Validation of Predialysis Erythropoietin Administration on the End-Stage Renal Disease Medical Evidence Report

Fischer MJ, Stroupe KT, Hynes DM, Browning M, Huo Z, O'Hare A, Kaufman J. Validation of Predialysis Erythropoietin Administration on the End-Stage Renal Disease Medical Evidence Report. Poster session presented at: American Society of Nephrology Renal Week Annual Meeting; 2007 Nov 2; San Francisco, CA.

Related HSR&D Project(s)

• CDA 06-018 – Role of Kidney Disease in an Elderly High-risk Population
• IIR 02-244 – Quality and Cost of VA and Medicare Covered Care for Veterans with ESRD
• RCS 98-352 – Research Career Scientist Award

The End-Stage Renal Disease Medical Evidence Report (Form 2728) is often used to identify erythropoietin stimulating agent (ESA) use and to assess anemia management during the predialysis period. USRDS Annual Reports have cited these data to bolster the claim that low levels of predialysis ESA use ( 30%) are contributing to ongoing suboptimal anemia among incident dialysis patients (50% of patients with hemoglobin values < 10 g/dl). We sought to validate predialysis ESA use data on Form 2728 against VA and Medicare pharmacy datasets. We conducted a retrospective analysis of veterans who initiated hemodialysis between 2000 and 2001 and were eligible for both VA and Medicare coverage (age 66 years) in the 12 months preceding dialysis initiation (predialysis period). Veterans who were not actively receiving outpatient care under VA or Medicare were excluded. ESA prescription records were confirmed in VA (Pharmacy Benefits Management) and ESA claims in Medicare (Carrier/Inpatient/Outpatient/DME files) datasets during the predialysis period and compared to the questionnaire reply on Form 2728 (yes/no question). Among 8715 veterans in the final analytic cohort, ESA use was noted in 2720 (31%) by Form 2728 and in 4082 (47%) by VA/Medicare datasets (p < .05). Using VA/Medicare datasets as the gold standard, the accuracy of Form 2728 for ESA use was characterized as follows: sensitivity 48%, specificity 84%, positive predicative value 72%, negative predictive value 65%, and a kappa statistic of 0.32. These results were unchanged when the analysis was restricted to VA only users, who may generally receive their medications from the VA. Form 2728 appears to significantly underreport ESA use during the predialysis period. Investigators should take into account this shortcoming and its introduction of possible bias for future research efforts and national reports.