Computers & Operations Research 54 (2015) 129–141

Contents lists available at ScienceDirect

Computers & Operations Research

journal homepage: www.elsevier.com/locate/caor

Stochastic integer programming models for reducing wastages

and shortages of blood products at hospitals
Serkan Gunpinar n, Grisselle Centeno
University of South Florida, Department of Industrial and Management Systems Engineering, 4202 E. Fowler Avenue, Tampa, FL 33620, USA

art ic l e i nf o a b s t r a c t

Available online 16 September 2014 Major challenges in the management of the blood supply chain are related to the shortage and wastage
Keywords: of the blood products. Given the perishable characteristics of this product, storing an excessive number
Supply chain of blood units on inventory could result on the wastage of this limited resource. On the other hand,
Health care application having shortages may result in cancellations of critical health related activities and as a result a potential
Perishable inventory increase on fatality rates at hospitals.
Blood products This paper presents integer programming models to minimize the total cost, shortage and wastage
Integer programming levels of blood products at a hospital within a planning horizon. The primary focus is on the red blood
Optimization cells and the platelet components of the whole blood cells. The stochastic and deterministic models
included consider uncertain demand rates, demand for two types of patients, and crossmatch-to-
transfusion ratio. Results show wastage rates decreasing from 19.9% to 2.57% on average. In addition, the
shortages and total cost are reduced 91.43% and 20.7% respectively for a given capacity increases.
Computational results are included and discussed.
& 2014 Elsevier Ltd. All rights reserved.

1. Introduction
Human blood is a scarce resource. It is only produced by human
beings and there are currently no other products or alternative
chemical process that can be used to generate blood. The blood
carries substances such as nutrients and oxygen to the cells and
delivers waste away from the cells.
Blood is usually drawn as “whole blood” but then it could be
mechanically separated into other useful components. These
components are then used to meet the specific transfusion
demands of patients. One unit of whole blood can be divided into
five different blood products: red blood cells (RBCs), plasma, white
blood cells, serum or platelets. Red blood cells are the most
abundant cells in blood and contain a protein called hemoglobin
that moves oxygen to our cells. Plasma is a yellowish liquid
component and is obtained by removing RBCs from whole blood.
White blood cells are part of the immune system and defend the
body against infectious agents. Serum is a blood plasma without
clotting factor, white and red blood cells. Finally, platelets are the
clotting factors that are contained in the plasma and relate to the
process of coagulation which repairs the body when a wound and
bleeding occurs.
n
Corresponding author.
E-mail addresses: serkan.gunpinar@gmail.com (S. Gunpinar),
gcenteno@usf.edu (G. Centeno).
Platelets can also be drawn directly from a person through the
use of an apheresis device. All blood components except for
plasma can become outdated. Platelets, especially, are considered
highly perishable since they can only be stored up to five days
before deteriorating. The second most perishable blood compo-
nent, RBC, can be kept for up to forty two days on inventory.
Fig. 1 shows the general process related to the supply chain of
blood products. It starts with the collection process. Blood units for
transfusion purposes are collected from donors either at a blood
center or through mobile units on remote locations. Potential
donors are evaluated through screening process to make sure that
diseases, including HIV and viral hepatitis, cannot be transmitted
through blood transfusion. After a thorough process of rules and
regulations for compliance of donors, units are tested. Then, the
whole blood units are either stored or mechanically separated
(extracted) into components. Hospitals place orders to blood
center based on the forecasted demand for the various procedures
scheduled. A recipient's blood is tested against a donor's blood
(this process is known as crossmatching) and, when compatible,
blood units are reserved for the specific patient for the period
known as crossmatch release period.
When comparing blood products with any other item, several
differences directly connected to the supply chain become evident.
First, supply of blood is volunteer-based whereas there is a cost
associated with most products. Second, the structure of the blood
supply chain is considered as reverse to the majority of traditional
products since the whole blood produced by the living beings is

http://dx.doi.org/10.1016/j.cor.2014.08.017
0305-0548/& 2014 Elsevier Ltd. All rights reserved.
130 S. Gunpinar, G. Centeno / Computers & Operations Research 54 (2015) 129–141
Fig. 1. Supply chain of blood products.
Fig. 2. Blood inventory model.
mechanically separated into components. However, in traditional
supply chain, parts are manufactured and then assembled to
create a finished product. Third, the price associated with the
acquisition of the blood is always linear, that is, no economies of
scale are present. Finally, the most significant difference is about
the inventory issuing policy as shown in Fig. 2. When the hospital
blood bank receives a blood request for a specific patient, the
crossmatched blood is moved from unassigned (free) inventory to
assigned (reserved) inventory and kept for this patient until the
blood is transfused or the crossmatch release period is over. If the
blood is not transfused and the crossmatch release period is over,
it could be returned to the unassigned inventory to be used for
other purposes. Since the amount of blood needed for a medical
procedure is uncertain, physicians tend to overestimate units
required for safety issues. Approximately 50% of blood units
requested by physicians are returned to the unassigned inventory
without being transfused [1]. Depending on the patient, organiza-
tional policy and types of procedures, the crossmatch-to-
transfusion ratio (C/T ratio) varies. This ratio is typically higher
for the cases in emergency rooms [2].
Minimizing shortage and wastage of the blood is the major
challenge related to the management of blood at a hospital. Due to
the perishable characteristic of blood (it becomes outdated if not
used during its predetermined shelf life) it is critical to avoid
storing an excessive number of blood units. At the same time,
insufficient number of blood products on inventory may increase
cancellations of the scheduled activities at a hospital and as a
result increase fatality rates. In 2004, 17% of platelet units that
were collected in the U.S. were outdated before being transfused
[3] (wastage); and a total of 492 reported cancellations of elective
surgeries on one or more days were due to blood shortages at 1700
U.S. hospitals participating in a survey in 2007 [4]. Thus, outdates
and shortages of blood products have been and continue to be an
issue for hospitals.
Given the evident need to better utilize this scarce resource,
this paper primarily focuses on the red blood cells and the platelet
components of whole blood cells. Integer programming (IP)
models to minimize the shortage and wastage levels and also
the total cost related to various procedures at a hospital are
presented. The remainder of the paper is organized as follows. In
Section 2, relevant literature is discussed. The model formulation
is shown in Section 3. Numerical results of the models are given in
Section 4. Concluding remarks and opportunities for future work
are presented in Section 5.
2. Literature review
The research related to supply chain management of perishable
products in general and blood products in particular was initiated
in the 1960s by van Zyl [5]. The paper written by Nahmias [6] in
1982 focuses on the perishable inventory and provides a brief
review for the applications of the models to the blood bank
management. In 1984, Prastacos [7] overviews the theory and
practice of blood inventory management. Since then, close to one
hundred blood related publications have become available in the
literature. Two peaks in the publication history of blood products
are observed [8]; one in the period between 1976 and 1985 and
more recently in the period between 2001 and 2010.
Supply chain problems of blood products have been analyzed using
wide range of methodologies. In particular, simulation methodology
[9–15, 17–19], mathematical proofs and derivations [20–23], dynamic
programming [11,24] and integer programming [25–29] are some of
the most common solution methods in the literature. These
approaches are either used alone or in combination with other
methods to analyze and solve real-life problems. Studies consider
deterministic demand [20,22,26,29–32], stochastic demand [9–
15,17,18,21,24–26,33–36], and can be classified at individual hospital
level [10–12,18,24,34,36–41] and regional blood center level
[11,12,15,17,18,25,26,29,33–36,40,42–44]. Furthermore, models incor-
porate different characteristics of blood supply chain and indicate
variety of decision variables, objective function and constraints.
2.1. Integer programming models
Given that integer programming models for blood inventory
management are most relevant to this paper, impact of related IP
models is discussed in detail next. Hemmelmayr et al. [25] develop
integer programming models to decide which hospitals a vendor
(through vehicles from blood centers) should visit each day given
that the routes are fixed for each region. Authors consider recourse
action in order for hospitals to be hedged against the uncertainty
associated with blood product usage. Both, integer programming
and variable neighborhood search approaches are used and
compared in terms of their efficiencies. Sahin et al. [26] formulate
three problems using integer programming to address the
location-allocation aspects of regionalization of blood services.
The experimental results obtained using real data for Turkish Red
Crescent blood services were reported. Jacobs et al. [27] build two
integer programming models to investigate a facility relocation
problem for the mid-Atlantic region of the American Red Cross in
Norfolk, Virginia. They provide insights into the current scheduling
activities of blood collections and distributions. The integer pro-
gramming model explained in [28] considers the orders for fresh
blood separately and allocates blood units from regional blood
transfusion service to the hospitals. The objective is to minimize
the total expected number of units that are sent back to the blood
transfusion service. Ghandforoush and Sen [29] formulate a non-
linear integer programming model to determine the minimum
cost platelet production schedule for the regional blood center.
Since the initial formulation carries a non-convex objective func-
tion that is difficult to solve and would not guarantee convergence
to optimality, the formulation is simplified to achieve a better
structure. As both objective function and constraints of the revised
formulation include quadratic terms, a two-step transformation
 S. Gunpinar, G. Centeno / Computers & Operations Research 54 (2015) 129–141 131

