OBJECTIVES: Many patients on antidepressants are not responsive to first-line therapy (‘treatment-resistant' depression [TRD]) and can undergo switches and optimizations to discover a beneficial therapeutic regimen. While patients with more complex forms of TRD have higher costs than non-TRD patients, little is known about the cost effects for patients along a gradient of TRD classifications (from moderate to complex). METHODS: Patients aged 18–64 years in employer-sponsored plans with at least three years of continuous medical and prescription coverage and at least one antidepressant prescription were found in the 2000–2006 MarketScan Database (n=78,476). An MGH TRD scale value (range from 0 to 16.5) was calculated for each patient and a value exceeding 3.5 indicated TRD. Twelve-month direct medical and prescription drug expenditures for patients with TRD (n=22,593) were compared to expenditures among an equal number of propensity-score matched patients with non-TRD depression. Propensity scores were estimated via demographic characteristics and case-mix. Generalized linear models (gamma family and log link) controlled for demographic and case-mix factors. RESULTS: Average 12-month direct medical care and prescription drug expenditures were almost 40% higher for TRD ($9470) compared to matched non-TRD patients ($6813) (p<0.01). A one-unit increase in TRD score was associated with a $772 increase in annual costs (p<0.01). Compared with a matched group of non-TRD patients, annual costs for patients were higher in each MGH score catergory: 3.5–4, 23.6%; 4.5–5, 32.9%; 5.5–6, 44.6%; 6.5+, 61.1% (all p<0.01). CONCLUSIONS: TRD is a costly disorder and merits consideration as interventions are developed to manage the burden of disease and improve productivity. Even patients with less complex forms of TRD have costs far in excess of those without TRD. Dichotomous definitions of TRD may not be adequate; a gradient from moderate to complex TRD may be more useful for providers and insurers.