BACKGROUND The Diagnostic workup of patients with fever of unknown origin is a challenge, due to the great number of possible etiologies. After we studied the etiologic spectrum of fever of unknown origin in Romania, we tried to evaluate the diagnostic procedures used and their efficiency. METHODS A multicenter cohort study of two years, with another two years of follow-up was carried out on 164 consecutive patients who met the classic, modified criteria of fever of unknown origin. We used a standardised diagnostic protocol. METHODS The role of every diagnostic procedure in establishing the final diagnosis. RESULTS The diagnosis was made by microbiology and serology in 41 cases (25%), by histopathology in 22 cases (18%), with the help of imaging techniques in 30 cases (1.3%), based on the clinical evolution and response to treatment in 54 cases (33%) and by other methods in 12 cases (7.3%). The abdomino-pelvic ultrasonography had a sensitivity of 60%, a specificity of 70%, a positive likelihood ratio of 2.02 and a negative likelihood ratio of 0.57, while the scanner had a sensitivity of 81%, a specificity of 64%, a positive likelihood ratio of 2.23 and a negative likelihood ratio of 0.29. CONCLUSIONS Of all the diagnostic procedures used, none had a good sensitivity/specificity. The clinical evolution and the treatment response had an important role in the diagnostic workup.