Wound infections are an infrequent but serious complication of total knee arthroplasty. Between January 1984 and November 1987, 20 of 243 (8.2%) patients at two affiliated hospitals developed surgical wound infections following 259 total knee arthroplasty procedures performed in clean-air operating rooms. Eighteen (90%) of the patients had deep infections; nine required removal of the prosthesis. A single surgeon (surgeon X) was associated with 18 of the procedures that had subsequent infection (risk ratio (RR) = 9.4, 95% confidence interval (CI) 2.2-39), and an investigation was carried out in an effort to explain the difference in infection rates between surgeon X and other surgeons. In a cohort study, stratified analyses identified a preoperative American Society of Anesthesiologists (ASA) physical status class greater than or equal to 3, surgeon X, and early postoperative use of a continuous passive motion device as risk factors associated with surgical wound infection following total knee arthroplasty procedures. Logistic regression analyses identified being a patient operated on by surgeon X with an ASA class greater than or equal to 3 as the only significant independent risk factor for total knee arthroplasty-associated surgical wound infections (RR = 9.3, 95% CI 2.8-31). The effect due to surgeon X could not be explained by receipt or timeliness of administration of antimicrobial prophylaxis, type of prosthesis inserted, duration of operation, postoperative use of continuous passive motion, or underlying etiology of joint disease. The authors conclude that surgical technique and patient's severity of illness were the primary determinants of surgical wound infection after total knee arthroplasty. This study demonstrates the complexity of epidemiologic investigation of surgical wound infections and the importance of considering patient severity of illness when interpreting surgeon-specific infection rates.