Background and objectives: Planning platelet collection and inventory must rely not only on adequate forecasts of transfusion demand but also sophisticated mathematical modeling techniques. This research aims to develop a better demand forecasting model of apheresis platelets and a mathematical programming model to determine the best target amounts of apheresis platelet collection.
Materials and methods: Time series data of apheresis platelets collected from donors and platelets supplied to hospitals daily in Taipei Blood Center from January 2014 to December 2015 was used to fit a forecasting model which combines a regression-type model for formulating the deterministic trends and seasonal variation and an autoregressive moving average model (ARMA) for explaining remaining serial correlations. A seasonal autoregressive integrated moving average (SARIMA) model was also used for benchmarking the prediction performance. A linear programming model was then formulated to solve for the optimal daily target collection volumes that maximize the total social benefits.
Results: The time series model achieved good predictive power with a mean absolute percentage error less than 10%. The appropriateness of the proposed target collection volumes was also verified by using a simulation model, and the proportion of the total platelets requested by hospitals that can be filled by collected apheresis platelets can increase significantly by using the new policy.
Conclusion: The methods proposed in this study can be easily implemented to enhance the management efficiency of blood collecting and supplying of a blood center, and to decrease the costs of the blood outdates and shortages.
Keywords: Apheresis platelet; generalized least square (GLS); linear programming; seasonal autoregressive integrated moving average model (SARIMA); target collection volume; time series.
© 2019 International Society of Blood Transfusion.