A machine learning based framework to analyze blood product transactional data for reducing discards

Blood products and their derivatives are perishable items that require detailed inventory management techniques to ensure their persistence and availability at the time of need. Wastage is a key factor affecting blood product availability. Products may waste either due to expiry (i.e., outdates) or discards (e.g., due to unsafe temperature, damaged bags, recalls, etc.). Reducing wastage due to discards is a challenging problem in stochastic inventory management. While the underlying causes for discards have not been fully characterized, models for management of blood products are common in the literature. An understudied, yet important, aspect of wastage that critically impacts the efficiency of blood services is that of discards; products that are within their rated shelf life, but that are wasted without having been transfused to patients.

The objective of this project is to investigate the underlying causes of blood product discards in hospital-based blood transfusion service. We plan to analyze multi-dimensional blood transfusion data collected over the last 5 years at our institution, using machine learning based data analysis methods to identify the causative patterns of blood product discards. We will operationalize the analysist to help blood transfusion staff detect potential discard trends in real time enabling them to proactively respond to reduce blood product wastage. The project's outcomes will inform the formulation of efficient institutional workflows and transfusion policies that reduce blood product discards. The research methods and outcomes are scalable and can be applied at other blood banks in Canada to reduce blood product discards.

CHENG, Calvino

Dalhousie University

Blood Efficiency Accelerator Program

Ontario

29950.0