called linear 0–1 integer alternative is proposed to guarantee
optimality.
2.2. Model characteristics
Haijema et al. [45] study a real life case of a Dutch blood bank. The
decision elements considered are the production size of platelet units
and pools where these units are taken from. The goal is to minimize
production, outdating, inventory, shortage and mismatch cost over a
period. Ghandforoush and Sen [29] present a platelet production and
bloodmobile scheduling problem for a regional blood center and the
objective of the model described minimizes total system cost in a
regional blood center including production, transportation and yield
loss costs. The problem is formulated using two critical decision
variables that are associated with the number of platelet concen-
trates produced at the blood center and the number of shuttle round
trips from the blood center to bloodmobile. The model limits the
total number of shuttle trips for all sites and platelet units produced
per shuttle trip. Zhou et al. [46] consider inventory management of
platelets in hospitals and the model identifies the regular order
quantity and the expedited order-up-to level for each cycle. The
expected cost that is minimized indicates the cost of expediting,
shortage and outdate. Sapountzis [28] develops a method that
allocates blood units from a regional blood transfusion service to
the hospitals. The decision variables considered are the number of
whole blood and packed cells which are shipped to given hospitals
and the goal is to minimize the total number of expiries. Nagurney
et al. [4] analyze the complex supply chain of human blood
consisting of collection sites, testing and processing facilities, storage
facilities, distribution centers, as well as demand points. Authors
develop a generalized network optimization model where multi-
criteria system-optimization approach enables decision makers to
minimize both total operational cost and total risk function.
2.3. Contributions to literature
Aforementioned advancements in addition to others have been
implemented to make further improvements in the supply chain
of blood products. However, a gap found in the literature is the
lack of models that incorporate the age of blood units into the

The age of blood units received from the blood center is known
and varies over time.

The lifetime of platelets is limited to five days including two
days of testing [3].

The lifetime of red blood cells is limited to forty two days
including two days of testing [17].

General blood issuing policy for the hospital is FIFO where
oldest units on inventory are issued first when the blood units
are requested by physicians for patient needs [6].

If demand is not satisfied due to the unavailability of blood
units, a shortage cost is incurred.

If a blood unit expires, a wastage cost is incurred associated
with discarding blood units.
3.1. Model 1: stochastic integer programming model
Hospitals usually face two types of uncertainties associated
with the use of blood products. The first relates to the uncertainty
of emergency cases which are difficult to anticipate. Unlike
scheduled procedures, emergency cases are unexpected and ran-
dom. Thus, the amount of blood units needed is unknown in
advance. The second uncertainty relates to the C/T ratio. Prior to a
procedure, blood is requested by the physician for a specific
patient and the number of blood units cross-matched is typically
overestimated for safety issues. Thus, some blood may be returned
back to inventory after the crossmatch release period is over. To
address these challenges, we use stochastic programming to
handle demand uncertainty and build integer programming mod-
els that explicitly consider age of blood on inventory. The decision
support tool developed will aid in the decision making process
associated with inventory management of blood supply.
Tables 1–3 summarize the indices, the parameters and the
variables that are used in the models.
Using the indices, parameters and decision variables in
Tables 1–3, the non-linear stochastic integer programming model
is formulated as follows:
!
T S I T S T S T
min ∑ c  x t þ ps  ∑ ∑ ∑ h  vsit þ ∑ ∑ w  ust þ ∑ ∑ b  r st
t¼1 s¼1i¼3t ¼1 s¼1t ¼1 s¼1t ¼1

formulations. In addition, the majority of the models do not ð1Þ

differentiate demand among patient groups with specific blood
age requirements. In our study, we develop stochastic integer s:t: xt r CAP t 8t ð2Þ
programming models that explicitly consider age of blood units on
inventory as well as the demand for two types of patients (one yit ¼ 0; i ¼ 1; 2; 8 t ð3Þ
which requires fresh blood). Furthermore, although many studies
assume that blood demand is uncertain, our paper differs from yit ¼ xt θit ; i ¼ 3; 4; …; I; 8 t ð4Þ
others in the methodology and uses stochastic programming to
zsit Z zsði  1Þt i ¼ 3; 4; …; I; 8 s; t ð5Þ
capture the uncertain nature of blood demand. Finally, C/T ratio
and crossmatch release period which are the unique character-
I
istics of blood supply chain are also taken into consideration in our s
dt ¼ ∑ ððvsði  1Þðt  1Þ þyit Þzsit  msit Þ þ r st ; 8 s; t ð6Þ
model formulations. i¼3

ðzsit  zsði  1Þt Þðvsði  1Þðt  1Þ þ yit Þ Zmsit

3. Problem description and model formulation i ¼ 3; 4; …; I; 8 s; t ð7Þ

We consider a two-level supply chain of blood products zs2t ¼ 0 8 s; t ð8Þ

consisting of one hospital and one blood center. The hospital faces
blood demands that need to be satisfied in order to perform its
Table 1
daily operation related to blood supply. Thus, optimal blood order Indices for models.
levels should be identified over multiple periods. In this research,
the following assumptions have been made: s Demand scenario, s ¼ 1,2,…,S
i Age of blood, i¼1,2,…,I (days)
Time period, t¼ 1,2,…,T (days)a

The capacity of the blood center is limited. t
a Age group of blood (‘young’ (0) or ‘any’ (1)), a A f0; 1g

Lead times for blood supply are zero. In other words, the
hospital orders for blood products are fulfilled in no time. a
t¼ 1 refers to a Monday.
132 S. Gunpinar, G. Centeno / Computers & Operations Research 54 (2015) 129–141

Table 2
Parameters for models.

