Purpose Current concepts in the treatment of prosthetic joint infections include prosthetic retention and exchange strategies according to published recommendations. A useful algorithm should fit for each type of prosthetic joint infection, even the most complicated situations. We present the outcome of 147 patients with prosthetic joint infections of the hip or the knee joint in an unselected population in clinical routine. Methods Between November 2006 and November 2009, 147 consecutive patients with prosthetic joint infections of the hip or knee were treated according to an algorithm based on the concept published by Zimmerli et al. in 2004. Causative organism, duration of infection, patient comorbidities, surgical treatment, antibiotic treatment, and outcome of treatment were analysed retrospectively. According to the criteria duration of infection, stability of prosthesis, local and systemic risk factors, and susceptibility of the causative pathogen, patients were treated either with debridement and retention or a longinterval two-stage procedure. Results A pathogen could be detected in 82.8 % of the patients, gram-positive cocci being most common. Twentyseven patients were treated with debridement and retention and 120 were treated with a two-stage procedure. In 68 cases difficult-to-treat pathogens could be detected, a polymicrobial infection was found in 51 patients. Definitely free of infection were 71.6 % after a two-stage procedure, and 70.4 % after debridement and retention. Conclusions Our data indicates that the applied algorithm is suitable to be applied as a day-to-day routine, and we confirmed that published results from the literature can be reproduced in an inhomogeneous patient cohort.