Despite the broad consensus that native arteriovenous fistula is the access of choice for hemodialysis, national-level information about vascular access at dialysis initiation has been unavailable in the United States. For incident hemodialysis patients, June 2005 to October 2007 (n=220,157), vascular access type was determined from the new Centers for Medicare & Medicaid Services Medical Evidence Report (form CMS-2728). Proportions with each type at first dialysis, demographic and clinical associations of each type, and associations between initial access type and survival were assessed. The mean patient age was 63.6 years; 29.4% of patients were African American, and for 44.5%, end-stage renal disease was due to diabetes. Vascular access proportions were: fistula, 13.2% of patients; graft, 4.3%; catheter/maturing fistula, 16.0%; catheter/maturing graft, 3.3%; and catheter alone, 63.2%. Adjusted odds ratios (vs. fistula) of catheter use alone were > or = 1.50 for lack of insurance (1.62 [95% confidence interval 1.62-1.68]), nephrologist care for 0 to 12 months (2.75 [2.69-2.81]), other (2.19 [2.09-2.29]), or unknown (1.53 [1.44-1.63]) cause of renal disease, institutional residence (1.51 [1.45-1.57]), and 7 of 18 end-stage renal disease networks. Over a mean follow-up of 1 year, 26.0% of the study population died. Compared with fistula, adjusted mortality hazards ratios were 1.39 (1.32-1.47) for grafts, 1.49 (1.44-1.55) for catheters/maturing fistulas, 1.74 (1.65-1.84) for catheters/maturing grafts, and 2.18 (2.11-2.26) for catheters alone. While geographic variability is pronounced, vascular access at dialysis inception is typically suboptimal; suboptimal access exhibits a graded association with mortality. Lack of timely access to specialty care appears to limit optimal access.