BACKGROUND: Treatment resistant depression (TRD) is common in daily practice. An empirical, widely accepted and applicable measure to quantify TRD is lacking. Previously, the Maudsley Staging Method (MSM) showed good validity. We aimed to improve the MSM by refining and extending its items resulting in the Dutch Measure for quantification of TRD (DM-TRD). METHODS: In addition to duration, severity and failed treatments in the current depressive episode, we added items for functional impairment, comorbid anxiety, personality disorders and psychosocial stressors. We extended the augmentation section and added items for failed psychotherapy and intensified treatment. We examined psychometric properties of the DM-TRD and tested prediction of future depressive symptoms and remission after 16 weeks in 274 (DSM-IV) depressed in- and outpatients entering naturalistic treatment. RESULTS: The DM-TRD showed excellent inter-/intra-rater reliability. Higher scores were associated with more symptoms and less remission during follow-up. The DM-TRD outperformed the MSM in prediction of future depressive symptomatology. Remission was predicted equally well by both measures. Longer duration of the current episode, larger functional impairment and larger baseline symptom severity were the strongest predictors of symptomatology at follow-up. Longer duration and larger functional impairment were negatively associated with remission. LIMITATIONS: Longer follow-up could have increased predictive power. Addition of items for somatic co-morbidity, childhood adversity and psychotic features must be investigated further. CONCLUSION: The DM-TRD has excellent psychometric properties and better predictive validity for clinical outcome than other sophisticated measure published to date. Its use in clinical practice and research will improve treatment planning in TRD-patients.