Using Simulation for Strategic Blood Supply Chain Design in the

Canadian Prairies

John Blake

1,2

and Ken McTaggart

Department of Industrial Engineering, Dalhousie University, Halifax, Nova Scotia, Canada

Centre for Innovation, Canadian Blood Services, Ottawa, Ontario, Canada

Keywords: Blood Supply Chain, Facilities Location, Transportation and Logistics, Simulation Modelling.

Abstract: Since 2010, Canadian Blood Services has been modernizing its facility infrastructure. Current plans call for

the amalgamation of production sites in the Prairie region by 2019. Under this plan existing production centres

in Alberta and Saskatchewan will be consolidated into a single site in Calgary. Because of the potential impact

to the distribution network, a simulation model of the logistics network was constructed in Visual Basic.Net,

using an established simulation framework. Experiments were conducted to estimate the robustness of the

network under varying assumptions for delivery interruptions and inventory reserves. Results suggest that,

given reasonable assumptions on road network reliability, product demand, and inventory staging, either no,

or very modest, changes to product wastage and product reliability should be expected after facility. This

work demonstrates the application of a generic simulation modelling framework to resolve important policy

questions.

1 INTRODUCTION

Canadian Blood Services is one of two organizations

in Canada whose mission is to manage the supply of

blood and blood products. In 2013/14, Canadian

Blood Services distributed more than 800,000 units

of red blood cells and 114,000 units of platelets. The

cost of operations in 2013/14 was $1.02B (Canadian

Blood Services, 2014).

Since 2010, Canadian Blood Services has been

modernizing its facility infrastructure. It has replaced,

or will replace, 14 local production and testing centres

with two national testing laboratories, and three

regional production and distribution centres, while

retaining four local sites in remote locations. Plans

call for the amalgamation of production sites in the

Prairie region of the country (Alberta and

Saskatchewan), which will consolidate local sites in

Edmonton, Calgary, and Edmonton. Once facilities

are consolidated, blood will either be shipped directly

to customers from Calgary or via two distribution

centres and stock holding units (SHU) to be located

in Regina and Edmonton. See Figure 1.

Consolidation of local facilities into regional hubs

allows for economies of scale, increased process

standardization, and improved productivity through

Figure 1: Map of facilities and associated customer

enhanced equipment utilization and greater

production intensity. However, facility consolidation

is not without critics, particularly in provinces that

lose a local centre; debate over the consolidation of

services generally garners considerable attention in

the popular press (CanadaEast.com, 2011) and

amongst political parties of all stripes. Accordingly,

to address stakeholder concerns, a simulation based

study of the proposed Prairie distribution network

was built. The model was constructed in Visual

Basic.Net, using an established simulation

framework (Blake and Hardy, 2014). The model

Blake, J. and McTaggart, K.

Using Simulation for Strategic Blood Supply Chain Design in the Canadian Prairies.

DOI: 10.5220/0006041703450352

In Proceedings of the 6th International Conference on Simulation and Modeling Methodologies, Technologies and Applications (SIMULTECH 2016), pages 345-352

ISBN: 978-989-758-199-1

345

reproduces the processes of blood supply chain

management in the Prairies, including collection of

raw materials, production and testing, distribution,

order arrivals, order completion and dispatch. In

addition, it incorporates hospital order, inventory, and

transfusion practices. The framework includes

routines to simulate both planned and unplanned

closures of the distribution network. The model was

validated against a series of benchmarks using both

qualitative assessments and statistical tests. Once

validated, the model was used to conduct a series of

experiments in which red blood cell (RBC) outdates

and availability were measured under differing

assumptions regarding the level of inventory held at

the SHUs and reliability of the distribution network.

2 LITERATURE REVIEW

There is an extensive operational research literature

dealing with blood and blood products. Osorio,

Brailsford, and Smith (2015) note that the blood

supply chain has motivated researchers since the

1960’s. Much of the early literature in the field

focused on inventory management policies for red

blood cells. The work in blood formed the basis for

much of the theoretical development of perishable

inventory theory. See Nahmias (1982) for a seminal

review.

