OBJECTIVES: Depression symptoms in medically ill patients are known to adversely impact treatment compliance, healthcare costs and outcomes. Thus, this study aims to compare costs of desvenlafaxine, escitalopram and duloxetine in medically ill inpatients with associated major depressive disorder (MDD) from the Brazilian public general hospital perspective. METHODS: A decision-tree model was built to assess costs of desvenlafaxine 50 mg/d, escitalopram 10 mg/d and duloxetine 60 mg/d to treat patients with MDD in Brazilian public hospitals. Four health states were considered: remission, dose escalation due to lack of response, switching to another antidepressant after discontinuation caused by adverse event, and treatment failure. Assessment assumed similar efficacy among alternatives due to lack of head-to-head clinical trials. Cost estimation assumed an 8-week time horizon. Relative risk of dropouts compared to placebo was obtained through literature systematic review. Since the model assumed that hospitalization are not due to MDD, only drug acquisition costs were considered in the base case. Unit costs were: 3.09BRL, 4.04BRL and 6.91BRL for desvenlafaxine, escitalopram and duloxetine, respectively. Official prices were obtained from 2010 CMED/ANVISA list. One-way sensitivity analyses were performed using drug prices and proportion of duloxetine patients requiring drug titration as key variables. RESULTS: Desvenlafaxine avoided 12% of estimated dropouts compared to competing alternatives. The estimated treatment costs were 144BRL, 250BRL and 364BRL per patient treated with desvenlafaxine, escitalopram and duloxetine, respectively. Therefore, desvenlafaxine exhibited savings of -106BRL and -220BRL versus escitalopram and duloxetine, respectively. Within the sensitivity analyses, the lower and higher savings were observed when Brazilian public procurement prices were adopted (-82BRL versus escitalopram and -216BRL versus duloxetine). CONCLUSIONS: Desvenlafaxine would be a cost-saving strategy compared to escitalopram and duloxetine in medically ill inpatients treated for MDD, due its lower price and lower dropout rates as a consequence of side effects.