Abstract
Recent reviews of the effectiveness of residential treatment for adolescents have attempted to identify predictors of outcome in the residential setting. In general, variables which predict outcome relate to the pathology of the adolescent, factors relating to their family and aspects of the treatment itself. However, it has not proved possible to identify a consistent set of predictors of outcome across residential settings or treatment programmes. The heterogeneity of the outcome data has been attributed to serious methodological flaws present in much of the recent research. Strategies used in conventional psychotherapy outcome trials to overcome methodological weaknesses have been suggested as remedies to improve the quality of the outcome data. This paper discusses why such strategies may be difficult to implement in residential settings. Alternative designs, which are theoretically motivated, are then considered. These designs may assist in overcoming some of the problems in conducting good quality research in residential settings.