ASN's Mission

To create a world without kidney diseases, the ASN Alliance for Kidney Health elevates care by educating and informing, driving breakthroughs and innovation, and advocating for policies that create transformative changes in kidney medicine throughout the world.

learn more

Contact ASN

1401 H St, NW, Ste 900, Washington, DC 20005

email@asn-online.org

202-640-4660

The Latest on X

Kidney Week

Abstract: TH-PO043

Machine Learning Approach to Predict Hemoglobin Levels for Erythropoietin Dosing in Hemodialysis Patients

Session Information

Category: Augmented Intelligence, Digital Health, and Data Science

  • 300 Augmented Intelligence, Digital Health, and Data Science

Authors

  • Zhao, Siwei, Sanderling Renal Services, Nashville, Tennessee, United States
  • Yang, Jason W., University of California San Diego, La Jolla, California, United States
  • Zhang, Jingyao, University of California San Diego, La Jolla, California, United States
  • Lee, Pei Lun, University of California San Diego, La Jolla, California, United States
  • Dasgupta, Subhasis, University of California San Diego, La Jolla, California, United States
  • Tannenbaum, Jerome S., Sanderling Renal Services, Nashville, Tennessee, United States
  • Ix, Joachim H., University of California San Diego, La Jolla, California, United States
  • Malhotra, Rakesh, University of California San Diego, La Jolla, California, United States
Background

Erythropoiesis-stimulating agents (ESA) are commonly used to treat anemia in HD patients. However, prediction of patient-specific hemoglobin (Hgb) response to ESA remains challenging. Here we used machine learning techniques to predict Hgb response to ESAs in HD.

Methods

We included patients undergoing HD in Sanderling clinics who were receiving intravenous (IV) ESA (Mircera ®) and IV iron and had ≥5 Hgb measurements. ESA dose, iron dose, demographics and clinical variables were collected. Data was preprocessed to create a patient-specific bi-weekly dataset. We used a range of machine learning models, including long short-term memory networks, random forest, XGBoost and support vector machine. Each model was implemented individually for each patient with K-fold cross-validation. Performance of models to predict future Hgb levels was defined using mean absolute percentage error (MAPE) (average absolute percentage error between the predicted future Hgb and the observed future Hgb).

Results

Among 427 HD patients, the mean age was 65 ± 14 yrs and 44% were female. Patients received, on average, 93 ± 77 mg per week IV iron, and mean Mircera dose of 103 ± 183 mcg per month. The mean Hgb was 10.1 ± 1.4 mg/dl, and transferrin saturation 32 ± 15 %. The average MAPE for the model over the study cohort was 5.9% vs. Hgb variation of 13.9%. Figure 1 provides distribution of the MAPE for patients included in the analysis.

Conclusion

Our results showed the promising performance of machine learning models in predicting future Hgb levels within 6% of observed levels in HD patients. Future studies should focus on refining these models with the goal of personalized ESA dosing to maximize on-target Hgb levels.

Histogram of MAPE for Hgb prediction.