S Number of scenarios
I Lifetime of blood product
T Length of planning horizon
b Unit shortage cost of blood at the hospital
c Unit purchasing cost of blood
ca Unit purchasing cost of ‘young’ blood (0) and ‘any’ blood (1)
h Unit holding cost of blood at the hospital
M Big M (Big Number)
ps Probability of scenario s, ∑Ss ¼ 1 ps ¼ 1
w Unit wastage cost of blood at the hospital
θit Proportion of i days old blood in blood shipments from blood center in time period t, 0 r θit r 1, ∑Ii ¼ 3 θit ¼ 1 8 t
θait Proportion of i days old blood in ‘young’ blood shipments (a ¼0) and in ‘any’ blood shipments (a ¼1)
from blood center in time period t, 0 r θait r 1, θ03t ¼ 1, ∑Ii ¼ 4 θ1it ¼ 1 8 t
ðsÞ Blood demand at the hospital in time t (for scenario s)
dt
ðsÞ
dat ‘Young’ blood (a ¼0) demand and ‘any’ blood (a ¼1) demand at the hospital in time t (for scenario s)
CAPt Capacity of the blood center (allocated to the hospital) in time period t
CRP Crossmatch release period at the hospital
CT Average C/T ratio at the hospital
TOL Small number to be used in the linearization of floor functions

Table 3
Decision variables for models.

mðsÞit
Auxiliary variable associated with age group i in time t (for scenario s). It captures the number of blood
units in an age group left to be used for the next period if all available blood in this age group
is not fully used to satisfy the demand in current period
r ðsÞ Number of blood shortage at the end of time t (for scenario s) at the hospital
t
π ðsÞ Number of ‘old’ blood demand in time t (for scenario s) that are not satisfied
t
by older blood units on inventory due to unavailability
uðsÞ
t
Number of blood wastage at the end of time t (for scenario s) at the hospital
vðsÞ
it
Inventory level of i days old blood at the end of time t (for scenario s) at the hospital
xt Number of blood ordered by the hospital from the blood center at the beginning of time t
xat Number of ‘young’ blood (0) and ‘any’ blood (1) ordered by the hospital from the blood center at the beginning of time t
yit Number of i days old blood received by the hospital at the beginning time t
zðsÞ
it
1 if i days old blood used to satisfy the demand in time period t (for scenario s), 0 otherwise
βðsÞ
it
Number of i days old blood returned from assigned inventory to unassigned inventory at the beginning of time t (for
scenario s)

s
I group. Constraint (5) guarantees the FIFO blood issuing policy.
dt  ∑ ðvsði  1Þðt  1Þ þ yit Þ r r st 8 s; t ð9Þ
i¼3
Constraint (6) requires demand to be fully satisfied when blood
supply exceeds demand. Otherwise, the hospital faces a shortage
vsit ¼ ð1 zsit Þðvsði  1Þðt  1Þ þ yit Þ þ ðzsit  zsði  1Þt Þmsit issue. msit is an auxiliary variable that captures the number of blood
i ¼ 3; …; I; 8 s; t ð10Þ units in an age group left on inventory when at least one unit is
absorbed from inventory during the given time period. Otherwise,
vsð2Þt ¼ 0 8 s; t ð11Þ it would be equal to zero. Note that there is at most one age group
having msit with non-zero value in each period. Constraint (7)
vsið0Þ ¼ 0 8 s; i ð12Þ assures that the values of auxiliary variable, msit, do not exceed the
number of blood units available in their age groups. Constraint (8)
ust ¼ vsðIÞt 8 s; t ð13Þ ensures that two days old blood units are not used to satisfy the
demand as the hospital only receives blood units older than two
xt A Z þ 8t ð14Þ days old from the blood center. Constraints (6) and (9) capture the
number of blood shortages. Constraint (10) updates end-period
r st ; ust A Z þ 8 s; t ð15Þ blood inventory levels for each age group. Constraint (11) assures
that two days old blood is never available on inventory. Constraint
yit A Z þ 8 i; t ð16Þ (12) states that there is no inventory available at the beginning of
the analysis period. Constraint (13) identifies the wastage levels of
msit ; vsit A Z þ 8 s; i; t ð17Þ the hospital at the end of each period. Constraints (14)–(17) show
rst , ust , xt, msit, vsit and yit are non-negative discrete variables since
zsit A f0; 1g 8 s; i; t ð18Þ
the blood units are received in blood bags. Constraint (18) states
The objective function (1) seeks to minimize the purchasing that zsit is a binary variable.
cost and the expected inventory, wastage and shortage costs Due to the interactions between binary and discrete variables,
during the planning horizon. Constraint (2) is the capacity con- the optimization problem includes non-linear terms in the above
straint of the blood center (supplier). Constraint (3) ensures that formulation. After the first linearization technique is applied
the hospital never receives one or two days old blood units from which is detailed in Appendix A, the interactions between v, y,
the blood center as two days are required for testing after the m and z variables in constraints (6), (7) and (10) are replaced with
blood is collected. Constraint (4) allocates blood units to each age the corresponding linearization variables from which constraints
 S. Gunpinar, G. Centeno / Computers & Operations Research 54 (2015) 129–141 133

(19)–(21) are obtained. In addition, constraints (65)–(84) are s

I
d1t ¼ ∑ ððvsði  1Þðt  1Þ þ yit Þ  zsit  msit Þ þ π st ; 8 s; t ð29Þ
added into the new formulation. i¼4
I
s
dt ¼ ∑ ðγ sit þ αsit  msit Þ þ r st 8 s; t ð19Þ s s
I
i¼3 d0t þ d1t  ∑ ðvsði  1Þðt  1Þ þ yit Þ rr st 8 s; t ð30Þ
i¼3
γ sit þ αsit μsði  1Þt  ψ sit Z msit
i ¼ 3; 4; …; I; 8 s; t ð20Þ xat A Z þ a A f0; 1g; 8 t ð31Þ

vsit ¼ vsði  1Þðt  1Þ þ yit  γ sit αsit þ λsit  δsit π st A Z þ 8 s; t ð32Þ

i ¼ 3; 4; …; I; 8 s; t
In summary, constraints (2)–(5), (8) and (9), (11)–(21), (65)–(84)
are present in the model discussed in this section.
3.2. Model 2: stochastic integer programming model with two
patient types
Freshness of the blood products (a.k.a. young blood) may be
critical and required for certain types of patient groups. Fresh blood
(specifically RBC) is highly preferable for operations such as open-
heart surgeries [47]. Also, according to the study from University of
Texas Southwestern [48], it is suggested to have younger blood units
for the transfusion of pediatric heart surgery patients, open heart
pediatric and cardiopulmonary bypass surgeries. In the absence of
fresh blood units, some of the blood related procedures may need to
be postponed to a later date. Similarly, young platelets are highly
preferred for oncology and hematology patients. However, platelets
of any age (up to the shelf life) can be used for traumatology or other
general surgery patients [45].
Based on these requirements, two types of patients are defined:
Type 1 which requires fresh/young blood and Type 2 which could
use blood of any age (young or old). Young platelets are defined as
being younger than 3 days old [45]. Young red blood cells are
younger than 5 days old [47].
The general procedure to order and manage blood units at a
hospital is as follows:
The hospital places separate orders for units of any age and/or
specifically young units to the blood center. If no age is specified,
blood centers will typically send units based on FIFO policy (this
may contain fresh blood units). Upon arrival, units are grouped
and stored based on their age group; then, they are allocated to the
procedures using the FIFO policy. The demand for Type 1 patients
can only be satisfied with young products. If demand for young
products exceeds supply, the hospital faces a shortage issue. On
the other hand, demand for Type 2 patients can be satisfied by
units of any age. That is, young units of blood can also be used to
satisfy demand for Type 2 patients but only when older blood
units are not available on inventory. To incorporate the demand for
two types of patients into our formulation, Constraints (4)–(7) and
(9) of the initial model are replaced with the following constraints:
x0t  θ0it þ x1t  θ1it ¼ yit i ¼ 3; 8 t
x1t  θ1it ¼ yit i ¼ 4; …; I; 8 t
zsit Z zsði  1Þt i ¼ 3; 8 s; t
zsit Z zsði  1Þt i ¼ 5; …; I; 8 s; t
ðzsit zsði  1Þt Þðvsði  1Þðt  1Þ þ yit Þ Z msit i ¼ 3; 8 s; t
ðzsit zsði  1Þt Þðvsði  1Þðt  1Þ þ yit Þ Z msit
i ¼ 5; …; I; 8 s; t
3
s designing efficient solution procedures for the stochastic problem
d0t þ π st ¼ ∑ ððvsði  1Þðt  1Þ þyit Þ  zsit  msit Þ þ r st
i¼3
8 s; t
ð21Þ Constraints (22) and (23) allocate blood units to each age
group. Constraints (24) and (25) enforce the FIFO policy. To
guarantee that young units are not prevented from being allocated
to meet demand for fresh blood, some links are removed from the
formulations; otherwise, young units cannot be used before all of
the older units on inventory are used. Constraints (26) and (27)
guarantee that the values of msit do not exceed the number of units
available in their age groups. Constraints (28) and (29) are demand
constraints and ensure the usage of young units for Type 2 patients
when all older units on inventory are depleted. Constraint (28) in
conjunction with constraint (30) identifies the shortage levels.
Finally, constraints (31) and (32) force xat and πst to take non-
negative discrete values.
Since the purchasing cost of young units exceeds the cost of any
units, the first term of the objective function in the first formula-
tion is replaced by the following revised term:
T T
∑ c0  x0t þ ∑ c1  x1t ð33Þ
t¼1 t¼1
Similar to the model described by (1)–(18), the first lineariza-
tion technique is applied to the constraints (26)–(29) as they
indicate non-linear terms. After the interactions between v, y and
z are replaced with the corresponding linearization variables,
constraints (34)–(37) are obtained as follows:
8
s
d0t þ π st ¼ ∑ γ sði  1Þðt  1Þ þ αsit  msit þ r st 8 s; t ð34Þ
i¼3
s
I
d1t ¼ ∑ γ sði  1Þðt  1Þ þ αsit  msit þ π st 8 s; t ð35Þ
i¼9
γ sit þ αsit  μsði  1Þt  ψ sit Zmsit
i ¼ 3; 4; …; 8; 8 s; t ð36Þ
γ sit þ αsit  μsði  1Þt  ψ sit Zmsit
i ¼ 10; 11; …; I; 8 s; t ð37Þ
In summary, constraints (2) and (3), (8), (10)–(18), (22)–(25),
(30)–(32), (34)–(37), (65)–(84) are present in the model discussed
in this section.
ð22Þ 3.3. Model 3: deterministic integer programming model with C/T
ratio and crossmatch release period
ð23Þ
According to [49], crossmatch-to-transfusion ratio should ide-
ð24Þ ally be 1:1. They also mentioned a C/T ratio less than 2.5 to be
acceptable; that is, out of five units crossmatched only two are
ð25Þ used. In order to incorporate C/T ratio and crossmatch release
period into our model, a deterministic IP model had to be used and
ð26Þ assumed that the blood demand is known. This route was taken
after not being able to successfully run the stochastic version of
the model. That is, a CPLEX error 1001 – diagnosed as a run-out-of-
ð27Þ memory issue – was encountered. The size of branch and bound
tree required more memory to run to completion. Given that
is beyond the scope of this paper, a deterministic demand was
ð28Þ assumed to be able to solve the problem and to reach a solution
134 S. Gunpinar, G. Centeno / Computers & Operations Research 54 (2015) 129–141

