical factors demonstrated limited ability to explain variations in patient-reported rehabilitation outcomes in schizophrenia. This suggests the need for targeted treatment programs and personalized service provision beyond standard psychiatric practice to enhance rehabilitation care.