ISTANBUL—Investigators have developed an algorithm using routinely available patient data that accurately predicts subsequent transfusion risk in dialysis patients, according to a report at the 50th Congress of the European Renal Association-European Dialysis and Transplant Association.
Preliminary results show that the algorithm can accurately identify the need for transfusion in 73% of patients.
The algorithm combines demographic, biochemical, clinical, and transfusion data from the DaVita Clinical Data Warehouse. Thirty-two variables were identified and included in the final model.
“We have never had a quantifiable method of predicting which dialysis patients were likely to require a transfusion,” said Steven Brunelli, MD, Senior Director of DaVita Clinical Research in Minneapolis, Minnesota, said. “In general, we felt that patients with low hemoglobin were at increased risk. But there’s a bell curve of hemoglobin level, and the question is where do you draw the line on the bell curve? We want to be able to rigorously quantify risk and say, for example, that for a particular patient, the risk of transfusion is ‘X’ or the risk of transfusion is ‘Y.’”
The team tested the algorithm in 103,350 patients, 1,756 of whom had a transfusion during the first three months of 2011. Transfusion data were drawn from hospital medical records by independent data services.
Patients with chronic kidney disease who undergo dialysis are at risk of persistent anemia, Dr. Brunelli explained. Transfusion is an option of last resort of anemia management and may decrease survival, patient well-being, and transplant candidacy.
The algorithm uses data commonly available in electronic health records, he said, but he could not cite specific variables that are included in the model “because of its proprietary and confidential nature.”
Following recent changes in the regulatory environment (changes in the FDA Black Box warning) and reimbursement (the introduction in January 2011 of Medicare’s prospective payment system for dialysis care), anemia management practices changed substantively throughout the U.S, he noted. “Specifically, the amount of erythropoiesis-stimulating agents that patients were getting went down, the amount of iron that patients were getting went up, and hemoglobin levels decreased,” Dr. Brunelli said. “And the big fear was that transfusion rates were going to go up, which could lead to bad outcomes for patients, for example, loss of transplant candidacy.”
In response, DaVita initiated a registry that tracks transfusion rates in its in-center hemodialysis patients. Because of the large volume of data compiled by the registry, Dr. Brunelli and colleagues believed they had an invaluable opportunity to develop a predictive model that could accurately and reproducibly risk-stratify patients on the basis of future transfusion risk.
Finally, Dr. Brunelli pointed out that the algorithm has broad potential applications. It could enable real-time clinical assessment of patient risk, identify opportunities for prevention efforts, and identify high-risk patients in whom prophylactic interventions can be tested.
“This is not an answer unto itself,” he said, “but we hope that it is a small step in a positive direction for patient care.”