Kidney Week

Abstract: SA-PO820

Development and Validation of a Transfusion Risk Score

Session Information

• Dialysis: Anemia and Iron Metabolism
  November 04, 2017 | Location: Hall H, Morial Convention Center
  Abstract Time: 10:00 AM - 10:00 AM

Category: Dialysis

• 605 Dialysis: Anemia and Iron Metabolism

Authors

• Gilbertson, David T., Chronic Disease Research Group, Minneapolis, Minnesota, United States

David T. Gilbertson,

• Yan, Heng, Chronic Disease Research Group, Minneapolis, Minnesota, United States

Heng Yan,

• Xu, Hairong, Astrazeneca, Westlake Village, California, United States

Hairong Xu,

• Sinsakul, Marvin V., AstraZeneca, Bethesda, Maryland, United States

Marvin V. Sinsakul,

• Peng, Yi, Chronic Disease Research Group, Minneapolis, Minnesota, United States

Yi Peng,

• Wetmore, James B., Hennepin County Medical Center, Minneapolis, Minnesota, United States

James B. Wetmore,

• Liu, Jiannong, Minneapolis Medical Research Foundation, Minneapolis, Minnesota, United States

Jiannong Liu,

• Li, Suying, Chronic Disease Research Group, Minneapolis, Minnesota, United States

Suying Li,

Background

Following changes to CMS payment for dialysis services in Jan 2011 and the ESA label revision 6 months later, a decline in hemoglobin (Hb) levels and an increase in transfusions were observed in dialysis patients. Transfusions have decreased from their 2012 peak, and transfusion avoidance is the preferable option in dialysis patients. We sought to develop a predictive model for transfusions using comorbidity, markers of inflammation, previous transfusion, vitamin D use, IV iron use, ESA dose, Hb, ferritin and TSAT.

Methods

USRDS/Crownweb data from 2012-13 were used for model development. Point prevalent hemodialysis (HD) patients on 11/1/12 with ≥ 6 months Medicare A/B coverage and mean Hb < 10 g/dL were included. Aug-Oct were used to assess anemia-related variables (Hb, ESA, TSAT, ferritin, IV iron use, vitamin D, and transfusion), and May-Oct were used to assess comorbidity from claims. Logistic regression with Lasso for variable selection was used to predict transfusion during the next 3 months. For model validation, similar cohort construction was used, with point prevalent HD patients on 8/1/13.

Results

Variables retained in the final model included Hb, ESA dose, ferritin, TSAT, IV iron, vitamin D, prior transfusion, and interactions of these variables. In the validation dataset, a calibration plot showed good agreement between observed/predicted transfusions (Figure): c-statistic = 0.74.

Conclusion

The addition of ferritin and TSAT, along with inflammatory comorbidities, aided in prediction of transfusions in patients with Hb levels < 10 g/dL. Optimal anemia management strategies involve balancing CV risk on the high end of Hb and ESA exposure with transfusion risk on the low end. The ability to identify patients at risk for transfusion may lead to improved anemia management and outcomes.

Funding

• Commercial Support – AstraZeneca