To delineate the pathogenesis and epidemiology of catheter-related infection with Swan-Ganz pulmonary artery (PA) catheters, a prospective clinical study of hospitalized adult medical and surgical patients was done. Role of catheter material was assessed by randomizing insertions to heparin-bonded PA catheters made of polyvinylchloride or polyurethane. Sources of infection and pathogenesis were studied by culturing skin, the introducer, the PA catheter tip, all hubs, infusate from each lumen, and the extravascular portion of the PA catheter beneath the external protective plastic sleeve. Concordance between isolates from sources and infected catheters was determined by speciation, antibiogram, and for coagulase-negative staphylococci, plasmid profile analysis. Risk factors for infection were determined by stepwise logistic regression. Overall, 65 (22%) of 297 Swan-Ganz catheters showed local infection of the introducer (58 catheters) or the intravascular portion of the PA catheter (20 catheters); only two catheters (0.7%) caused bacteremia. Eighty percent of infected Swan-Ganz catheters (the introducer or PA catheter) showed concordance with organisms cultured from skin of the insertion site, 17% with a contaminated hub and 18% with organisms contaminating the extravascular portion of the PA catheter beneath the sleeve. Isolates from infected PA catheters were most likely to show concordance with concomitantly infected introducers (71%). Cutaneous colonization of the insertion site with greater than 10(2) cfu/10 cm2 (relative risk [RR] 5.5; p less than 0.001), insertion into an internal jugular vein (RR 4.3; p less than 0.01), catheterization greater than 3 days (RR 3.1; p less than 0.01), and insertion in the operating room using less stringent barrier precautions (RR 2.1; p = 0.03) were each associated with a significantly increased risk of catheter-related infection. The risk of bacteremic infection with Swan-Ganz catheters is now low, in the range of 1%, with reasonable care. Swan-Ganz catheters are vulnerable to contamination from multiple sources, but the patient's skin is the single most important source of organisms causing invasive infection, which in most cases involves the introducer rather than the PA catheter. Heavy colonization of the insertion site, percutaneous insertion in the internal jugular vein rather than subclavian vein, catheterization longer than 3 days, and insertion with less stringent barrier precautions significantly increase the risk of catheter-related infection. These findings hold promise for application to management of Swan-Ganz catheters and research in catheter design to reduce the risk of catheter-related infection.