Abstract: SA-PO254
Using a Predictive Algorithm to Provide Decision Support in Anemia Management
Session Information
- Anemia and Iron Metabolism: Clinical
November 09, 2019 | Location: Exhibit Hall, Walter E. Washington Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: Anemia and Iron Metabolism
- 202 Anemia and Iron Metabolism: Clinical
Authors
- Dunn, Lori, University of Virginia Health System, Charlottesville, Virginia, United States
- Bowman, Brendan T., University of Virginia School of Medicine, Charlottesville, Virginia, United States
- Lobo, Benjamin, University of Virginia, Charlottesville, Virginia, United States
Background
Anemia in hemodialysis is very common with most receiving erythropoietin stimulating agent (ESA) therapy. Most ESA dosing protocols are based on manufacturer recommendations and clinical experience. They are standardized across entire dialysis populations and do not account for individual response. To improve anemia management, we developed a predictive algorithm to forecast 1, 2 and 3 month hemoglobin (HGb) values as a way of providing decision support for monthly ESA dose adjustments.
Objectives: To determine if the addition of future HGb predictions to the information currently provided to anemia managers will result in a reduction in HGb variability, an increase in the number of HGb observations within the target range, and a decrease in the average per treatment ESA use.
Methods
We developed a predictive algorithm utilizing historic electronic medical record (EMR) data collected during the 2009-2017 time period. The data set included dialysis sessions, ESA administered, intravenous iron, HGb, ferritin, and transferrin saturation (Tsat %). Following algorithm development, we conducted a 10 month (July 2018 to April 2019) QI project with approximately 20 patients in an academic dialysis center. Each month, future hemoglobin predictions were provided for 1, 2, and 3 months based on the current monthly lab draw and patient’s own historical data. These predictions were then provided to anemia managers along with interpretive guidance from the development team.
Results
Incorporation of this form of decision support into the anemia management process saw the number of HGb readings below 10 g/dL decrease from 23.7% pre intervention to 19.5% post intervention (p=0.44). There was an increase from 10.8 g/dL to 11.2 g/dL in the average patient mean HGb (p=0.09) and a 77.3% decrease in the average patient HGb variance (p<0.01). Overall, there was a 32.7% decrease in the average per treatment ESA dose (p<0.01).
Conclusion
Incorporating decision support from a predictive algorithm into an existing anemia management process resulted in a decrease in overall ESA use while decreasing patient HGb variability. Small study population was a limitation to this study.