Vascular access (VA) complications account for 16 to 25% of hospital admissions. This study tested the hypothesis that the type of VA in use is correlated with overall mortality and cause-specific mortality. Data were analyzed from the U.S. Renal Data System Dialysis Morbidity and Mortality Study Wave 1, a random sample of 5507 patients, prevalent on hemodialysis as of December 31, 1993. The relative mortality risk during a two-year observation was analyzed by Cox-regression methods with adjustments for demographic and comorbid conditions. Using similar methods, cause-specific analyses also were performed for death caused by infection and cardiac causes. In diabetic mellitus (DM) patients with end-stage renal disease, the associated relative mortality risk was higher for those with arteriovenous graft (AVG; RR = 1.41, P < 0.003) and central venous catheter (CVC; RR = 1.54, P < 0.002) as compared with arteriovenous fistula (AVF). In non-DM patients, those with CVC had a higher associated mortality (RR = 1.70, P < 0.001), as did to a lesser degree those with AVG (RR = 1.08, P = 0.35) when compared with AVF. Cause-specific analyses found higher infection-related deaths for CVC (RR = 2.30, P < 0.06) and AVG (RR = 2.47, P < 0.02) compared with AVF in DM; in non-DM, risk was higher also for CVC (RR = 1.83, P < 0.04) and AVG (RR = 1.27, P < 0.33). In contrast to our hypothesis that AV shunting increases cardiac risk, deaths caused by cardiac causes were higher in CVC than AVF for both DM (RR = 1.47, P < 0.05) and non-DM (RR = 1.34, P < 0.05) patients. This case-mix adjusted analysis suggests that CVC and AVG are correlated with increased mortality risk when compared with AVF, both overall and by major causes of death.