Recovery Curves for the PHQ-8 and GAD-7: Clinical Utility and Implementation of Measurement-based Care in a Real-world Outpatient Setting

Abstract Measurement-based care (MBC) relies on repeated outcome assessment and feedback to guide clinical decision-making. Although recovery curves are central to many MBC systems, they have primarily been developed using session-by-session data. This study developed and validated recovery curves for the PHQ-8 and GAD-7 using monthly assessments within a naturalistic outpatient setting. We generated our recovery curves using mixed-effects models with symptom change over four months of treatment as the dependent variable ( N = 1124) and cross-validated the results in a second sample (PHQ-8 N = 734; GAD-7 N = 758). Models were evaluated for predictive performance and calibration. We examined additional baseline predictors and evaluated confidence interval thresholds for identifying patients at risk of treatment failure. Predictive performance was strong in the derivation sample (PHQ-8 R² = 0.44; GAD-7 R² = 0.47) and remained robust in validation (PHQ-8 R² = 0.37; GAD-7 R² = 0.40) with acceptable calibration. Baseline symptom severity was the only clinically meaningful predictor of recovery trajectories. A 60% one-sided confidence interval provided the optimal balance of sensitivity and alert frequency, identifying approximately 24–28% of patients as at risk of treatment failure at one month. We conclude that recovery curves for the PHQ-8 and GAD-7 can be effectively derived using monthly assessments and applied in real-world MBC settings.