AI and Blood Banking: Predicting Transfusion Demand - A Systematic Review of Forecasting Approaches

Area of Science:

• Medical Informatics
• Health Services Research
• Data Science in Healthcare

Background:

• Accurate blood product demand prediction is crucial for efficient healthcare logistics, balancing supply with patient needs and minimizing wastage.
• Artificial intelligence (AI) and machine learning (ML) offer advanced computational approaches for improving the accuracy of transfusion demand forecasting.
• Existing AI/ML applications in blood banking span diverse scales, from individual patient needs to regional supply chain management.

Purpose of the Study:

• To systematically review and categorize AI/ML applications in blood banking based on prediction scope and clinical objectives.
• To evaluate the performance and identify optimal AI/ML methodologies for different transfusion demand forecasting tasks.
• To provide a structured overview of the current landscape and future directions for AI in blood supply management.

Main Methods:

• A systematic review adhering to PRISMA 2020 guidelines was conducted, searching PubMed and Scopus databases.
• Studies published between January 2015 and December 2025 applying AI or ML to transfusion demand prediction were included.
• Findings were organized into a four-tier framework: patient-individual, procedure-specific, facility-level, and regional forecasting, with systematic comparison of technical approaches.

Main Results:

• Twenty-four studies met inclusion criteria, analyzed across the four-tier framework.
• Patient-individual prediction achieved AUCs of 0.45-0.97; procedure-specific models showed AUCs of 0.70-0.91.
• Facility-level optimization reduced platelet outdating rates by approximately 50% and achieved low error rates (MAPE as low as 4.18%); regional forecasting achieved R² up to 0.85.
• Deep learning excelled in patient-level temporal patterns, while traditional methods like ARIMA were effective for regional forecasting; prediction horizons varied by blood product.

Conclusions:

• AI-driven demand prediction significantly enhances blood supply chain management across all levels.
• The selection of AI/ML modeling approaches should align with specific prediction scopes, data infrastructure, and clinical goals.
• Standardized external validation is essential for the successful clinical deployment of these AI tools.