within reasonable computation time using the commercial soft- blood units returned back to unassigned inventory in each period.
ware. The deterministic model is discussed next. The number of returned units is computed by multiplying the
Running out of memory message is appeared as the size of the number of blood units in assigned inventory with the blood return
branch and bound tree is large in stochastic formulation and ratio derived from subtracting the inverse of hospital's average C/T
CPLEX needs more memory to run to completion. ratio from one. For each age group it is assumed that the same
T I T T T proportions of blood units in assigned inventory are returned to
min ∑ c  xt þ ∑ ∑ h  vit þ ∑ w  ut þ ∑ b  r t ð38Þ unassigned inventory. Since it takes CRP time periods for reserved
t¼1 i¼3t ¼1 t¼1 t¼1
blood units to be returned to the unassigned inventory, the value
s:t:xt r CAP t 8t ð39Þ of βit in Constraint (52) is set equal to zero for the first CRP periods.
Similarly, as the youngest blood units received by the hospital are
yit ¼ 0; i ¼ 1; 2; 8 t ð40Þ three days old, the value of βit is set equal to zero for blood units
that are younger than 3 þ CRP days old in Constraint (53). Con-
yit ¼ xt θit ; i ¼ 3; 4; …; I; 8 t ð41Þ straint (54) identifies the wastage levels of hospital at the end of
each period. Constraints (55)–(58) are the non-negativity con-
zit Z zði  1Þt i ¼ 3; 4; …; I; 8 t ð42Þ straints. Finally, Constraint (59) forces zit to assume binary values.
I
Similar to the first two models, linearization techniques
dt ¼ ∑ ððvði  1Þðt  1Þ þ yit Þzit  mit Þ þr t 8t ð43Þ (Appendix A) are applied to the non-linear terms impacting
i¼3 Constraints (43), (44) and (47) and resulting on modified Con-
straints (60)–(62). In addition, Constraints (65)–(84) are added
ðzit zði  1Þt Þðvði  1Þðt  1Þ þyit Þ Z mit
into the new formulation.
i ¼ 3; 4; …; I; 8 t ð44Þ
I
z2t ¼ 0 8t ð45Þ dt ¼ ∑ ðγ it þαit  mit Þ þ r t 8t ð60Þ
i¼3

I
dt  ∑ ðvði  1Þðt  1Þ þ yit Þ r r t 8t ð46Þ γ it þ αit  μði  1Þt  ψ it Zmit i ¼ 3; 4; …; I; 8 t ð61Þ
i¼3

vit ¼ ð1 zit Þðvði  1Þðt  1Þ þ yit Þ þ ðzit  zði  1Þt Þmit þβit vit ¼ vði  1Þðt  1Þ þ yit  γ it  αit þλit δit þ βit
i ¼ 3; …; I; 8 t ð47Þ i ¼ 3; 4; …; I; 8 t ð62Þ

vit ¼ βit i ¼ I þ 1; …; I þ CRP; 8 t ð48Þ Finally, the floor function shown in Constraint (51) is modified
using techniques described in Appendix B, resulting in Constraints
vð2Þt ¼ 0 8t ð49Þ (63) and (64).
við0Þ ¼ 0 8i ð50Þ βit Zððγ ði  CRP  1Þðt  CRP  1Þ αði  CRPÞðt  CRPÞ Þ

βit ¼ ⌊ððvði  CRP  1Þðt  CRP  1Þ þ yði  CRPÞðt  CRPÞ Þ  mði  CRPÞðt  CRPÞ Þ  ð1  CT  1 Þ  1 þTOL
zði  CRPÞðt  CRPÞ mði  CRPÞðt  CRPÞ Þ  ð1  CT  1 Þc i ¼ 3 þ CRP; …; I þCRP; t ¼ CRP þ 1; …; T ð63Þ
i ¼ 3 þ CRP; …; I þ CRP; t ¼ CRP þ 1; …; T ð51Þ
βit rððγ ði  CRP  1Þðt  CRP  1Þ αði  CRPÞðt  CRPÞ Þ
βit ¼ 0
i ¼ 3 þ CRP; …; I þ CRP; t ¼ 1; …; CRP ð52Þ  mði  CRPÞðt  CRPÞ Þ  ð1  CT  1 Þ þ TOL
i ¼ 3 þ CRP; …; I þCRP; t ¼ CRP þ 1; …; T ð64Þ
βit ¼ 0 i ¼ 1; …; 3 þ CRP  1; 8 t ð53Þ
In summary, the model presented in this paper contains three
CRP
ut ¼ ∑ vðI þ nÞt 8t ð54Þ types of constraints: completely new/original, common/general or
n¼0 extensions to existing models. For instance, terms in the objective
function are mostly new formulations introduced to the literature
xt A Z þ 8t ð55Þ of blood supply chain. Similarly, constraints (4), (5), (7), (9), (10)
and (51) are completely original. The formulation of capacity
r t ; ut A Z þ 8t ð56Þ
constraint (2) is common to the existing model that is described
yit A Z þ 8 i; t ð57Þ in [25]. Demand constraint (6) is formulated similar to the
inventory balance and daily demand constraints as in [25] and
mit ; vit ; βit A Z þ 8 i; t ð58Þ [29] respectively. Although models in [4,28,45,46] consider inven-
tory, wastage and shortage costs using different methodological
zit A f0; 1g 8 i; t ð59Þ approaches, the formulations of aforementioned costs presented
The objective function (38) captures the deterministic version in our paper differ from these models. However, we use the similar
of the objective function presented in (1). Similarly, it seeks to formulation as in [28] to model wastage cost. Finally, constraint
minimize the purchasing, inventory, wastage and shortage costs (47) is the extension of constraint (10).
during the planning horizon. Constraints (39)–(46) and (49) and
(50) are the deterministic formulations of Constraints (2)–(9) and
(11) and (12) respectively. Constraints (47) and (48) update end- 4. Computational study
period blood inventory levels for each age-group. Furthermore,
right hand-side of Constraint (48) only indicates the variable In this section, we present the data used in our analysis and
associated with returned blood units from assigned inventory discuss the numerical results obtained from above models using
since the hospital would not receive or keep blood units that are IBM ILOG CPLEX 12.1 on Dell OPTIPLEX 755 with 2.20 GHz CPU
expired. Constraints (51)–(53) are used to compute the number of and 2GB of RAM.
 S. Gunpinar, G. Centeno / Computers & Operations Research 54 (2015) 129–141 135

