A blood component production–inventory problem with preparation method combination and ABO compatibility

We consider the component dependence resulting from each preparation method consistently outputting a particular component combination in the integrated blood component production–inventory problem in a blood center. We first formulate a Markov decision process that comprehensively considers the alternative preparation methods, ABO compatibility, varying age-based inventories, stochastic supply of whole blood, and stochastic demand for components. Then, an approximate dynamic programming algorithm with the interval–adaptive and myopic-learning acceleration approaches is proposed to solve the problem. It performs well in the improvements of both the precision and learning speed. The numerical study displays the sophisticated optimal inventory levels and issuance policies of the blood components. We show that an integrated, operational production-inventory modeling is more capable of dealing with the interaction among the dependent outputs and their differentiated compatibilities and perishabilities. Moreover, the different substitution strategies (i.e., push and pull) are verified to provide similar overall supply levels while causing slight differences in other aspects. Further suggestions for specific components are also provided.