In This Issue
Winter Bridge on Frontiers of Engineering
December 15, 2023 Volume 53 Issue 4
This issue features articles by 2023 US Frontiers of Engineering symposium participants. The articles cover pressing global issues including resilience and security in the information ecosystem, engineered quantum systems, complex systems in the context of health care, and mining and mineral resource production.

Improving Blood Collection Operations at the American Red Cross

Wednesday, December 13, 2023

Author: Turgay Ayer

Blood collection operations at the American Red Cross were improved through a dynamic programming approach that systematized the selection of collection sites.

Approximately 29,000 units of red blood cells are needed every day in the United States, accounting for a total of 16 million annual blood component transfusions (American Red Cross 2023). The demand for blood ­transfusions continues to rise due to an increasing prevalence of chronic diseases, an aging population, and recent advancements in major therapies such as heart ­surgeries and organ transplants (Kasraian and Maghsudlu 2012). The margin between blood need and transfusable blood product availability is critically tight (Free et al. 2023), and warnings of blood shortages have recently received extensive media coverage.

How Blood Collection Works

The American Red Cross (ARC) is the primary provider of blood products within the United States, delivering over 40% of the nation’s blood supplies to over 3,000 hospitals. Furthermore, on an annual basis, it plays a critical role in responding to and assisting with the management of over 67,000 disasters worldwide. Operating through an extensive nationwide network, it relies on the dedication of over one million volunteers, employs 30,000 individuals, operates across 650 chapters, and serves 36 major regions.

Due to the limited supply and perishable nature of blood products, ­effective management of blood collection is critical for high-quality healthcare ­delivery. Blood collection procedures involve intricate processes. Donors ­contribute blood through either automated blood collection, known as apheresis, or standard whole blood donation. In the case of apheresis, a machine extracts specific blood ­components, like plasma or platelets, from the donor. Conversely, the more prevalent method is regular whole blood donation, where donors provide whole blood (Eder and Sebok 2007). Once collected, whole blood can be further divided into various components, including red blood cells, platelets, plasma, and cryoprecipitate.

Blood collection operations are complex and involve mobile collection sites. The selection of these locations and their associated collection timeframes is planned months in advance and determined based on factors like projected demand, the recency of prior visits to mobile sites, and the convenience of these locations for hosts. The duration of blood collection windows varies but generally spans between four and eight hours. The mobile collection vehicles are dispatched from the production facility early enough to commence collection at the start of the designated window. Once the collection period concludes, the mobile collection vehicle is packed up and returns to the production facility.

Cryo Collection: An Intricate and Costly Process

Cryoprecipitate (cryo) collection stands out as the most demanding among blood products. To generate cryo, the collected whole blood must be transported to a production facility. The plasma must be separated from red blood cells and then rapidly frozen into fresh-frozen plasma within eight hours of collection. In contrast, most other blood products allow a processing window of at least twenty-four hours from collection. This tight collection-to-processing timeframe for cryo complicates blood collection and production planning, presenting managerial complexities and practical challenges. For instance, to ensure compliance with the eight-hour constraint, additional courier services are often required to transport the collected whole blood back to the production facility promptly. Consequently, collecting blood for cryo is a more intricate and costly process compared to other blood products. Additionally, to produce cryo, special bags, known as cryo bags or triple bags, are utilized to collect whole blood. For other noncryo applications, less expensive bags, referred to as noncryo or double bags, can be used.

Mobile collection sites are typically designated as either cryo or noncryo sites for a given collection day, with the number of cryo sites depending on the Red Cross ­Headquarters’ weekly cryo collection target. The decision to designate sites for cryo collection does not impact the production of other blood products. Blood collected at mobile sites is transported back to the manufacturing facility by the end of the day, a process termed “end-of-day delivery.” However, for cryo sites, if the end-of-day ­delivery cannot meet the eight-hour collection-to-­processing constraint, an additional midday pickup, typically provided by a courier company, is scheduled, incurring extra transportation expenses.

Cryoprecipitate collection stands out as the most demanding among blood products.

