The Medicare payment policy for recombinant human erythropoietin (rHuEPO) treatment for dialysis patients changed in January 1991 from a relatively fixed payment per treatment (allowed charge of $40 per < or = 10,000 units injected) to a more variable payment based on the amount of rHuEPO administered with each treatment (allowed charge of $11 per 1,000 units injected). This change provided an opportunity to examine how payment policy can effect the use, cost, and health outcome of a biotechnology product used in the dialysis population. In cross-sectional (n = 71,880 Medicare-entitled dialysis patients) and longitudinal (n = 29,088 Medicare-entitled dialysis patients) study designs, we used Medicare end-stage renal disease program and claims data in bivariate and multivariate analyses to examine the effect of the change in payment policy for rHuEPO on access to the biotechnology, dosing, costs, and hematocrit, including the prescribing patterns at for-profit versus not-for-profit providers. The observation period included several months before (July 1989 to December 1990) and 6 months after (January to June 1991) the change in Medicare payment policy. The mean dose per treatment during the initial and fourth month of therapy was low (2,742 [95% confidence interval, 2,703 to 2,781] units and 2,632 [95% confidence interval, 2,598 to 2,667] units, respectively, in June 1990) and increased 3.4% and 5.0%, respectively, in the next 6 months prior to the change in Medicare payment policy compared with 14.6% and 14.8%, respectively, in the 6 months following the change in payment policy. The average monthly allowed charge for rHuEPO per dialysis patient receiving rHuEPO decreased from $455 before the policy change to $349 immediately following the policy change, because the allowed charge per unit of rHuEPO was lower when payment became more dependent on the amount of rHuEPO administered with each treatment than when the payment was fixed at $40 per treatment. The average monthly allowed charge for rHuEPO increased to $375 in the sixth month following the change in payment policy as a result of the increase in dose and the new variable payment. The unadjusted and adjusted changes in mean hematocrit 6 months after the payment change were positive but clinically very small (0.3 and 0.2 percentage points, respectively).(ABSTRACT TRUNCATED AT 400 WORDS)