Optimization Of Blood Inventory Management In Central Blood Banks Using Artificial Intelligence

Abstract

Introduction: Managing inventory in central blood banks is still an important issue owing to its perishability, unpredictable demand, and serious implications of shortages. Traditional inventory control approaches often lead to either excess inventory, which causes wastage, or insufficient inventory, which poses a threat to patient health. However, there are innovative ways to solve this problem using artificial intelligence (AI) and machine learning (ML). Objectives: The present research explores how various AI models such as LSTM, Random Forest (RF), XGBoost, and Reinforcement Learning (RL) can be used to optimize demand prediction, minimize wastage by classifying expired blood products, and streamline RBCs, Platelets, and Fresh Frozen Plasma (FFP) distribution in regional blood banks. Methods: A dataset collected during 48 months (from 2020 to 2023) in three central blood banks was used. Different AI models were developed and tested to predict demand, identify wastage, and optimize delivery routes. Evaluation metrics included MAPE, F1-score, and wastage reduction. Results: The LSTM approach yielded the best performance in terms of MAPE, with an average MAPE value of only 6.3% for RBCs. The proposed hybrid model consisting of XGBoost and RL algorithms resulted in a decrease of wastage of all blood components by 34.7%, and their on-time delivery by 28.5%. Wastage of platelets, the most problematic among the three due to its relatively short shelf-life (5-7 days), decreased from 21.4% to 13.8%. Conclusion: The use of artificial intelligence in blood bank inventory optimization holds immense promise.