4.1. Data 2. To analyze the extension of stochastic integer programming

Table 4 summarizes the values of cost parameters that are used
in our models. Most of the cost parameters are obtained from the
literature as shown in [4,29,45,46]. Holding cost was obtained
from real data from a Regional Medical Center (RMC).
Demand distributions from [45] were used for estimating daily
demand values of platelet units. Two studies [45,46] provide daily
demand distribution of platelets at a hospital blood bank. Two
types of demand are considered in [45]. Only one type of demand
for weekly platelet production is considered in [46]. The mean
values of platelet demands obtained from Sanquin Blood Bank and
used in both references are 24, 16, 32, 16, 24, 0, 8 for Monday
through Sunday respectively. The demand data are assumed to be
purely random and distributed around the mean; thus, the authors
used a Poisson distribution in [45] and gamma distribution in [46].
In [12], daily fluctuations of red blood cells in terms of
percentages are provided for the months between April 2003
and March 2004 at Southampton Center. They are 100%, 93%, 56%,
59%, 44%, 18% and 17% for Monday through Sunday respectively
where 100% value relates to the day of the week with highest
average number of units. Mean demand values for each day were
computed by multiplying percentages by 100.
Matlab 2010a was used to generate a total of 35 datasets for
numerical study. Three categories were selected to group them
based on the purpose they serve.
1. To test the stochastic integer programming model 15 datasets
of various sizes were generated. For each of the demand scenarios
in a given time period (t ¼30 days), data are generated from a
Poisson distribution using the mean values of platelet discussed
above. There are five datasets (F1–F5) with 4 scenarios, five
datasets (E1–E5) with 8 scenarios and five datasets (S1–S5) with
16 scenarios.
Table 4
Cost parameters.
Parameters Value Units Reference
Purchase cost (Platelet)-c 538 $/unit [29]
Purchase cost (RBC)-c 180 $/unit [4]
Shortage cost-b 1500 $/unit [46]
Wastage cost-w 150 $/unit [45]
Holding cost-h 1.25 $/unit*day RMC
model that considers the blood demand for two types of patients,
another 15 datasets of three groups with same sizes were
generated. For all groups in a given time period (t¼30 days),
scenario data were generated from a Poisson distribution for
8 scenarios. Five of these datasets (SY1–SY5) were generated
similar to the datasets in first category, but 1/8 and 7/8 of the
average daily demand was taken as the mean daily demand of
young and any platelet respectively. The remaining ten datasets
(SY6–SY10 and SY11–SY15) were generated taking 1/4 and 1/2 of
average daily demand as the mean daily demand of young platelet.
3. To evaluate the effect of C/T ratio and crossmatch release
period 5 datasets (D1-D5) were generated for each time period
(t¼ 10 days) from a Poisson distribution using the average demand
values of red blood cells discussed above.
4.2. Numerical results
The following four subsections detail experiments conducted
for the models described previously. Due to space limitation, for
each point discussed one figure (out of several possible scenarios)
is presented.
4.2.1. Solution times for all models
The problem described by (1)–(18) was solved with datasets
F1–F5, E1–E5, and S1–S5 using CPLEX 12.1. Fig. 3 summarizes the
solution times for solving these 15 problems given daily blood
center (platelet) capacity of 30 units and average platelet age of
4 days old in blood shipments. Solution time in Fig. 3 is the time it
takes CPLEX to solve the problem within an optimality gap which
are selected as 1.5% in the stochastic formulation and 2.5% in its
extension. In addition, the gap is chosen as 3.0% in the determi-
nistic formulation.
The average solution times of stochastic model by problem size
are 14.600, 83.419 and 449.837 s for datasets with 4, 8 and 16
scenarios respectively. As the number of scenarios increases, the
problem size grows and the solution time to reach the solution
increases.
Fig. 4 shows the solution times of the problem explained by
(38)–(59) given the average C/T ratio of 4/3 and the total blood
center (RBC) capacity of 120-110-80-80-60-40-40 units for Mon-
day through Sunday (where the beginning of the analysis period,
t¼1, starts with Monday). Length of CRP refers to the length of

Fig. 3. Solution times of datasets for stochastic model.

136 S. Gunpinar, G. Centeno / Computers & Operations Research 54 (2015) 129–141

Fig. 4. Solution times of datasets for deterministic model.

Table 5
The abbreviations of the daily capacity levels of the
blood center.

Abbreviation Daily capacity levels (units)

CAP1 20-10-30-10-20-5-5
CAP2 25-15-35-15-25-10-10
CAP3 30-20-40-20-30-15-15
CAP4 35-25-45-25-35-20-20
CAP5 40-30-50-30-40-25-25
CAP.E1 2-1-3-1-2-1-1
CAP.E2 3-2-4-2-3-1-1
CAP.E3 5-4-6-4-5-3-3
CAP.E4 7-6-8-6-7-5-5
CAP.E5 9-8-10-8-9-7-7
Fig. 5. Effect of blood center capacity on total expected shortages given average CAP.E6 12-11-13-11-12-10-10
platelet age of 4.5 in blood shipments. CAP.F1 4-2-6-2-4-1-1
CAP.F2 6-4-8-4-6-2-2
crossmatch release period (in days) for RBC units that is deter- CAP.F3 8-6-10-6-8-4-4
CAP.F4 10-8-12-8-10-6-6
mined by hospital policy. The average solution times of determi-
CAP.F5 13-11-15-11-13-9-9
nistic model by length of CRP are 265.651, 16.508 and 57.110 s. The CAP.F6 16-14-18-14-16-12-12
longer the crossmatch release period, the shorter the time CAP.T1 9-5-13-5-9-1-1
required to solve the problem. CAP.T2 12-8-16-8-12-4-4
CAP.T3 15-11-19-11-15-7-7
CAP.T4 18-14-22-14-18-10-10
CAP.T5 21-17-25-17-21-13-13
CAP.T6 24-20-28-20-24-15-15
4.2.2. Model 1 parameter effects
A variety of situations were considered to determine the
sensitivity of outcomes to the model parameters. We considered level does not have a major impact on the outcomes. Same
the effect of the average platelet age in blood shipments and daily behavior was noted for the total expected cost ($).
blood center capacity as well as the effect of unit shortage cost on It was noted that the average platelet age in blood shipments
the solution of the problem described by (1)–(18). This was done has some effect on both total cost and wastage levels at the
through the use of datasets E1–E5. The results are shown in Fig. 5. hospital (Fig. 6). Receiving older platelet units (average age of
In addition, the abbreviations shown in Table 5 are used in the rest 4.5 days) causes higher levels of wastage and increases total cost.
of the paper for the daily platelet capacity of the blood center Whereas raising age of platelet units in blood shipments from
(Monday through Sunday). 3.0 to 3.5 does not have an apparent impact. In the latter case, the
Capacity of the blood center allocated to the hospital has a changes in wastage levels show both downward and upward trend
noteworthy effect on the solution values. Since similar trend is depending on the dataset used for testing. However, when CAP1 or
observed for each given average platelet age (3.0, 3.5 and 4.5), the CAP2 capacity is selected, the total cost for the 3.5 average age is
numerical results associated with only one of them is discussed. higher than the 3.0 average age. For other capacity levels (CAP3,
Fig. 5 displays the effects of blood center capacity for given average CAP4 and CAP5), raising age of platelet units in blood shipments
platelet age of 4.5 in blood shipments. As the capacity is increased from 3.0 to 3.5 does not have any effect on the total cost.
to the CAP2 level, blood shortages related to blood operations The effects of unit shortage cost are analyzed in Fig. 7 as this
decrease significantly. However, increasing capacity beyond this may vary based on different patient types (trauma, heart and
 S. Gunpinar, G. Centeno / Computers & Operations Research 54 (2015) 129–141 137

