This paper critiques current models for evaluation of mental health services provided to children and adolescents with emotional disturbances and offers an alternative model. The most popular model, Donabedian's tripartite model, is described. This model separately evaluates three domains: (1) structure (characteristics of the care providers, their tools and resources, and the physical/organizational setting); (2) process (both interpersonal and technical aspects of the treatment process); and (3) outcome (change in the patient's symptoms and functioning). The paper then notes three approaches used to assess these domains: the use of standards and guidelines, peer review, and outcomes monitoring. Problems with these approaches are identified including: lack of validity data on the importance of structure and process indicators, the poor reliability of peer reviews, and difficulty in relating outcomes directly to services. The paper proposes a model in which quality is viewed strictly as the relationship between quality indicators (i.e., structure and process variables) and quality validators (i.e., outcomes). This approach is in contrast to viewing quality as existing separately within each component. Benefits of this approach include better services, use of quality indicators with known validity, and better program planning. (Contains 19 references.) (DB)