Citation
Mathematical programming model for optimal design of sustainable vaccine supply chain in developing countries

Material Information

Title:
Mathematical programming model for optimal design of sustainable vaccine supply chain in developing countries
Creator:
Mousavi, Bahador ( author )
Language:
English
Physical Description:
1 electronic file (97 pages) : ;

Subjects

Subjects / Keywords:
Vaccines -- Developing countries ( lcsh )
Business logistics -- Mathematical models ( lcsh )
Materials management -- Mathematical models ( lcsh )
Inventory control -- Mathematical models ( lcsh )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Review:
The enormous health and economic damage of epidemic diseases and the introduction of new vaccines, such as pneumococcal and rotavirus, - which need significantly higher cold storage capacity and other infrastructures - make designing a sustainable vaccine supply chain essential. The health policy makers in developing countries as well as low-and middle income- countries should be aware of the current and future needs of immunization infrastructure based upon population growth rate and the introduction of these vaccines. Thus the availability of a model with the capability to recognize and design the vast vaccine infrastructure expansion is critical for health policy makers to realize future immunization plans. Vaccine supply chain is a large, inter-continental network with enormous cost and environmental implications. Therefore the design of not only a cost effective supply chain but also an environmental friendly one, is important. This Study addresses the optimal design and planning of vaccine supply chain based on environmental and economic objectives. The environmental objective is measured by a life cycle assessment methodology. With this approach, the environmental emissions associated with all echelons of vaccine supply chain, including vaccine manufacturing and packaging, transportation, cold chain and disposal was measured from a cradle to grave perspective. Further, the economic objective was evaluated by considering the total investment cost of each echelon of the supply chain. In order to optimize the vaccine supply chain by considering environmental and economic issues simultaneously, a multi-objective continuous linear mathematical programming model is developed that accounts for major characteristics of the vaccine supply chain, including geographical and industrial diversity of suppliers, various shipment types, different refrigeration options, multiple disposal methods, vaccine demand distribution, cold chain and transportation capacity. The resulting Pareto-optimal curves display the tradeoff between the environmental and economic dimensions of vaccine supply chain.
Thesis:
Thesis (M.S.) - University of Colorado Denver.
Bibliography:
Includes bibliographic references
System Details:
System requirements: Adobe Reader.
General Note:
Department of Civil Engineering
Statement of Responsibility:
by Bahador Mousavi

Record Information

Source Institution:
|University of Colorado Denver
Holding Location:
|Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
945985514 ( OCLC )
ocn945985514

Downloads

This item has the following downloads:


Full Text
MATHEMATICAL PROGRAMMING MODEL FOR OPTIMAL DESIGN OF
SUSTAINABLE VACCINE SUPPLY CHAIN IN DEVELOPING COUNTRIES
By
BAHADOR MOUSAVI
B.Sc., Mechanical Engineering at Yazd University
A thesis submitted to the
Faculty of the Graduate School of the
University of Colorado in partial fulfillment
of the requirements for the degree of
Master of Science
Civil Engineering Department
2015


This thesis for the Master of Science degree by
Bahador Mousavi
has been approved for the
Civil Engineering Program
by
Arunprakash Karunanithi, Advisor
Bruce Janson, Chair
Azadeh Bolhari
November 20th, 2015


Mousavi, Bahador (MS, Civil Engineering)
Mathematical Programming Model for optimal Design of Sustainable Vaccine Supply
Chain in Developing Countries
Thesis directed by Associated Professor Arunprakash Karunanithi.
ABSTRACT
The enormous health and economic damage of epidemic diseases and the introduction of
new vaccines, such as pneumococcal and rotavirus, which need significantly higher cold
storage capacity and other infrastructures make designing a sustainable vaccine supply
chain essential. The health policy makers in developing countries as well as low-and
middle income- countries should be aware of the current and future needs of
immunization infrastructure based upon population growth rate and the introduction of
these vaccines. Thus the availability of a model with the capability to recognize and
design the vast vaccine infrastructure expansion is critical for health policy makers to
realize future immunization plans. Vaccine supply chain is a large, inter-continental
network with enormous cost and environmental implications. Therefore the design of not
only a cost effective supply chain but also an environmental friendly one, is important.
This Study addresses the optimal design and planning of vaccine supply chain based on
environmental and economic objectives. The environmental objective is measured by a
life cycle assessment methodology. With this approach, the environmental emissions
associated with all echelons of vaccine supply chain, including vaccine manufacturing
and packaging, transportation, cold chain and disposal was measured from a cradle to
grave perspective. Further, the economic objective was evaluated by considering the total
investment cost of each echelon of the supply chain. In order to optimize the vaccine


supply chain by considering environmental and economic issues simultaneously, a multi-
objective continuous linear mathematical programming model is developed that accounts
for major characteristics of the vaccine supply chain, including geographical and
industrial diversity of suppliers, various shipment types, different refrigeration options,
multiple disposal methods, vaccine demand distribution, cold chain and transportation
capacity. The resulting Pareto-optimal curves display the tradeoff between the
environmental and economic dimensions of vaccine supply chain.
The form of and content of this abstract are approved. I recommend its publication.
Approved: Arunprakash T. Karunanithi
IV


ACKNOWLEDMENTS
It is difficult to explain how much I owe to my wife. Leila Aghili, who sacrificed her
career and took care of everything during my MS. I specially want to thank my mother,
father and brother for their love and support. They have supported every decision in my
life and encouraged me to pursue my MS. This thesis is dedicated to them.
I would like to thank my advisor, Professor Karunanithi, for guiding my research.
Finally, I want to thank my friends and colleague who provided me with this career
recommendation and inspiration.
v


TABLE OF CONTENTS
1- INTRODUCTION.............................................................1
1.1- Motivation.........................................................1
1.2- Contribution.......................................................6
2- IS SOLAR REFRIGERATION ALWAYS A SUSTAINABLE OPTION......................8
2.1- Introduction......................................................8
2.2- Process description and system boundaries........................10
2.3- Life cycle inventories 12
2.4- Impact assessment................................................15
2.5- Uncertanity......................................................19
2.6- Normalized damage analysis.......................................20
2.7- Sensitivity analysis.............................................22
2.8- Geographical delineation of impacts..............................24
2.9- Discussion.......................................................27
3- COMPARATIVE El OF ALTERNATIVE TECHNOLOGIES IN VACCINE
SUPPLY CHAIN ECHELONS......................................................31
3.1- Goal and scope...................................................31
3.2- Vaccine manufacturing and packaging..............................31
3.3- Vaccine transportation...........................................33
3.4- Off-grid refrigeration alternatives..............................35
3.5- Process description and system boundaries........................35
3.6- Faunctional unit.................................................38
3.7- LCI of absorption refrigeration systems..........................38
vi


3.8- Impact assessment....................................................39
3.9- Results and discussion...............................................40
3.10- Medical waste treatment system......................................44
3.10.1 -Incinerati on.................................................44
3.10.2-Autoclaving....................................................45
3.11. Refrence Flow.......................................................46
4-MATHEMATICAL PROGRAMMING MODEL FOR OPTIMAL DESIGN OF
SUSTAINABLE VACCINE SUPPLY CHAIN...............................................48
4.1- Motivation...........................................................48
4.2- Introduction.........................................................48
4.3- Problem formulation..................................................49
4.4- Mathematical model...................................................52
4.4.1- Production, refrigeration and transportation capacities........52
4.4.2- EI quantification methodology..................................55
4.4.3- Cost objective function........................................59
4.5- Computational experiments............................................60
4.6- Parameter setting....................................................61
4.7- Vaccine Supply Chain in Ghana........................................61
4.7.1- Network superstructure.........................................61
4.7.2- Results........................................................63
4.7.3- Other objective function.......................................65
4.7.3.1-Open vial waste..............................................65
4.7.4- Percentage of vaccine met demand...............................66
vii


4.7.5-Contribution of each echelon in cost and El of vaccine SC..68
4.8- Time schedualing for vaccine transportation.......................70
4.8.1-Optimization resulats based on different objective functions.72
4.9- Conclusion........................................................73
REFRENCES..............................................................75
APPENDIX...............................................................81
viii


LIST OF TABLES
Table
ECharacteeristics of the vaccine refrigeration technologies ...... 11
2:Distribution of human health impact for SPSB refrigerator in ghana ..............25
3 features and characters of various off-grid vaccine refrigerators................36
4:EI of various off-grid vaccine refrigerators.....................................41
5:EI and cost of alternative medical waste treatment technologies..................46
6:Ghana immunization program for under five childeren and pregnant women...........62
Table 7: contribution of life cycle C02e emission and cost of various echelons of vaccine
SC.................................................................................69
8:Comparison of current and optimum vaccine shipment frequency.....................71
IX


LIST OF FIGURES
Figures
1: Schematic picture of vaccine supply chain.......................................4
2: Solar power supplimented with battery vaccine refrigerator......................11
3: Cradle to grave system boundary for vaccine refrigerator.........................12
4: Comparetive El of solar and conventional refrigerator in Ghana...................16
5: Unit process with breakdown of GWP impact for the case of Ghana .................17
6: Comparative impacts of 3 refrigeration options for the use in South Africa ......18
7 Comparison El of EP used in Ghana and South Africa .............................19
8 Monte-Carlo simulation results for HH cancer, HH non cancer and ecotoxicity impacts
...................................................................................20
9: Normalized impacts of SPSB vaccine refrigeration system.........................21
10: Comparative impacts based on various parameters................................24
11: Transition plot for net GHG reductions ........................................28
12: Dominant sources of electricity generation in African Countries................29
13: Comparative GWP impact of manufacturing and packaging of one dose of various
types of vaccine....................................................................33
14: Comparative GWP of transportation alternatives..................................34
15: Fossil fuel and SAR refrigerator system boundary................................38
16: GWP breakdown for various refrigeration technologies...........................43
17: Total cost of various refrigeration systems for functional unit................44
x


18: Incineration process system boundary............................................45
19: Autoclaving process system boundary.............................................46
20: Vaccine SC reference flow.......................................................47
21: Schematic graph of vaccine network super structure..............................51
22: Vaccine supply chain map for Ghana..............................................62
23: Efficient set of Pareto solutions for the vaccine SC in Ghana...................63
24: Network configuration for multi-objective optimization..........................65
25:3D Pareto front for multi-objective optimization (OWV as a third objective function)
.....................................................................................67
26: 3D Pareto front for multi-objective optimization (Percentage of vaccine met demand
as a third objective function).......................................................68
27: Effect of different objective functions on C02e..................................72
28: Effect of different objective function on total cost............................73
xi


NOMENCLATURES
Sets and Descriptions
indices
ter Time periods of the investment horizon, t=l, T
j e/ Potential vaccine supplier sites, j=l, J
keK Potential vaccine UNICEF storages, k=l, K
xEX Potential vaccine production technologies, x=l, X
l E L Potential vaccine national Storages, 1=1, L
i E / Potential vaccine shipment modes from UNICEF storages to national Levels i=l, I
m e M Potential vaccine regional level storages, m=l,... .M
nn e NN Potential number of clinics nn=l,... .NN
ss e SS Potential vaccine refrigeration technologies in clinics, ss=l,... .SS
0 E 0 Potential number of disposal facilities, o=l,... .0
ffEFF Potential disposal methods, ff=l,... .FF
b E B Emissions and waste substances(burdens),b=l,... .B
e E E Environmental impact indicator categories ,e=l,... .E
nE N Normalized impact categories, n=l,... .N
pe E PE Life Cycle assessment stages, pe=l,... .PE
Pen Classification of impact e under normalized category n
hEH Potential various types of vaccines, h=\,....H
Xll


Continuous Descriptions
variables
yhjkxt Production rate of liter of vaccine h by supplier j to UNICEF storage k with technology x during interval t [liter/month]
Zhklit Amount of liter of vaccine h which is shipped from UNICEF storage k to national store / with shipment mode i during interval t [liter/month]
Hhlmt Amount of liter of vaccine h which is shipped from national store / to regional store m during interval t [liter/month]
Vhmnnssst Amount of liter of vaccine h which is shipped from regional level m to clinic nn with refrigeration technology ss during interval t [liter/month]
Wfi nnofft Amount of syringes and needles which is disposed from clinic nn to disposal location o with disposal technology ff during interval t [Kg/month]
Dpe uet Environmental impact in terms of category e resulting from life cycle stage pe during interval^ [impact/month]
iipe nt Normalized environmental impact in terms of category n resulting from life cycle stage pe during interval t [points/month]
Xlll


Parameters Descriptions
pe Emissions inventory of burden b resulting from the unit reference flow of life stage pe
^be Characterization factor used to convert burden b into impact category e
Yn Normalization factor for impact categories belonging to group n
$n Weighting factor for each normalized impact category n
r1 Unit production cost of vaccine using technology x during interval t
cxt [$/liter vaccine]
r2 cjkt Unit transportation cost based on the distance between j and k nodes during interval t
r3 Unit ice production cost during transportation stage 2 in time t
ct [$/liter vaccine]
4 cit Unit refrigeration cost
r12 cfft Unit cost of disposing of immunization wastes [$/Kg of immunization waste]
RDkt Transportation capacity of UNICEF vaccine storage k during interval t
Pjt Production capacity of manufacturer j during interval t
Sit Refrigeration capacity of national level store / during interval t
SDlt Transportation capacity of national store 1 during interval t
Ti-mt Refrigeration capacity of regional level store m during interval t
XIV


TiDmt Transportation capacity of regional store m during interval t
Dnnt YF vaccine demand in clinic nn during interval t
DDnnf Transportation capacity of clinic nn during interval t
Qnnt Refrigeration capacity of clinic nn during interval t
Bi0t Capacity of disposal device located in location o during interval time t
BiDot Transportation capacity of disposal location o during interval t
ajkxt Distance between supplier j with technology x and UNICEF store k during interval t
aklit Distance between UNICEF store k and national store / with shipment mode i during interval t
rf'" ulmt Distance between national store / and regional store m during interval t
XV


CHAPTER 1
INTRODUCTION
1.1 Motivation
Epidemic diseases kill millions of people each year, especially in low and middle income
countries. Therefore, preventing the spread of these diseases, and controlling them, can
prevent significant social and economic damage. Vaccination is a very cost effective
method to control epidemic diseases, especially for children under the age of five.
Delivery of vaccines to developing countries is facilitated by a large intercontinental
supply chain. In addition, most vaccines must be maintained within a specific
temperature range during transportation. This requires a strictly controlled cold storage
infrastructure. It is obvious then that the supply chain management of pharmaceutical
products and especially of vaccines, is very important. This is why there is a great deal of
research that has been done and is on-going.
The introduction of new types of vaccines each year causes the vaccine supply chain to
change in each country on a frequent basis. New vaccines need more storage capacity,
and hence health policy makers need the flexibility to modify their immunization
infrastructure at any time as changes dictate. In the last few years, great advances have
been made in introducing new vaccines and expanding the reach of immunization
programs globally [1], New vaccines such as pneumococcal conjugate- for pneumonia-
and rotavirus -for diarrhea- are currently being rolled out or planned to be introduced
globally with a particular emphasis in low- and middle- income countries. These
1


programs are extremely important as 18% and 12% deaths of all children under the age of
five in low-income countries occur due to pneumonia and diarrhea respectively [2],
GAVI alliance is at the forefront of vaccine program plans to support 40 low- and
middle- income countries in introducing rotavirus vaccines and to immunize 50 million
children by 2015 [3], It is estimated that by 2013 about 2.4 million deaths of children can
be averted globally through accelerated introduction of rotavirus vaccines [4], Similarly,
GAVI also plans to introduce pneumococcal vaccines to nearly 60 countries potentially
averting 500,000 deaths by 2015 and up to 1.5 million deaths by 2020 [3]; unfortunately
the global numbers for death are significantly higher. In 2010, only a tiny fraction of the
0.5 billion dollars spent on vaccine programs in low- and middle-income countries has
been allocated to rotavirus and pneumococcal vaccines. It is projected that by 2015, half
of the total of 1.2 billion dollars will be spent towards these two vaccines alone [5], The
volume of a dose of these two vaccines is higher than other traditional vaccines, so
providing cold rooms with sufficient capacity for handling these vaccines through their
supply chain is a challenging task for public health officials in low- and middle- income
countries.
Vaccine supply chain (SC) is necessary for distributing vaccine from manufacturers to
the final consumers and it includes various steps such as production and packaging,
transportation, cold chain and disposal. Vaccine demand fluctuations with location and
time and an unfixed production rate of manufacturers are challenging points in this SC. In
the low- and middle-income countries, due to ascending birth rates, vaccine demand is
rapidly increasing. Equity and efficiency are two critical parameters that must be
considered in vaccine Supply Chain Management (SCM). This means that the vaccine
2


should be provided uniformly for each recipient in the system while the infectious
diseases are averted as much as possible. Chen [6] optimized the national level vaccine
SC based on maximizing fully immunized child and vaccine met demand. Lee et al. [7]
analyzed the impact of introduction of Pneumococcal and Rotavirus vaccines on total
vaccine demand of Niger and its consequences on Niger immunization infrastructure.
Assi et al. [8] conducted a study on the effects of removing one echelon of Niger vaccine
supply chain on vaccine met demand. Additionally, some models were developed to
forecast the spread of infectious diseases, and to forecast the special vaccine demand in
each region. For example, Ferrari et al. [9] developed a method to estimate the epidemic
effects of infectious deceases with the capability to predict the special vaccine demand in
a given location. In the past, SCM was used to design networks that led to the maximum
economic performance, but more recently, environmental impact issues have become one
of the significant concerns in SCM. Today, there are design methods and tools available
that can suggest and advance the most eco-friendly options in each echelon of a given
SC. These efforts show that integration of environmental issues with other SCM
objectives such as economic functions, could be an important topic of future research.
The implementation of environmental targets in SCM is commonly referred to as Green
SCM [10],Immunization supply systems of low- and middle- income countries are not
yet optimized to reduce environmental impacts. People in low- and middle-income
countries are most vulnerable to environmental, climate change and human health
impacts. It is proven that strategies to reduce greenhouse gas (GHG) emissions indirectly
result in wide ranging public health co-benefits so any efforts related to the use of a
sustainable infrastructure within vaccine supply chain can avert negative human health
3


impacts. Environmental impact (El) can be minimized in a vaccine supply chain in terms
of energy and resource efficiency and waste management For example it is possible to
reduce environmental impacts by replacing kerosene refrigerators or inefficient old
refrigerators with solar powered refrigerators, or by using thermostable vaccines that do
not require ice packs during their transportation. Selection of sustainable transportation
options such as electric vehicles or ocean freights can make the SC more eco-friendly
[11],By providing economic incentives for innovation in equipment manufacturing,
WHO and PATH encourage vaccine refrigerator production companies to use solar
technology for their future products. In order to initiate further new innovation and new
directions, WHO established new standards for refrigeration options which did not
previously exist such as direct drive solar refrigerators and stationary passive cooling
containers [11],
By implementing some funding and incentives for innovation in equipment
manufacturing, WHO and PATH encourage vaccine refrigerator production companies to
use solar technology for their future products. On the other hand, WHO published new
standards for products which had not previously existed like direct derive solar
refrigerators and stationary passive cooling containers [11],
Vaccine SC consists of several echelons, all of which have significant cost and
environmental burdens (Figure 1). Furthermore, there are several alternative options
available in each section of the SC. Therefore, designing and planning a sustainable
vaccine SC becomes important for immunization programs.
4


Figure 1: Schematic picture of Vaccine Supply Chain
Chapter 2 discusses whether solar vaccine refrigeration is always a sustainable option for
all African countries from a life cycle perspective. The electricity portfolio of each
country is analyzed and a criteria is provided for selection of solar refrigeration option
based on a countries electricity portfolio. In addition, the geographical distribution of
environmental impacts of the cold chain is established.
In chapter 3 we calculate the environmental impacts, pollutant emissions and energy use
of each echelon of this SC using Life Cycle Assessment methodology (LCA) from cradle
to grave and a comparative study is performed for different alternative technologies that
can be used in each part of the SC such as packaging, transportation, cold chain and
disposal. The most eco-friendly alternative technology in each part is recognized by
LCA, the method that considers all of the upstream processes.
In Chapter 4, Vaccine SC is simulated with a generic mathematical model. This modeling
is based on continuous linear programming (CLP). The ultimate goal is to optimize this
linear algebraic system to achieve the most economic and eco-friendly results. Therefore,
a multi-objective optimization approach, Non-Dominated Sorting Genetic Algorithm II
(NSGA II), which is a stochastic optimization methodology, was used within the
MATLAB programming language environment. The developed code is capable of
optimizing the vaccine SC not only for traditional objectives (such as vaccine met
5


demand, SC cost and open vial waste) considered by other researchers but also for
environmental objectives.
This model is applicable for designing a sustainable system by selecting the optimal
alternative technology in each part of SC and also for rearrangement of SC by optimal
scheduling time for vaccine shipment between various parts of SC. The significant
contributions of this research are presented below.
1.2 Contributions
Quantifying the environmental impacts of all SC echelons with a comprehensive
environmental assessment methodology (LCA)
Creation of geographically dependent methodology for establishing the
sustainability of solar vaccine refrigerators in developing countries.
Creation of a continuous linear programming optimization model for vaccine
supply chain which is solved using NSGA II.
Creation of a multi-objective optimization code which has the capability to
optimize the supply chain in an environmentally conscious way, in addition to the
traditional objectives associated with vaccine SC.
Designing new optimal solutions for SC infrastructure and adjusting the schedules
and rearrangement of shipment frequency of existing vaccine SC.
6


