Abstract
Across 5 decades, hundreds of randomized trials have tested psychological therapies for youth internalizing (anxiety, depression) and externalizing (misconduct, attention deficit and hyperactivity disorder) disorders and problems. Since the last broad-based youth meta-analysis in 1995, the number of trials has almost tripled and data-analytic methods have been refined. We applied these methods to the expanded study pool (447 studies; 30,431 youths), synthesizing 50 years of findings and identifying implications for research and practice. We assessed overall effect size (ES) and moderator effects using multilevel modeling to address ES dependency that is common, but typically not modeled, in meta-analyses. Mean post-treatment ES was 0.46; the probability that a youth in the treatment condition would fare better than a youth in the control condition was 63%. Effects varied according to multiple moderators, including the problem targeted in treatment: Mean ES at posttreatment was strongest for anxiety (0.61), weakest for depression (0.29), and nonsignificant for multiproblem treatment (0.15). ESs differed across control conditions, with "usual care" emerging as a potent comparison condition, and across informants, highlighting the need to obtain and integrate multiple perspectives on outcome. Effects of therapy type varied by informant; only youth-focused behavioral therapies (including cognitive-behavioral therapy) showed similar and robust effects across youth, parent, and teacher reports. Effects did not differ for Caucasian versus minority samples, but more diverse samples are needed. The findings underscore the benefits of psychological treatments as well as the need for improved therapies and more representative, informative, and rigorous intervention science.