Forecasting demands of blood components based on prediction models

Area of Science:

• Biomedical Engineering
• Health Informatics
• Data Science

Background:

• Maintaining an adequate blood supply is critical for healthcare.
• Uncertainty in blood demand and supply necessitates effective management strategies.
• Accurate forecasting of blood resource demand is essential for optimizing blood supply chains.

Purpose of the Study:

• To develop and evaluate predictive models for blood component demand forecasting.
• To improve blood stock control and reduce wastage and shortages.
• To integrate forecasting models into a blood station's operational system.

Main Methods:

• Implemented a hybrid prediction system combining SARIMAX and LSTM models.
• Integrated real-time blood component collection and supply data for automatic model updates.
• Forecasted daily demands for apheresis platelets, washed RBCs, suspended RBCs, and plasma from January to June 2023.

Main Results:

• Models demonstrated strong forecasting performance across all blood components.
• Apheresis platelet model achieved R2 up to 87.6% and MAPE as low as 0.0037.
• Washed RBCs, suspended RBCs, and plasma models showed good fit with MAPE values of 0.0121, 0.0372, and 0.0394 respectively, with R2 over 66% for suspended RBCs and over 90% for plasma.

Conclusions:

• Forecasting models accurately predict future blood component demands based on historical data.
• These models offer a valuable tool for blood stock management, mitigating challenges.
• Effective forecasting leads to reduced blood waste and fewer instances of blood shortages.