CHAPTER 2
IS SOLAR REFRIGERATION FOR VACCINE COLD STORAGE IN LOW AND
MIDDLE INCOME COUNTRIES ALWAYS A SUSTAINABLE OPTION? A
COMPARETIVE LIFE CYCLE ASSESSMENT STUDY
2.1 Introduction
An important component of vaccine programs is the cold chain, which is a collection of
cold rooms, refrigerators, freezers and cold boxes that keep vaccines at an appropriate
temperature. Vaccines require cold storage (between 2C and 8C) to remain stable and
be potent when administered. Heat sensitive vaccines lose potency after exposure to
temperatures above the recommended range and freeze sensitive vaccines lose potency
when exposed to freezing temperatures. The influx of new vaccines has outstripped the
capacity of current cold-chain systems in many low- and middle- income countries [1],
Successful implementation of these new vaccine programs require augmentation of
existing capacity with additional cold storage infrastructure as the new vaccines need
significantly higher storage volumes than traditional vaccines. For example, Rotavirus
vaccines on an average require a total storage volume of 85.95-334.3 cm3 per fully
immunized child while traditional vaccines (such as vaccines for measles) on an average
require only 15.4 cm3 [12], Therefore, in the next few years there is an urgent need to
streamline and expand efficient and reliable cold chain infrastructure in these countries.
An additional 390,000 refrigeration units will be required by 2015 to sustain introduction
of new vaccines in the 40 GAVI alliance countries alone. Refrigeration systems have
significant environmental impacts due to several factors: They are energy intensive
requiring significant amount of electricity or heat and leakage of refrigerants and sealing
7


agents have impact on global warming and ozone depletion. In this context, the
importance of building new cold chain infrastructure that minimizes the environmental
impact of energy, materials and processes both within the country and globally cannot be
overstated. This provides an opportunity, as replacement of existing infrastructure with
eco-friendly alternatives is not feasible in low- and middle- income countries due to the
costs involved, but eco-friendly options can be considered for the proposed new
expansions. This has been recognized by important stake holders and policymakers [1]
and there is a drive towards eco-friendly cold chain infrastructure, such as solar
refrigeration [13], in the developing world. It is believed that solar refrigeration offers
lower environmental footprint as it does not bum hydrocarbons -for electricity
production- and hence does not produce carbon dioxide. However, to legitimately
evaluate the sustainable features of solar vaccine refrigeration we need to consider the
life cycle impacts rather than just the use phase C02 emissions. In addition consideration
of health and ecological impacts is also critical. To our knowledge there has been no
study done to quantify the life cycle impacts of vaccine cold storage. Analysis specific to
vaccine refrigerators is very important as they are very different from domestic and
commercial refrigerators due to their unique characteristics, stringent design
specifications, constraints [14], and complex supply chains involved. In this study we
examine the life cycle environmental impacts of solar refrigeration systems for vaccine
storage and compare it with conventional vaccine refrigeration. In this study we examine
the life cycle environmental impacts of solar refrigeration systems for vaccine storage
and compare it with conventional vaccine refrigeration. This analysis will be critical for
stakeholders such as World Health Organization (WHO), Global Alliance for Vaccines
8


and Immunization (GAVI), Program for Appropriate Technology in Health (PATH),
United Nations International Children's Emergency Fund (UNICEF), Gates foundation,
and Governments and health agencies of developing countries who will be involved in
decision making processes related to immunization programs.
2.2 Process description and system boundaries
This analysis considers small scale off grid solar refrigeration systems and compares it
with electrical powered refrigerators for vaccine storage in Ghana, which introduced the
two new vaccines (Pneumococcal and Rotavirus vaccines) in 2012 and would require
augmentation of their current cold storage capacity to sustain their immunization
program. Also, Ghana is interested in solar refrigeration for vaccines. The following three
options are considered: a) Solar Powered Supplemented Battery (SPSB) [15]; b) Solar
Powered Battery Free (SPBF) [16]; and c) Electricity powered (EP) refrigerators [17],
WHO [2] approved common commercial vaccine refrigerators (See Figure 2 and Table 1)
were selected for this study. The life span time of SPSB and SPBF were considered as 15
years based on the study by Sayigh and MCViegh [18] while the life span of EP was also
fixed at 15 years based on the Energy Star report [19],
The system boundaries for the three refrigeration options are shown in Figure 3. The unit
processes included are: solar panel production (for SPSB, SPBF); battery production (for
SPSB); refrigerant (R134-a) production; blowing agent (HCFC141-b) production; other
materials production; refrigerator assembling; transport; grid electricity consumption (for
EP), refrigerant and blowing agent leakage during use and disposal.
9


(a) Figure 2: (a) Solar Powered Supplemented Battery Vaccine Refrigerator (SPSS),
the battery used as a electrical storage [15](b) Solar Powered Battery Free vaccine
Refrigerator (SPBF), this type uses ice thermal storage of energy instead of battery
(c) Electricity Powered Vaccine Refrigerator (EP)
Table 1: Characteristics of the vaccine refrigeration technologies
Technical Characteristics SPSB SPBF EP
Vaccine Storage net Capacity(Liter) 38.7 54.4 103
Shipping weight (kg) 91 118 175
Energy Consumption (12 volt) at ambient temperature of 43 C 0.61 kWh/day 0.86 kWh/day 2.6 kWh/day
Autonomy(the number of days that you need the system to operate when there is no power produced by PV panels) 7 days 7 days NA
Refrigerant R134-a R134-a R134-a
Manufacturing Location US US UK
Life span time (years) 15 15 15
10


2.3 Life cycle inventories
A functional unit of 1 liter cold storage capacity per year was selected. A life cycle
inventory (LCI) was compiled for the three different refrigeration options. Most of the
Figure 3- Cradle to grave system boundary for vaccine refrigeration systems
11


LCI data was secondary and were derived from literature sources and standard LCI
databases such as U.S. life cycle inventory database (USLCI) [20] and Ecoinvent [21]
with a particular focus on geographical specificity. All data which are reported in this
section are related to the selected functional unit.
The amount of refrigerant was calculated as 0.17g for all the types. [22], The LCI related
to refrigerant (R134a) production were derived from McCulloch and Lindley [23] which
provided the cradle-to-gate emission factors for refrigerant production. The process tree
for the refrigerant included production of the reactants (trichloroethane, hydrogen
fluoride, chlorine and alkali) needed for the refrigerant, production of the precursors
needed for these reactants (ethylene, chlorine & alkali, sulfuric acid and lime), and
extraction of raw materials (elementary flows) needed to produce these precursor
materials (crude oil, solution mined brine, fluorspar mined, sulfur mined and limestone
quarried). In addition, the direct global warming impact associated with leakage of the
refrigerant during the use and disposal phases was considered with the characterization
factor for R134a, fixed at 1357 kg C02e per kg of R134a [24,25] The amount of blowing
agent needed was estimated to be equal to 0.56 and 1.22 g for SPSB, SPBF, and EP
respectively [26], All LCI data related to blowing agent (HCFC141-b) production and
loss during installation were derived from Little et al [24] and the characterization factor
for leakage of HCFC141-b was 704 Kg C02e per kg of HCFC-141b. Twelve month
average solar irradiation for Ghana was estimated to be equal to 200 w/ m2 [27] .We
selected mono crystalline photovoltaic (PV) modules [28] and the number of modules
needed to meet the requirement of 5.75 kWh (at 12 volt and 43C ambient temperature)
for SPSB and SPBF was calculated as 0.18 for the selected functional unit. The ambient
12


temperature of 43C hottest day of Ghana- i.e. worst case scenario was considered for
design purposes. Cradle-to-gate emissions related to production of PV modules were
derived from Fthenakis al [29], To meet the requirement of 7 days of autonomy (the
number of days that the system needs to operate when there is no power produced by PV
panels) 0.74 AH deep cycle parallel lead acid batteries [30] (for SPSB) were considered.
Cradle-to-gate emissions related to battery manufacturing were derived from Sullivan et
al.[31], The battery is only used in the case of SPSB and for EP, it is assumed that the
cooler box is used during blackout. For the solar options, we considered emissions related
to truck and ocean freight transportation from California (manufacturing location) to
Ghana. This equated to 2.90 tkm and 2.56 tkm ground transportation through diesel
powered truck and 22.20 and 19.61 tkm ocean freight transportation for SPSB and SPBF
respectively for the selected functional unit. For the EP refrigerator we considered ground
transportation of 0.83 tkm from United Kingdom to Ghana. The assembly of refrigeration
units included door assembly, resin finishing, cabinet assembling and refrigeration cycle
assembly [32] and associated energy consumption was obtained through personal
communication with a vaccine refrigerator manufacturing company. [15] The total input
energy for assembling phase was 0.32, 0.34 and 0.30 kWh for SPSB, SPBF and EP
respectively. Emissions associated with the production of all other materials needed such
as resins, aluminum, copper etc. were derived from Boustani et al [33], and the total input
energies for their production was 1.28, 1.31 and 1.23 kWh for SPSB, SPBF and EP
respectively. Ghanas electricity profile (60 % hydro power and 40% combined cycle
natural gas and steam power) [34] was used to derive emission factors for electricity used
in the EP option [35], All emission factors related to grid electricity and natural gas
13


combustion for US and UK (manufacturing locations) were derived from USLCI [20] and
Ecoinvent [21] respectively. The total electricity consumption of EP was 9.21 kWh. We
assume that the entire refrigerant and blowing agent is lost to the atmosphere during the
use phase or at the end of life without any recovery. Emissions associated with the
disposal of the refrigerator unit were assumed to be negligible and recovery and recycle
of solar modules and refrigeration unit was not considered to be realistic in low-and
middle-income countries. The Complete Life Cycle inventory data for each unit process
are presented in appendix (See Tables Al, A2).
2.4 Impact assessment
Impact assessment methods are used to translate the inventory data into environmental
impacts through the use of characterization factors. In this study we adopted the impact
assessment methodology TRACI [36] developed by the U.S. Environmental Protection
Agency. The following impact categories were considered: (1) global warming; (2)
acidification; (3) eutrophication; (4) smog formation; (5) human health criteria; (6)
human health cancer; (7) human health non-cancer; and (8) ecotoxicity. In addition to
midpoint metrics we also considered damage indicators using TRACI 2.1 methodology
[37] for the limited purpose of normalized damage analysis and for total health impact
analysis.
This section summarizes the key findings for the case of Ghana. The total scores related
to all of the 9 impact categories for the three different options were calculated for the
functional unit of 1 liter cold storage/year .The impact profiles related to SPSB, SPBF,
and EP are shown in Figure 4, with SPSB impact set at 100% and the other two displayed
at a level relative to the former. Comparison between the three options show that in all of
14


the categories, except ozone depletion potential, the impact of SPSB was higher than
SPBF which in turn was higher than EP. The two solar refrigeration options had more
than two times higher climate impact than the conventional option. SPSB had two times
more human health- criteria impact (HH Criteria) than SPBF which implies that battery
production plays a big role in criteria emissions (PM10, PM2.5, NOx, SOx). Further
human health criteria for EP is five times less compared to both solar refrigeration
options. The human health cancer impacts of EP is insignificant in comparison to the
two solar refrigeration options. For human health non cancer and ecotoxicity categories,
the impacts of SPBF is only about 3% of the impacts of SPSB. This huge difference can
be attributed to emissions associated with lead acid battery production. The ozone
depletion indicator can entirely be attributed to the leakage of blowing agent in all three
options and has the following trend: EP > SPBF > SPSB (see Figure 5). With respect to
photochemical smog, acidification potential, and eutrophication potential the impacts
followed the same trend (SPSB > SPBF > EP) and were similar in relative magnitudes
(SPBF and EP had roughly 75% and 15% impact as that of SPSB).
Figure 4: Comparative impacts of solar (SPSB and SPBF) and conventional vaccine
refrigeration in Ghana
15


Figure 5-a expands on the above results by providing the breakdown of individual
contributions of different unit processes towards GWP for SPSB. We can see that mono
crystalline PV module production has by far the highest contribution towards GWP but in
the case of EP, electricity consumption of refrigerator during use phase has the highest
contribution towards GWP. Unit process contribution towards GWP for EP is shown in
Figure 5-b.
Refrigerant
Production and
Leakage
PV Module
Production
Battery
Production
Blowing Agent
Production and
Leakage
Mattreial
Processing
Assembly
Transportation
Figure 5: a) Unit process wide breakdown of GWP impact for the case of Ghana:
for SPSB b) Unit process wide breakdown of GWP impact for the case of Ghana
for EP
The findings in the previous section contradicts the general belief that solar refrigeration
systems are always more eco-friendly than conventional electric refrigeration. The results
suggest that life cycle impacts of EP systems depend greatly on the electricity generation
profile of the country where the refrigerator is used. In order to investigate the influence
of the electricity generation mix on the results we developed a scenario where we
compare the two refrigeration options for use in South Africa. The electricity production
profile of South Africa is very different from that of Ghana (93% coal and 7% nuclear)
[38], The solar irradiation in South Africa is 230 W/m2and the number of solar panel
SPSB
2.8
EP
I Refrigerant
Production and
Leakage
l Blowing Agent
Production and
Leakage
i Electricity
Consumption
During Use Phase
i Mattreial
Processing
l Assembly
Transportation
16


required remained the same. The results for South Africa- as seen in Figure 6 -suggests
that for most of the impact categories the electric powered refrigeration option results in
significantly higher impacts than solar refrigeration option. This finding is completely
opposite to the results of Ghana. We find that for most of the categories the Ghana
impacts are less than 20% of the South African impacts (Figure 7), This is a very
significant difference which suggests that solar refrigeration offers environmental
benefits in countries with a high contribution of fossil energy to their electricity mix and
electric refrigerators seem to be better in countries having relatively low contribution of
fossil energy to their electricity mix.
120
m
c£
m
tat)
a
e
O)
u
u
0)
CL
80
60
> 40
w
0)
a. 20
O
U

i
i J
L
1
ii min 11 ii
& ^





SPSB
SPBF
EP-South Africa
Figure 6: Comparative impacts of the three refrigeration options for use in South
Africa
17


Figure 7: Comparison of environmental impacts of EP used in Ghana and South
Africa
2.5 Uncertainty
The developers of USEtox model which is used to derive human health and ecotoxicty
characterization factors for TRACI caution about the use of the characterization factors
for metals due to the uncertainty present in the data. Therefore, a Monte Carlo simulation
was performed by varying the characterization factors (CF) of all metals by an order of
magnitude [39] Uniform distribution was considered for this analysis. The impact
categories associated with metal emissions include Human Health Cancer, Human Health
Non Cancer and ecotoxicity. The Figures 8(a),(b) and (c) show the mean value of the
impacts with error bars (uncertainty) from Monte Carlo simulation for the cases of SPSB
and SPBF used in Ghana (the EP health impacts are negligible). These findings indicate
the importance of metal emissions towards human health and ecosystem impacts. Many
heavy metals such as mercury, lead, chromium and nickel emissions during lead acid
18


battery production were the main cause of higher health impacts of SPSB in comparison
to SPBF.
8.00E-08
7.00E-08 -
g 6.00E-08 -
c
re
u 5.00E-08 -
D
H
^ 4.00E-08 -
$ 3.00E-08 -
X
a 2.00E-08 -
E
X 1.00E-08 -
0.00E+00
O
z
5.00E-06
4.00E-06
3.00E-06
2.00E-06
1.00E-06
0.00E+00
SPSB SPBF
(a)
(b)
5.00E+00
uj 4.00E+00
D
H
U
2 3.00E+00
>
o 2.00E+00
o
u
1.00E+00
O.OOE+OO
SPSB SPBF
(c)
Figure 8: Monte Carlo simulation results for human health cancer, human health
non-cancer and Ecotoxicity impacts
2.6 Normalized damage analysis
In this section the normalized impacts of total climate change, total human health and
total ecosystem quality are presented for SPSB option. Normalization allows us to
compare different types of impact categories, through a common unit (e.g. person year),
19


in order to identify the most important categories and their relative impacts.
Normalization is accomplished by considering the life cycle emissions of a kind from a
given process (e.g. vaccine refrigeration) in relation to the total emissions of the same
kind by a given population. In this analysis we used damage categories rather than
midpoint categories. The normalized damage factors were derived from TRACI 2.1 [37]
this indicator is used for United States and Canada and the population was used for the
normalization of SPSB vaccine refrigerator is population of US in 2008 (due to most unit
processes of SPSB happen in US) Figure 9 shows that the climate change impact is an
order of magnitude higher than both Ecotoxicity and Human Health impacts due to lower
person-year scores. Unlike GHG emissions, health related emissions are a local
phenomenon and the impacts are greater close to the source of emissions. Therefore a
detailed spatial assessment of distribution of local effects (health) is essential.
Figure 9: Normalized impacts of SPSB vaccine refrigeration (Vertical axis is in
logarithmic scale)
20


2.7 Sensitivity analysis
The results related to solar refrigeration option are sensitive to parameters such as solar
irradiation and ambient temperature that have an influence on number of solar modules
required. Further, the type of panel (monocrystalline vs cadmium Telluride) will have a
significant influence on upstream emissions and related impacts. On the other hand, the
manufacturing and assembly location of photovoltaic modules and refrigerators has
significant effects on the environmental impacts of vaccine cold storage in developing
countries. (Ghana in this case study) To analyze the variability of results we performed a
sensitivity analysis around these parameters. This study compares the GWP and HH
criteria impacts of SPBF type between average data derived from Ghana and different
conditions such as ambient temperature of 21c Solar radiation of 600 W/ m2, CdTe
photovoltaic modules and manufacturing and assembly location of solar modules and
refrigerators in Sweden.
The Figure 10-a shows that with deacresing the ambient temperature that vaccine
refrigerator works at it about 22c the GWP and HH Criteria impacts decrese 46% and
48% respectively.the energy consumption of vaccine refrigerators at 21C reduces
dramatically and the number of monocrystalline modules that needed for running the
refrigerator reduce from 10 to 4 module. On the other hand, the solar PV production has
the significant role in GWP and HH Criteria and changing the number of them changes
these impacts dramaticaly so in this case the huge decrese in impacts seen. With incresing
the solar raddiation the number of pv moudules for running the refrigerators decreses so
in Figure 10-b the same trend was seen. Figure 10-c shows that using Cadmium Telluride
instead of Monocrystalline cause to decresing the GWP impact about 30% because the
21


production process of Cadmium Telluride has the lower GWP impact than
monocrysstalline moudule. but the HH Criteria of Cadmium Telluride is more than
monocrysstalline about 42% Although the energy for production of one module of
CdTe is 40% of monocrysstalline but because of lower efficiency of CdTe, the number of
CdTe PV moudule for runinig the refrigerator is 1.4 times of monocrystalline. Figure 10-
d shows the LCA of same PV modules and refrigerators used in ghana but manufactured
and assembled in US and Sweden. The difference between GWP and HH Criteria is due
to different electricity profile of US and sewden that cuase to US location has 22% more
GWP impacts and 34% less HH criteria impacts.these results show that LCA study of
cold chain of vaccine suuply is function of many parameters and is dramatically sensitive
to these parameters so the LCA Scenario for refrigerator that manufactured in US and
used in ghana will be totally different for different manufacturing and using locations.
22


> a)
M Cfi
01
01
re ex
g-a
E a
o 0)
u a
0)
Q.
120
100
80
60
40
20
0



&
&
Ambient
Temperature
: 43C
Ambient
Temperature
: 21C
120
Solar
Radiation:
200 W/m2
Solar
Radiation:
600 W/m2
(a)
(b)
120
Vi
5 100
> $ 80
S* 60
40
S g 20
u a o
0)
0.
Monocrystalline
Module
Cadmium
Tellur ide
Module
Vi

