The aim of the present study was to analyse the diagnostic yield of the rose Bengal test for the rapid diagnosis of human brucellosis in an emergency department in an area where the disease is endemic. The study included 711 patients diagnosed initially with brucellosis and 270 controls. Brucellosis patients were divided into three groups: group I, individuals with no regular exposure to or history of brucellosis; group II, individuals exposed repeatedly to Brucella infection; and group III, individuals infected with Brucella who had received appropriate treatment during the previous 12 months. Blood cultures were positive for 445 (62.6%) brucellosis patients, while the remaining 266 (37.4%) patients were diagnosed according to clinical and serological criteria. The overall sensitivity of the rose Bengal test was 92.9%. The specificities for groups I, II and III were 94.3%, 91.7% and 76.9%, respectively, with positive likelihood ratios of 16.5, 10.4 and 4.2, respectively. The diagnostic gain after the performance of the rose Bengal test was good or very good in patients with no previous exposure to Brucella or history of brucellosis, but poor in patients who were exposed repeatedly to Brucella or had a history of brucellosis and a low pre-test probability. Use of the rose Bengal test as the sole technique for the diagnosis of brucellosis in endemic areas should be considered very carefully in the context of patients who are exposed repeatedly to Brucella or have a history of brucellosis.