The selection of cryo-collection sites is usually made two days before the collection day, taking into account courier requirements, production planning, and staffing needs. If a site initially selected for cryo collection no longer requires cryo blood (e.g., the weekly production target has been met), cryo collection at that site can be canceled to avoid the midday pickup and its associated transportation cost. Cancellation may result in penalties and necessitate the use of more expensive cryo bags for noncryo products. The number of cryo units collected at a site is uncertain due to factors like donor no-shows, walk-in donors, and the success of cryo production from blood collected in cryo bags.

Meeting demand for cryo is very important because cryo plays a critical role in clotting and controlling massive hemorrhaging, and cryo is often used in the treatment of massive trauma and many major diseases, including metastasized cancers, cardiac diseases, hepatic failures, and organ transplants (Ness and Perkins 1979). Due to the advancements in major surgeries and treatments, the demand for cryo has been rapidly increasing (Curry et al. 2022). For example, from 2000-2010, the use of fresh-frozen plasma, the raw material for cryo production, increased tenfold and grew to more than 2.4 million units used annually in the United States (Gupte 2011).

Solving the Cryo Collection Problem

Our research team at Georgia Tech, in collaboration with colleagues at the American Red Cross, conducted a study in which we analyzed a regional-level cryo collection problem faced by the ARC’s Southern Regional Manufacturing and Service Center (RMSC).[1] The ARC’s Southern RMSC is one of the largest manufacturers and ­suppliers of blood and blood products, serving more than 120 hospitals in the southern United States. In particular, we focused on determining when and from which mobile collection sites to collect blood for cryo production and how to schedule the courier services such that the collection targets are met, and the total collection costs are minimized. A regional-level cryo collection problem imposes several challenges: i) if the blood collected is to be processed into cryo units, it has to be processed within eight hours after collection; ii) the collection quantities are uncertain due to no-shows, random walk-ins, and random yields in production; and iii) collection schedules need to be made in advance and may need to be dynamically adjusted depending on the realizations of uncertainties.

Future research avenues might include simultaneous optimization of site scheduling and production decisions,
as well as donor optimization strategies to maximize the efficiency and flexibility of blood component production.

Historically, in the ARC’s southern region, cryo and noncryo site designations were determined at the end of the previous week. Cryo sites collected whole blood in cryo bags throughout the day, with all collected blood intended for cryo production. These sites typically required a midday pickup to meet the eight-hour processing window for blood units collected early in the day. In contrast, noncryo sites collected whole blood for noncryo production using regular blood bags, following a model termed the “nonsplit model,” where all or none of the blood collected at a site is designated for cryo collection. In our project, an alternative collection model, referred to as the “split model,” was proposed and analyzed. In the split model, the collection window of each site is divided into two intervals, allowing cryo collection in both intervals or solely in the second interval. If only the second interval is designated for cryo collection, eliminating the need for cryo bags in the first interval, the mobile collection vehicle can transport the whole blood collected during the second interval to the production facility in time for cryo processing without incurring extra transportation costs.

To formally analyze the problem and optimize the collection schedules, we first formulated a large-scale ­Markov decision process (MDP). The model considered all ­collection sites and determined the sites for cryo collection dynamically over the week, given the day of the week, the type of blood collection, and the collection ­window. However, given the size of the problem, this MDP was computationally intractable. We analyzed this MDP ­model structurally, proved several structural ­properties, and developed a near-optimal solution algorithm. We also built a fast heuristic algorithm to solve the model and compare its performance with the MDP-based solution. To facilitate implementation, we further developed a decision support tool (DST) to systematize the selection of the collection sites, which is now used in practice in the entire southern US region.

Using historical real data from the ARC, we estimated the potential benefit of our proposed solution approach as a 70% reduction in total collection costs, leading to a roughly $100,000 cost reduction per regional center per year and a $4 million annual reduction in blood collection operations at the national level. The actual implementation of the DST in the ARC’s southern region resulted in an increase in the number of whole blood units that can be shipped back to the production facility and processed within eight hours after collection. During the fourth quarter of 2016, this facility processed about 1,000 more units of cryo per month (an increase of 20%) at a slightly lower collection cost, resulting in an approximately 40% reduction in the per-unit collection cost for cryo. Based on the successful implementation in the southern region, the ARC expanded the implementation of the DST, first to its St. Louis facility, and ultimately rolled it out to the entire nation.