Over the years, the literature has expanded from

pure inventory policy into the broader range of issues.

Beliën and Forcé (2012) classify the literature along

multiple dimensions of blood product, solution

method, system hierarchy, supply chain type, and

modelling structure.

Strategies for designing network topology have

been extensively studied. Pierskalla (2004) presents

a review of models for determining the number, size,

and location of regional blood distribution nodes.

Brodheim and Prastacos (1979) use a statistical

analysis to create a piece-wise linear relationship

between target inventory and daily demand.

Prastacos (1981) compares myopic policies based on

immediate data with optimal policies and shows that

myopic policies cannot be “too different” from

optimal.

Transhipment and rotation policies have been

studied by Gregor, Forthofer, and Kapadia (1982)

who show that lower outdates and greater product

availability are associated with these policies when

compared to traditional no-redistribution policies.

Katsaliaki and Brailsford (2007) describe the use of a

large scale simulation model to evaluate a blood

supply chain. They show that for a single producer

and single consumer system, the amount of inventory

stored can be reduced if improved ordering and cross-

matching policies are implemented. Lang (2010)

uses simulation based optimization to set study

inventory policy for a two-echelon system consisting

of a single supplier and seven hospitals in which

transhipment is allowed along with product

substitution.

While planning for inbound and outbound

logistics is frequently addressed in the literature,

fewer consider the operational impact of disruptions

to the delivery network. Sha and Huang (2012)

describe a location-allocation model to site donor

collection facilities in Beijing following an

earthquake. Jabbarzadeh et al. (2014) present a

robust version of the Sha and Yue model, which

incorporates terms for both cost and undersupply.

However, neither the Sha and Ye model, nor the

Jabbarzadeh et al. models explicitly consider failure

of transportation links, nor do they model period to

period inventory decisions or the aging of a

perishable product. Perhaps most applicable to this

problem are studies by Blake and Hardy (2013) which

describe a simulation methodology to evaluate

inventory decisions in a regional network subject to

periodic failures of the logistics network and a later

paper (Blake and Hardy, 2014) which details the

development of a generic simulation framework for

modelling regional blood networks. Recently Blake,

Hardy, and McTaggart published a simulation study

of a blood supply chain subject to period delivery

failures, but that study was limited to evaluation of a

single stock holding unit (Blake, Hardy, and

McTaggart, 2015)

We conclude that while there is an established

literature on location/allocation problems in blood

supply chains, there are few papers describing

methods for evaluating network logistics under

operational conditions. There are no papers, that we

are aware of, in which a generic simulation

framework has been applied to an instance of a large

regional blood supply distribution network with

several production and distribution hubs and hundred

hospitals linked via a logistics network subject to

periodic failures. We note, finally, that proof of

operational implementation of blood supply chain

study results is generally absent from the existing

literature.

3 PROCESS DESCRIPTION

Hospital customers in Alberta and Saskatchewan, and

are presently serviced from three production and

SIMULTECH 2016 - 6th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

346

distribution sites located in Calgary, Edmonton, and

Regina. After consolidation, blood will continue to

be collected from both fixed and mobile collection

sites in regions centred on Calgary, Edmonton, and

Regina. In the new network, however, whole blood

will be consolidated at a regional site and then

transported for processing in Calgary. Production

and testing activities are expected to require between

one and two days to complete and thus blood is

expected to become available for release as follows:

Table 1: Expected daily distribution of units becoming

available for release.

Day Units/Year Units/Day

Sunday 19,642 377.73

Monday 23,466 451.27

Tuesday 19,326 371.66

Wednesday 26,005 500.10

Thursday 36,676 705.31

Friday 34,167 657.06

Saturday 24,586 472.81

Total 183,868 3,535.92

The blood type profile of units collected in the

Prairie region is expected to be as defined in Table 2

Table 2: Expected distribution of blood units collected in

the Prairie region.

