Researchers from Emory University have developed a robust AI-powered meta-model designed to predict the likelihood of blood transfusion in non-traumatic ICU patients. Published in Health Data Science, the study addresses key challenges in predicting transfusion needs by utilizing machine learning algorithms trained on clinical data from over 72,000 ICU patient records collected over five years. The model demonstrated outstanding performance, achieving an AUROC of 0.97, an accuracy of 0.93, and an F1 score of 0.89​.

Unlike traditional decision-support systems that focus on specific patient subgroups, this AI model leverages a wide range of clinical biomarkers, including hemoglobin and platelet levels, to provide precise 24-hour transfusion predictions. The research team emphasizes the model's potential to optimize transfusion decisions, reduce complications, and improve resource management in ICU settings. Future plans include integrating the AI model into real-world clinical workflows to further validate its performance and enhance its impact on patient care.

Evolution is complex and difficult to study, but a new software package developed by the Arkansas Agricultural Experiment Station offers researchers a better way to simulate how organisms change over time. The new software, called TraitTrainR, builds on work in the field of comparative biology to provide an efficient and effective framework for replicating the evolutionary process many times over. An ultimate goal is to use this software to better understand the diversity of life forms on our planet.