Mortality of prevalent chronic hemodialysis patients remains high. The potential effect of the dialysis membrane on this mortality has not been previously investigated in a large population of chronic hemodialysis patients. Using data from the United States Renal Data System (USRDS), we analyzed a random sample of 6,536 patients receiving hemodialysis on December 31, 1990. The study design was a historical prospective study. By limiting the study to patients dialyzed for at least one year with bicarbonate dialysate, in whom the dose of dialysis could be calculated, and in whom dialysis membrane and co-existing morbidities were defined, the sample size was reduced to 2,410 patients. A Cox proportional hazards model was used to estimate relative mortality risk. The types of dialysis membranes used were broadly classified into three categories: unsubstituted cellulose, modified cellulose (generally cellulose membranes that have been modified by substitutions of some or most of their hydroxyl moieties) and synthetic membranes that are not cellulose-based. The results of the study suggest that after adjusting for the dose of dialysis and the presence of co-morbid factors, the relative risk of mortality of patients dialyzed with modified cellulose or synthetic membranes was at least 25% less than that of patients treated with unsubstituted cellulose membranes (P < 0.001). To account for the possibility that these differences were due to regional practice patterns, we further stratified the data for nine different regions. There was still a 20% difference in relative risk of mortality between membrane groups with the mortality statistically significantly less in patients treated with synthetic membranes (P < 0.045) compared to patients dialyzed with unsubstituted cellulose membranes. The results of this study suggest that the dialysis membrane plays an important role in the outcome of chronic hemodialysis patients. However, more definitive studies are needed before a cause and effect relationship can be proven.