Blood

Type

Units/Year Units/Day

A- 12,569 34.53

A+ 54,493 149.71

AB- 1,250 3.43

AB+ 5,841 16.05

B- 3,878 10.65

B+ 16,763 46.05

O- 21,456 58.95

O+ 67,618 185.76

Total 183,868 505.13

After production and testing are complete, units

are released for distribution. Following consolidation

of production in Calgary, some customers will be

serviced directly from Calgary, while others will

receive blood products from their regional stock

holding unit. SHUs are expected to function in two

modes – on most days the SHU will operate as a local

distribution hub for materials transhipped from

Calgary. However, on days during which shipments

cannot be completed from Calgary, the SHUs will

serve as a forward store for all customers in an area.

Since some large customer sites in Edmonton and

Saskatoon will be supplied from Calgary, rather than

their regional SHU, the anticipated volume of

demand to be met at the regional SHU depends on the

status of the network. For this study, it is assumed

that the regional SHUs will hold sufficient inventory

to meet the greater of 6 days of regular demand from

hospitals in their regular catchment area or 3 days’

emergency demand from all hospitals in their

catchment area.

Table 3: Demand and inventory levels at supplier sites.

(Reg = regular daily demand, Emer = daily demand during

a delivery network failure, and Inv = target inventory).

Type Calgary Edmonton Regina

Reg Emer Inv Reg Emer Inv Reg Emer Inv

A-

17.7 7.3 142 4.8 13.2 44 3.4 5.5 17

A+

82.8 36.9 662 22.6 59.5 203 20.2 29.2 88

AB-

1.3 0.5 11 0.3 0.9 3 0.3 0.6 2

AB+

7.4 2.3 60 1.1 5.3 16 2.2 3.2 10

B-

4.5 1.6 36 0.6 3.0 9 1.0 1.5 5

B+

26.4 11.6 211 5.0 16.8 51 4.0 7.0 21

O-

30.4 16.8 243 11.5 22.5 103 6.3 8.8 27

O+

107.5 45.8 860 28.3 78.3 255 21.6 33.3 100

3.1 Simulation Framework

The blood distribution network in the Prairie

provinces is represented by a simulation model

derived from a generic framework developed by

Blake and Hardy (2014) Within the framework a

fixed sequence of events is assumed to occur daily.

The production and distribution site (supplier), the

SHUs (SHU), and hospitals (consumers) are

modelled as separate software classes. Each class has

a series of properties that define the state of the object

and a set of methods that can be called to change or

update the object’s state. The objects are linked

together through a simulation control algorithm. This

algorithm implements a special case of the next-

event, time-advance inventory model outlined in Law

(2006) in which a set of events are executed

sequentially and a single, daily update is made to the

simulation clock.

This implementation of the generic framework

assumes that one or more distribution centres and

several SHUs exist within a network of consumer

objects. The supplier object contains methods that

simulate the process of collecting, producing,

inventorying, aging and distributing blood to SHUs

and customers. SHU objects are implemented as a

sub-class of supply objects; they inherit supplier

methods that allow for inventory ordering, product

receipt, product aging, and distribution of products to

consumers. Unlike the supplier object, SHU objects

Using Simulation for Strategic Blood Supply Chain Design in the Canadian Prairies

347

do not collect their own products. Rather they are

“connected” to a supplier object at which they place

orders. Operationally, SHUs are assumed to function

like distribution centres: They hold stock and fill

requests for products from hospitals in their

catchment area. According to operational plans in the

Prairie region, some hospital consumers will be

supplied directly from the Calgary DC during regular

day-to-day operations. However, in the event of an

interruption to the logistics network, all consumer

sites within a region will order and receive product

from their local SHU.

Hospitals, in the framework, are modelled as a

consumer object that encapsulates methods for

ordering, receiving, inventorying, and aging blood, in

addition to methods for simulating patient demand.

3.2 Process Description

At the beginning of each run, the system is initialized.

All objects are instantiated and their properties are set

according to data read in from a transaction database.

Initialization is completed by assigning a starting

inventory, by blood group and type, to each supplier

object, based on the anticipated demand for products

from consumers and any associated SHUs.

