Abstract Background Clinical assessment of depression is an important part of pre-surgical assessment among individuals with morbid obesity. However, there is no agreed-upon instrument to identify mood psychopathology in this population. We examined the reliability and criterion validity of the clinician-administered Montgomery-Åsberg Depression Rating Scale (MADRS) and the utility of a short version for bariatric surgery candidates. Methods The sample was 374 patients with obesity, consecutively recruited from the waiting list of a bariatric surgery clinic of University Hospital, Brazil: women 80 %, mean BMI 47 kg/m2, mean age 43.0 years. The 10-item MADRS was analyzed against the SCID-I. Items that showed small relevance to sample’s characteristics and contribution to data variability were removed to develop the short 5-item version of scale. We calculated the sensitivity and specificity of cutoff points of both versions MADRS, and values were plotted as a receiver operating characteristic curve. Results For the 10-item MADRS, the Cronbach’s alpha coefficient was 0.93. When compared against SCID-I, the best cut-off threshold was 13/14, yielding sensitivity of 0.81 and specificity 0.85. Following items were removed: reduced appetite, reduced sleep, concentration difficulties, suicide thought and lassitude. The 5-item version showed an alpha coefficient of 0.94 and a best cut-off threshold of 10/11, yielding sensitivity of 0.81 and specificity 0.87. Similar overall ability to discriminate depression of almost 90 % was found for both 10-item and 5-item MADRS. Conclusion The MADRS is a reliable and valid instrument to assess depressive symptoms among treatment-seeking bariatric patients. Systematic application of the abbreviated version of the MADRS can be recommended for enhancing the clinical detection of depression during perioperative period.