How do we know whether treatment has failed? Paradoxical outcomes in counseling with young people

Background In both routine practice contexts and research studies, evidence from standardized self-report symptom measures, administered pre- and post-treatment, is predominantly used to determine whether psychotherapy has been successful. Understanding the nature of unsuccessful psychotherapy requires an ability to evaluate the credibility of outcome data generated by such techniques. An important body of research has identified discrepancies between outcomes assessed through symptom measures and those obtained from other sources. However, not enough is known about the extent to which such paradoxical outcomes exist.Objective This study analyzes the relationship between outcomes, as assessed by a standardized self-report measure, and as assessed by ratings of young people's descriptions of change at post-counseling interviews.Methods Participants were 50 young people (13-16 years old) who had taken part in a trial of up to 10 weeks of school-based humanistic counseling. Our primary standardized measure was the Young Person's CORE (YP-CORE). To assess young people's experiences of counseling change, three independent raters scrutinized transcripts of post-counseling interviews, and scored levels of helpfulness on a 1 (Not at all helpful) to 10 (Extremely helpful) scale. Inter-rater reliabilities were 0.94 (Cronbach's Alpha) and 0.96 (McDonald's Omega). Sensitivity analyses were conducted to explore relationships between helpfulness ratings and other outcome measures, i.e., satisfaction with counseling (ESQ) and the Goal-Based-Outcome Tool (GBO), and process measures, i.e., the Working Alliance Inventory (WAI-S) and the Barret Lennard Relationship Inventory (BLRI).Results Multilevel analysis indicated that helpfulness ratings were not significantly associated with changes in YP-CORE scores. Analyzed categorically, 38% of those showing reliable improvement on the standardized measure were below the median for self-described helpfulness, and 47% of those not showing reliable change were at or above the median for self-described helpfulness. Sensitivity analyses demonstrated closer correlations between helpfulness ratings and other outcome measures (ESQ and GBO), and between helpfulness ratings and process measures (WAI-S and BLRI).Discussion Our results raise questions about reliance on symptom change outcome measures for defining treatment success and failure, given their disparity with clients' own descriptions of the helpfulness of therapy. Implications for practice and research are discussed.