Once initialized, the model is run for some

replications of a specified number of days. On each

simulated day, the model steps through a fixed

sequence of events. The day begins with a call to

advance suppliers’ inventory. This ages the stock on

hand at each supplier by one day and causes any stock

with -1 days of shelf-life remaining to be outdated; a

similar call is then made to advance inventory at all

SHU objects.

After advancing the inventory age, both the

supplier object and SHU objects make a call to have

inventory arrive. The supplier object samples from a

day-of-week specific distribution to determine the

number of units to arrive from testing. Each unit is

assigned a blood group and type and a shelf-life

drawn from empirical distributions. Reductions in

the rated shelf-life of arriving units are primarily

intended to represent delays in the testing process, but

are also used to represent mandated reductions in

shelf-life when units are irradiated.

SHUs are assumed to evaluate their inventory

each morning against a two-level (s,S) inventory

policy and place an order for product, rather than

observing a randomly distributed product collection.

Thus, each morning, every SHU object evaluates its

inventory. If the inventory level for a particular blood

type is less than s, an order is placed with the supplier

to have exactly enough stock (S-s, where S≥s) arrive

to return the inventory level to S. The supplier is

assumed to fill requests for product using a FIFO

inventory policy in most instances. Inventory is

assumed to be delivered instantaneously from the

supplier to the SHU.

Once all incoming inventory is in place at

suppliers and SHUs, the simulation loops through

each of the consumer objects and makes a call to

advance the inventory. This causes the stock on hand

to age by one day. Any units with -1 days of shelf-

life remaining are counted as outdated and exit the

system. Each consumer object then determines if an

order is required, in a manner similar to that of the

SHU. Consumers are assumed to order from either

from their regional SHU or directly from the Calgary

DC. In instances where a consumer object is

associated with a SHU, but regularly receives

products from the DC, orders are usually placed with

the DC. However, when an emergency order is

issued, or if the logistics network has been disrupted,

all consumers are assumed to place an order with their

local SHU.

Once orders are received and entered into

inventory, customer sites experience demand for

blood products from patients. Each day the model

issues a call to each consumer object. The call

generates requests for blood, using a zero-inflated

Poisson distribution with a day-of-week specific

mean value. Once the number of units required is

known, blood group and type are assigned to the

demand items via empirical distributions. Demand is

filled at the consumer FIFO from available units. If

no unit is available, the consumer site issues a

demand for additional units, on an emergency basis,

from its supplier (either the SHU or the DC). If

available, emergency units are transferred to the

consumer object, using the same logic as regular

demand. If no unit is available, the demand is

considered to be lost and counted as a shortage. The

day ends, the simulation clock is then advanced by

one day and the cycle repeats.

4 DATA

Data for the simulation was derived from Canadian

Blood Services’ operational database which provided

transaction level data for all units of packed red cells

collected in, distributed in, or disposed from any CBS

facility in Alberta and Saskatchewan during fiscal

2013/2014. A total of 277,000 records were

retrieved.

The transaction level data was processed using a

set of custom routines to format the data and to

SIMULTECH 2016 - 6th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

348

prepare a set of pre-defined input data lists and

distributions for the regional simulation framework

(Blake and Hardy, 2014). Automatically derived

from the data are simulation inputs such as lists of

CBS facilities and hospitals as well as distributions

describing hospital demand, whole blood collection,

and product testing parameters, amongst other

elements.

5 VA L I D AT I O N

Extensive verification activities were undertaken to

ensure the model functions as intended. Once inputs

were verified, the model was validated by comparing

simulated output against historical data.

5.1 Verification

At a macro level, the distribution network is similar

to a queuing network with random arrivals

(collections) and random services (demand).

Drawing upon that analogue, the key elements

dictating system performance are inputs (collections)

and outputs (filled demand or outdates)

5.1.1 Verifying Inflows

Inflows of materials within the simulation model are

comprised primarily of collections.

Table 4 provides

a summary of model results for collections against the

historical value. This table provides a 95% prediction

interval, based on 10 replications of the simulation

model for a 10-year period, following a 364-day

warm up. The results suggest that there is no

evidence that historical data is inconsistent with the

simulation framework.

