Depression is present in about 20-30% of hemodialysis patients and is associated with morbidity and mortality. However, depression is inadequately diagnosed and treated among dialysis patients. This is due in part to the overlap between depressive symptoms (e.g. appetite change, trouble sleeping, feeling tired) and symptoms related to persistent metabolic derangements in hemodialysis patients (e.g. nausea, nocturnal cramps, feeling washed out after treatment). The overlap between depressive symptoms and dialysis-related complications makes it difficult to diagnose and therefore to treat depression. In addition, prescription of antidepressant medication may increase an already high pill burden and result in poor adherence. Moreover, the evidence base to guide depression treatment among hemodialysis patients is limited. In our previous work, we developed methods to use latent variables and structural equation modeling to isolate depressive symptoms. Other investigators have demonstrated that directly observed treatment enhances the effectiveness of tuberculosis and HIV treatment. We now propose a cross-sectional study followed by a randomized controlled trial at 17 dialysis facilities. The cross-sectional study will involve assessments of depressive symptoms (using the PHQ-9 screening instrument) as well as dialysis-related complications. We will then use structural equation modeling to develop and validate a hemodialysis-specific PHQ-9 (hdPHQ-9) that will isolate depressive symptoms. The trial will involve 216 patients with confirmed depression who will be randomly assigned to (a) directly observed weekly antidepressant treatment with fluoxetine or (b) referral to their nephrologists, their primary care physicians, or nearby mental health providers. The primary outcome of the trial will be remission of depression at 12 weeks. The trial results will also be used to compare the responsiveness of the PHQ-9 and the hdPHQ-9. We anticipate that the hdPHQ-9 will be a valid and responsive instrument that will isolate depressive symptoms in hemodialysis patients and ultimately improve the screening and diagnosis of depression. We also expect that directly observed weekly fluoxetine treatment will be an effective way to manage depression among hemodialysis patients. Innovative features of the proposed project include the use of latent variables to address overlap, administration of a long acting weekly antidepressant, directly observed treatment, and a rigorous randomized controlled trial design. The project has the potential not only to improve the diagnosis and management of depression among hemodialysis patients but also to improve their morbidity and mortality. Furthermore, it may serve as a model for future studies to isolate symptoms among overlapping medical conditions.