The incidence of early post-operative wound infection was studied prospectively in 1053 patients undergoing elective orthopaedic procedures over a 3-month period. The study was repeated in 1131 patients a year later, 6 months after the hospital had moved to new premises. A clinical definition of wound infection identified disturbingly high sepsis rates. A total of 44% of all infections occurred after discharge from hospital. Despite the large sample, there was no significant difference in the rate of early wound infection between the two periods (7.85 and 6.82%). There was no significant difference in infection rates between theatres with and without laminar air flow. In the majority (35 out of 53) of minor infections, bacteriological confirmation was not available because no microbiological specimens were received. Conversely, there were five negative swabs out of 16 wounds defined clinically as major infections. We conclude that, where the rate of bacteriological confirmation of wound infection is low, the use of a clinical definition gives higher audited sepsis rates. Large audit samples are required to demonstrate differences as a result of a changed practice.