Profiles of mental ill-health and health within a large, representative UK school-based sample of adolescents

To support effective healthcare delivery, alternative frameworks are needed that capture the real-world complexity of mental health and ill-health in young people. We applied an unsupervised machine learning algorithm (Growing Hierarchical Self-Organizing Map; GHSOM) to data from a representative sample of over 26,000 adolescents (11-14 years) from 85 UK schools. Input features included anxiety, depression, conduct issues, hyperactivity, interpersonal challenges, and markers of mental well-being. Eight reproducible subgroups were identified for males and females. Profiles ranged from flourishing adolescents to those with widespread social-emotional and behavioural difficulties and low mental well-being. Other profiles showed moderate well-being despite minimal symptoms, or focal patterns of internalising or externalising. Regarding sex, females were more likely to experience patterns of emotional distress and depression, while males tended to display depression, behavioural, and socialization challenges. Overall, this study presents a data-driven, multidimensional framework for classifying youth mental health and informing targeted intervention and healthcare solutions.