“…In addition, comparable research in adults suggests that empirically derived algorithms for predicting treatment failure (those using empirical outcome data and average change trajectories for patients with similar symptom levels) tend to be more accurate than those that use rationally derived methods (Lambert, Whipple, Bishop et al, 2002;Spielmans, Masters, & Lambert, 2006). Subsequently, Bybee, Lambert, and Eggett (2007) tested an empirically derived method (based on multilevel modeling of empirically based change trajectories) for identifying youth at risk for treatment failure in an outpatient managed care setting, accurately identifying 72% of deteriorators. However, several factors limited the interpretability of these findings, such as the combination of parent and youth self-report data in the analyses and the neglect of potentially important predictors in the models such as patient age, diagnosis, and treatment setting variables.…”