A stochastic programming appoach for optimizing cryoprecipitate collection schedules

Abstract

In this thesis, we investigate the problem of generating weekly collection schedules for cryoprecipitate, a vital blood product as the main source of brinogen. As cryoprecipitate requires special equipment for collection, a two-day notice is needed before a mobile site can be assigned to cryoprecipitate collection. Due to the perishable nature of cryoprecipitate, we consider its eight hours collection-to-completion time constraint, in addition to the daily processing capacity of host sites. We aim to minimize the total collection cost while determining which mobile sites should be assigned as cryoprecipitate collection sites to satisfy weekly collection targets. We formulate the problem as an integer programming problem and propose a robust and a stochastic programming approach to model the uncertain nature of blood supplies. We analyze these two approaches in which the rst one focuses on feasibility by meeting the weekly demand and the second approach aims to minimize the expected penalty due to the unsatis ed demand. Our results show that stochastic approach performs better with lower total collection cost, whereas robust approach presents a more cautious schedule with less amount of unsatis ed demand. Furthermore, we compute the value of the stochastic solution, which results in a signi cant improvement in the results as the weight assigned to the penalty of unmet demand amount increases.