Table 4: Comparison of daily collections by blood group

within the Prairie region.

Blood

Group

Daily Collections Prediction Interval

Historic

Value

Average Variance

Lower

Limit

Upper

Limit

A- 34.557 0.010 30.547 38.566 34.530

A+ 149.66 0.035 141.316 158.004 149.706

AB- 3.525 0.000 2.244 4.805 3.434

AB+ 16.042 0.003 13.31 18.774 16.047

B- 10.681 0.001 8.452 12.91 10.654

B+ 46.076 0.022 41.446 50.705 46.052

O- 58.915 0.014 53.679 64.15 58.945

O+ 185.774 0.120 176.478 195.07 185.764

5.1.2 Verifying Outflows

Outflows of materials in the simulation model consist

of items provided to patients.

Table 5 provides a

summary of model results for demand, as observed in

the simulation model, against the historical value for

fiscal 2013/2014. This table provides a 95%

prediction interval, again based on 10 replications of

the simulation model for a ten-year period, following

a 364-day warm up period. The results show that

there is no evidence to suggest that historical demand

data is not consistent with the results of the simulation

model.

Table 5: Comparison of daily demand by blood group

within the Prairie region.

Blood

Group

Daily Demand Prediction Interval

Historic

Value

Average Variance

Lower

Limit

Upper

Limit

A- 26.025 4.6E-03 22.546 29.505 26.066

A+ 125.884 6.5E-02 118.231 133.536 125.824

AB- 2.0992 8.6E-04 1.111 3.088 1.901

AB+ 10.994 2.4E-03 8.732 13.255 10.739

B- 6.042 1.5E-03 4.365 7.719 6.104

B+ 35.254 6.6E-03 31.204 39.304 35.404

O- 48.020 1.0E-02 43.294 52.747 48.228

O+ 157.869 5.7E-02 149.299 166.439 157.885

5.2 Validation

To confirm that the model represents reality, output

values for outdates were compared against historical

values. Since outdates are not an input to the

simulation, but rather a result of the differences

between simulated inflows and outflows, a basic test

of validity is to ensure that the number of outdating

units matches the historical record. The simulation

model was run for ten replications of 10-year’s

duration, using a 364-day warm-up period under the

method of batch means. The average number of

outdates for RBC was recorded and a 95% prediction

interval was constructed. A comparison, shown in

Table 6, indicates there is no reason to suggest that

outdates recorded by the simulation model are not

consistent with those observed in the 2013/14

historical data.

Table 6: Comparison of model outdates within the Prairie

region as recorded in the simulation against historical data.

Blood

Group

Daily Outdates Prediction Interval

Historic

Value

Average Variance

Lower

Limit

Upper

Limit

A- 0.293 0.004 0.000 0.098 0.063

A+ 1.898 0.100 2.865 5.685 4.578

AB- 0.015 0.000 0.000 0.259 0.058

AB+ 4.275 0.008 3.148 6.078 4.841

B- 0.074 0.001 0.242 1.524 0.258

B+ 4.613 0.049 1.074 3.027 1.211

O- 0.883 0.002 0.000 0.098 0.063

O+ 2.050 0.002 2.865 5.685 4.578

Using Simulation for Strategic Blood Supply Chain Design in the Canadian Prairies

349

6 EXPERIMENTS AND RESULTS

The simulation framework assumes the distribution

system will function as a network of stock holding

units, each linked to the Calgary DC. The SHUs in

Edmonton and Regina will provide regular shipments

to some facilities and emergency shipments to all

customers in their catchment area. Accordingly, the

function of the SHUs is to provide regular deliveries

and to serve as a buffer against network interruptions.

A basic test of technical feasibility therefore involves

evaluation of network operations as delivery

interruptions are introduced into the system.

6.1 Delivery Interruptions

Road network reliability data was obtained from the

Alberta Department of Transportation. The data covers

the period between 03 Mar 13 and 04 Jan 16 and

describes closures to the Trans-Canada Highway. In

total, the data shows 23 road closures over 34 months.

