Recombinant erythropoietin, first approved for Medicare reimbursement in June 1989, was prescribed at initial doses for dialysis patients of 2,500 to 2,700 U per administration independent of hematocrit level. By 1997, however, patients with hematocrits less than 30% were administered 6,000 U/dose, compared with 4,500 U administered to patients with hematocrits of 33% to 36%. Since 1990, the percentage of patients with hematocrits less than 30% decreased from 60% to 22% in 1997, whereas the percentage of patients with hematocrits of 33% to 36% increased from 10% to 30%. In 1997, Medicare initiated the Hematocrit Measurement Audit (HMA) policy, which was directed at reducing the percentage of claims for hematocrits greater than 36% and increasing the stability of the hematocrit levels. The policy change achieved the initial effect but resulted in a reduction of the mean hematocrit as well. The policy was changed in 1998 in response to patient and provider concerns. Mortality studies show that hematocrits less than 30% (or hemoglobin levels < 110 g/L) are associated with an 18% to 40% increased associated risk for death. Higher hematocrits of 33% to 36% appear to be associated with a 7% reduced risk for death. The risk for hospitalization parallels that of mortality. Patients with sustained hematocrits of 33% to 36% over 1 year appear to have the best outcome compared with patients with hematocrits that decrease. The latter are at greater risk than those patients in whom the hematocrits increase. In conclusion, dramatic improvements in hemodialysis patient hematocrits have occurred since 1989. Mortality and hospitalization studies support the National Kidney Foundation Dialysis Outcomes Quality Initiative (NKF DOQI) target hematocrit range of 33% to 36% as providing the best associated outcomes.