Fig. 6. Effect of average platelet age.

elective surgeries). When the unit cost of shortage is higher, the

hospital carries more units on inventory. As a result of this, the
number of shortages at the hospital is reduced. Increased inven-
tory levels result in higher wastage rates and increases total
expected cost at the hospital.

4.2.3. Model 2 parameter effects

Datasets SY1–SY15 were used to examine the effects of blood
center capacity for young platelets on the solution of the problem
discussed in Section 3.1.1. Fig. 8 shows the results of these trials.
The total capacity of blood center is assumed to be 30-20-40-20-
15-15 units for Monday through Sunday. Young Platelet Capacity Fig. 7. Effect of shortage cost given blood center capacity of CAP2 and average
platelet age of 3.5.
refers to the maximum amount of blood units that the blood
center has agreed to supply to the hospital. For given young
platelet demands, which are 1/8, 1/4 and 1/2 of total platelet As can be noted, young platelet capacity of blood center has a
demand, CAP.E1-E6, CAP.F1-F6 and CAP.T1-T6 capacities of young significant impact on the solution of the problem. Increasing the
platelet are selected in this analysis respectively. Total Expected capacity up to CAP:E3, CAP:F3 and CAP:T3 results in significant cost
π  r shows the expected number of young platelet units on savings and reduction of shortage levels at the hospital. Further
inventory used to satisfy platelet demand. increases from these capacity levels provide slight improvements on
138 S. Gunpinar, G. Centeno / Computers & Operations Research 54 (2015) 129–141
Fig. 8. Effect of blood center capacity for young platelet.
aforementioned outcomes. Nevertheless, average values of total
expected wastages for five datasets of given young platelet capacities
are presented. As anticipated, these values show, in general, a down-
ward trend when increasing young platelet capacity. When young
Fig. 9. Effect of crossmatch release period.
platelet capacity is higher, as a result of increasing availabilities for
young platelet units, the number of young platelet units that are used
to satisfy any platelet demand increases. Moreover, due to higher
purchase cost associated with young platelet units, when more young
platelet units are needed, the cost increases.
4.2.4. Model 3 parameter effects
The problem described by (38)–(59) was solved with datasets
D1–D5 and the results are displayed in Figs. 9 and 10 to show the
effects of different C/T ratios (4/3, 8/5, 2/1) and crossmatch release
periods (1, 2, 3 days) on outcomes. The total RBC capacity of blood
center is assumed to be 120-110-80-80-60-40-40 units for Monday
through Sunday and the average age of RBC units in blood
shipments is 39.5 days old. C/T Ratio refers to the average value
of C/T ratio in blood related procedures at the hospital.
Finally, the numerical study shows that the changes in C/T ratio
and length of CRP have significant effects on the solution values.
Longer length of CRP may lead a long stay of RBC units in reserved
inventory without being transfused. As a result, the lifetime of RBC
units diminishes and increased RBC wastages cause higher cost in
blood related operations (Fig. 9). In addition, higher C/T ratios
increase the number of RBC units returned to free inventory and,
at the same time, cause higher levels of wastages at the hospital.
The increase on returning units decrease the amount of RBC placed
to blood center and thus results in reduced total cost (Fig. 10).
5. Concluding remarks and opportunities for future work
In this paper, optimization models for hospitals were devel-
oped to manage blood resources more efficiently which will
ultimately result in both cost reduction and improved service to
hospitals' patients. The focus of this study is on red blood cells and
platelets as they have short lifetimes and are the most scarce
products among the whole blood components. Integer program-
ming techniques have been used to analyze the problems. Both
deterministic and stochastic models have been discussed. The
models explicitly account for the age of blood units on inventory
 S. Gunpinar, G. Centeno / Computers & Operations Research 54 (2015) 129–141 139

Fig. 10. Effect of C/T ratio.

and consider the demand for two types of patients, uncertain

demand rates and crossmatch-to-transfusion ratio. In addition, an
extensive computational study is provided to analyze the effects
of several factors such as average age of blood in blood shipments,
C/T ratio, the length of crossmatch release period, and capacity
levels of blood center. The obtained results will be beneficial to
hospital administrators and will aid in the process of determining
adequate order sizes to minimize shortage, wastage and total costs.
As mentioned earlier, 17% of platelet units that were collected
in 2004 were wasted (outdated) in the United States. In addition,
19.9% of platelet units received by Stanford University Medical
Center (SUMC) during January through April 2006 are outdated Fig. 11. The model described by (1)–(18) for given instances.
[3]. The age distribution of these units was 22.5%, 30.1%, 28.3% and
19.1% for 2, 3, 4 and 5 days old respectively. Thus, average age of results; that is, decrease wastage. Second, C/T ratio is incorporated
platelet units received by SUMC was 3.44 days old. Furthermore, into the model using hospital's average value. It would be valuable to
there was no capacity restriction for the blood center in SUMC explore impact on the model when C/T requirements for specific
case. Fig. 11(a) shows wastage rates of the model in terms of patient-groups (or procedures) are incorporated. Third, varying
percentages. Although the results highly depend on the average shortage cost depending on the type of the patient groups will be
age in shipments, the figure only listed the wastage results for the a worthwhile extension into our models. Fourth, Model 3 becomes
average platelet age of 3.5 days old in blood shipments given the computationally challenging in stochastic formulation as indicated
capacity of CAP5. Comparing this part of our test results with before. To designing an efficient solution procedure for the stochastic
SUMC case makes more sense since the parameters are more or problem, an algorithm that could help reduce the size of branch and
less the same. From the model results, we conclude that wastage bound tree and improve the runtime should be explored. An
rates can be reduced from 19.9% to 2.57% on average. In addition, additional consideration would be to expand the model to analyze
Fig. 11(b) and (c) displays average values of shortage and total cost a centralized system for the supply chain of blood products where a
for given capacity levels respectively. Increasing capacity levels number of hospitals and a blood center form a network. Current
from CAP1 to CAP3 reduces the shortages and total cost 91.43% and models where only tested using data for one hospital. Finally,
20.7% respectively. Changes beyond CAP3 level do not have a big considering multiple blood centers competing over the demand
impact on the outcomes. points will be an interesting study to analyze.
As most quantitative models addressing an applied problem, the
models presented in our study have several opportunities for
expansion. First, some hospitals use double crossmatching policy Acknowledgments
where same unit of blood is crossmatched for more than one patient.
We only consider single crossmatching policy. It is anticipated that The authors would like to thank Prof. Dawood Sultan and Prof.
incorporating a double crossmatch policy will further improve John Large from Public Health Department at University of South
140 S. Gunpinar, G. Centeno / Computers & Operations Research 54 (2015) 129–141

Florida for their help to introduce us to several hospital adminis- δðsÞ ðsÞ ðsÞ
it Z M  ðzði  1Þt  1Þ þ mit
trators in the Tampa Bay region in Florida. Additionally, we wish to i ¼ 3; 4; …; I; 8 ðsÞ; t ð76Þ
express our sincere appreciation to the reviewers for their valuable
comments and feedback throughout the submission process. zðiðsÞ 1Þt yit ¼ ψ ðsÞ i ¼ 3; 4; …; I; 8 ðsÞ; t
it

