Objective: Some cases are thought to be more complex and difficult to treat, although there is little consensus on how to define complexity in psychological care. This study proposes an actuarial, data-driven method of identifying complex cases based on their individual characteristics. Method: Clinical records for 1,512 patients accessing low- and high-intensity psychological treatments were partitioned in 2 random subsamples. Prognostic indices predicting post-treatment reliable and clinically significant improvement (RCSI) in depression (Patient Health Questionnaire-9; Kroenke, Spitzer, & Williams, 2001) and anxiety (Generalized Anxiety Disorder-7; Spitzer, Kroenke, Williams, & Löwe, 2006) symptoms were estimated in 1 subsample using penalized (Lasso) regressions with optimal scaling. A PI-based algorithm was used to classify patients as standard (St) or complex (Cx) cases in the second (cross-validation) subsample. RCSI rates were compared between Cx cases that accessed treatments of different intensities using logistic regression. Results:St cases had significantly higher RCSI rates compared to Cx cases (OR = 1.81 to 2.81). Cx cases tended to attain better depression outcomes if they were initially assigned to high-intensity (vs. low intensity) interventions (OR = 2.23); a similar pattern was observed for anxiety but the odds ratio (1.74) was not statistically significant. Conclusions: Complex cases could be detected early and matched to high-intensity interventions to improve outcomes.