> s Vi
at.
u
re bX
a S 3 G
o
u u u

CL
120
100
80
60
40
20
0
Manufacturing
in US
Manufacturing
in Sweden
(c)
(d)
Figure 10(a) comparative impact based on ambient temperature 10(b) comparative
impact based on solar radiation 10(c) comparative impact based on material of
module 10(d) comparative impact based on manufacturing location.
2.8 Geographical delineation of impacts
Emissions and impacts of vaccine cold storage infrastructure occur at multiple spatial
levels. Typically, emissions of manufacturing the refrigerators, solar panels, and batteries
occur in developed countries (North America and Europe) while the emissions during the
use phase occur in the developing world (e.g. Ghana and South Africa). We
disaggregated the human health impacts geographically based on manufacturing location
and place of use as shown in Tables 2(a), 2(b) and 2(c). We clearly see that human health
23


impacts related to solar refrigeration option happens outside (in the place of
manufacturing) rather than the place of use (Ghana or South Africa) while for
conventional refrigeration option there is a wide spatial spread of the impacts. The unit of
human health damage in this section is DALY (The disability-adjusted life year (DALY)
is a measure of overall disease burden, expressed as the number of years lost due to
mortality and morbidity)
Table 2(a): Distribution of Human Health Impact for SPSS use in Ghana
Total Manufacture Location (US) % Contribution (US) % Contribution Intercontinental transport Use Location (Ghana) % Contribution (in Ghana)
Total HH
effects 3.64E-6 2.41E-6 66% 32% 7.09E-8 2%
(DALY)
Table 2(b): Distribution of Human Health Impact for EP use in Ghana
Total Manufacture Location (UK) % Contribution (UK) % Contribution Intercontinental transport Use Location (Ghana) % Contribution (in Ghana)
Total HH
effects 6.1 E-7 3.08E-7 50% 1% 3.01 E-7 49%
(DALY)
24


Table 2(c): Distribution of Human Health Impact for EP use in South Africa
Total Manufacture Location (UK) % Contribution (UK) % Contribution Intercontinental transport Use Location (South Africa) % Contribution (in South Africa)
Total HH
effects 1.38E-5 1.2E-7 1% 1% 1.36E-5 98%
(DALY)
Table 2(a) shows that for solar refrigeration option (SPSB), 66% of all health impacts
happen in the country where the refrigerator is manufactured (i.e. United States). This
implies that stakeholders, policy makers and local governments should consider the
choice of the country where the solar refrigerators are manufactured in order to minimize
the life cycle human health impacts. This is also important because the medical treatment
costs associated with these human health impacts are significantly higher in the
developed countries (i.e. the manufacturing location) resulting in greater societal and
economic costs to these countries. Further, people who live near the battery and PV
production locations in the developed countries are most vulnerable to these human
health impacts due to heavy metals such as Arsenic, Mercury and Lead emissions. In
the case of electric power refrigeration (Table 2(b) and Table 2(c)) for both Ghana and
South Africa we see that the health impacts varied spatially. About two-third of the
impacts in the case of Ghana and almost all of the impacts in the case of South Africa
happens at the location of use. This implies that policy makers in the developing world
should consider the additional health burdens that could be avoided in their countries by
replacing conventional refrigerators with solar refrigerators. It is interesting to see that,
25


although the total health impacts of EP refrigerators (6.10E-7 DALYS) used in Ghana is
less than SPSB (3.64E-6 DALYS) (Table 2(a) and 2(c)) the local health impacts that
happen specifically in Ghana for the EP system (3.01E-7DALYS) is much more than
SPSB (7.09E-8 DALYS). For the case of EP refrigeration in South Africa (Table 2c) the
local health impacts that happen in South Africa is 44 time higher than the respective
impact in Ghana and this can be attributed to the differences in the electricity profiles of
the two countries. In this case, people who live downwind of the thermal power plants in
the developing countries (where the refrigerant systems are used) are most vulnerable to
human health impacts associated with criteria pollutant such as PM10, SOx and NOx.
2.9 Discussion
An important finding is that use of solar vaccine refrigeration in low-and middle- income
countries does not always translate into environmental benefits. Results indicate that
under South African average GHG intensity of electricity solar refrigeration options were
found to reduce life cycle GHG emissions by 55% in comparison to electric powered
refrigeration while under Ghana average GHG intensity of electricity solar refrigeration
options were found to actually increase life cycle GHG emissions by 65%. The potential
for solar vaccine refrigerators to be a sustainable alternative to electric powered
conventional vaccine refrigerators is highly dependent on energy sources of electricity in
the country of use. This result is counter-intuitive to public health officials as it is
commonly assumed that solar refrigeration systems have minimal impact on the
environment and are more environmentally friendly than electric refrigeration systems.
This study disproves that notion and provides a mechanism for public health officials to
26


make informed cold storage choices based on country specific data (particularly
electricity mix).
Figure 11 illustrates the conditions (based on a countries electricity mix) under which
solar refrigeration would offer greenhouse gas reductions. The horizontal line indicates
the threshold at which transition to solar option will yield benefits. For countries with
electricity mix [40] that produces more than 0.75 Kg C02e/KWh (break even line) the
use of solar systems will yield net GHG reductions. Figure 11 shows that the carbon
intensity of South Africa with dominant coal power plants and Nigeria with dominant
natural gas power plants in their electricity profile is above the break even line but in the
case of Congo and Ghana with dominant hydroelectricity power in their electricity profile
the carbon intensity is under the break even line.
Figure 11: Transition plot for net greenhouse gas (GHG) reductions
Figure 12 shows the dominant electricity source for most of the African countries [41]
based on the type of power plant contributing the most electricity. Countries near oil and
27


natural gas fields have more non-renewable power plants while hydropower (renewable)
dominates in countries near high stream rivers. Information from Figures 11 and 12 will
enable stakeholders and policymakers to select appropriate vaccine refrigeration option
independently for each country in Africa without assuming any one type to be more
sustainable. From Figure 12 we see that solar vaccine refrigeration can be an eco-friendly
option in many of the African countries that have significant contribution of coal, natural
gas and oil to their electricity mix (such as Nigeria, Algeria, and Morocco). Assuming
most of the new vaccine storage capacity is required in countries using significant
amount of fossil fuel we estimate that use of solar vaccine refrigeration can cut carbon
emissions equivalent to taking 65,000 cars off the road or preventing over 13,800 acres of
tropical forest from deforestation.
Figure 12: Dominant Source of Electricity Production in African Countries
The spatial disaggregation of localized health impacts reveal that for the solar options
almost all of the health impacts happen in the manufacturing location (USA or Europe)
28


while for conventional option the majority of health impacts happen in the location of use
(low-and middle- income countries). It is recommended that decision makers consider the
spatial distribution of health impacts while making choices related to type of refrigerator
and the manufacturing location.
Finally, we would like to add that even though solar refrigeration option is not the
sustainable choice in Ghana there might be situations where selection of the sustainable
option might not be possible. As an example, the use of conventional electrical option in
locations which do not have access to electricity is not possible. Similarly, conventional
electrical vaccine refrigeration systems are not portable and hence their use is possible
only in health clinics but not for mobile immunization campaigns. In these situations
solar refrigerators (or kerosene refrigerators) are the only option. Therefore, the
comparative results from this study are only relevant for the situation where both
electrical and solar refrigeration can be used (such as in central storage, distribution
facilities and health clinics).
29


Chapter 3
COMPARETIVE ENVIRONMENTAL OMPACT OF ALTERNATIVE
TECHNOLOGIES IN VACCINE SUPPLY CHAIN ECHELONS
3.1 Goal and scope
The goal of this cradle-to-grave LCA study is to compare the environmental impacts of
different vaccine manufacturing and packaging technologies, transportation, cold storage
and medical waste treatment methods in the context of low-and middle-income countries.
Four different types of vaccines, three different types of vaccine shipment modes, five
different types of off-grid vaccine refrigeration options and two medical waste treatment
methods with different system boundaries and characteristics are considered. Using this
comparative LCA policy makers would have enough information to make informed
choices about the eco-friendly SC options for a given situation, and context.
3.2 Vaccine manufacturing and packaging
Life cycle assessment (LCA) tracks the physical, material, and energy flows through each
stage or unit process and determines the resulting emissions and impacts.. Using LCA
requires direct knowledge of the materials used to make a product. Due to the complex
makeup of vaccines, such direct knowledge would require extensive pharmaceutical
training to accurately estimate the composition of various vaccines. Economic
input/output data provides a way to estimate the impact of manufacturing a product based
on its cost. The theory of EIOLCA [42] practice involves the assumption that emissions
are directly related to economic value of products within a given sector. Thus, a vaccine
which is twice as expensive demands twice as much energy and resources and therefore,
30


generates twice as much emission. While this assumption does incur some uncertainty,
this method is the only practical and viable option for this section. We selected sectors
which appropriately described the unit processes involved in the vaccine manufacturing
and packaging. The database provides outputs in several categories, but categories of
interest for our study were: Conventional Air pollutants, Greenhouse Gases, and TRACI
Impact Assessment.
By running the model for these categories, impacts were calculated in terms of emission
per $1M of vaccine. Using these results and scaling according to the cost of each vaccine
produced a life cycle inventory of the emissions associated with the production of one
dose of a given vaccine is estimated. The costs of the vaccines in question were obtained
from private sector prices [43] from 2014 and were used in our calculations as the most
appropriate estimation of the economic activity generated by the vaccines production. It
was assumed that these private sector prices included the cost of the vials, tubes or
syringes in which the vaccines were contained as well as the packaging. From various
sources, appropriate estimates for the cost of these supplies were determined and
subtracted from the private sector prices to arrive at a cost for just the vaccine
medication. Figure 13-a and 13-b, illustrate the global warming and human health
impacts of manufacturing and packaging of one dose of different types of pneumococcal
and rotavirus vaccines.
Figure 13-a and 13-b, illustrate the global warming and human health impacts of
manufacturing and packaging of one dose of different types of pneumococcal and
rotavirus vaccines.
31


35.0
Pneumococcal Pneumococcal Rotavirus, Live, Rotavirus, Live,
(Pediatric) Polysaccharide Oral, Oral,
Pentavalent
Figurel3-a) Comparative global warming impact of manufacturing and packaging
of one dose of various types of vaccine
0.0500
0.0400
m
S 0.0300
S
^ 0.0200
* 0.0100
0.0000
Pneumococcal Pneumococcal Rotavirus, Live, Rotavirus, Live,
(Pediatric) Polysaccharide Oral, Oral,
Pentavalent
Figure 13-b) Comparative human health impact of manufacturing and packaging of
one dose of various types of vaccine
3.3 Vaccine transportation
Transportation part in vaccine supply chain plays an important role due to large distances
between manufacturing and demand locations. Also during transportation, vaccines
should be kept within an appropriate temperature range. Therefore, providing a
sustainable transportation system is very important in this SC. In this study, all of inter-
country transportations are considered with diesel trucks but for inter-continental
shipments three alternatives are made available: 1-diesel truck, 2-ocean freight and 3-
airplane. In the following section, (Figure 14) a comparative environmental impact and
32


cost assessment of these options is presented. It is worth mentioning that all of the data is
derived from SIMAPRO software [44] and Transportation European Commission [45],
global warming
Diesel Truck Ocean Airplane
Freight
(a)
Human Health
4.50E-06 4.06E-06
0) 4.00E-06
o 3.50E-06
CL 3.00E-06
cud 2.50E-06
w n 2.00E-06
a 1.50E-06
E 1.00E-06
X X 5.00E-07 1.87E-07 1.32E-07 |
0.00E+00
Diesel Ocean Airplane
Truck Freight
(b)
Transportation Cost
Diesel Truck Ocean Freight Airplane
(c)
Figure 14-a) Comparative GWP of transportation alternatives, 14-b) Comparative
HH impact of transportation alternatives, 15-c) Cost analysis for 1 ton-km
It is worth mentioning that the functional unit for this comparison study is 1 ton-km.
Figure 14-c) illustrates that diesel truck shipment cost is 8% and global warming
potential (GWP) is almost 10% of the airplane shipment mode. But ocean freight cost is
27% of diesel truck and GWP of ocean freight El is 17% of diesel truck. So it is obvious
33


that ocean freight is the cheapest and most sustainable option for inter-continental
vaccine transportation.
3.4 Off grid refrigerators alternatives
The goal of this section is to compare the environmental impacts of different off-grid
vaccine cold storage options in the context of low- and middle-income countries. Five
different types of vaccine refrigeration systems with different system boundaries and
characteristics are considered.
3.5 Process description and system boundaries
In this analysis, we compare the environmental impacts of small scale off grid solar
refrigeration systems with absorption refrigerators which use fossil fuels for vaccine
storage. Different refrigeration options that are most relevant and related to existing cold
storage infrastructure in low-and middle- income countries are considered: a) Solar
Powered Supplemented Battery (SPSB) ; b) Solar Powered Battery Free (SPBF); c)
Kerosene Refrigerators (KR); d) LPG Refrigerators (LR); and e) Solar Absorption
Refrigerators (SAR). Most of these types are commercially available brands of vaccine
refrigerators. The characteristics of these refrigerators are depicted in the Table 3.
34


Table 3: Features and characters of various off-grid vaccine refrigerators
Technical Characters SPSB SPBF KR LR SAR
Vaccine Storage net Capacity(Liter) 38.7 54.4 102 102 102
Shipping weight (kg) 91 99 118 86 118
Energy Consumption at ambient temperature of 43 C (kWh/day) 0.61 0.86 6.3 6.3 6.3
Fuel Consumption NA NA 0.9 Lit Kerosene in 24 hours 500 g LPG in 24 hours NA
Autonomy 7 days 7 days 3.10 hours 3.02 hours 3.10 hours
refrigerant R134-a R134-a ammonia ammonia ammonia
Blowing Agent HCFC141-b HCFC141-b HCFC141-b HCFC141-b HCFC141-b
Refrigeration Cycle Vapor- Compression Vapor Compression Absorption Absorption Absorption
Manufacturing Location US US Sweden Sweden Sweden
Life span time (years) 15 15 5 5 5
The system boundary of the two types of photovoltaic vaccine refrigerators is illustrated
in Figure 3 (Chapter2). For photovoltaic refrigeration options, the unit processes
included: photovoltaic production, battery production (for SPSB), refrigerant production
(R134-a), blowing agent production (HCFC141-b) other materials needed for
35


refrigerator production, assembly of refrigerator; transportation and disposal of
refrigerators in landfill that includes the leakage of R134a and HCFC141b. The system
boundaries for the two solar options included no significant impacts during the use phase.
The system boundary for fossil fuel refrigerators (Figure 15) includes ammonia
(refrigerant) production, other materials for refrigerator production, assembly of
refrigerator, transportation, in use phase, (See Figure 15) the kerosene and LPG (fuel)
production and combustion is considered and in disposal section, leakage of R134a and
HCFC141b are included. Most of the unit processes of SAR are the same with KR and
LR but SAR has the production and manufacturing of solar thermal plate in
manufacturing phase and does not have fossil fuel production and combustion during use
phase.
36


Ammonia production
Copper coil production Aluminum roll production
Cold roll steel production Motor compressor production
V_________________J
Material Process olRefrigerator
HCFC141 -b Production
r ,,
Door Assembly Resin Finishing
Cabinet Assembly Refrigeration Cycle assembly
Assembly of Refrigerator
Water Tank
Production
V________________________________________J
Support
Production
Collector
Production
Solar Collector Production (for SAR)
......I........
Transportation
Manufacturing
Phase
(For KR and LR)
Use Phase
Leakage of
Ammonia
Leakage of
HCFC141-b
Disposal to
Landfill
Disposal
Phase
Figure 15: Fossil fuel and SAR refrigerators system boundary
3.6 Functional unit
A functional unit of 1 liter of cold storage capacity per year was selected for this study.
3.7 Life cycle inventory of absorption refrigeration systems
The LCI has already been presented in Chapter 2. Therefore, in this section the LCI of
absorption refrigeration systems is presented. The refrigerant for absorption refrigeration
cycles is ammonia. Based on the Boudehenn et al. study [46], 0.33g of ammonia is
required for a KR, LR, or SAR options for the functional unit. All of the data related to
37


LCI of production of ammonia were derived from the Mendivil et al. [47] study. The
amount of blowing agent (2.03g) was derived based on the work of James et al. (Ref) for
1 liter cold storage for one year. The energy needed for material process (1.95 kWh) and
assembling of absorption systems (0.48 kWh) was estimated based on the work of
Boustani et al. [33], Crude oil for the production and combustion of kerosene and LPG is
assumed to be extracted from oilfields in Nigeria. The crude oil is transported to Ghana,
and the refining process is assumed to take place in Ghana, and e kerosene and LPG were
transported inside Ghana. In order to calculate the emissions associated with the
production of a SAR unit, the size of the solar collector plate must be determined. Based
on the solar radiation of Ghana (200 w/m2) and the amount of thermal energy needed to
operate absorption refrigerator (120 MJ based on functional unit), the net surface area of
the required solar collector plate was estimated as 70 cm2. All of these calculations were
based on data from a solar collector company [48] and Ardente et al. [49] work.
3.8 Impact assessment
The life cycle impact assessment methods are used to translate the inventory emissions
data into environmental impacts through the use of characterization factors. These
characterization factors represent relative impacts of different chemicals (as in the case of
global warming potential) or are a function of fate, exposure and effect of the chemicals
(as in the case of human health and ecotoxicity). In this study, we adopted the life cycle
impact assessment methodology (LCIA) based on Tools for the Reduction and
Assessment of Chemical and other Environmental Impacts (TRACI) developed by the
U.S. Environmental Protection Agency. This study considered only midpoint impacts
(derived from TRACI) as it was deemed that they provide sufficient details needed for
38


our analysis and the use of endpoint damage modeling would bring in further uncertainty.
The following impact categories were considered: (1) global warming; (2) acidification;
(3) eutrophication; (4) smog formation; (5) human health criteria; and (6) ozone depletion
.We applied classification and characterization steps to relate individual elementary flows
in the inventory to the relevant impact categories and to identify characterization factors
based on the media where the emissions occurred. We did not consider the normalization
step as emissions at different stages in the life cycle occurred in different geographical
locations.
3.9 Results and discussion
This section summarizes the key findings comparing the different off-grid vaccine
refrigeration options. The total scores related to all of the 6 impact categories for the five
different options are shown in Table 4. Results are based on functional unit of 1 liter
vaccine cold storage per year. The impact profiles related to SPSB, SPBF, KR, LR and
SAR are shown in Table 4. In the category of Global Warming Potential, the KR has the
maximum and SAR has the minimum GWP. Burning the LPG produces 45% fewer CChe
emissions than kerosene so
LPG is a more eco-friendly fuel than kerosene. Production of PV modules in SPSB also
results in considerable GWP but less than KR and LR.
39


Table 4: El of various off-grid vaccine refrigerators
Impact GWP (kgC02 eq) HH Criteria kg PMlOeq Acidification Kg H+ moles Eutrophication Kg N eq Smog air Kg 03 eq Ozone Depletion (kgCFC-11)
SPSB 8.39 0.0039 1.38 9.28E-4 0.52 6.76E-05
SPBF 6.98 0.0022 1.07 7.12E-4 0.40 1.03E-04
KR 24.03 0.012 3.60 1.69E-3 1.28 2.45E-4
LR 13.56 0.0031 1.19 8.33E-4 0.58 2.45E-4
SAR 5.72 0.0045 1.24 7.15E-4 0.39 2.45E-4
GWP of SPBF is less than SPSB because SPBF does not require lead acid battery
production. The SAR type has the minimum GWP impact among all of the vaccine
refrigerators. This system uses an absorption refrigeration cycle, similar to KR and LR.
By replacing kerosene with a solar thermal collector, the CCRe emissions decreases by
about 78% in comparison to KR. Policymakers concerned about GWP impact, should
consider SARs as the best option. In the category of Human Health (HH) Criteria impact,
KR has the highest scores, and it is apparent that the ranking of other options is
completely different to that of GWP ranking. Burning of kerosene fuel in the use phase of
KR generates significant amount of PM10, SOx and NOx which results in higher HH
impact score. While the SAR unit scored the lowest in GWP, it generated the second
highest score in the HH impact category. This is due to the emissions caused during the
production of solar collectors, water tanks and external supports. It is also interesting to
note that LPG refrigeration is a cleaner option in terms of HH impact than SAR. The
ozone depletion impact scores results entirely from the discharge of the blowing agent
40


(HCFC141-b) to the atmosphere. The amount of blowing agent in the types of KR, LR
and SAR are the same, thus the ozone depletion impacts are equal for these three systems
but the SPSB and SPBF systems use much less of the blowing agent resulting in a lower
ozone depletion score. In the eutrophication, acidification, and smog air categories, the
trends can easily be seen from Table 4.
Figure 16 expands on the above results by providing the breakdown of individual
contributions of different unit processes towards GWP, and HH criteria for all of the
refrigerator types. For the KR and LR option, it can be seen that kerosene production and
combustion has by far the highest contribution towards both GWP and HH criteria
impacts. For the SPSB, PV module production has the highest effect on GWP, while in
the category of HH criteria the lead acid battery production and PV module production
have an equal contribution. So it is clear that HH impact of SPBF is less than SPSB. In
the case of SAR, production of solar collectors has the highest contribution on both GWP
and HH impact. It is obviously seen that for KR and LR types, most of these impacts take
place in use phase (located in developing countries) while for the solar options most of
the impacts are attributable to production which occurs in developed countries.
41


