Work

Machine learning analysis of clinical, psychological and sociodemographic factors predicting mental well-being in patients with severe mental illness: Insights from the French REHABase cohort

Year: 2025

Type: article

Abstract: Mental well-being is a cornerstone of recovery for people with mental disorders. Unfortunately, despite many studies on the topic, the literature still lacks results from large samples processed using advanced models capable of taking numerous variables into account. This study aims to leverage robust machine learning techniques on data from the REHABase cohort to identify key predictors of mental well-being in patients with severe mental disorders, with a particular focus on schizophrenia. In all, 47 clinical, psychological, and sociodemographic variables were selected. Three distinct samples were analyzed: all patients ( n = 2206), schizophrenia patients ( n = 1136) and non-schizophrenia patients ( n = 1070). The performances of 3 machine learning algorithms were compared: regularized linear regression, random forest, and gradient boosting with tenfold cross-validation. The regularized linear regression model was found to be the best performing in terms of generalizability. Results highlighted the influence of the internalized stigmatization subscales of resistance, resilience, and alienation alongside the resilience dimension of quality of life in predicting mental well-being across all of the disorders. For schizophrenia patients, mental well-being was primarily predicted by social withdrawal. Using advanced machine learning methodology on a large sample of patients with transdiagnostic exploration, our results showed both the crucial role of resilience and resistance to stigma in overall mental well-being and the critical importance of social contact in schizophrenia. These insights should help improve care trajectories for individuals with severe psychopathologies, particularly by promoting social cognitive remediation in schizophrenia. • Large-scale analysis using advanced machine learning helps overcome previous limitations in mental well-being research. • Three leading machine learning algorithms were compared, with regularized linear regression showing the best generalizability for predicting mental well-being. • Internalized stigma and resilience are identified as major predictors of mental well-being across all severe mental disorders. • Among patients with schizophrenia, social withdrawal stands out as the most influential factor for mental well-being. • The findings highlight the need to focus on resilience, stigma reduction, and social cognitive remediation—especially for schizophrenia care.

Source: Comprehensive Psychiatry

DOI: 10.1016/j.comppsych.2025.152651