To examine the effects of recombinant human erythropoietin (rHuEPO) on hospital utilization, hospital costs, and Medicare reimbursements for hospital care, a longitudinal, matched cohort study was conducted using Medicare claims data of 23,806 Medicare-eligible, dialysis patients who received rHuEPO, did not have a transplant, and were alive for 18 mo or longer and 22,720 controls matched on age, sex, race, cause of ESRD, and dialysis modality. The relative odds (rHuEPO versus control) of admission for all causes and for specific causes over 9 mo, adjusted for admission in the prior 9 mo and the per patient change in total admissions, inpatient days, hospital costs, and Medicare hospital payments between the prior 9-mo period and the subsequent 9-mo period was examined. The adjusted relative odds (95% confidence interval) of admission (rHuEPO versus control) was: higher and statistically significant for all causes, 1.08 (1.03 to 1.14); seizure, 1.52 (1.28 to 1.75); vascular access revision, 1.11 (1.06 to 1.17), and heart failure, 1.17 (1.09 to 1.26); higher but not statistically significant for angina, 1.09 (0.99 to 1.20) and stroke, 1.08 (0.86 to 1.31); and lower but not statistically significant for myocardial infarction, 0.91 (0.72 to 1.10); peripheral vascular disease, 0.81 (0.60 to 1.02); anemia, 0.86 (0.56 to 1.17); and depression, 0.89 (0.37 to 1.40). The mean change per 1,000 patients in admissions was less by 38 (P = 0.03) because of fewer readmissions, and in days was 1,309 less (P < 0.001), for patients treated with rHuEPO versus controls. The mean change per patient in hospital costs was $371 less and was statistically significant (P = 0.03) and in Medicare hospital payments was $132 less but was not statistically significant (P = 0.43) for patients treated with rHuEPO versus controls. rHuEPO was associated with an increase in the probability of hospital admission (particularly admissions potentially related to adverse effects) but a decrease in readmissions, overall admissions, hospital days, and cost to hospitals in this cohort of patients surviving for 18 mo. Although not realized short term, Medicare savings from potential rHuEPO-related reductions in hospital care may be long term through future adjustments in diagnosis-related group-based hospital payment.