R134a Production
ammonia Production
R134aloss
blowing agent loss
LPG production and combustion
kerosene production and combustion
transportation
solar collector production
battery production
PV module production
assembly of refrigerator
(a)
1.40E-02
1.20E-02
1.00E-02
8.00E-03
6.00E-03
4.00E-03
2.00E-03
0.00E+00
KR LR SPSB SPBF SAR
LPG production and combustion
kerosene production and combustion
transportation
solar collector production
battery production
PV module production
assembly of refrigerator
material process of Refrigerator
(b)
Figure 16(a): GWP breakdown for various refrigeration technologies. 16(b) HH
breakdown for various refrigeration technologies.
In Figure 17 the total cost (capital cost + operating cost) of the four types of off-grid
vaccine refrigeration options are shown. The interesting point is that although the capital
cost of solar options are much higher than that of the fossil fuel options, the total cost of
solar options are less than fossil fuels. Therefore, the solar options are not only
42


sustainable in many locations but also are more economical when operating costs are
considered.
Figure 17: Total cost of various refrigeration systems for functional unit
3.10 Medical waste treatment system
In this section, an analysis is done around the disposal echelon of vaccine supply chain
and two alternative technologies are considered: 1-incineration, and 2-autoclaving.
3.10.1 Incineration
Vaccine waste disposal is considered as an infectious medical waste. Therefore, to
decrease the hazards the medical wastes are exposed to high temperature for long period
of time (incineration). However, there are certain drawbacks such as ash disposal and
emission of very toxic pollutants like dioxin.
For quantifying the LCA emissions of incineration the system boundary shown in Figure
18 is considered [50],
43


/ >
Ancillary Materials production External energy production (electricity,
(cement and etc) natural gas)
a

Figure 18: Incineration process system boundary
3.10.2 Autoclaving
Autoclaving is another disposal method. In this method, syringes and needles are
sterilized by high temperature and pressure of steam. This method is costly but has very
low pollutant emissions to the atmosphere. Disinfection of needles, syringes and open
vials is guaranteed with autoclaving. The needles could sometimes carry HIV or other
infectious viruses. Therefore, sterilizing them is vital. The system boundary of
autoclaving process is shown in Figure 19 [50],
44


Ancillary Materials production
External energy production
Steam autoclave sterilization
Sanitary landfill (leachate treatment, residues landfill, electricity
production)
Waste water treatment
Figure 19: Autoclaving process system boundary
In Table 5 the various environmental impacts and cost of both disposal technologies for 1
kg of immunization waste is illustrated.
Table 5: El and Cost of alternative medical waste treatment technologies
Impact GWP (kgC02 eq) HH Criteria kg PMlOeq Acidification Kg H+ moles Eutrophication KgN eq Smog air Kg 03 eq Cost (USD 2013)
Incineration 1.07 0.0035 0.145 9.17E-5 0.051 2.28
Autoclaving 0.35 0.0019 0.058 6.24E-4 0.036 6.68
Table 5 shows that global warming and human health impacts of incineration is much
higher than autoclaving but the cost of incineration is lower than autoclaving.
3.11 Reference flow
For quantifying the total life cycle impacts of vaccine SC, it is required to sum up all of
the echelons LCA results, but the functional units for each of them is different. So, we
should establish an equivalence relationship between all echelons. This relationship is
45


called reference flow. The reference flow for pneumococcal vaccine SC is displayed in
Figure 20.
Equivalent to
Manufacturing of 1 dose of vaccine
0.984 tkm transportation
Equivalent to
I
Equivalent to
^ 0.5 mL vaccine cold storage capacity
0.110 Medical waste treatments
Figure 20 : Vaccine SC reference flow
46


CHAPTER 4
MATHEMATICAL PROGRAMMING MODEL FOR OPTIMAL DESIGN OF
SUSTAINABLE VACCINE SUPPLY CHAIN
4.1 Motivation
This chapter addresses the optimal design and planning of vaccine SC with
environmental and economic objectives. The environmental objective is measured by the
LCA methodology; the economic objective is measured by the total investment cost of
supply chain. A multi-objective continuous linear programing model is developed that
accounts for major characteristics of vaccine supply chains, including geographical and
industrial diversity of suppliers, various shipment methods, multiple refrigeration
technologies, different disposal treatment options, demand distribution, cold chain and
transportation capacity. The resulting Pareto-optimal curves display the tradeoff between
the environmental and economic dimensions of vaccine supply chain.
4.2 Introduction
Vaccine supply chain (SC) which can be defined as the network necessary for
distributing vaccine from manufacturers to the final consumers includes various sections
such as production and packaging, transportation, cold chain and disposal. Vaccine
demand fluctuations with location and time and an unfixed production rate of
manufacturers are challenging points in this SC. In the low and middle income countries
due to higher birth rates, vaccine demand is continuously increasing. Equity and
efficiency are two parameters that should be considered in vaccine Supply Chain
Management (SCM). It means that vaccine should be provided uniformly for each
recipient in the system while the infectious diseases are averted as much as possible.
47


Vaccine SC consists of several echelons all of which have significant cost and
environmental burdens. Furthermore, there are several alternative options available in
each section of SC. Therefore, design and planning of a sustainable vaccine SC could
offer significant benefits to immunization programs. In this study, we propose a
methodology to minimizing the environmental impact (El) and cost of the vaccine SC by
developing a generic mathematical model. The Continues Linear Programming problem
is optimized using stochastic multi-objective function method called Non Dominated
Sorting Genetic Algorithm II (NSGA II) [51], For quantifying the El of the SC, LCA
methodology is utilized [52, 53, 54], Finally, achieving long range planning and
designing the most sustainable SC are the goals of this study.
4.3 Problem formulation
The economic and environmentally-conscious alternative selection problem for the
planning and design of vaccine SC can be expressed as follow.
A set of clinics and their demands for various types of vaccine over a given long-
term future (planning horizon)
A set of candidate medical waste treatments technologies
A set of candidate vaccine refrigeration options in clinics
A set of potential geographical sites for the various types of manufacturing
locations.
A set of production and packaging options for the desired vaccines,
The target is the following:
Design the supply chain network of vaccine that would meet the demand over the
entire planning horizon.
Such that both the
(1) The overall capital and operating investment cost at the end of the planning
horizon is minimized
(2) El of the entire SC is minimized.
48


As illustrated in Figure 22, the model suggested here to solve the problem is based upon a
network structure consisting of a set of h type of vaccines with a set of J manufacturing
facilities (sites) with X types of packaging options. Regarding the locations and
availability of UNICEF walk-in cold rooms, K UNICEF refrigeration warehouses are
taken into account. / different shipment modes from UNICEF warehouses to national
cold storage facilities is considered. From there vaccines are shipped to M existing
regional cold storage facilities. From there the vaccines move towards NN clinics at the
next level. We consider SS refrigeration types in clinics and 0 locations for vaccine vial
and needle disposal with FF medical waste treatment options.
Optimal decisions include selection of the optimum combination of manufacturing and
packaging methods and locations from available options. In addition, the optimal
structure of transportation links between the chosen sites and existing clinics needs to be
designed. All these choices are made within a finite number of T time periods.
Unlike traditional methodologies, which just consider economic objective functions, the
methodology developed here tries to reach the optimum superstructure configuration and
schedule a planning strategy that minimizes both environmental burdens and the cost of
the SC. The solutions of this multi-objective continues linear programming problem lead
to trade-offs between conflicting economic and environmental objectives.
49


<
Shipment Mode 1
^
Yhikxt UNICEF Zhkht Shipment Mode 2 National Uhlmt Regional
Store S o o Level > Level
Shipment Mode / Storage Storage
^
Vhmnnsst
V
Whnnofft
Disposal Method
Technology 1
Disposal Method
Technology 2
Disposal Method
Technology FF
Figure 21: Schematic graph of vaccine network superstructure
For quantifying the total investment of the SC, the procurement cost of various vaccine
types, transportation cost, cold box cost, energy consumption cost, disposal cost, capital
and operating cost of the clinic refrigerators are all used as part of the economic objective
function. On the other hand, the environmental objective function is based upon the El
50


outcomes from the operation of the entire SC throughout the entire time horizon. This is
evaluated through the concept of LCA and using IMPACT2002+ methodology [55],
4.4 Mathematical model
4.4.1 Production, refrigeration and transportation capacities and material balances
Equation (1) implies that the volume of vaccines stored in the UNICEF cold room k in
time period t should be less than that of cold room capacity in time t.
(1) Zy=i It=i l£=i2h=i yhjkxr "2f= 1 2t=i £i=1 £h=i zhkiiT - Vfc, t A T
(2) 'ZJj=i'Zx=i'Zh=iyhjkxt RR)kt vk,t
Equation (2) ensures that vaccines received by UNICEF store k during time interval t is
less than its transportation capacity.
(3) l£= i l£=i ynjkxt < Phjt V/, t, h
Equation (3) ensures that vaccine h produced in manufacturing location j with technology
x should be less than its production capacity during time interval t.
(4) ynjkxt = 0 Y/, k, x,t (t > T)
(5) All Variables > 0
Equation (6) ensures the mass balance between input and output vaccines for UNICEF
cold room k.
(6) HJj=iHl=lllx=l^h=iyhjkxt =
Constraint (7) ensures that the history of product quantities shipped by each production
plant to the UNICEF centers from beginning until the time period t does not exceed the
product outward flow rate.
51


Vk.t^T
(7) Z/=i Ht=i Hx=i Hft=i ytijkxr Z[=i Ht=i Hi=i Hft=iz
hklir
Constraint (8) ensures that the difference between the outgoing and incoming quantities,
from and to the UNICEF warehouses, in each period is confined to the holding capacity
of that warehouse.
(8) Zfc=1 Ht=1 Zi=1 Zft= 1 zftfcZiT _Zm= 1 Ht=1 Zft=l uhlmr
It
Vl,tT
Constraint (9) ensures that the product amounts which are received by national
warehouse / do not exceed the transportation capacity of it in time interval t.
(9) llk=l^i=l^h=lzhklit ^ SDit Vl,t
Equation (10) sets up a balance between the total incoming and outgoing items, to and
from the national warehouse cold room / respectively.
(10) Zfc=lZt=lZz=iZft=izftA:Zif: = Zm=lZt=lZft=iuftZmt ^
Equation (11) assures that the history of product amounts sent by each production plant to
the national warehouse center / from the beginning until the time period t does not exceed
the product outgoing flow rate.
(11) lk=lTr=lli=llh=lzhklir > lm=lTr=llh=l^hlmr Vl,t T
Constraint (12) assures that the difference between the outgoing and incoming amounts,
from and to the regional warehouses m respectively, in each period is confined to the
holding capacity of that warehouse.
(12) Zi=iZr=lZh=luhImr-Znn=lZr=lZss=lZh=lVmnn5sr ^ Timt Vm,t =£ T
Constraint (13) assures that the vaccine amounts which are received by regional cold
room warehouse m do not exceed the transportation capacity of it.
(13) Zz=i Zh=i uhimr < TiDmt Vm, t
52


Equation (14) sets up a balance between the total incoming and outgoing items, to and
from the regional warehouse cold room m.
(14) D[= 1 Xt=1 Eh=1 uhlniT Ymn=1 Zt= 1 Xss=l Y*h=1 Vhmnnssr
>55 vH
Vm
Constraint (15) assures that the history of vaccine amounts sent by each production plant
to the regional warehouse center m from beginning until the time period t does not exceed
the vaccine outgoing flow rate.
(15) Yn=l It=1 1 Uhlmr > Inn=1 It=1 Sff=l Zh=i Vhmnnssr Vm.t^T
Equation (16) represents that the vaccine provided for clinic nn is 1.1 of the total demand
of it during the planning horizon. 10 percent vaccine wastage through SC is considered in
this study.
(16) Im=l Zr=l Iff=l Vhmnnsst = (1-1) It=1 Dfirmr Vnn, t T,h
Constraint (17) assures that the vaccine amounts which are received by clinic nn should
not exceed the transportation capacity of the vehicle which delivers vaccine to clinic nn.
(17) I£=1 Zff=i iLl Vhmnnssr < DD
nnt
Vnn, t
Constraint (18) represents that the amount of vaccine inventory in clinic nn should be less
than total various types of refrigerating capacity of clinic nn.
(18) Zm=l Zt=1 Zss=l Hh=l Vhmnnssr Zr=l Zft=l Oftrmr Qnnt
Vnn, t
Equation (19) displays the vaccine volume balance for vaccine volume inputs to clinics
and empty vaccine vials volume outgoing to disposal locations for each clinic nn.
(19) Im = l It=l Zss=l Ih=l Vhmnnssr = Zo=l Zt=l Zff=l Zh=l whnnofft Vnn
Constraint (20) represents that the vaccine wastage amounts in each vaccine disposal
location o have not been allowed to violate medical wastage treatment device capacity.
53


Vo, t A T
(20) 2nn=l 2x=l 2ff=i ^hnnofft ^ Biot
Constraint (21) assures that the vaccine vials and needle amounts which are received by
disposal devices do not exceed the transportation capacity of the delivery vehicle in each
time interval t.
(21) Inn=i Iff=l Zh=l whnnofft < BiDot Vo,t
4.4.2 El quantification methodology
For measuring the El with LCA approach, the definition of system boundary and
functional unit are essential. For the model implemented here, the system boundary of SC
network (Figurel) covers the 12 sets of life cycle stages. The functional unit for the
overall system is defined as the delivery of various types of vaccines to the clinics over
the entire planning horizon. The impacts due to operation of each of these stages were
calculated separately. The LCA method categorizes the emission inventories of each life
cycle stage in terms of appropriate impact category. The IMPACT 2002+ indicator
estimates the environmental damages based upon three main categories: Global Warming
(GW), Human Health (HH) and Ecosystem Quality (EQ). Equation (22) illustrates the El
measuring process with LCA method [53],
(22) DEgte = £b=1 flbe MEfe is the reference flow required for life cycle stage pe during time t (such as a liter of
vaccine in the manufacturing echelon or the weight and distance of vaccine transported in
transportation echelon etc.) specified for each life cycle stage. Secondly, oope which is the
emissions factor per unit of reference flow of life cycle stage pe is specified from various
resources (Such as SEVLAPRO and EIO [42]). It is worth mentioning that b G B is the
54


set of environmental burdens. At the third step, t)be, which is the characterization factor of
substance b contributing to the impact indicator, e is derived from IMPACT2002+
indicator. (e G E = {gw, hh, ec} ).The DEgj? quantity are represent the human health
impacts with the unit of Disability Adjusted Life Years (DALY), or PDF m2yr for the
ecosystem quality category which represents the potentially disappeared fraction per square
meter per year, or Kg CChe for the global warming potential category for the life cycle
stage pe during time interval t.
For the vaccine production and packaging stage, Equation (22) is rewritten as Equation
(23)
(23) DEe = Zb=i^bebZk=iZy=iZt=iZ^=iZh=iyhyfcXT Ve,pe =
1: vaccine Production Stage
At the next level, the El resulting from transporting vaccines from manufacturing
locations to UNICEF cold rooms needs to be assessed. The functional unit for
transportation is based upon both the distance and weight of vaccine carried. Therefore,
knowing the distance, otbjkxt, between the location of the vaccine supplier at site j and
UNICEF warehouse k and using the transportation emissions inventory, oob, the impact
associated with transportation can be assessed through Equation 24.
(24) DE| = Zb=l^be Zk=lZy=iZt=lZ£=iZh=i(yhyfceT X ahjkxt)
Ve, pe = 2: Transportation Stage From Supplier to UNICEF store
El resulting from the ice production step for ice needed for cold boxes during
transportation is calculated based on Equation 25. The functional unit for this stage is
based on the amount of ice needed to carry 1 liter of YF vaccine. (Equation 25)
55


(25) DEf =Zb=i^bebIk=iZy=iZt=iZ^=iZh=iyhyfcxt
Ve, pe = 3: Ice Production for Transportation Stage 2
El resulting from the refrigeration process of vaccines in walk-in cold rooms at UNICEF
warehouse is derived from equation 26. The functional unit for this stage is refrigeration
of 1 liter of vaccine per year. go£, is the emissions inventory of the refrigeration process
based on electricity profile of the UNICEF warehouse location. The El score is
proportional to the amount of vaccine inventory.
(26) DEf Zg=ibeWbZiL=iZ|=iZj=iZk=iZx=iZT=i Zf=i [(ytijkxt
zhklit)] x [] Ve, pe = 4: UNICEF Store Refrigeration Impact
Equations (27-33) are used for determining El scores for refrigeration, transportation and
ice production in various echelons of the supply chain.
(27) DEf = Zb=l^be wb Z!=iZk=iZ!=i 'Zt=l'Zh=l(zhklit x ahklit)
Ve, pe = 5: Transportation Stage From UNICEF store to National Store
(28) DEf = Zb=l^be wb ZlL=iZk=lZi=iZu=l Hh=lzhklit
Ve, pe = 6: Ice Production for transportation stage 5
(29) DEl = Zb=l^be WbZlL=iZ!=iZm = iZk=iZx=l Zh=l [(Zhklit uhlmt) X (^)]
Ve, pe = 7: National Store Refrigeration Impact
(30) DEf = Zb=l^bebZ{=iZm = lZ?=lZh=l(Uhlmt X ahlmt)
Ve, pe = 8: Transportation Stage From National store to Regional Store
(31) DEf = Zb=l^beWbZiL=iZm = iZ?=iZft=l uhlmt
Ve, pe = 9: Ice Production for transportation stage 8
(32) DE£ =2g=16be (x/12)] Ve, pe = 10: Regional Store Refrigeration Impact
56


.mnnsst
whnnofft) x (x/12)] Ve, pe = 11: Clinic Refrigeration Impact
Equation 34 illustrates how the impact score of the waste treatment stage is calculated.
The functional unit is disposal of 1 liter vaccine glass vial and oo^f is emission inventory
of waste treatment with ff technology method.
By using the normalization factors ( yn ) of each category, the set of impact indicators for
the various life cycle stages are normalized as follows (Equation 35):
At the final step, the scalar IMPACT 2002+ score can be achieved by using the
weighting factor of relative importance to the normalized categories (which are
represented as 0n )[53], This factor in this study is considered 0.2, 0.6 and 0.2 for human
health, global warming and ecosystem quality respectively.
(36) Environmental Impact Score = £,T=1 Hn=1 Hnt
4.4.3 Cost objective function
The economic performance of the SC network is relates to the total cost of investment at
the end of the planning horizon. Similar to the environmental objective function, each
stage of vaccine SC has a specified investment cost that is determined individually by
Equations (37-48).
hnnofft
Production Stage
57


Transportation
(38) Cost2 = Z/=1Zfc=iZ£=iZt=i
from Supplier to UNICEF Store
Tih=l(yhjkxt x ahjkxt) X c
jkt
(39) Cost3 = Z/=1Zfc=iZ£=iZt=i lh=i(ynjkxt X u3)
Ice Production for Transformation stage 2
(40) Cost4 = li=1li=1lJj=1lt=ilx=ilr=i lh=i([(ynjkxt ~ Zhkiit) x (t/12)] x
Cfct) Refrigeration in UNICEF Store
(41) Cost5 = Z{=iZfc=iZi=iZt=i 'Lh=i((Zhkiit x a'hkiit) x cfciit) Transportation
from UNICEF Store to National Store
(42) Cost6 = Zi'=iZfc=iZ{=iZt=i lh=i(zhkiit x c ) Ice Production for
Transformation stage 5
(43) Cost7 = Zl,= lZi=iZm=lZfc=lZr=l Zfc=l[Ofclit imt) x 0712)] X C,7t
Refrigeration in National Store
(44) Cost8 = Zi=iZm=iZt=i X a'hlmt) x Qmt) Transportation
from National Store to Regional Store
(45) Cost9 = Zi'=iZm=iZt=i 'Zh=i(uhimt X Ct9) Ice Production for
Transformation stage 8
(46) Cost10 = Zf=iZm=lZ£=lZ£n=lZff=lZfc=i([OWt Vhmnnsst) X (t/12)] X
O Refrigeration in Regional Store
(47) Cost11 = Z?=1 Zm=l Zt=! Znn=i Zff=l Z//=i
(t/12)] X c^t Refrigeration in Clinic
Zft=i [(fftmnnsst tVhnnofft)
X
(48) Cost12 = Zo=iZ//=iZnn=iZt=i Zh=i Whnnafft x Disposal Stage
In all of the above equations, C represents the unit cost at each SC stage. Unlike the
environmental function, there is another stage in cost function that represents the cost of
Open Vial Waste of vaccines that is measurable by equation (49).
58