Road closures were observed to occur approximately

once every 54 days. However, of the 23 closures

recorded in the dataset, 20 involved a closure of less than

7 hours’ duration, while the remaining three resulted in

an average closure duration of 9.46 hours. Thus, in the

simulation framework it is assumed that the time

between road failures is exponentially distributed with a

mean of 54 days and that 87% (20/23) of all road failures

result in a delay that is too short to disrupt deliveries.

6.2 Experimental Framework

All scenarios for the simulation model of the Prairie

region assume the existence of the Calgary distribution

centre and stock holding units located in Edmonton and

Regina. Deliveries to SHUs are assumed to take place

six days per week (Tuesdays through Sundays).

Inventory policies at all ordering nodes in the

model (SHUs, and customer sites) are assumed to

follow an (s, S) type inventory policy. In all runs of

the model it is assumed that facilities review their

inventory daily, placing an order for stock on any day

of the week that deliveries are feasible. Inventory

targets for the three SHUs are as defined in

Table 3

and, for hospital customers, it is assumed that S is

equal to 6 days’ of average patient demand.

It is assumed that every hospital is connected to a

local SHU. Some larger facilities in Edmonton and

Regina may bypass their regular SHU and instead

order directly from Calgary for their day-to-day

needs. Nevertheless, in the event of a disruption to the

delivery network, all hospitals are assumed to draw

from their local SHU.

Finally, it is assumed that the mean time between

network road failures is 54 days, exponentially

distributed, but that only 13% of the failures result in

a significant delay to deliveries.

Two parameters were varied in the experimental

framework:

1) The mean time to repair the delivery link

between Calgary and the SHUs. Repair time was

tested at 9.46, 15.46, and 21.46 hours to repair,

with all times to repair assumed to be

exponentially distributed.

2) The amount of inventory held at the SHUs. In

the base case, it is assumed that the amount of

inventory held at the SHUs is as defined in

Table

. In experiments, the amount of inventory held

is adjusted by -1, 0, +1, or +2 days’ demand to

determine the impact of safety stock at the SHUs

to buffer out disruptions to the logistics network.

7 RESULTS

Table 7 shows the results of the experiments in terms

of units outdated per day. Results are shown as the

mean time to recover from a network failure ranges

from 9.46 hours to 21.46 hours and as SHU

inventories are varied from -1 days’ demand on hand

to +2 days’ demand on hand from the base case.

Table 7: Average daily number of RBC units outdated vs.

mean time to road failure and changes to SHU inventory.

Change in SHU Inventory

MTTR -1 0 1 2

0.394 14.392 14.720 15.637 25.158

0.644 14.643 14.389 15.598 25.155

0.894 14.756 14.535 15.905 25.105

The results in Table 7 indicate outdates are not

affected by the mean time to recover the network after

a road failure. However, the results suggest that

increases to the inventory held at the SHU increases

the amount of outdating in the system. This

conclusion is supported by an analysis of variance

(ANOVA) as shown in

Table 8.

Table 8: ANOVA for units wasted per day against time to

recover from a network failure (MTTR) and changes in the

SHU inventory.

Analysis of Variance for Daily Wastage

Source DF SS MS F P

MTTR 2 0.04 0.02 0.75 0.51

Change in

SHU

Inventory

3 236.08 78.7 3246.35 0.00

Error 6 0.15 0.02

Total 11 236.26

SIMULTECH 2016 - 6th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

350

Table 9: Average daily number of units short vs. mean time

to road failure and change to SHU inventory.

Change in SHU Inventory

MTTR -1 0 1 2

0.394 5.96E-03 4.67E-03 2.18E-02 1.98E-02

0.644 4.95E-03 5.66E-03 2.02E-02 2.16E-02

0.894 7.45E-03 6.04E-03 2.22E-02 2.14E-02

Table 9 shows that product shortages increase as

the mean time to recover from a network failure

increases. In all cases, however, shortages were

observed to be rare events, with instances occurring

less than once per 45 days to once per every 168 days,

depending on the scenario. Reductions in the SHU

inventory below baseline values, did not significantly

change shortage rates. However, when SHU

inventory was increased, a significant increase in

shortage was observed. In the simulation, product

substitutions are explicitly disallowed. Thus, if

demand appears for a particular blood type and none

is available at the customer hospital, its associated

SHU, or at the supplier site in Calgary, the demand is

lost. The majority of shortages recorded in the

simulation are due to requests for rarer AB- and B-

blood. Shortages of these blood types occur when

additional inventory is held at the SHUs since this

policy sequesters more of these rare types to the

SHUs and away from the DC in Calgary.