ψ itðsÞ r zðiðsÞ 1Þt  M i ¼ 3; 4; …; I; 8 ðsÞ; t ð77Þ

Appendices: Linearization techniques
ψ itðsÞ r yit i ¼ 3; 4; …; I; 8 ðsÞ; t ð78Þ
The model formulations indicate two types of non-linear terms
thus two different linearization techniques that are exact approx- ψ itðsÞ Z M  ðzðsÞ
ði  1Þt 1Þ þyit
imations will be explained in this section.
i ¼ 3; 4; …; I; 8 ðsÞ; t ð79Þ

Appendix A. First linearization technique zðiðsÞ 1Þt vðsÞ ðsÞ

ði  1Þðt  1Þ ¼ μði  1Þt i ¼ 3; 4; …; I; 8 ðsÞ; t
μðsÞ ðsÞ
ði  1Þt r zði  1Þt M i ¼ 3; 4; …; I; 8 ðsÞ; t ð80Þ
The first linearization technique is focused on the interactions
between binary and discrete variables and assigns new discrete μðsÞ ðsÞ
ði  1Þt r vði  1Þðt  1Þ i ¼ 3; 4; …; I; 8 ðsÞ; t ð81Þ
variables to replace the products of interacting variables.
As in [50], y is called the linearization variable and reflects the μsði  1Þt Z M  ðzsði  1Þt  1Þ þ vsði  1Þðt  1Þ
products of x and d in the linearization process where x is a i ¼ 3; 4; …; I; 8 ðsÞ; t ð82Þ
discrete variable and d is a binary variable. Lower bound and
upper bound of x are assumed to be known and take the values of γ ðsÞ ðsÞ ðsÞ ðsÞ þ
it ; αit ; ψ it ; μði  1Þt A Z i ¼ 3; 4; …; I; 8 ðsÞ; t ð83Þ
L and U respectively. Then, integer programming formulation after
linearization process is as follows:
λðsÞ ðsÞ
it ; δit A Z
þ
i ¼ 3; 4; …; I; 8 ðsÞ; t ð84Þ
Ld r yr Ud
As both stochastic and deterministic models indicate nonlinear
Lð1 dÞ r ðx  yÞ rUð1  dÞ
terms with the same interacting variables (for example, v and z), in
order to save some space, we show their linearization in one
To linearize our models, linearization technique mentioned formulation and differentiate them using (s) at the top corner of
above is applied to the non-linear terms and the interactions the variables.
between binary variable z and discrete variables v, y and m in the
original formulation are replaced with their products (called
linearization variables) as shown below. Furthermore, the follow- Appendix B. Second linearization technique
ing constraints (that are numbered) are added into the new
formulation. The second linearization technique focuses on the floor func-
tion (x ¼ ⌊yc) used to determine the number of blood units
zðsÞ ðsÞ
it vði  1Þðt  1Þ ¼ γ itðsÞ i ¼ 3; 4; …; I; 8 ðsÞ; t returned to unassigned inventory as follows:
γ ðsÞ ðsÞ
it rzit  M i ¼ 3; 4; …; I; 8 ðsÞ; t ð65Þ x Z y 1 þTOL
x r yþ TOL
γ ðsÞ ðsÞ
it rvði  1Þðt  1Þ i ¼ 3; 4; …; I; 8 ðsÞ; t ð66Þ xAZþ
yZ 0
γ ðsÞ ðsÞ ðsÞ
it ZM  ðzit  1Þ þ vði  1Þðt  1Þ When the second linearization technique is applied to Con-
i ¼ 3; 4; …; I; 8 ðsÞ; t ð67Þ straint (51), the following set of constraints are obtained and used
to replace Constraint (51).
zðsÞ ðsÞ
it yit ¼ αit i ¼ 3; 4; …; I; 8 ðsÞ; t
βit Zððvði  CRP  1Þðt  CRP  1Þ yði  CRPÞðt  CRPÞ Þ  zði  CRPÞðt  CRPÞ
αðsÞ ðsÞ
it r zit  M i ¼ 3; 4; …; I; 8 ðsÞ; t ð68Þ
 mði  CRPÞðt  CRPÞ Þ  ð1  CT  1 Þ  1 þTOL
i ¼ 3 þ CRP; …; I þCRP; t ¼ CRP þ 1; …; T
αðsÞ
it r yit i ¼ 3; 4; …; I; 8 ðsÞ; t ð69Þ
βit rððvði  CRP  1Þðt  CRP  1Þ yði  CRPÞðt  CRPÞ Þ  zði  CRPÞðt  CRPÞ
αðsÞ ðsÞ
it Z M  ðzit  1Þ þ yit
 mði  CRPÞðt  CRPÞ Þ  ð1  CT  1 Þ þ TOL
i ¼ 3; 4; …; I; 8 ðsÞ; t ð70Þ i ¼ 3 þ CRP; …; I þCRP; t ¼ CRP þ 1; …; T