(49) Cost13 = 02=11^=1 Iff=ilLil£=i(0VWhnnsstx)) x
(&x=lYJj=iYJk=l^=l^h=iyhjkxt )/(Iiy=iIifc=iIix=lIit=lIift=iyftyA:xt )) x chxt
Cost of Open Vial Waste
0VWhnnsstx represents the percentage of open vial waste of vaccine h at clinic nn, with
refrigeration technology ss, during time interval t with vaccine production technology x.
0VWhnnsstx is determined by equation (50). It is worth mentioning, that Dhnnsst, is
vaccine h demand at clinic nn which is kept in refrigerator type ss in time interval t.
(50) OVW,
hnnsstx
= 1-0
Jhnnsst
Jhnnsst
f^hx
/ fthx)
Percentage of open vial vaccine waste
In the resulting section, it will be shown that OVWhnnsstx can be considered as a separate
objective function for the optimizing of this SC [56],
The overall cost function is shown in Equation (51).
(51) CostFunction = Y,}=iCostl
4.5 Computational experiments
We have coded this model in MATLAB and NSGAII algorithm within MATLAB was
used to solve the model. The algorithm was tested with real data of a vaccine supply
chain. To run each condition it took 8 hours of 1 CPU to reach a solution with a 0% total
gap on a 3.4 GHz Intel Core i7 with 16 GB physical memory.
4.6 Parameter setting
This algorithm was run with different value of generation number with an initial
population size of 50 and with the following parameters for NSGAII. The probability of
crossover was set as 0.7 and the probability of mutation was set at 0.02. In addition
the genetic operations such as the Pareto dominance ranking procedure and the
elitist selections were used.
59


4.7 Vaccine supply chain in Ghana
4.7.1 Network superstructure
This section will be used to illustrate the capability of the supply chain model which
consists of different types of vaccine needed for children under the age of five and for
pregnant women in an immunization program for Ghana (the list of vaccines and their
dosage are available in Table 6) where the vaccines are demanded at twenty province
clinics (nn = 20). Vaccines are produced in 3 UNICEF approved locations (J = 3),
Russia, France and Senegal with 3 different product formulations (x = 3), 5 doses, 10
doses and 20 doses per vial [43], All of these vaccines are first gathered at UNICEF
walk-in cold room in Copenhagen (k = 1). Then these products are transported to
national cold room in Accra, (/ = 1). A total of three potential transportation modes are
considered, (t = 3), Diesel Truck, Ocean Freight and Airplane. Next the vaccines are
shipped to 10 regional cold rooms [57], at Ashanti, Brong Ahafo, Central, Eastern,
Greater Accra, Northern, Upper East, Upper West, Volta and Western Regions(m = 10).
From they are transported to twenty province clinics (nn = 20). There are four potential
refrigeration options in the clinics (ss = 4) namely, Electrical Ice-lined Refrigerators,
Solar Refrigerators Supplemented with Battery, Solar Refrigerators Battery Free and
Kerosene [58], The final echelon in this supply chain is vial, needle and syringe waste
treatment. Twenty disposal locations (o = 20) with two technologies, Incineration and
Autoclaving [19], are considered in this study. (// = 2). The geographic locations of the
SC echelons are shown in Figure 22.
Table 6: Ghana immunization program for under five ages children and pregnant
women
60


Vaccines Doses per person Volume per one dose (ml)
BCG 1 dose 1.3
DTP-HepB-Hib 3-doses 13.1
Yellow Fever 1-dose 3.6
Oral polio Virus (OPV) 4-doses 1
Tetanus Toxoid (TT) 5-doses 2.6
Measles 2-doses 5.2
Pneumococcal (PCV) 3-doses 21
Rotavirus 2-doses 156
Figure 22 [59]: Vaccine SC map for Ghana
4.7.2 Results
Multi (two) objective optimization of this SC model leads to a set of solutions (Pareto
Front) as presented in Figure 23. From the Pareto Front solutions, it is obvious that a
conflict exists between a structure design that leads to a minimum El and a minimum
total cost. The figure shows that reduction of El is only possible if policy makers
compromise on the cost. The extremes of the Pareto front reach the optimum structural
61


design that leads to minimum El and minimum total cost. The point which is located at
the middle of the Pareto front represents the multi-objective optimum network
superstructure which is shown in Figure 23.
Furthermore, it is worth mentioning that the discontinuity in Pareto front (Figure 23)
comes from the switching of one structure to another. This switch provides a chance to
achieve a significant improvement in environmental performance at a marginal increase
in cost by rebuilding the network and implementing various sustainable policies. For
instance, around the cost value of 4.76 x 107 USD, a compromise in the total cost of
only 0.21% can reduce the El by 2%. This is achieved by rebuilding the network
structure from configuration B to A.
+->
c
Q
c
o
u
PN
c
u
5.15E+08
5.10E+08
O
O 5.05E+08
u
C/5 5.00E+08
+->
% 4.95E+08
g* 4.90E+08
w 4.85E+08
4.80E+08
4.50E+07 4.60E+07 4.70E+07 4.80E+07 4.90E+07
Feasible : Solutions
Re igion


Multi-Objective Configur; ition B
Optimum Poini
Configu ration
Cost (USD)
Figure 23: Efficient set of Pareto solutions to the vaccine SC in Ghana
Fig 24 shows a configuration located in the middle of the Pareto space where the
combination of 5-doses-per-vial packaging technology, located in Russia, and 20-doses-
per-vial packaging technology, located in Senegal, may result in an optimum alternative.
The cost and El of 20-doses-per-vial packaging technology are less than other options, so
the major portion of vaccine flow rate was passed through this manufacturing location.
On the other hand, the cost due to open vial waste of 5 doses per vial is less than the
62


others and hence the remaining portion of this vaccine flow rate passes through 5 doses
per vial manufacturer in Russia. For intercontinental transportation the ocean freight is
the optimum option due to less cost and EI. The combination of Electrical Ice-Lined
refrigerator and Solar direct derive refrigerator is the optimum option in terms of EI and
total cost functions. Autoclaving medical waste treatment is a more sustainable method
for disposal section because this technology does not have hazardous emissions such as
Dioxin [50], but incineration is much cheaper. Therefore, a combination of 23% of
autoclaving and 77% of incineration from total vaccine wastage flow rate were obtained
from multi-objective optimization. As it can be observed, the multi-objective
optimization results in a set of Pareto solutions; among these sets of non-dominated
solutions the policy makers should select one. The stakeholders chosen solutions will be
dependent on the weights that they give to each of the objectives.


Figure 24: Network Configuration for multi-objective optimization
4.7.3 Other important objective functions for vaccine supply chain
4.7.3.1 Open Vial Waste (OVW)
Open vial wastage occurs due to the fact that when multi-dose vials are opened during a
vaccine session and if the vial is not completely used then the remaining doses have to be
discarded and cannot be used for subsequent sessions. So for efficient vaccine SC,
minimizing the OVW is very important. This fact illustrates that although, using multi-
dose vials is sometimes an economic option it also increases the risk of unaccounted cost
due to OVW. The definition of OVW is brought in equation (50).
64


4.7.4 Percentage of vaccine met demand
Larger vial sizes can improve the percentage of vaccine met demand. If the clinics are
equipped with large vial sizes, they can easily provide sufficient amount of vaccine for
unpredicted demands (more than estimated demand). For instance, if our estimation is 14
doses of a vaccine and there are: 14 one dose or three 5 dose or two 10 dose packaging
options available, then with a more than estimated arrival rate large vial sizes could meet
the demand better. So maximizing this objective function could be worthy in vaccine SC
optimization. Therefore, optimization based on opposite objective functions (like vaccine
met demand and OVW) is necessary to balance the trade-offs. Therefore, multi-objective
optimization is completely relevant in this case. With curve fitting process, equation (51)
is derived from Rajgopal et al. study [60], This correlation displays the percentage of
vaccine met demand based on vial sizes.
If the percentage of open vial waste of vaccine (OVW) which is defined in equation (50)
is considered as a separate objective function, a 3D Pareto front will be achieved (Figure
25) that not only tries to provide the minimum cost and El supply chain network, but also
attempts to reduce the open vial waste. Figure 26 displays the 3D Pareto front when
percentage of vaccine met demand is replaced with OVW. The points that the surface
passes through are optimum solutions of vaccine SC with each of them being a detailed
design and scheduling solution that leads to the Pareto fronts. In the next section the
optimum scheduling pattern is discussed.
(51)
65


Percentage of met vaccine demandt
( /
\
92.153 X t00108 x

+
95.975 x t0 0071 x

+
95.975 x t0 0071 x

+
98.04 x t 0044 x

K
K X

yjkxt
-1 k=1t=1
j=l k=1 x=l t=l
if
Jkxt
i = 1 1>=1 f=1 i = 1 1>=1 v=1 f=1
yjkxt
K
Jkxt
i = 1 1>=1 f=1 i = 1 L-=1 v=1 f=1
yjkxt
/
J k t Jkxt
Z Z Z w/(Z z z z
y=ifc=it=i j=ik=ix=it=i
yjkxt
\
Figure 25 : 3D Pareto front for multi-objective optimization (OVW as a third
objective function)
66


Figure 26 : 3D Pareto front for multi-objective optimization (Percentage of vaccine
met demand as a third objective function)
4.7.5 Contribution of each echelon in cost and El of vaccine SC
Table 7 shows the contribution of each supply chain echelon to global warming and
investment cost. The vaccine production and packaging has a significant role in CC>2e
emissions (96.7%); the second echelon that generates the most CC>2e is vaccine
intercontinental transportation, but it is far less than the first echelon (2.6%). This
suggests that the activities for CC>2e reductions should be focused on the vaccine
production echelon.
67


Table 7: Contribution of life cycle CChe emission and cost of various echelons of
vaccine SC
vaccine echelon CChe Contribution investment Cost Contribution
Vaccine Production and Packaging 96.7% 84.31%
Transportation from Manufacturing Location to UNICEF Store in Copenhagen! 1) 0.17% 0.36%
Ice Production for Cooling Vaccines During Transportation (1) 0.03% 0.02%
Refrigeration Process in UNICEF Store in Copenhagen 0.005% 0.001%
Transportation from UNICEF Store in Copenhagen to National Store in ACCRA(2) 2.60% 0.17%
Ice Production for Cooling Vaccines During Transportation (2) 0.29% 0.02%
Refrigeration Process in National Store in ACCRA 0.006% 0.001%
Transportation from National Store in ACCRA to Regional Stores(3) 0.009% 0.02%
Ice Production for Cooling Vaccines During Transportation (3) 0.03% 0.02%
Refrigeration Process in Regional Stores 0.02% 0.005%
Refrigeration Process in Clinics 0.10% 0.08%
Transportation from Regional Stores to Clinics(4) 0.008% 0.02%
Ice Production for Cooling Vaccines During Transportation (4) 0.06% 0.57%
Vaccination Waste Treatment 0.18% 0.18%
Open Vial Waste (OVW) " 14.36%
Similar to CC>2e emissions, vaccine production stage represents 84.12% of total vaccine
supply chain investment cost, but the cost associated with Open Vial Waste is next with
68


14.36% of total cost. Thus, solutions that decrease the rate of wastage could result in
significant cost savings.
As mentioned before, the cost and environmental emissions per dose of large vial sizes is
the lowest. On the other hand, the higher dosage per vial will result in an increase in rate
of OVW and consequently the OVW cost will increase. Multi-objective optimization will
find a trade-off solution (optimum mixture of vaccine production/packaging technology).
4.8 Time scheduling for vaccine transportation:
Table 8 illustrates the current vaccine transportation plan and Optimum results from
NSGAII code for Ghana. Currently, the vaccine transportation from national level to
regional level takes place 4 times per year (every 3 months) [57], It is obvious that this
schedule is not efficient since it requires high capacity for cold chain infrastructure. But
in our optimum scenario, the transportation frequency is monthly, and this frequency
prevents the high vaccine accumulation in cold rooms and reduces the inventory cost.
69


Table 8: Comparison of current and optimum vaccine shipment frequency
Optimum Shipment Schedule Current Shipment Schedule in Ghana
L M ^ ^ uhimi = 3532 lit L M y y Uhlmi = 8618 lit
1 = 1 m=1 1=1 m=l
L M ^ ^ ^hlm2 3064 lit L M ^ ^ ^hlm2 ~ 0 Ht
1 = 1 m=l 1 = 1 m=l
L M uhim-3 = 3335 lit L M y y uhim3=ou
1 = 1 m=l 1 = 1 m=l
L M y uhlm4 = 3069 lit L M y y uhlm4 = 8618 lit
1 = 1 m=l 1=1 m=l
L M y y uhlm5 = 3086 lit L M ^ ^ ^hlm.5 ~ 0 Ht
1 = 1 m=l 1 = 1 m=l
L M y y uhlm6 = 2847 lit L M ^ ^ Ufilm6 0 Ht
1 = 1 m=l 1 = 1 m=l
L M y y uhlm7 = 2844 lit L M y y uhlm7 = 8618 lit
1 = 1 m=l 1=1 m=l
L M y y ulm8 = 2790 lit L M ^ ^ ^lm8 0 Ht
1 = 1 m=l 1 = 1 m=l
L M y y ulm9 = 3274 lit L M ^ ^ M-lm 9 OUt
1=1 m=l 1 = 1 m=l
L M y y Ulmio = 2363 lit L M y y Ulmio = 8618lit
1=1 m=l 1 = 1 m=l
L M y y ulmll=2191 lit L M ^ ^ ^imll 0 Ht
1 = 1 m=l 1 = 1 m=l
L M y y ulml2 = 2363 lit L M ^ ^ ^lml2 0 Ht
1=1 m=l 1 = 1 m=l
70


4.8.1 Optimization results based on different objective functions:
Figure 28 shows the life cycle C02e emissions (indirect + direct emission) through
vaccine SC in Ghana, the current SC in Ghana produces (2.2 x 107 kg) C02e. As it is
shown in Fig 28, El optimization causes the highest reduction for C02e emission (24%).
This reduction is equal to carbon emissions of 1004 passenger vehicle during one year
[61],It is depicted in Figure 29 that optimization based on total investment cost causes a $
13, 158,000 reduction in vaccine supply chain in Ghana (22% reduction). When this cost
is averted, policy makers will be able to invest more towards new infrastructure.
2.50E+07
2.00E+07
O)
g 1.50E+07
U
W) 1.00E+07
*
5.00E+06
0.00E+00
Cost Optimization Environmental Multi-Objective Current Situation
Impact Optimization
Optimization
Figure 27: Effect of different objective functions on CChe
7.00E+07
n- j.
2 "Z 6.00E+07
1/5 -S 5T
3 rH 5.00E+07
M
a N 4.00E+07
0> X* Q
S 5-S5 3.00E+07
M p
in 3 w
g £ 2.00E+07
e m c
S a 1.00E+07
+2 u ^5 0.00E+00
o
E > Cost Optimization Environmental Multi-Objective Current Situation
Impact Optimization
Optimization
Figure 28: Effect of different objective functions on total Cost
71


4.9 Conclusion:
In this work, a method to design environmentally friendly and profitable vaccine SC is
presented. The model consisted of multi-period Continues linear programming that
accounts for multi-objective optimization of economics and environmental
considerations. The model considers the long term strategic decisions (e.g. selection
suppliers, shipment modes, refrigeration technologies) with mid-term time schedule
planning for SCs. The vaccine SC was modeled as a multi-objective optimization model
and a stochastic methodology (NSGAII) was adopted as solution methodology. Design
results indicate that a combination of 20 and 5 dose vaccine in production/packaging
section, ocean freight in inter-continental transportation, combination of electrical ice-
lined and solar direct drive vaccine refrigerators in cold storage, and combination of
incineration and autoclaving for medical waste disposal results in optimum configuration
network based on economic and El objectives. The solution to the scheduling problem
suggests that although increasing the shipment frequency between different echelons
leads to increased shipment cost and El, it results in significant reductions to the
inventory cost and El associated with cold room warehouses. Therefore, high shipment
frequency leads to optimum results. By implementing both the designing and time
scheduling optimization scenarios, 24% carbon emission and 22% total investment cost
reductions can be achieved.
72


REFERENCES
1- WHO world health statistics 2012: Home Page-
http://www.who.int/gho/publications/world health statistics/2012/en/
(Accessed June 17, 2013)
2- WHO world health statistics 2012: Home Page-
http://www.who.int/gho/publications/world health statistics/2012/en/
(Accessed June 17, 2013)
3- Gavi Alliance, Home Page- http://gavialliance.org/support/nvs/pneumococcal/
(accessed June 26, 2013).
4- Atherly, D. E., Lewis, K. D., Tate, J., Parashar, U. D., & Rheingans, R. D.
(2012). Projected health and economic impact of rotavirus vaccination in
GAVI-eligible countries: 2011-2030. Vaccine, 30, A7-A14.
5- Saxenian, H., Cornejo, S., Thorien, K., Hecht, R., & Schwalbe, N. (2011). An
analysis of how the GAVI alliance and low-and middle-income countries can
share costs of new vaccines. Health affairs, 30(6), 1122-1133.
6- Chen, Sheng-I. Modeling the WHO-EPI vaccine supply chain in low and
middle income countries. PhD Dissertation. University of Pittsburgh, 2012.
7- Lee, Bruce Y., et al. "Impact of introducing the pneumococcal and rotavirus
vaccines into the routine immunization program in Niger." American journal of
public health 1022 (2012): 269-276.
8- Assi, T. M., Brown, S. T., Kone, S., Norman, B. A., Djibo, A., Connor, D. L.,
... & Lee, B. Y. (2013). Removing the regional level from the Niger vaccine
supply chain. Vaccine, 31(26), 2828-2834.
9- Ferrari, Matthew J., Ottar N. Bjornstad, and Andrew P. Dobson. "Estimation
and inference of Ro of an infectious pathogen by a removal method."
Mathematical biosciences 198A (2005): 14-26.
10-Bojarski, A. D., Lainez, J. M., Espuna, A., & Puigjaner, L. (2009).
Incorporating environmental impacts and regulations in a holistic supply chains
modeling: An LCA approach. Computers & Chemical Engineering, 33(10),
1747-1759.
73


11- Zaffran, M., Vandelaer, J., Kristensen, D., Melgaard, B., Yadav, P., Antwi-
Agyei, K. O., & Lasher, H. (2013). The imperative for stronger vaccine supply
and logistics systems. Vaccine, 31, B73-B80.
12- Introduction of rotavirus vaccines into national immunization programs,
Immunization, Vaccines and Biologicals, WHO. Home Page
http://whqlibdoc.who.int/ha/2009AVHO IVB 09,09 eng.pdf (accessed June
17, 2013).
13- Burton, A. (2007). Solar Thrill: Using the sun to cool vaccines. Environmental
health perspectives, 115(4), A208.
14- WHO, Home Page
http://who.int/immunization standards/vaccine qualitv/pqs e03 fridges freez
ers/en/ (accessed 17 June 2013).
15- RFVB134a Solar Vaccine Refrigerator, Sunfrost Company. Areata, California,
US. Home page http://www.sunfrost.com/vaccine refrigerators.html (accessed
17 June 2013).
16- BFRV55, Solar Battery Free Vaccine Refrigerator. Sundanzer company,
Texas, US. Home Page- http://sundanzer.com (accessed 26 June 2013).
17- BLF100, Electricity Powered Vaccine Refrigerator, Surechill Company, UK
Home Page- http://surechill.com (accessed 26 June 2013).
18- Solar Air Conditioning and Refrigeration; Sayigh, A.A.M, MCViegh, J.C,
Pergamon Press, 1992.
19- Innovation Performance Savings. ENERGY STAR Report, Refrigerators
2007 Partner Resource Guide. Home Page
www.howards.com/eStar/2007Refrigerator prg.pdf (accessed 1
December 2013)
20- US Life Cycle Inventory: National Research Laboratory, (NREL), US Home
Page http://www.nrel.gov/lci/ (accessed 26 June 2013).
21- Life Cycle Inventories of Energy Systems, Results for Current Systems in
Switzerland and other UCTE Countries. Ecoinvent Report NO 5, December
2007. Home page-
http://www.ecolo.org/documents/documents in english/Life-cvcle-analvsis-
PSI-05.pdf.
74