In collaboration with the ARC, we tackled the challenge of optimizing blood collection schedules for perishable products, a daily struggle for blood collection agencies. Our innovative approach introduced a collection model that divided each site into two intervals, accommodating different types of collections, departing from the conventional cryo or noncryo categorization.

To evaluate the value of this split model and optimize cryo collection schedules, we formulated the problem as a stochastic dynamic program, developing a near-optimal solution algorithm. Our numerical analysis, based on real data, demonstrated significant cost reductions compared to the previous approach. Post-implementation data confirmed our model’s effectiveness in increasing cryo collection quantity while keeping costs in check, benefiting the ARC’s production capacity expansion and cost management.

As a nonprofit organization reliant on blood donations, the ARC found these improvements substantial and implemented our proposed methods in multiple manufacturing facilities. This success underscores the potential impact of our collection model and solution approach for other blood collection organizations engaged in cryo unit production from whole blood.

Our innovative collection model, solution approach, and insights into cryo collection scheduling have broader implications for blood collection organizations involved in cryo unit production from whole blood. We anticipate that our research will encourage further exploration of various aspects of blood collection and procurement operations. Future research avenues might include simultaneous optimization of site scheduling and production decisions, as well as donor optimization strategies to maximize the efficiency and flexibility of blood component production. These endeavors hold the potential to enhance the overall performance and sustainability of blood supply chains.


American Red Cross. Blood Facts & Statistics. Accessed ­November 13, 2023. Online at and-statistics#donor-facts.

Ayer T, Zhang C, Zeng C, White CC III, Joseph VR. 2019. Analysis and Improvement of Blood Collection Operations. Manufacturing & Service Operations Management 21(1):29–46.

Ayer T, Zhang C, Zeng C, White CC III, Joseph VR, Deck M, Lee K, Moroney D, Ozkaynak Z. 2018. American Red Cross Uses Analytics-Based Methods to Improve Blood-­Collection Operations. INFORMS Journal on Applied Analytics 48(1):24–34.

Curry N, Davenport R, Lucas J, Deary A, Benger J, Edwards A, Evans A, Foley C, Green L, Morris S, and three others. 2022. The CRYOSTAT2 trial: The rationale and study ­protocol for a multi-Centre, randomised, controlled trial evaluating the effects of early high-dose cryoprecipitate in adult patients with major trauma haemorrhage requiring major haemorrhage protocol activation. Transfusion Medicine 33(2):123–131.

Eder AF, Sebok MA. 2007. Plasma components: FFP, FP24, and thawed plasma. Immunohematology 23(4):150–7.

Free RJ, Sapiano MRP, Ortiz JL, Stewart P, Berger J, ­Basavaraju SV. 2023. Continued stabilization of blood collections and transfusions in the United States: Findings from the 2021 National Blood Collection and Utilization Survey. ­Transfusion 63(S4):S8–S18.

Gupte S. 2011. Recent Advances in Pediatrics. New Delhi, India: Jaypee Brothers Medical Pub.

Kasraian L, Maghsudlu M. 2012. Blood donors’ attitudes towards incentives: influence on motivation to donate. Blood Transfusion 10(2):186–90.

Ness PM, Perkins HA. 1979. A Simple Enzyme-Immunoassay (EIA) Test for Factor VIII-Related Antigen (VIIIAGN). ­Journal of Thrombosis and Haemostasis 42(03):848–854.


[1]  Original research articles based on this collaboration were published in Manufacturing & Service Operations Management (Ayer et al. 2019) and INFORMS Journal on Applied Analytics (Ayer et al. 2018).


About the Author:Turgay Ayer is Virginia C. and Joseph C. Mello Chair and professor of industrial and systems engineering, the H. Milton Stewart School of Industrial and Systems Engineering, Georgia Institute of Technology.