Objective: To compare the predictive validity of the clinical-diagnostic and the empirical-quantitative approach to assessment of childhood psychopathology, and to investigate the usefulness of combining both approaches. Method: A referred sample (N=96), aged 6 to 12 years at initial assessment, was followed up across - on average - a period of 3.2 years. It was assessed to what extent DISC/DSM-III-R diagnoses - representing the clinical-diagnostic approach, and CBCL scores - representing the empirical-quantitative approach, predicted the following signs of poor outcome: outpatient/inpatient treatment, or parents' wish for professional help for the child at follow-up, disciplinary problems in school, and police/judicial contacts. Results: Both diagnostic systems added significantly to the prediction of poor outcome, and neither of the two systems was superior. Use of both systems simultaneously provided the most accurate estimation of the prognosis, reflected by the occurrence of future poor outcome. Even diagnostic concepts that are generally regarded as relatively similar, such as ADHD (DSM) and attention problems (CBCL), or conduct disorder (DSM) and delinquent behavior (CBCL), appeared to differ in their ability to predict poor outcome. Conclusions: The present study supports the use of the empirical-quantitative approach and the clinical-diagnostic approach simultaneously, both in research and in clinical settings, to obtain a comprehensive view of the prognosis of psychopathology in children. © Association for Child Psychology and Psychiatry, 2004.