22- Refrigerant Charge Calculator (Version 1.1). Guidance on Impact of the
Fluorinated Gas Regulations for Users of Stationary Refrigeration, Air-
Conditioning and Heat Pump January 2007.
23- McCulloch, A., & Lindley, A. A. (2003). From mine to refrigeration: a life
cycle inventory analysis of the production of HFC-134a. International Journal
of Refrigeration, 26(8), 865-872.
24- Global Comparative analysis of HFC and alternative Technologies
Refrigeration, Air Conditioning, Foam, Solvent, Aerosol Propellant and Fire
Protection Applications; Arthur D. Little, Inc., Cambridge Massachusetts,
2002.
25- Agarwal, R; Anderson, J; Bivens, D; Colboume, D; Hundy, G; Konig, H;
Lundqvist, P; Mclnemey, E; Neksa, P. IPCC/TEAP Special Report:
Safeguarding the Ozone Layer and the Global Climate System (Chapter 4:
Refrigeration). Home page- http://www.ipcc.ch/pdf/special-
reports/sroc/sroc04.pdf (accessed 26 June 2013).
26- Sand, J. R., Fischer, S. K., & Baxter, V. D. (1997). Energy and global warming
impacts of HFC refrigerants and emerging technologies (pp. 5-60). Oak Ridge,
TN, USA: Oak Ridge National Laboratory.
27- Boxwell, M. Solar Electricity Handbook. 2013 Edition. Home page-
http ://solarelectri cityhandbook.com (accessed 26 June 2013)
28- Sunvia Company. OPTXXX-60-4-1B0 (60 CELL MODULE). Home page
http://sunvia.com (accessed 17 June 2013)
29- Fthenakis, V. M., Kim, H. C., & Alsema, E. (2008). Emissions from
photovoltaic life cycles .Environmental science & technology, 42(6), 2168-
2174.
30- Trojan Company. Deep Cycle Battery. US Home Page-
http://trojanbattery.com (accessed 26 June 2013)
31- Sullivan, J. L., & Gaines, L. (2010). A review of battery life-cycle analysis:
state of knowledge and critical needs (No. ANL/ESD/10-7). Argonne National
Laboratory (ANL).
32- Horie, Y. A. Life Cycle Optimization of Household Refrigerator-Freezer
Replacement. MS Dissertation (Report No. CSS04-13), University of
Michigan, August 2004.
75


33- Boustani, A; Gutowski, T; Graves, S. Appliance Remanufacturing and Energy
Savings. January 2010. Home Page-
http://web.mit.edu/ebm/www/Publications/MITEI-l-a-2010.pdf (accessed 26
June 2013)
34- Guide to Electric Power in Ghana, University of Ghana. 2005 Home Page-
http://www.beg.utexas.edu/energyecon/IDA/USAID/RC/Guide to Electric%2
OPower in Ghana.pdf (accessed 26 June 2013)
35- Riberio, F. M; Da Silva, G. A. Life Cycle Inventory for Hydroelectric
Generation: A Brazilian Case Study. J. Cleaner Production 2010, 18(1), 44-54.
36- U.S. Environmental Protection Agency (EPA), Tools for the Reduction
and Assessment of Chemical and other Environmental Impacts (TRACI) Home
Page http://epa.gov/nrmrl/std/traci/traci.html (accessed June 26 2013)
37- Ryberg, M., Vieira, M. D., Zgola, M., Bare, J., & Rosenbaum, R. K.
(2014). Updated US and Canadian normalization factors for TRACI 2.1. Clean
Technologies and Environmental Policy, 16(2), 329-339.
38- Southern Africa Grid Map, Eskom Holdings Limited Annual report
2009,HomePagehttp://www.financialresults.co.za/eskom_ar2009/ar_2009/dow
nloads/0l_profile.pdf (accessed June 26 2013)
39- Eckelman, M.J,Mauter, M.S, Isaacs,J.A, Elimelech, M. New Perspectives on
NanoMaterial Aquatic Ecotoxicity: Production Impacts Exceed Direct
Exposure Impacts for Carbon Nanotoubes. Journal Environmental Science and
Technology 2012, 46, 2902-2910
40- US Energy Information Administration (EIA), Home Page:
http://www.eia.gov/countrie (Accessed 30 November 2013)
41- Electrical Power in Africa, Home Page:
http://mbendi.eom/indv/power/af/p0005.htm#country profiles (accessed 24
June 2013).
42- Hendrickson, C., Horvath, A., Joshi, S., & Lave, L. (1998). Peer reviewed:
economic input-output models for environmental life-cycle assessment.
Environmental science & technology, 32(1), 184A-191A.
43- UNICEF vaccine price data, home page,
http://www.unicef.org/supplv/index 57476.html (accessed January 2015).
44- SIMAPRO 8 LCA software. Home page: (http://simapro.com/)
76


45- Transportation European Commission,
http://ec.europa.eu/transport/index en.htm
Home Page
46- Boudehenn, F., Demasles, H., Wyttenbach, J., Jobard, X., Cheze, D., &
Papillon, P. (2012). Development of a 5 kW cooling capacity ammonia-water
absorption chiller for solar cooling applications. Energy Procedia, 30, 35-43.
47- Mendivil, R., Fischer, U., Hirao, M., & Hungerbiihler, K. (2006). A New LCA
Methodology of Technology Evolution (TE-LCA) and its Application to the
Production of Ammonia (1950-2000)(8 pp). The International Journal of Life
Cycle Assessment, 11(2), 98-105.
48- Solar Collector Inc. Home Page: (http://www.solarcollectorinc.com)
49- Ardente, F., Beccali, G., Cellura, M., & Brano, V. L. (2005). Life cycle
assessment of a solar thermal collector. Renewable Energy, 30(1), 1031-1054.
50- Zhao, W., van der Voet, E., Huppes, G., & Zhang, Y. (2009). Comparative life
cycle assessments of incineration and non-incineration treatments for medical
waste. The International Journal of Life Cycle Assessment, 14(2), 114-121.
51- Kalyanmoy, et al. "A fast and elitist multi-objective genetic algorithm: NSGA-
II." Evolutionary Computation, IEEE Transactions on 6.2 (2002): 182-197.
52- Baumann, H., & Tillman, A. M. (2004). The Hitch Hiker's Guide to LCA. An
orientation in life cycle assessment methodology and application. External
organization
53- Hugo, A., & Pistikopoulos, E. N. (2005). Environmentally conscious long-
range planning and design of supply chain networks. Journal of Cleaner
Production,! 3(15), 1471-1491.
54-You, F., Tao, L., Graziano, D. J., & Snyder, S. W. (2012). Optimal design of
sustainable cellulosic biofuel supply chains: multi objective optimization
coupled with life cycle assessment and input-output analysis. AIChE
Journal, 55(4), 1157-1180.
55- Jolliet, O., Margni, M., Charles, R., Humbert, S., Payet, J., Rebitzer, G., &
Rosenbaum, R. (2003). IMPACT 2002+: a new life cycle impact assessment
77


methodology. The International Journal of Life Cycle Assessment, 5(6), 324-
330.
56- Assi, T. M., Brown, S. T., Djibo, A., Norman, B. A., Rajgopal, J., Welling, J.
S., ... & Lee, B. Y. (2011). Impact of changing the measles vaccine vial size on
Niger's vaccine supply chain: a computational model. BMC Public
Health, 17(1), 425.
57- Ministry of Health Ghana, Immunization Program Comprehensive Multi-year
Plan (2010-2014) Home Page, file:///C:/Users/Bahador/Downloads/Ghana-
Comprehensive%20multi-vear%20plan%20for%202010-2014%20-
%20Year%20Unknown%20(2).pdf (accessed January 2015).
58- Kilfoyle, D., & Ventre, G. G. (1988). Test and evaluation of vaccine
refrigeration systems. In Photovoltaic Specialists Conference, 1988.,
Conference Record of the Twentieth IEEE (pp. 1200-1205). IEEE.
59- Google Map, Home Page, https://www.google.com/maps (accessed in August
2105)
60- Rajgopal, J., Connor, D. L., Assi, T. M., Norman, B. A., Chen, S. I., Bailey, R.
R., ... & Lee, B. Y. (2011). The optimal number of routine vaccines to order at
health clinics in low or middle income countries. Vaccine, 29(33), 5512-
5518.Deb,
61- EPA, Office of Transportation and Air Quality, Home Page,
http://www.epa.gov/otaa/consumer/420f08024.pdf (accessed in January 2015).
78


APPENDIX
These unit processes happen in US
Transportation Ohana
Jl _A_
Unit Processes Refrigerant Production PV Panel Production Battery Production Material Processing Assembly Transportation Refrigerant Leakage Blowing agent Leakage
Electricity input (KWh) 2.1E-4 2.51 0.53 0.32 0.17 - - -
Heat input (KWh) 1.33E-3 7.54 1.58 0.96 0.15 - - -
Emissions to air (g)
Arsenic 3.72E-8 0.000316 0.007855 5.94E-5 3.1E-5 " " "
Benzene 1.38E-5 0.094388 0.007209 0.0123 0.003322 " " "
Cadmium 3.13E-9 9.46E-5 0.007855 4.99E-6 2.6E-6 " " "
Carbon dioxide 0.49 3348.137 3749.91 489.507 162.702 915.204 261.816 441.653
Carbon monoxide 7.41E-5 0.660 0.2397 0.086373 0.035 1.595 - -
Chromium 7.74E-8 0.000474 0.007855 0.000124 6.45E-5 " "
Hydrogen chloride 2.99E-5 0.363 0.076192 0.04768 0.025 - - -
79


Lead 5.1E-8 0.000592 0.00024 8.14E-5 4.25E-5 " " "
Mercury 3.45E-8 0.000158 6.25E-6 3.89E-6 2.03E-6 " " "
Methane 9.82E-4 7.225 0.9192 1.029 0.29456 0.03157 " "
Hydrogen fluoride 3.72E-6 0.0452 0.0095 0.005938 0.0031 - - -
NOx 0.00596 6.606 0.6993 0.7349 0.315 12.472 " "
PM10 1.67E-5 0.097 1.139 0.012617 0.002157 0.2898 " "
PM 2.5 " " " " " " " "
sox 0.000559 7.98 1.198 0.890644 0.4647 0.180274 " "
NMVOC 0.000517 0.699 0.599 0.091 0.021 0.57536 " "
Calcium Sulfate 0.0052 - - - - - - -
N20 8.5E-6 0.10336 0.025424 0.012113 7.056E-3 0.0201376 " "
Nickel 8.25E-8 0.001471 0.000247 0.000132 6.87E-5 " " "
Emissions to water
Arsenic " " 0.00646 " " " " "
Cadmium " " 0.0060 " " " " "
Chromium " " 0.0050 " " " " "
Lead " " 0.000959 " " " " "
Mercury 6.89E-9 " " " " " " "
BOD 3.3E-6 0.021 0.001307 0.002788 0.000672 " " "
COD 8.24E-6 0.069 0.009275 0.00906 0.003435 " " "
Table Al: LCI data of different unit process for the case of SPSB manufactured in US and used in Ghana
for functional unit of 1 liter cold storage per year
80


South Ghana or
UK Transportation Ghana
J^ Africa South^frica
f > ( \l\ (\ ( )
Unit Processes Refrigeran t Production Material Processing Assembly Transportation Electricity consumption in Ghana Electricity consumption in South Africa Refrigeran t Leakage Blowing Agent Leakage
Electricity input (KWh) 2.1E-4 0.32 0.17 - 9.21 9.21 - -
Heat input (KWh) 1.33E-3 0.96 0.15 - - - - -
Emissions to air (g)
Arsenic 3.72E-8 " " " 2.91E-7 0.000474 " "
Benzene 1.38E-5 " " " 9.08E-5 " " "
Cadmium 3.13E-9 " " " " " " "
Carbon dioxide 0.49 360.2621 118.205 141.8524 1485.08 14363.99 233.764 853.41
Carbon monoxide 7.41E-5 0.2863 0.04507 0.2036 4.7541 - - -
Chromium 7.74E-8 " " " 1.04E-7 " " "
Hydrogen chloride 2.99E-5 - - - - - - -
Lead 5.1E-8 " " " " " " "
Mercury 3.45E-8 " " " 4.58E-14 " " "
81


Methane 9.82E-4 0.02624 0.00413 0.00342 4.3850 3.4651 " "
Hydrogen fluoride 3.72E-6 - - - - - - -
NOx 0.0059 0.4122 0.18091 1.0104 1.2345 28.57161 " "
PM10 1.67E-5 0.003991 0.00062 0.01951 0.137863 2.4893 " "
PM 2.5 " 0.03027 0.01625 " 0.40602 3.8387 " "
sox 0.0005 0.40268 0.20260 0.031249 0.5744 134.1568 " "
NMVOC 0.0005 0.0642 0.02065 0.069801 0.6664 0.450654 " "
Calcium Sulfate 0.0052 - - - - - - -
N20 8.5E-6 0.01945 0.00030 0.005129 3.26E-6 0.25537 " "
Nickel 8.25E-8 " " " 1.32E-7 " " "
Emissions to water
Arsenic " " " " " " " "
Cadmium " " " " " " " "
Chromium " " " " " " " "
Lead " " " " " " " "
Mercury 6.89E-9 " " " " " " "
BOD 3.3E-6 0.08868 0.04760 " 0.014764 " " "
COD 8.24E-6 " " " 1.36E-6 0.38341 " "
Table A2: LCI data of different unit processes of EP manufactured in UK and used in Ghana or South
Africa for functional unit of 1 liter of cold storage per year
82


Full Text

PAGE 1

i MATHEMATICAL PROGRAM M ING MODEL FOR OPTIMAL DESIGN OF SUSTAINABLE VACCINE SUPPLY CHAIN IN DEVELOPING COUNTRIES By B AHADOR M OUSAVI B .Sc ., Mechanical Engineering at Yazd University A thesis submitted to the Faculty of the Graduate School of the Univer sity of Colorado in partial fulfillment of the requirements for the degree of Master of Science Civil Engineering Department 2015

PAGE 2

ii This t hesis for the Master of Science degree by Bahador Mousavi h as been approved for the Civil Engineering Program by Arunprakash Karunanithi Advisor Bruce Janson Chair Azadeh Bolhari November 20th 2015

PAGE 3

iii Mousavi Bahador ( MS Civil Engineering ) Mathematical Program m ing M odel for optimal Design of Sustainable Vaccine Supply Chain in Developing Countries Th esis directed by Ass ociated Professor Arunprakash Karunanithi ABSTRACT The enormous health and economic damage of epidemic diseases and the introduction of new vaccines, such as pneumococcal and rotavirus, which need significantly higher cold storage ca pacity and other infrastructures make designing a sustainable vaccine supply chain essential. The health policy makers in developing countries as well as low and middle income countries should be aware of the current and future needs of immunization inf rastructure based upon population growth rate and the introduction of these vaccines. Thus the availability of a model with the capability to recognize and design the vast vaccine infrastructure expansion is critical for health policy makers to realize fut ure immunization plans. Vaccine supply chain is a large, inter continental network with enormous cost and environmental implications. Therefore the design of not only a cost effective supply chain but also an environmental friendly one, is important. This Study addresses the optimal design and planning of vaccine supply chain based on environmental and economic objectives. The environmental objective is measured by a life cycle assessment methodology. With this approach, the environmental emissions associat ed with all echelons of vaccine supply chain, including vaccine manufacturing and packaging, transportation, cold chain and disposal was measured from a cradle to grave perspective. Further, the economic objective was evaluated by considering the total inv estment cost of each echelon of the supply chain. In order to optimize the vaccine

PAGE 4

iv supply chain by considering environmental and economic issues simultaneously, a multi objective continuous linear mathematical programming model is developed that accounts for major characteristics of the vaccine supply chain, including geographical and industrial diversity of suppliers, various shipment types different refrigeration options multiple disposal methods vaccine demand distribution, cold chain and transportat ion capacity. The resulting Pareto optimal curves display the tradeoff between the environmental and economic dimensions of vaccine supply chain. The form of and content of this abstract are approved. I recommend its publication. Approved: Arunprakash T Karunanithi

PAGE 5

v ACKNOWLEDMENTS It is difficult to explain how much I owe to my wife. Leila Aghili, who sacrificed her career and took care of everything during my MS. I specially want to thank my mother, father and brother for their love and sup port. They have supported every decision in my life and encouraged me to pursue my MS. This thesis is dedicated to them. I would like to thank my advisor, Professor Karunanithi, for guiding my research. Finally, I want to thank my friends and colleague who provided me with this career recommendation and inspiration.

PAGE 6

vi TABLE OF CONTENTS 1 INTRODUCTION ................................ ................................ ................................ ...... 1 1 .1 Motivation ................................ ................................ ................................ ....... 1 1.2 Contribution ................................ ................................ ................................ ...... 6 2 IS SOLA R REFRIGERATION ALWAYS A SUSTAINABLE OPTION ................ 8 2.1 Introduction ................................ ................................ ................................ ...... 8 2.2 Process description and system boundaries ................................ .................... 10 2.3 Life cycle inventories 12 2.4 Impact assessment ................................ ................................ .......................... 15 2.5 Uncertanity ................................ ................................ ................................ ..... 19 2.6 Normalized d amage a nalysis ................................ ................................ .......... 2 0 2.7 Sensitivity a nalysis ................................ ................................ ......................... 22 2.8 Ge ographical delineation of impacts ................................ .............................. 24 2.9 Discussion ................................ ................................ ................................ ....... 27 3 COMPARETIVE EI OF ALTERNATIVE TECHNOLOGIES IN VACCINE SUPPLY CHAIN ECHELONS ................................ ................................ .................... 31 3 .1 Goal and scope ................................ ................................ ................................ 31 3.2 Vaccine m anufacturing and packaging ................................ ........................... 31 3.3 Vaccine transportation ................................ ................................ .................... 33 3.4 Off grid refrigeratio n alternatives ................................ ................................ .. 35 3.5 Proce ss description and system boundaries ................................ .................... 35 3.6 Faunctional u nit ................................ ................................ .............................. 38 3.7 L CI of absorption refrigeration systems ................................ ......................... 38

PAGE 7

vii 3.8 Impact assessment ................................ ................................ .......................... 39 3.9 Results and discussion ................................ ................................ .................... 40 3.10 Medical waste treatment system ................................ ................................ ... 44 3.10.1 Incinerati on ................................ ................................ ......................... 44 3.10.2 Autoclaving ................................ ................................ ........................ 45 3.11. Refrence Flow ................................ ................................ .............................. 46 4 MATHEMATICAL PROGRAMMING MODEL FOR OPTIMAL DESIGN OF SUSTAINABLE VACCINE SUPPLY CHAIN ................................ .......................... 4 8 4.1 Motivation ................................ ................................ ................................ ...... 4 8 4.2 Introduction ................................ ................................ ................................ ..... 4 8 4 .3 Problem f ormulation ................................ ................................ ....................... 49 4.4 Mathematical model ................................ ................................ ....................... 52 4.4.1 Production refrigeration and transp ortation capacities ........................ 52 4.4.2 EI quantification methodology ................................ .............................. 55 4.4.3 Cost objective function ................................ ................................ ......... 5 9 4.5 Computational experiments ................................ ................................ ........... 60 4 .6 Parameter setting ................................ ................................ ............................ 61 4.7 Vaccine Supply Chain in Ghana ................................ ................................ .... 61 4.7.1 Network superstructure ................................ ................................ ......... 61 4.7.2 Results ................................ ................................ ................................ ... 63 4.7.3 Other objective function ................................ ................................ ........ 65 4.7.3.1 Open vial waste ................................ ................................ ............... 6 5 4.7. 4 Percentage of vaccine met demand ................................ ...................... 66

PAGE 8

viii 4.7. 5 Contribution of each echelon in cost and EI of vaccine SC ................. 68 4. 8 Time schedualing for vaccine transportation ................................ ................. 70 4. 8.1 Optimization resulats based on different objective functions .............. 7 2 4. 9 Conclusion ................................ ................................ ................................ ..... 7 3 REFRENCES ................................ ................................ ................................ ........ 7 5 APPENDIX ................................ ................................ ................................ ........... 81

PAGE 9

ix LIST OF TABLES Table 1 : Characteerist ics of the vaccine refrigeration technologies 11 2 : Distribution of human health impact for SPSB refrigerator in ghana ........................ 2 5 3 : Features and character s of various off grid vaccine refrigerators ................................ 36 4 : EI of various off gri d vaccine refrigerators ................................ ................................ ... 41 5 : EI and cost of alternative medical waste treatment technologies ................................ .. 46 6 : Ghana i mmunization program for under five childeren and pregnant women .............. 62 Table 7 : contribution of life cycle CO2e emission and cost of various echelons of va ccine SC. ................................ ................................ ................................ ................................ ..... 69 8 : Comparison of current and optimum vaccine shipment frequency ............................... 7 1

