A study was undertaken to determine the sensitivity of a computerized surveillance system for postoperative wound infection. From 1987 to 1990, 5177 orthopaedic operations were recorded. Four hundred and forty-five randomly selected patients received a questionnaire. Three hundred and eighty-eight (87.2%) replied. Of 355 (79.8%) usable answers, 75 (21.1%) stated to have had signs of infection, 50 (14.1%) had been seen by a doctor, 25 (7.0%) had not found this necessary. We found seven patients with deep wound infection of which three were not recorded. The sensitivity of the surveillance system concerning deep wound infection was found to be 57.1% (18.4-90.1), and for superficial wound infection to be 9.1% (1.9-24.3). We conclude that the surveillance system is not useful for the intended purpose. We propose to record a few selected types of operation, where all should be seen as out-patients at a suitable time following operation, and thereby get a more valid recording of the frequency of postoperative wound infection.