OBJECTIVE: Major depressive disorder (MDD) is a debilitating condition with significant economic consequences. Estimates indicate that up to 30% of individuals with MDD are treatment-resistant (TRD). The study objectives were (a) evaluation of studying TRD using claims data; (b) estimation of cost differences between TRD and non-TRD patients. METHODS: Data source was administrative claims data from a Fortune 100 manufacturer. Claims included medical, pharmaceutical and disability claims for 1996-1998 (n > 100,000). The sample was restricted to claims for MDD (NMDD = 4,186). Using a treatment algorithm, patients were classified into TRD-likely and TRD-unlikely groups (NTRD = 487). Resource utilization was compared between TRD-likely, TRD-unlikely patients, and a sample of the overall population. RESULTS: The algorithm classified twelve percent of the MDD sample as TRD. Average annual costs were $10,954 for TRD-likely patients, $5,025 for TRD-unlikely patients, and $3,006 for average beneficiaries. The average number of health claims among TRD-likely patients were one and a half times greater than that of TRD-unlikely patients. CONCLUSION: Resource utilization by TRD-likely patients is substantial, not only for direct treatment of depression but also for treatment of co-morbid medical conditions. Additionally, TRD imposes substantial indirect costs on employers, primarily resulting from high rates of depression associated disability.