PAGE 10

x LIST OF FIGURES Figure s 1 : Schematic picture of vaccine supply chain ................................ ................................ .... 4 2 : Solar power supplimented with battery vaccine refrigerator ................................ ........ 11 3 : Cradle to grave system boundary for vaccine refrigerator ................................ ........... 12 4 : Comparetive EI of solar and conventional refrigerator in Ghana ................................ 16 5 : Unit process with breakdown of GWP impact for the case of Ghana ...................... 1 7 6 : Comparative impacts of 3 refrigeration options for the use in South Africa ............ 1 8 7 Comparison EI of EP used in Ghana and South Africa ................................ .............. 19 8 Monte Carlo simulation results for HH cancer, HH non cancer and ecotoxicity impacts ................................ ................................ ................................ ................................ ........... 2 0 9 : Normalized impacts of SPSB vaccine refrigeration system ................................ ......... 2 1 10 : Compar ative impacts based on various parameters ................................ .................... 24 1 1 : Transition plot for net GHG reductions ................................ ................................ ..... 28 1 2 : Dominant sources of electricity generation in African Countries .............................. 29 1 3 : Comparative GWP impact of manufacturing and packaging of one dose of various types of vac cine ................................ ................................ ................................ ................. 33 1 4 : Comparative GWP of transportation alternatives ................................ ....................... 34 1 5 : Fossil fuel and SAR refrigerator system boundary ................................ ..................... 38 1 6 : GWP breakdown for various refrigeration technologies ................................ ............ 43 1 7 : Total cost of various refrigeration systems for functional unit ................................ ... 44

PAGE 11

xi 1 8 : In cineration process system boundary ................................ ................................ ........ 45 19 : Autoclaving process system boundary ................................ ................................ ....... 46 2 0 : Vaccine SC reference flow ................................ ................................ ......................... 47 2 1 : Schematic graph of vaccine network super structure ................................ ................. 5 1 2 2 : Vaccine supply chain map for Ghana ................................ ................................ ......... 6 2 2 3 : Efficient set of Pareto solutions for the vaccine SC in Ghana ................................ .... 6 3 2 4 : Network configuration for multi objective optimization ................................ ............ 6 5 2 5 :3D Pareto front for multi objective optimization (OWV as a third objective function) ................................ ................................ ................................ ................................ ........... 6 7 2 6 : 3D Pareto front for multi objective optimization (Percentage of vaccine met demand as a third objective function) ................................ ................................ ............................. 6 8 2 7 : Effect of different objective functions on CO2e ................................ ......................... 72 2 8 : Effect of different objective function on total cost ................................ ..................... 73

PAGE 12

xii NOMENCLATURES Sets and indices Descriptions T J Potential vaccine UNICEF stor ages K X Potential vaccine national Stor age s L Potential vaccine shipment modes from UNICEF storages to national Levels I Potential vaccine regional level stor ages M Potential number of clinics NN Potential vaccine refrigeration te SS O FF B E N Life Cycle assessment stage s PE Classification of impact e under normalized category n Potential various types of vaccines h H

PAGE 13

xiii Continuous variables Descriptions Production rate of liter of vaccine h by supplier j to UNICEF storage k with technology x during interval t [liter/month] Amount of liter of vaccine h which is shipped from UNICEF storage k to national store l with shipment mode i dur ing interval t [liter/month] Amount of liter of vaccine h which is shipped from national store l to regional store m during interval t [liter/month] Amount of liter of vaccine h which is shipped from regional level m to clinic nn with refrigeration technology ss during interval t [liter/month] Amount of syringes and needles which is disposed from clinic nn to disposal location o with disposal technology ff during interval t [Kg/month] Environmental impact in terms of category e resulting from life cycle stage pe during interval t [impact/month] Normalized environmental impact in terms of category n resulting from life cycle stage pe during interval t [points/month]

PAGE 14

xiv Parameters Descriptions Emissions inventory of burden b resulting from the unit reference flow of life stage pe Characterization factor used to convert burden b into impact category e Normalization factor for impact cate gories belonging to group n Weighting factor for each normalized impact category n Unit production cost of vaccine using technology x during interval t [$/liter vaccine] Unit transportation cost based on the distance be tween j and k nodes during interval t Unit ice production cost during transportation stage 2 in time t [$/liter vaccine] Unit refrigeration cost Unit cost of disposing of immunization wastes [$/Kg of immunization was te] Transportation capacity of UNICEF vaccine storage k during interval t Production capacity of manufacturer j during interval t Refrigeration capacity of national level store l during interval t Transportat ion capacity of national store l during interval t Refrigeration capacity of regional level store m during interval t

PAGE 15

xv Transportation capacity of regional store m during interval t YF vaccine demand in clinic nn duri ng interval t Transportation capacity of clinic nn during interval t Refrigeration capacity of clinic nn during interval t Capacity of disposal device located in location o during interval time t Tran sportation capacity of disposal location o during interval t Distance between supplier j with technology x and UNICEF store k during interval t Distance between UNICEF store k and national store l with shipment mode i dur ing interval t Distance between national store l and regional store m during interval t

PAGE 16

1 CHAPTER 1 INTRODUCTION 1.1 Motivation Epidemic diseases kill millions of people each year, especially in low and middle income countries. Therefo re, preventing the spread of these diseases, and controlling them, can prevent significant social and economic damage. Vaccination is a very cost effective method to control epidemic diseases, especially for children under the age of five. Delivery of vacc ines to developing countries is facilitated by a large intercontinental supply chain. In addition, most vaccines must be maintained within a specific temperature range during transportation. This requires a strictly controlled cold storage infrastructure. It is obvious then that the supply chain management of pharmaceutical products and especially of vaccines, is very important. This is why there is a great deal of research that has been done and is on going. The introduction of new types of vaccines each y ear causes the vaccine supply chain to change in each country on a frequent basis. New vaccines need more storage capacity, and hence health policy makers need the flexibility to modify their immunization infrastructure at any time as changes dictate. In t he last few years, great advances have been made in introducing new vaccines and expanding the reach of immunization programs globally [1] New vaccines such as pneumococcal conjugate for pneumonia and rotavirus for diarrhea are currently being rolled out or planned to be introduced globally with a particular emphasis in low and middle income countries. These

PAGE 17

2 programs are extremely important as 18% and 12% deaths of all children under the age of five in low income countries occur due to pneumonia and diarrhea respectively [2]. GAVI alliance is at the forefront of vaccine program plans to support 40 low and middle income countries in introducing rotavirus vaccines and to immunize 50 million children by 2015 [3]. It is estimated that by 2013 about 2.4 million deaths of children can be averted globally through accelerated introduction of rotavirus vaccines [4]. Similarly, GAVI also plans to introduce pneumococcal vaccines to nearly 60 countries potentially averting 500,000 deaths by 2015 and up to 1.5 m illion deaths by 2020 [3]; unfortunately the global numbers for death are significantly higher. In 2010, only a tiny fraction of the 0.5 billion dollars spent on vaccine programs in low and middle income countries has been allocated to rotavirus and pneum ococcal vaccines. It is projected that by 2015, half of the total of 1.2 billion dollars will be spent towards these two vaccines alone [5]. The volume of a dose of these two vaccines is higher than other traditional vaccines, so providing cold rooms with sufficient capacity for handling these vaccines through their supply chain is a challenging task for public health officials in low and middle income countries. Vaccine supply chain (SC) is necessary for d istributing vaccine from manufacturers to the fin al consumers and it includes various steps such as production and packaging transportation cold chain and disposal. Vaccine demand fluctuations with location and time and an unfixed production rate of manufacturers are challenging points in this SC In t he low and middle income countries due to ascending birth rate s vaccine demand is rapidly increas ing Equity and efficiency are two critical parameters that must be considered in vaccine Supply Chain Management (SCM) This means that the vaccine

PAGE 18

3 should be provided uniformly for each recipient in the system while the infectious diseases are averted as much as possible Chen [6] optimized the national level vaccine SC based on maximizing fully immunized child and vaccine met demand. Lee et al. [7] analyzed the impact of introduction of Pneumococcal and Rotavirus vaccines on total vaccine demand of Niger and its consequences on Niger immunization infrastructure. Assi et al. [8] conducted a study on the effects of removing one echelon of Niger vaccine supply chain on vaccine met demand. Additionally, some models were developed to forecast the spread of infectious diseases, and to forecast the special vaccine demand in each region. For example, Ferrari et al. [9] developed a method to estimate the epidemic effe cts of infectious deceases with the capability to predict the special vaccine demand in a given location. In the past SCM was used to design networks that led to the maximum economic performance, but more r ecently environmental impact issues have become one of the significant concerns in SCM. Today, t here are design methods and tools available that can suggest and advance the most eco friendly options in each echelon of a given SC T hese efforts show that integration of environmental issues with other SCM objectives such as economic functions, cou ld be an important topic of future research. The i mplemen tation of environmental targets in SCM is commonly referred to as Green SCM [10] Immunization supply systems of low and middle income countries are not ye t optimized to reduce environmental impacts. People in low and middle income countries are most vulnerable to environmental, climate change and human health impacts. It is proven that strategies to reduce greenhouse gas (GHG) emissions indirectly result i n wide ranging public health co benefits so any efforts related to the use of a sustainable infrastructure within vaccine supply chain can avert negative human health

PAGE 19

4 impacts Environmental impact (EI) can be minimized in a vaccine supply chain in terms of energy and resource efficiency and waste management For example it is possible to reduce environmental impacts by replacing kerosene refrigerators or inefficient old refrigerators with solar powered refrigerators, or by using thermostable vaccines that do not require ice packs during their transportation. Selection of sustainable transportation options such as electric vehicles or ocean freights can make the SC more eco friendly [11].By providing economic incentives for innovation in equipment manufacturin g, WHO and PATH encourage vaccine refrigerator production companies to use solar technology for their future products. In order to initiate further new innovation and new directions, WHO established new standards for refrigeration options which did not pre viously exist such as direct drive solar refrigerators and stationary passive cooling containers [11]. By implementing some funding and incentives for innovation in equipment manufacturing, WHO and PATH encourage vaccine refrigerator production companies t o use solar technology for their future products. On the other hand, WHO published new standards for products which had not previously existed like direct derive solar refrigerators and stationary passive cooling container s [ 11 ]. Vaccine SC consists of sev eral echelons, all of which have significant cost and environmental burdens (Figure 1). Furthermore, there are several alternative options available in each section of the SC. Therefore, designing and planning a sustainable vaccine SC becomes important for immunization programs.

PAGE 20

5 Figure 1: Schematic picture of Vaccine Supply Chain Chapter 2 discusses whether solar vaccine refrigeration is always a sustainable option for all African countries from a life cycle perspective. The electricity portfolio of e ach country is analyzed and a criteria is provided for selection of solar refrigeration option based on a countries electricity portfolio. In addition, the geographical distribution of environmental impacts of the cold chain is established. In chapter 3 w e calculate the environmental impacts, pollutant emissions and energy use of each echelon of this SC using Life Cycle Assessment methodology (LCA) from cradle to grave and a comparative study is performed for different alternative technologies that can be used in each part of the SC such as packaging, transportation, cold chain and disposal. The most eco friendly alternative technology in each part is recognized by LCA, the method that considers all of the upstream processes. In Chapter 4, Vaccine SC is si mulated with a generic mathematical model. This modeling is based on continuous linear programming (CLP). The ultimate goal is to optimize this linear algebraic system to achieve the most economic and eco friendly results. Therefore, a multi objective opti mization approach, Non Dominated Sorting Genetic Algorithm ( NSGA which is a stochastic optimization methodology was used with in the MATLAB programming language environment The developed code is capable of optimizing the vaccine SC not only for tr aditional objective s (such as vaccine met Vaccine Manufacture Packaging Transportati on Cold Storage Use Disposal

PAGE 21

6 demand, SC cost and open vial waste) considered by other researchers but also for environmental objectives. This model is applicable for designing a sustainable system by selecting the optimal alternative technolo gy in each part of SC and also for rearrangement of SC by optimal scheduling time for vaccine shipment between various parts of SC. The significant contribution s of this research are presented below 1.2 Contributions Quantifying the environmental impacts of all SC echelons with a comprehensive environmental assessment methodology (LCA) Creation of geographically dependent methodology for establishing the sustainability of solar vaccine refrigerators in developing countries. Creation of a continuous linear programming optimization model for vaccine supply chain which is solved using NSGA Creation of a multi objective optimization code which has the capability to optimize the supply chain in an environmentally conscious way, in addition to the traditional objectives associated with vaccine SC. Designing new optimal solutions for SC infr astructure and adjusting the schedules and rearrangement of shipment frequency of existing vaccine SC.

PAGE 22

7 CHAPTER 2 IS SOLAR REFRIGERATION FOR VACCINE COLD STORAGE IN LOW AND MIDDLE INCOME COUNTRIES ALWAYS A SUSTAINABLE OPTION? A COMPARETIVE LIFE CYCLE A SSESSMENT STUDY 2.1 Introduction An important component of vaccine programs is the cold chain, which is a collection of cold rooms, refrigerators, freezers and cold boxes that keep vaccines at an appropriate temperature. Vaccines require cold storage (betw een 2 o C and 8 o C) to remain stable and be potent when administered. Heat sensitive vaccines lose potency after exposure to temperatures above the recommended range and freeze sensitive vaccines lose potency when exposed to freezing temperatures. The influx of new vaccines has outstripped the capacity of current cold chain systems in many low and middle income countries [1]. Successful implementation of these new vaccine programs require augmentation of existing capacity with additional cold storage infrast ructure as the new vaccines need significantly higher storage volumes than traditional vaccines. For example, Rotavirus vaccines on an average require a total storage volume of 85.95 334.3 cm 3 per fully immunized child while traditional vaccines (such as v accines for measles ) on an average require only 15.4 cm 3 [12]. Therefore, in the next few years there is an urgent need to streamline and expand efficient and reliable cold chain infrastructure in these countries An additional 390,000 refrigeration units will be required by 2015 to sustain introduction of new vaccines in the 40 GAVI alliance countries alone. Refrigeration systems have significant environmental impacts due to several factors: They are energy intensive requiring significant amount of electri city or heat and leakage of refrigerants and sealing

PAGE 23

8 agents have impact on global warming and ozone depletion. In this context, the importance of building new cold chain infrastructure that minimizes the environmental impact of energy, materials and proces ses both within the country and globally cannot be overstated. This provides an opportunity, as replacement of existing infrastructure with eco friendly alternatives is not feasible in low and middle income countries due to the costs involved, but eco fr iendly options can be considered for the proposed new expansions. This has been recognized by important stake holders and policymakers [1] and there is a drive towards eco friendly cold chain infrastructure such as solar refrigeration [13], in the develop ing world. It is believed that solar refrigeration offers lower environmental footprint as it does not burn hydrocarbons for electricity production and hence does not produce carbon dioxide. However, to legitimately evaluate the sustainable features of s olar vaccine refrigeration we need to consider the life cycle impacts rather than just the use phase CO2 emissions. In addition consideration of health and ecological impacts is also critical. T o our knowledge there has been no study done to quantify the l ife cycle impacts of vaccine cold storage. A nalysis specific to vaccine refrigerators is very important as they are very different from domestic and commercial refrigerators due to their unique characteristics, stringent design specifications constraints [14] and complex supply chains involved In this study we examine the life cycle environmental impacts of solar refrigeration systems for vaccine storage and compare it with conventional vaccine refrigeration. In this study we examine the life cycle envir onmental impacts of solar refrigeration systems for vaccine storage and compare it with conventional vaccine refrigeration. This analysis will be critical for stakeholders such as World Health Organization (WHO), Global Alliance for Vaccines

PAGE 24

9 and Immunizati on (GAVI) Program for Appropriate Technology in Health ( PATH ), United Nations International Children's Emergency Fund ( UNICEF ) Gates foundation, and Governments and health agencies of developing countries who will be involved in decision making processes related to immunization programs. 2.2 Process d escription and s ystem b oundaries This analysis considers small scale off grid solar refrigeration systems and compares it with electrical powered refrigerators for vaccine storage in Ghana, which introduced the two new vaccines (Pneumococcal and Rotavirus vaccines) in 2012 and would require augmentation of their current cold storage capacity to sustain their immunization program. Also, Ghana is interested in solar refrigeration for vaccines. The following thr ee options are considered: a) Solar Powered Supplemented Battery (SPSB) [15] ; b) Solar Powered Battery Free (SPBF) [16] ; and c) Electricity powered (EP) refrigerators [17]. WHO [2] approved common commercial vaccine refrigerators (See Figure 2 and Table 1) were selected for this study. The life span time of SPSB and SPBF were considered as 15 years based on the study by Sayigh and MCViegh [18] while the life span of EP was also fixed at 15 years based on the Energy Star report [19] The system boundaries fo r the three refrigeration options are shown in Fig ure 3 T he unit processes included are: solar panel production (for SPSB, SPBF) ; battery production (for SPSB); refrigerant ( R134 a) production; blowing agent ( HCFC141 b ) production; other materials product ion; refrigerator assembling; transport; grid electricity consumption (for EP), refrigerant and blowing agent l eakage during use and disposal.

PAGE 25

10 Figure 2 : (a) Solar Powered Supplemented Battery Vaccine Refrigerator (SPSB), the battery used as a electric al storage [15] (b) Solar Powered Battery Free vaccine Refrigerator (SPBF), this type uses ice thermal storage of energy instead of battery (c) Electricity Powered Vaccine Refrigerator (EP) Table 1 : Characteristics of the vaccine refrige ration technologies Technical Characteristics SPSB SPBF EP Vaccine Storage net Capacity(Liter) 38.7 54.4 103 Shipping weight (kg) 91 118 175 Energy Consumption (12 volt) at ambient temperature of 43 C 0.61 kWh/day 0.86 kWh/day 2.6 kWh/day Autonomy(the number of days th at you need the system to operate when there is no power produced by PV panels) 7 days 7 days NA Refrigerant R134 a R134 a R134 a Manufacturing Location US US UK Life span time (years) 15 15 15

PAGE 26

11 2.3 Life c ycle i nventories A functional unit of 1 lite r cold storage capacity per year was selected. A life cycle inventory (LCI) was compiled for the three different refrigeration options. Most of the Figure 3 Cradle to grave system boundary for vaccine refrigeration systems

PAGE 27

12 LCI data was secondary and wer e derived from literature sources and standard LCI databases such as U.S. life cycle inventory database (USLCI) [20] a nd Ecoinvent [21] with a particular focus on geographical specificity All data which are reported in this section are relate d to the sele cted functional unit. The amount of refrigerant was calculated as 0.17g for all the types. [22] T he LCI related to refrigerant ( R134a ) production were derived from McCulloch and Lindley [23] which provided the cradle to gate emission factors for refrigera nt production The process tree for the refrigerant included production of the reactants (trichloroethane, hydrogen fluoride, chlorine and alkali) needed for the refrigerant, production of the precursors needed for these reactants (ethylene, chlorine & alk ali, sulfuric acid and lime), and extraction of raw materials (elementary flows) needed to produce these precursor materials (crude oil, solution mined brine, fluorspar mined, sulfur mined and limestone quarried). In addition, the direct global warming imp act associated with leakage of the refrigerant during the use and disposal phases was considered with the characterization factor for R134a, fixed at 1357 kg per kg of R134a [24,25] The amount of blowing agent needed was estimated to be equal to 0. 56 and 1.22 g for SPSB, SPBF, and EP respectively [26] All LCI data related to blowing agent (HCFC141 b) production and loss during installation were derived from Little et al [24] and the characterization factor for leakage of HCFC141 b was 704 Kg per kg of HCFC 141b Twelve month average solar irradiation for Ghana was estimated to be equal to 200 w/ [27].W e selected mono crystalline photovoltaic (PV) modules [28] and the number of modules needed to meet the requirement of 5.75 kWh (at 12 volt and 43 O C ambient temperature) for SPSB and SPBF was calculated as 0.18 for the selected functional unit. The ambient

PAGE 28

13 temperature of 43 O C hottest day of Ghana i.e. worst case scenario was considered for design purposes. Cradle to gate emissions rel ated to production of PV modules were derived from Fthenakis al [29]. T o meet the requirement of 7 days of autonomy (the number of days that the system need s to operate when there is no power produced by PV panels) 0.74 AH deep cycle parallel lead acid bat teries [30] (for SPSB) were considered Cradle to gate emissions related to battery manufacturing were derived from Sullivan et al.[31] The battery is only used in the case of SPSB and for EP, it is assumed that the cooler box is used during blackout. For the solar options we considered emissions related to truck and ocean freight transportation from California (manufacturing location) to Ghana. This equated to 2.90 tkm and 2.56 tkm ground transportation through diesel powered truck and 22.20 and 19.61 tk m ocean freight transportation for SPSB and SPBF respectively for the selected functional unit. For the EP refrigerator we considered ground transportation of 0.83 tkm from United K ingdom to Ghana. The assembly of refrigeration units included door assembly resin finishing, cabinet assembling and refrigeration cycle assembly [32] and associated energy consumption was obtained through personal communication with a vaccine refrigerator manufacturing company. [15] The total input energy for assembling phase was 0.32, 0.34 and 0.30 k W h for SPSB, SPBF and EP respectively. E missions associated with the production of all other materials needed such as resins, aluminum, copper etc. were derived from Boustani et al [33]. and the total input energ ies for their producti on was 1.28, 1.31 and 1.23 kWh for SPSB, SPBF and EP natural gas and steam power) [34] was used to derive emission factors for electricity used in the EP option [35] All em ission factors related to grid electricity and natural gas

