The identification of Propionibacterium acnes in cultures of bone and joint samples is always difficult to interpret because of the ubiquity of this microorganism. The aim of this study was to propose a diagnostic strategy to distinguish infections from contaminations. This was a retrospective analysis of all patient charts of those patients with >or=1 deep samples culture-positive for P. acnes. Every criterion was tested for sensitivity, specificity, and positive likelihood ratio, and then the diagnostic probability of combinations of criteria was calculated. Among 65 patients, 52 (80%) were considered truly infected with P. acnes, a diagnosis based on a multidisciplinary process. The most valuable diagnostic criteria were: >or=2 positive deep samples, peri-operative findings (necrosis, hardware loosening, etc.), and >or=2 surgical procedures. However, no single criterion was sufficient to ascertain the diagnosis. The following combinations of criteria had a diagnostic probability of >90%: >or=2 positive cultures + 1 criterion among: peri-operative findings, local signs of infection, >or=2 previous operations, orthopaedic devices; 1 positive culture + 3 criteria among: peri-operative findings, local signs of infection, >or=2 previous surgical operations, orthopaedic devices, inflammatory syndrome. The diagnosis of P. acnes osteomyelitis was greatly improved by combining different criteria, allowing differentiation between infection and contamination.