Table 10

shows an ANOVA which indicates that road recovery

time does not significantly influence shortages, but

that changes in the SHU inventory level does.

Table 10: Average daily number of RBC units short vs.

mean time to road failure and change to SHU inventory.

Analysis of Variance for Daily Shortage x 100

Source DF SS MS F P

MTTR 2 0.0388 0.0194 2.56 0.157

Change in SHU Inventory 3 7.1018 2.3673 313.16 0.000

Error 6 0.0454 0.0076

Total 11 7.1859

8 CONCLUSIONS

Based on the results of the experiments, it may be

concluded that consolidation of production and

distribution facilities in the Prairie region will result

in either no, or very modest, change to product

wastage and shortage, given the assumptions on road

network reliability, product demand, and inventory

staging made in this study.

If one assumes the same types of operational

structures as in the simulation experiments, it would

be expected that 14.92 +/- 1.08 units per day of RBC

would be wasted, while product shortages would be

in the range of 0.01 +/- 0.008 units per day prior to

consolidation; after consolidation, and assuming the

most likely scenario for road closures, it is expected

that 14.72 +/- 0.43 units per day of RBC will be

wasted, while there will be 0.004 +/- 0.002 instances

of shortage per day. Since these results are not

statistically different from the base case, it may be

concluded that, under the most likely scenario for

network reliability, there should be no discernible

changes in product availability or system wastage

after consolidation.

Experimental results to test the effect of decreased

network reliability and changes in inventory staging

suggest that the network performance is reasonably

robust. Wastage rates were observed to increase from

14.4 units per day to 25.2 units per day as inventory

at the SHUs in Edmonton and Regina were varied

from their baseline amounts by -1 to +2 days of

demand, but were unaffected by changes in the road

network reliability. In all runs tested, shortages

ranged from 4.7x10

-3

units per day to 2.2x10

-2

units

per day, or roughly 1.7 to 7.8 units per year.

Shortages were observed to increase modestly as the

road network reliability decreased. However,

shortages were also observed to increase as more

inventory is held at the SHUs and less is held at

Calgary.

Overall, the results of the simulation experiments

suggest that, within the range of likely network

failures, the impact on customer service resulting

from facility consolidation in the Prairie region is

likely to be negligible.

9 DISCUSSION

The value of modelling changes in the blood

distribution network in the Prairie provinces extends

beyond a proof of technical feasibility. In Canada,

the provision of health care is a provincial

responsibility; provinces are responsible for the

regulation, function, and funding of health care.

Blood, because it is a biologic product, is federally

regulated. Thus, while provinces pay for the

provision of blood services, they are disallowed from

directing the operations of blood agencies. Issues can

arise when changes are suggested to blood

distribution networks, particularly when such change

involves the loss of facilities within a province.

Concerns over the loss of facilities sometimes leads

stakeholders to question whether the revised network

Using Simulation for Strategic Blood Supply Chain Design in the Canadian Prairies

351

will provide a similar level of service to that of the

existing network (Blake and Hardy, 2013).

To address stakeholder concerns and, as due

diligence for patient safety, Canadian Blood Services

has found it useful to develop detailed simulation

models of its regional networks. The models have

been instrumental in establishing proof of concept

and forecasting operational robustness. In this paper

we have reported on the development of a specific

instance of a simulation model created from a generic

framework to represent changes in the distribution

network in the Prairie provinces of Canada. This

model, and its derivatives, have been used to address

a specific series of policy questions and has served as

a vehicle for fostering discussions between the blood

agency and stakeholder groups in the Canadian

Prairie Provinces.

