Logistic regression analysis was applied to a sample of more than 12,000 hemodialysis patients to evaluate the association of various patient descriptors, treatment time (hours/treatment), and various laboratory tests with the probability of death. Advancing age, white race, and diabetes were all associated with a significantly increased risk of death. Short dialysis times were also associated with high death risk before adjustment for the value of laboratory tests. Of the laboratory variables, low serum albumin less than 40 g/L (less than 4.0 g/dL) was most highly associated with death probability. About two thirds of patients had low albumin. These findings suggest that inadequate nutrition may be an important contributing factor to the mortality suffered by hemodialysis patients. The relative risk profiles for other laboratory tests are presented. Among these, low serum creatinine, not high, was associated with high death risk. Both serum albumin concentration and creatinine were directly correlated with treatment time so that high values for both substances were associated with long treatment times. The data suggest that physicians may select patients with high creatinine for more intense dialysis exposure and patients with low creatinine for less intense treatment. In a separate analysis, observed death rates were compared with rates expected on the basis of case mix for these 237 facilities. The data suggest substantial volatility of observed/expected ratios when facility size is small. Nonetheless, a minority of facilities (less than or equal to 2%) may have higher rates than expected when compared with the pool of all patients in this sample. The effect of various laboratory variables on mortality is substantial, while relatively few facilities have observed death rates that exceed their expected values. Therefore, we suggest that strategies designed to improve the overall mortality statistic for dialysis patients in the United States would be better directed toward improving the quality of care for all patients, particularly high-risk patients, within their usual treatment settings rather than trying to identify facilities with high death rate for possible regulatory intervention.