An AI-Driven Predictive Modelling Framework to Analyze and Visualize Blood Product Transactional Data for Reducing Blood Products’ Discards

Abstract

Maintaining an equilibrium between shortage and wastage in blood inventories is challenging due to the perishable nature of blood products. Research in blood product inventory management has predominantly focused on reducing wastage due to outdates (i.e. expiry of the blood product), whereas wastage due to discards, related to the lifecycle of a blood product, is not well investigated. In this study, we investigate machine learning methods to analyze blood product transition sequences in a large real-life transactional dataset of Red Blood Cells (RBC) to predict potential blood product discard. Our prediction models are able to predict with 79% accuracy potential discards based on the blood product’s current transaction data. We applied advanced data visualizations methods to develop an interactive blood inventory dashboard to help laboratory managers to probe blood units’ lifecycles to identify discard causes.