AABB Foundation Alumni Publish Research in The Lancet Hematology

A new study led by a multinational research team — including two former recipients of AABB Foundation early-career scientific research grants — suggests that machine-learning models can accurately predict blood donors’ hemoglobin and ferritin levels at future donation visits, even across global settings.

The research team included corresponding author W. Alton Russell, PhD (2022 grant recipient), and Brian Custer, PhD, MPH (2005 grant recipient). Custer is also a 2019 AABB Foundation Hall of Fame inductee. The findings were published this month in The Lancet Hematology.

Model performance was consistent across datasets. For hemoglobin predictions, the models achieved a root-mean-square percentage error (RMSPE) of 6.78 in both U.S. datasets and did not exceed an 8% increase in external validations. However, ferritin predictions were more accurate when baseline ferritin was available (RMSPE of 14.9) compared to the hemoglobin-only dataset (RMSPE of 27.4).

According to investigators, the results suggest that machine learning may enable personalized donation intervals, potentially reducing iron deficiency and low hemoglobin deferrals while preserving the adequacy of the blood supply.