REFERENCES

Belien, J. and Force, H. (2012). Supply chain

managementof blood products: A literature review.

European Journal of Operational Research, 217(1), 1-

16.

Blake, J. (2012). A case study on the use of operations

research to evaluate changes in a blood supply chain.

Proceeding of the 4th Interanational Conference on

Information Systems, Logistics, and Supply Chain.

Quebec : ILS.

Blake, J., and Hardy, M. (2013). Using simulation to

evaluate a blood supply network in the Canadian

maritime provinces. Journal of Enterprise

Managements, 26((1/2)), 119-134.

Blake, J., and Hardy, M. (2014). A generic modelling

framework to evaluate network blood management

policies: The Canadian Blood Services experience.

Operations Research for Healthcare, 3(3), 116-128.

Blake, J., and Hardy, M. (2014). Evaluating the impact of a

shorter shelf life for red blood cells with a generic

simulation. In Y. Guan, and H. Liao (Ed.), Proceedings

of the 2014 Industrial and Systems Engineering

Research Conference. Montreal, QC: IIE.

Blake, J., and Hardy, M. (2015). Modelling exchanges in a

national blood system. Proceedings of the First

Karlsruhe Service Summit Workshop - Advances in

Service Research (pp. 73-80). Karlsruhe: Karlsruhe

Service Research Institute. doi:10.5445/KSP/100

0045634.

Blake, J., Hardy, M., and McTaggart, K. (2015). Modelling

a blood distribution network in the Prairies with a

generic simulation framework. INFOR, 53(4), 194-201.

Brodheim, E., and Prastacos, G. (1979). The Long Island

blood distribution system as a prototype for regional

blood management. Interfaces, 9(5), 3-20.

CanadaEast.com. (2011, June 27). NB to ensure top quality

of blood supply. Retrieved from CanadaEast.com:

http://timestranscript.canadaeast.com/newstoday/articl

e/1411744.

Canadian Blood Services. (2014). Delivering Value:

Annual Report 2013-2014. Ottawa, ON: Canadian

Blood Services.

Gregor, P., Forthofer, R., and Kapadia, A. (1982). An

evaluation of inventory and transportation policies of a

regional blood distribution system. European Journal

of Operational Research, 10(1), 106-113.

Jabbarzadeh, A., Fahimnia, B., and Seuring, S. (2014).

Dynamic supply chain network design for the supply of

blood in disasters: A robust model with real world

application. Transportation Research Part E: Logistics

and Transportation Review, 70, 225-244.

Katsaliaki, K., and Brailsford, S. (2007). Using simulation

to improve the blood supply chain. Journal of the

Operational Research Society, 58(2), 219-227.

Katsaliaki, K., Mustafee, N., T. S., and Brailsford, S.

(2009). Facilitating the analysis of a UK National Blood

Service supply chain using distributed simulation.

SIMULATION, 85(2), 113-128.

Lang, J. (2010). Blood bank inventory control with

transshipments and substitutions. In M. Beckmann, H.

Kunzi, G. Fandel, and W. Trockel, Production and

Inventory Management with Substitutions (pp. 205-

226). Berlin: Springer.

Law, A. (2006). Simulation Modeling and Analysis, 4th

Edition. New York: McGraw-Hill.

Nahmias, S. (1982). Perishable inventory theory: a review.

Operations Research, 30(4), 680-780.

Osorio, A., Brailsford, S., and Smith, H. (2015). A

structured review of quantitative models in the blood

supply chain: a taxonomic framework for decision

making. International Journal of Production Research.

doi:10.1080/00207543.2015.1005766.

Pierskalla, W. (2004). Blood banking supply chain

management. In M. Brandeau, F. Sainfort, and W.

Pierskalla, Operations Research and Health Care (pp.

103-146). Boston: Klewer.

Prastacos, G. (1981). Allocation of a perishable product

inventory. Operations Research, 29(1), 95-107.

Sha, Y., and Huang, J. (2012). The multi-period location-

allocation problem of engineering emergency blood

supply systems. Systems Engineering Procedia 5, 21-

28.

SIMULTECH 2016 - 6th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

352