Platelet Inventory Management in Blood Supply Chains

Restricted (Penn State Only)
Rajendran, Suchithra
Graduate Program:
Industrial Engineering
Doctor of Philosophy
Document Type:
Date of Defense:
May 25, 2017
Committee Members:
  • A. Ravi Ravindran, Dissertation Advisor
  • A. Ravi Ravindran, Committee Chair
  • Vittal Prabhu, Committee Member
  • Soundar Kumara, Committee Member
  • Terry Harrison, Outside Member
  • Jeanne Lumadue, Outside Member
  • Platelet wastage
  • stochastic integer programming
  • heuristic ordering policies
  • blood supply chain
  • genetic algorithm
It has been shown that a significant amount of platelets is wasted due to their short life after collection, just 5 days. Most of the previous work develop inventory policies based on a known demand pattern over a finite time horizon. Further, they assume that the entire platelet units received by the hospitals from the blood center are fresh (with all 3 days of transfusable life remaining). However, in practice, nearly 50% of the incoming platelets have one day shelf life. This research develops inventory models for the hospitals and the blood center under realistic settings (demand uncertainty, platelets with varying shelf life, finite supply at the blood center) with the objective of minimizing platelet shortage and wastage (due to outdating). In this dissertation, single objective deterministic inventory model is first developed to determine the number of platelet units to order and time between orders at the hospital. The model is extended to multiple objective inventory models at the hospital. The deterministic models are later extended stochastic programming models under demand uncertainty for hospital inventory management. Finally, inventory management along the entire blood supply chain is studied and platelet ordering policies are developed under demand uncertainty. Due to the computational complexity of the stochastic programming model developed for hospital inventory management, three heuristic rules are proposed for determining the platelet ordering policy at the hospital. The performance of these three ordering policies is compared against that of the traditional periodic review order-up-to policy, using real-life data obtained from a medical center. The shelf life of arriving platelets, coefficient of variation of demand and cost parameters are varied, and their impact is analyzed on the performance measures and the best rule with respect to each setting is determined. Based on the hospital setting and cost prioritization, the decision maker can decide the best performing rule. A new variant of the genetic algorithm, called modified stochastic genetic algorithm (MSGA) is proposed for determining the order-up-to level and re-order points at the various stages of the blood supply chain consisting of a blood center, which serves several hospitals. The performance of the MSGA algorithm is tested against an existing genetic algorithm. Using actual platelet demand data, it is shown that the MSGA algorithm performs well and can be easily scaled up to solve for larger supply chain problems. The proposed MSGA methodology is generic and can also be applied to determine ordering policies for other perishable supply chains, such as food or drug supply chains.