Center effects in anemia management of dialysis patients.

This study set out to determine whether there is a center effect on anemia management in hemodialysis patients. The US Renal Data System and Medicare standard analysis files were analyzed. Between-center variation and within-facility correlations in hematocrit values were examined in two separate data sets (years 2000 and 2001) and compared with simulated samples that were composed of random values that assumed no center effect. Mixed-effect models were used to adjust for multiple factors and quantify the within-facility correlation in hematocrit values. Expected hematocrit values were compared in patients who underwent dialysis at poor and superior performing facilities with fixed characteristics including epoetin alpha dosing. There was a wider center variation in hematocrit for the actual versus simulated data and a coefficient of variation of 4.1% for the former versus 1.7% for the latter, in both years. The within-facility correlation for hematocrit was 0.053 (95% confidence interval 0.050 to 0.056; P < 0.001) in 2000 with similar values in 2001 but no within-facility correlation in the simulated data. The impact of these findings was demonstrated with a difference in expected hematocrit for a patient who was treated with fixed-dosage epoetin alpha in the poorest versus best performing units (mean difference in expected hematocrit 3.06; 95% confidence interval 3.03 to 3.09; P < 0.001). Key attributes of a center effect on anemia management in hemodialysis have been identified. The presence of a center effect suggests that there are facility-specific processes that influence performance in dialysis anemia management and are independent of commonly titrated inputs, such as dosing of erythropoietic agents.