zðsÞ ðsÞ ðsÞ

it mit ¼ λit i ¼ 3; 4; …; I; 8 ðsÞ; t
Parameter TOL is a small number and critical in the lineariza-
λðsÞ ðsÞ
it rzit M i ¼ 3; 4; …; I; 8 ðsÞ; t ð71Þ tion process of Constraint (51). Without indicating this parameter,
β may take an incorrect value when the inner term (inside the
λðsÞ ðsÞ
it rmit i ¼ 3; 4; …; I; 8 ðsÞ; t ð72Þ floor function) of the right hand-side value in Constraint (51) is an
integer value. Even after the second linearization technique is
λðsÞ ðsÞ ðsÞ
it ZM  ðzit  1Þ þ mit applied, the resulting constraints still indicates non-linear terms
i ¼ 3; 4; …; I; 8 ðsÞ; t ð73Þ due to the interactions between binary and discrete variables.
When the first linearization technique is applied, the following
zðsÞ ðsÞ ðsÞ
ði  1Þt mit ¼ δit i ¼ 3; 4; …; I; 8 ðsÞ; t constraints are obtained as replacement for Constraint (51).
δitðsÞ r zðiðsÞ 1Þt  M i ¼ 3; 4; …; I; 8 ðsÞ; t ð74Þ βit Zððγ ði  CRP  1Þðt  CRP  1Þ αði  CRPÞðt  CRPÞ Þ
 mði  CRPÞðt  CRPÞ Þ  ð1  CT  1 Þ  1 þTOL
δitðsÞ r mðsÞ
it i ¼ 3; 4; …; I; 8 ðsÞ; t ð75Þ i ¼ 3 þ CRP; …; I þCRP; t ¼ CRP þ 1; …; T
 S. Gunpinar, G. Centeno / Computers & Operations Research 54 (2015) 129–141
βit rððγ ði  CRP  1Þðt  CRP  1Þ  αði  CRPÞðt  CRPÞ Þ
mði  CRPÞðt  CRPÞ Þ  ð1  CT  1 Þ þ TOL
i ¼ 3 þ CRP; …; I þ CRP; t ¼ CRP þ 1; …; T
References
[1] Jennings JB. Blood bank inventory control. Manag Sci 1973;19(6):637–45.
[2] Basnet RB, Lamichhane D, Sharma VK. A study of blood requisition and
transfusion practice in surgery at BIR hospital. Postgrad Med J 2009;9(2):14–9.
[3] Fontaine MJ, Chung YT, Rogers WM, Sussmann HD, Quach P, Galel SA, et al.
Improving platelet supply chains through collaborations between blood
centers and transfusion services. Transfusion 2009;49(10):2040–7.
[4] Nagurney A, Masoumi AH, Yu M. Supply chain network operations manage-
ment of a blood banking system with cost and risk minimization. Comput
Manag Sci 2012;9(2):205–31.
[5] van Zyl GJJ. Inventory control for perishable commodities. Dissertation,
University of North Carolina; 1964.
[6] Nahmias S. Perishable inventory theory: a review. Oper Res 1982;30
(4):680–708.
[7] Prastacos GP. Blood inventory management: an overview of theory and
practice. Manag Sci 1984;30(7):777–800.
[8] Belien J, Force H. Supply chain management of blood products: a literature
review. Eur J Oper Res 2012;217(1):1–16.
[9] Kamp C, Heiden M, Henseler O, Seitz R. Management of blood supplies during
an influenza pandemic. Transfusion 2010;50(1).
[10] Pereira A. Blood inventory management in the type and screen era. Vox Sang
2005;89(4):245–50.
[11] van Dijk N, Haijema R, van der Wal J, Sibinga CS. Blood platelet production: a
novel approach for practical optimization. Transfusion 2009;49(3):411–20.
[12] Katsaliaki K. Cost-effective practices in the blood service sector. Health Policy
2008;86:376–87.
[13] Rytila JS, Spens KM. Using simulation to increase efficiency in blood supply
chains. Manag Res News 2006;29(12):801–19.
[14] Madden E, Murphy EL, Custer B. Modeling red cell procurement with both
double red cell and whole blood collection and the impact of European travel
deferral on units available for transfusion. Transfusion 2007;47(11):2025–37.
[15] Katsaliaki K, Brailsford SC. Using simulation to improve the blood supply
chain. J Oper Res Soc 2007;58(2):219–27.
[17] Kopach R, Balcioglu B, Carter M. Tutorial on constructing a red blood cell
inventory management system with two demand rates. Eur J Oper Res
2008;185(3):1051–9.
[18] Mustafee N, Taylor SJE, Katsaliaki K, Brailsford S. Facilitating the analysis of a
UK national blood service supply chain using distributed simulation. Simula-
tion 2009;85(2):113–28.
[19] Alfonso E, Xie X, Augusto V, Garraud O. Modeling and simulation of blood
collection systems. Health Care Manag Sci 2012;15:63–78.
[20] Kaspi H, Perry D. Inventory systems of perishable commodities. Adv Appl
Probab 1983;15(3):674–85.
[21] Jagannathan R, Sen T. Storing crossmatched blood: a perishable inventory
model with prior allocation. Manag Sci 1991;37(3):251–66.
[22] Pierskalla WP, Roach CD. Optimal issuing policies for perishable inventory.
Manag Sci 1972;18(11):603–14.
[23] Prastacos GP. Optimal myopic allocation of a product with fixed lifetime. J
Oper Res Soc 1978;29(9):905–13.
[24] Blake JT. On the use of operational research for managing platelet inventory
and ordering. Transfusion 2009;49(3):396–401.
[25] Hemmelmayr V, Doerner KF, Hartl RF, Savelsbergh WP. Vendor managed
inventory for environments with stochastic product usage. Eur J Oper Res
2010;202(3):686–95.
[26] Sahin G, Sural H, Meral S. Locational analysis for regionalization of Turkish Red
Crescent blood services. Comput Oper Res 2007;34(3):692–704.
141
[27] Jacobs DA, Silan MN, Clemson BA. An analysis of alternative locations and
service areas of American Red Cross facilities. Interfaces 1996;26(3):40–50.
[28] Sapountzis C. Allocating blood to hospitals from a central blood bank. Eur J
Oper Res 1984;16(2):157–62.
[29] Ghandforoush P, Sen TK. A DSS to manage platelet production supply chain for
regional blood centers. Decis Support Syst 2010;50(1):32–42.
[30] Prastacos GP, Brodheim E. PBDS: a decision support system for regional blood
management. Manag Sci 1980;26(5):451–63.
[31] Hirsch RL, Cazal P. Recommendations of the task force on record-keeping and
blood distribution systems. Vox Sang 1981;40(3):216–21.
[32] Hirsch RL, Brodheim E. Blood distribution systems and the exchange of
information between hospital blood banks and regional blood centers. Vox
Sang 1981;40(3):239–44.
[33] Bosnes V, Aldrin M, Heier HE. Predicting blood donor arrival. Transfusion
2005;45(2):162–70.
[34] Haijema R, van Dijk N, van der Wal J, Sibinga CS. Blood platelet production
with breaks: optimization by SDP and simulation. Int J Prod Econ 2009;121
(2):464–73.
[35] Custer B, Johnson ES, Sullivan SD, Hazlet TK, Ramsey SD, Murphy EL, et al.
Community blood supply model: development of a new model to assess the
safety, sufficiency, and cost of the blood supply. Med Decis Mak 2005;25
(5):571–82.
[36] Delen D, Erraguntla M, Mayer RJ, Wu CN. Better management of blood supply-
chain with GIS based analytics. Ann Oper Res 2011;185:181–93.
[37] Erickson ML, Champion MH, Klein R, Ross RL, Neal ZM, Snyder EL. Manage-
ment of blood shortages in a tertiary care academic medical center: the Yale-
New Haven Hospital frozen blood reserve. Transfusion 2008;48(10):2252–63.
[38] Heddle NM, Liu Y, Barty R, Webert KE, Whittaker S, Gagliardi K, et al. Factors
affecting the frequency of red blood cell outdates: an approach to establish
benchmarking targets. Transfusion 2009;49(2):219–26.
[39] Perera G, Hyam C, Taylor C, Chapman JF. Hospital blood inventory practice: the
factors affecting stock level and wastage. Transfus Med 2009;19(2):99–104.
[40] Carden R, DelliFraine JL. An examination of blood center structure and hospital
customer satisfaction: What can centralized and decentralized blood centers
learn from each other?. Health Mark Q 2005;22(3):21–42.
[41] Novis DA, Renner S, Friedman R, Walsh MK, Saladino AJ. Quality indicators of
blood utilization: Three College of American Pathologists Q-Probes studies of
12,288,404 red cell units in 1639 hospitals. Arch Pathol Lab Med 2002;126
(2):150–6.
[42] Denesiuk L, Richardson T, Nahirniak S, Clarke G. Implementation of a
redistribution system for near-outdate red blood cell units. Arch Pathol Lab
Med 2006;130(8):1178–83.
[43] Davis R, Geiger B, Gutierrez A, Heaser J, Veeramani D. Tracking blood products
in blood centres using radio frequency identification: a comprehensive
assessment. Vox Sang 2009;97(1):50–60.
[44] Glynn SA, Busch MP, Schreiber GB, Murphy EL, Wright DJ, Tu Y, et al. Effect of
a national disaster on blood supply and safety: The September 11 Experience. J
Am Med Assoc 2003;289(17):2246–53.
[45] Haijema R, van der Wal J, van Dijk NM. Blood platelet production: optimiza-
tion by dynamic programming and simulation. Comput Oper Res 2007;34
(3):760–79.
[46] Zhou D, Leung LC, Pierskalla WP. Inventory management of platelets in
hospitals: optimal inventory policy for perishable products with regular and
optional expedited replenishments. Manuf Serv Oper Manag 2011;13
(4):420–38.
[47] Kendall KE, Lee SM. Formulating blood rotation policies with multiple
objectives. Manag Sci 1980;26(11):1145–57.
[48] Shulman IA, Stevens WT, Lezin EMS. A recent study from University of Texas
Southwestern. California Blood Bank Society Web. 〈http://www.cbbsweb.org/
enf/2001/freshwb.html〉 2004; Accessed 6 October 2012.
[49] Olawumi H, Bolaji B. Blood utilization in elective surgical procedures in Ilorin.
Trop J Health Sci 2006;13(1):15–7.
[50] FICOs Xpress Optimization Suite. MIP formulations and linearization quick
reference. FICO Decision Management Community Forum. 〈http://resear
ch-rules.fico.com/2011/01/mip-formulations-and-linearizations-quick-refer
ence.html〉 2009; Accessed 6 October 2012.