PAGE 29

14 combustion for US and UK (manufacturing locations) were derived from USLCI [20] and Ecoinvent [21] respectively. The total electricity consumption of EP was 9.21 kWh. We assume that the entire refr igerant and blowing agent is lost to the atmosphere during the use phase or at the end of life without any recovery. Emissions associated with the disposal of the refrigerator unit were assumed to be negligible and recovery and recycle of solar modules and refrigeration unit was not considered to be realistic in low and middle income countries. The Complete Life Cycle inventory data for each unit process are presented in appendix (See T able s A1 A2 ). 2.4 Impact a ssessment I mpact assessment methods are used to translate the inventory data into environmental impacts through the use of characterization factors. In this study we adopted the impact assessment methodology TRACI [36] developed by the U.S. Environmental Protection Agency The following impact catego ries were considered: (1) global warming; (2) acidification; (3) eutrophication; (4) smog formation; (5) human health criteria; (6) human health cancer; (7) human health non cancer; and (8) ecotoxic i ty. In addition to midpoint metrics we also considered da mage indicators using TRACI 2.1 methodology [37] for the limited purpose of normalized damage analysis and for total health impact analysis. This section summarizes the key findings for the case of Ghana The total scores related to all of the 9 impact cat egories for the three different options were calculated for the functional unit of 1 liter cold storage/year The impact profiles related to SPSB, SPBF, and EP are shown in Figure 4 with SPSB impact set at 100% and the other two displayed at a level relat ive to the former. C omparison between the three options show that in all of

PAGE 30

15 the categories, except ozone depletion potential, the impact of SPSB was higher than SPBF which in turn was higher than EP. T he two solar refrigeration options had more than two ti mes higher climate impact than the conventional option. SPSB had two times more human health criteria impact (HH Criteria) than SPBF which implies that battery production plays a big role in criteria emissions (PM10, PM2.5, NO x SO x ). Further human health criteria for EP is five times less compared to both solar refrigeration options. The human health cancer impacts of EP is insignificant in comparison to the two solar refrigeration options. For human health non cancer and ecotoxicity categories, the impacts of SPBF is only about 3% of the impacts of SPSB. This huge difference can be attributed to emissions associated with lead acid battery production. The ozone depletion indicator can entirely be attributed to the leakage of blowing agent in all thre e options and has the following trend: EP > SPBF > SPSB (see Figure 5 ). With respect to photochemical smog, acidification potential, and eutrophication potential the impacts followed the same trend (SPSB > SPBF > EP) and were similar in relative magnitudes (SPBF and EP had roughly 75% and 15% impact as that of SPSB). Figure 4 : Comparative impacts of solar (SPSB and SPBF) and conventional vaccine refrigeration in Ghana

PAGE 31

16 Figure 5 a expand s on the above results by providing the breakdown of individual contrib utions of different unit processes towards GWP for SPSB. W e can see that mono crystalline PV module production has by far the highest contribution towards GWP but in the case of EP, electricity consumption of refrigerator during use phase has the highest c ontribution towards GWP. Unit process contribution towards GWP for EP is shown in Figure 5 b. Figure 5 : a) Unit process wide breakdown of GWP impact for the case of Ghana: for SPSB b ) Unit process wide breakdown of GWP impact for the case of Ghana f or EP The findings in the previous section contradicts the general belief that solar refrigeration systems are always more eco friendly than co nventional electric refrigeration The results suggest that life cycle impacts of EP systems depend greatly on th e electricity generation profile of the country where the refrigerator is used. In order to investigate the influence of the electricity generation mix on the results we developed a scenario where we compare the two refrigeration options for use in South A frica. The electricity production profile of South Africa is very different from that of Ghana (93% coal and 7% nuclear) [38]. The solar irradiation in South Africa is 230 W/ and the number of solar panel SPSB EP

PAGE 32

17 required remained the same. The results for South Africa as seen in Figure 6 suggests that for most of the impact categories the electric powered refrigeration option results in significantly higher impacts than solar r efrigeration option. This finding is completely opposite to the results of Ghana We find that for most of the categories the Ghana impacts are less than 20% of the South African impacts (Figure 7 ), This is a very significant difference which suggests that solar refrigeration offers environmental benefits in countries with a high contribution of fossil energy to their electricity mix and electric refrigerators seem to be better in countries having relatively low contribution of fossil energy to their electr icity mix. Figure 6 : Comparative impacts of the three refrigeration options for use in South Africa 0 20 40 60 80 100 120 Comparetive Percentage Results SPSB SPBF EP-South Africa

PAGE 33

18 Figure 7 : Comparison of environmental impacts of EP used in Ghana and South Africa 2.5 Uncertainty The developers of USEtox model which is used to de rive human health and ecotoxicty characterization factors for TRACI caution about the use of the characterization factors for metals due to the uncertainty present in the data. Therefore, a Monte Carlo simulation was performed by varying the characteriza tion factors (CF) of all metals by an order of magnitude [39] Uniform distribution was considered for this analysis. The impact categories associated with metal emissions include Human Health Cancer, Human Health Non Cancer and ecotoxicity. The F igures 8 (a),(b) and (c) show the mean value of the impacts with error bars (uncertainty) from Monte Carlo simulation for the cases of SPSB and SPBF used in Ghana (the EP health impacts are negligible). These findings indicate the importance of metal emissions towa rds human health and ecosystem impacts. Many heavy metals such as mercury, lead, chromium and nickel emissions during lead acid 0 20 40 60 80 100 120 Comparetive Percentage Results EP manufactured in UK and Used in Ghana EP manufactured in UK and Used in South Africa

PAGE 34

19 battery production were the main cause of higher health impacts of SPSB in comparison to SPBF. (a) (b) (c) Figure 8 : Monte Carlo simulation results for human health cancer, human health non cancer and Ecotoxicity impacts 2.6 Normalized d ama ge a nalysis In this section the normalized impacts of total climate change, total human health and total ecosystem quality are presented for SPSB option. Normalization allows us to compare different types of impact categories, through a common unit (e.g. p erson year),

PAGE 35

20 in order to identify the most important categories and their relative impacts. Normalization is accomplished by considering the life cycle emissions of a kind from a given process (e.g. vaccine refrigeration) in relation to the total emissions of the same kind by a given population. In this analysis we used damage categories rather than midpoint categories. The normalized damage factors were derived from TRACI 2.1 [37] this indicator is used for United States and Canada and the population was u sed for the normalization of SPSB vaccine refrigerator is population of US in 2008 (due to most unit processes of SPSB happen in US) Figure 9 shows that the climate change impact is an order of magnitude higher than both Ecotoxicity and Human Health impact s due to lower person year scores. Unlike GHG emissions, health related emissions are a local phenomenon and the impacts are greater close to the source of emissions. Therefore a detailed spatial assessment of distribution of local effects (health) is esse ntial. Figure 9 : Normalized impacts of SPSB vaccine refrigeration (Vertical axis is in logarithmic scale)

PAGE 36

21 2.7 Sensitivity a nalysis The results related to solar refrigeration option are sensitive to parameters such as solar irradiation and ambient tempe rature that have an influence on number of solar modules required. Further, the type of panel (monocrystalline vs cadmium Telluride ) will have a significant influence on upstream emissions and related impacts. On the other hand, the manufacturing and assem bly location of photovoltaic modules and refrigerators has significant effects on the environmental impacts of vaccine cold storage in developing countries. (Ghana in this case study) To analyze the variability of results we performed a sensitivity analysi s around these parameters. This study compares the GWP and HH criteria impacts of SPBF type between average data derived from Ghana and different conditions such as ambient temperature of 21c Solar radiation of 600 W/ CdTe photovoltaic modules a nd manufacturing and assembly location of solar modules and refrigerators in Sweden. The F igure 10 a shows that with deacresing the ambient temperature that vaccine refrigerator works at it about 22c the GWP and HH Criteria impacts decrese 46% and 48% res pectively.the energy consumption of vaccine refrigerators at 21C reduces dramatically and the number of monocrystalline modules that needed for running the refrigerator reduce from 10 to 4 module. On the other hand, the solar PV production has the signifi cant role in GWP and HH Criteria and changing the number of them changes these impacts dramaticaly so in this case the huge decrese in impacts seen. With incresing the solar raddiation the number of pv moudules for running the refrigerators decreses so in F igure 10 b the same trend was seen. Figure 10 c shows that using C admium Telluride instead of Monocrystalline cause to decresing the GWP impact about 30% because the

PAGE 37

22 production process of C admium Telluride has the lower GWP impact than monocrysstalline mo udule. but the HH Criteria of C admium Telluride is more than monocrysstalline about 42% A lthough the energy for production of one module of CdTe is 40% of monocrysstalline but because of lower efficiency of CdTe, the number of CdTe PV moudule for runini g the refrigerator is 1.4 times of monocrystalline. Figure 1 0 d shows the LCA of same PV modules and refrigerators used in ghana but manufactured and assembled in US and sweden. The difference between GWP and HH Criteria is due to different electricity pr ofile of US and sewden that cuase to US location has 22% more GWP impacts and 34% less HH criteria impacts.these results show that LCA study of cold chain of vaccine suuply is function of ma n y parameters and is dramatically sensitive to these pararmeters so the LCA Scenario for refrigerator that manufactured in US and used in ghana will be totally different for different manufacturing and using locations.

PAGE 38

23 Figure 1 0 (a) comparative impact based on ambient temperature 1 0 (b) co mparative impact based on solar radiation 1 0 (c) comparative impact based on material of module 1 0 (d) comparative impact based on manufacturing location. 2.8 Geographical delineation of impacts Emissions and impacts of vaccine cold storage infrastructure oc cur at multiple spatial levels. Typically, emissions of manufacturing the refrigerators, solar panels, and batteries occur in developed countries (North America and Europe) while the emissions during the use phase occur in the developing world (e.g. Ghana and South Africa). We disaggregated the human health impacts geographically based on manufacturing location and place of use as shown in Tables 2(a), 2(b) and 2(c). We clearly see that human health 0 20 40 60 80 100 120 Comparetive Percentage Results Ambient Temperature : 43C Ambient Temperature : 21C 0 20 40 60 80 100 120 Comparetive Solar Radiation Solar Radiation: 200 W/m Solar Radiation: 600 W/m 0 20 40 60 80 100 120 Comparetive Percentage Results Monocrystalline Module Cadmium Telluride Module 0 20 40 60 80 100 120 Comparetive Percentage Results Manufacturing in US Manufacturing in Sweden (a) (b) (c) (d)

PAGE 39

24 impacts related to solar refrigeration option happens outs ide (in the place of manufacturing) rather than the place of use (Ghana or South Africa) while for conventional refrigeration option there is a wide spatial spread of the impacts The unit of human health damage in this section is DALY (The disability adju sted life year (DALY) is a measure of overall disease burden expressed as the number of years lost due to mortality and morbidity) Table 2(a): Distribution of Human Health I mpact for SPSB use in Ghana Total Manufacture Location (US) % Contribution (US) % Contribution Intercontinental transport Use Location (Ghana) % Contribution (in Ghana) Total HH effects (DALY) 3.64E 6 2.41E 6 66 % 32 % 7.09E 8 2 % Table 2(b): Distr ibution of Human Health Impact for EP use in Ghan a Total Manufacture Location (UK) % Contribution (U K ) % Contribution Intercontinental transport Use Location (Ghana) % Contribution (in Ghana) Total HH effects (DALY) 6.1 E 7 3.08E 7 50 % 1 % 3.01 E 7 49 %

PAGE 40

25 Table 2(c): Distribution of Human Health Im pact for EP use in South Africa Total Manufacture Location (UK) % Contribution (UK) % Contribution Intercontinental transport Use Location (South Africa ) % Contribution (in South Africa) Total HH effects (DALY ) 1.38E 5 1.2E 7 1 % 1% 1.36E 5 98 % Table 2 (a) shows that for sol ar refrigeration option (SPSB), 66 % of all health impacts happen in the country where the refrigerator is manufactured (i.e. United States). This implies that stakeholders, policy makers and local governments should consider the choice of the country where the solar refrigerators are manufactured in order to minimize the life cycle human health impacts. This is also important because the medical treatment costs associated with these human hea lth impacts are significantly higher in the developed countries (i.e. the manufacturing location) resulting in greater societal and economic costs to these countries. Further, people who live near the battery and PV production locations in the developed co untries are most vulnerable to these human health impacts due to heavy metals such as Arsenic, Mercury and Lead emissions. In the case of electric power refrigeration (Table 2 (b) and Table 2 (c)) for both Ghana and South Africa we see that the health im pacts varied spatially. A bout two third of the impacts in the case of Ghana and almost all of the impacts in the case of South Africa happens at the location of use. This implies that policy makers in the developing world should consider the additional hea lth burdens that could be avoided in their countries by replacing conventional refrigerators with solar refrigerators. It is interesting to see that,

PAGE 41

26 although the total health impacts of EP refrigerators ( 6.10E 7 DALYS) used in Ghana is less than SPSB ( 3.6 4E 6 DALYS) (Table 2(a) and 2(c)) the local health impacts that happen specifically in Ghana for the EP system ( 3.01E 7 DALYS) is much more than SPSB (7.09E 8 DALYS ) For the case of EP refrigeration in South Africa (Table 2c) the local health impacts that happen in South Africa is 44 time higher than the respective impact in Ghana and this can be attributed to the differences in the electricity profiles of the two countries. In this case, people who live downwind of the thermal power plants in the developin g countries (where the refrigerant systems are used) are most vulnerable to human health impacts associated with criteria pollutant such as PM10, SO x and NO x 2.9 Discussion An important finding is that use of solar vaccine refrigerat ion in low and middl e income countries does not always translate into environmental benefits. Results indicate that under South African average GHG intensity of electricity s olar refrigeration options were found to reduce life cycle GHG emissions by 55 % in comparison to elec tric powered refrigerat ion while under Ghana average GHG intensity of electricity solar refrigeration options were found to actually increase life cycle GHG emissions by 65 %. The potential for solar vaccine refrigerators to be a sustainable alternative to electric powered conventional vaccine refrigerators is highly dependent on energy sources of electricity in the country of use This result is counter intuitive to public health officials as it is commonly assumed that solar refrigeration systems have mini mal impact on the environment and are more environmentally friendly than electric refrigeration systems. This study disproves that notion and provides a mechanism for public health officials to

PAGE 42

27 make informed cold storage choices based on country specific d ata (particularly electricity mix). Figure 1 1 illustrates the conditions (based on a countries electricity mix) under which solar refrigeration would offer greenhouse gas reductions. The horizontal line indicates the threshold at which transition to solar option will yield benefits. For countries with electricity mix [40] that produces more than 0.75 Kg CO2e/KWh (break even line) the use of solar systems will yield net GHG reductions. Figure 1 1 shows that the carbon intensity of South Africa with dominant c oal power plants and Nigeria with dominant natural gas power plants in their electricity profile is above the break even line but in the case of Congo and Ghana with dominant hydroelectricity power in their electricity profile the carbon intensity is under the break even line. Figure 1 1 : Transition plot for net greenhouse gas (GHG) reductions Figure 1 2 shows the dominant electricity source for most of the African countries [41] based on the type of power plant contributing the most electricity. Countri es near oil and

PAGE 43

28 natural gas fields have more non renewable power plants while hydropower (renewable) dominates in countries near high stream rivers. Information from Figures 1 1 and 1 2 will enable stakeholders and policymakers to select appropriate vaccine refrigeration option independently for each country in Africa without assuming any one type to be more sustainable. From Figure 1 2 we see that solar vaccine refrigeration can be an eco friendly option in many of the African countries that have significant contribution of coal, natural gas and oil to their electricity mix (such as Nigeria, Algeria, and Morocco). Assuming most of the new vaccine storage capacity is required in countries using significant amount of fossil fuel we estimate that use of solar vac cine refrigeration can cut carbon emissions equivalent to taking 65,000 cars off the road or preventing over 13,800 acres of tropical forest from deforestation. Figure 1 2 : Dominant Source of Electricity Production in African Countries The spatial disaggr egation of localized health impacts reveal that for the solar options almost all of the health impacts happen in the manufacturing location (USA or Europe)

PAGE 44

29 while for conventional option the majority of health impacts happen in the location of use (low and middle income countries). It is recommended that decision makers consider the spatial distribution of health impacts while making choices related to type of refrigerator and the manufacturing location. Finally, we would like to add that even though solar refrigeration option is not the sustainable choice in Ghana there might be situations where selection of the sustainable option might not be possible. As an example, the use of conventional electrical option in locations which do not have access to electr icity is not possible. Similarly, conventional electrical vaccine refrigeration systems are not portable and hence their use is possible only in health clinics but not for mobile immunization campaigns. In these situations solar refrigerators (or kerosene refrigerators) are the only option. Therefore, the comparative results from this study are only relevant for the situation where both electrical and solar refrigeration can be used (such as in central storage, distribution facilities and health clinics)

PAGE 45

30 Chapter 3 COMPARETIVE ENVIRONMENTAL OMPACT OF ALTERNATIVE TECHNOLOGIES IN VACCINE SUPPLY CHAIN ECHELONS 3.1 Goal and s cope The goal of this cradle to grave LCA study is to compare the environmental impacts of different vaccine manufacturing and p ackaging technologies, transportation, cold storage and medical waste treatment methods in the context of low and middle income countries. Four different types of vaccines, three different types of vaccine shipment modes, five different types of off grid v accine refrigerat ion options and two medical waste treatment methods with different system boundaries and characte ristic s are considered. Using this comparative LCA policy makers would have enough information to make informed choices about the eco friendly SC options for a given situation, and context. 3.2 Vaccine m anufacturing and p ackaging Life cycle assessment (LCA) tracks the physical, material, and energy flows through each stage or unit process and determines the resulting emissions and impacts.. Usin g LCA requires direct knowledge of the materials used to make a product. Due to the complex makeup of vaccines, such direct knowledge would require extensive pharmaceutical training to accurately estimate the composition of various vaccines. Economic inpu t/output data provides a way to estimate the impact of manufacturing a product based on its cost. The theory of EIOLCA [42] practice involves the assumption that emissions are directly related to economic value of products within a given sector. Thus, a va ccine which is twice as expensive demands twice as much energy and resources and therefore,

PAGE 46

31 generates twice as much emission. While this assumption does incur some uncertainty, this method is the only practical and viable option for this section. We select ed sectors which appropriately described the unit processes involved in the vaccine manufacturing and packaging. The database provides outputs in several categories, but categories of interest for our study were: Conventional Air pollutants, Greenhouse Gas es, and TRACI Impact Assessment. By running the model for these categories, impacts were calculated in terms of emission per $1M of vaccine. Using these results and scaling according to the cost of each vaccine produced a life cycle inventory of the emissi ons associated with the production of one dose of a given vaccine is estimated. The costs of the vaccines in question were obtained from private sector prices [43] from 2014 and were used in our calculations as the most appropriate estimation of the econom was assumed that these private sector prices included the cost of the vials, tubes or syringes in which the vaccines were contained as well as the packaging. From various sources, appropriate estimates for the cost of these supplies were determined and subtracted from the private sector prices to arrive at a cost for just the vaccine medication. Figure 1 3 a and 1 3 b, illustrate the global warming and human health impacts of manufacturing and packaging of one dose of different types of pneumococcal and rotavirus vaccines. Figure 1 3 a and 1 3 b, illustrate the global warming and human health impacts of manufacturing and packaging of one dose of different types of pneumococcal and rotavirus vaccines.

PAGE 47

32 Figur e1 3 a) Comparative global warming impact of manufacturing and packaging of one dose of various types of vaccine Figure 1 3 b) Comparative human health impact of manufacturing and packaging of one dose of various types of vaccine 3.3 Vaccine t ransportati on Transportation part in vaccine supply chain plays an important role due to large distances between manufacturing and demand locations. Also during transportation, vaccines should be kept within an appropriate temperature range. Therefore, providing a su stainable transportation system is very important in this SC. In this study, all of inter country transportations are considered with diesel trucks but for inter continental shipments three alternatives are made available: 1 diesel truck, 2 ocean freight a nd 3 airplane. In the following section (Figure 1 4 ) a comparative environmental impact and 0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 Pneumococcal (Pediatric) Pneumococcal Polysaccharide Rotavirus, Live, Oral, Pentavalent Rotavirus, Live, Oral, Kg CO2e 0.0000 0.0100 0.0200 0.0300 0.0400 0.0500 Pneumococcal (Pediatric) Pneumococcal Polysaccharide Rotavirus, Live, Oral, Pentavalent Rotavirus, Live, Oral, Kg PM10e

PAGE 48

33 cost assessment of these options is presented. It is worth mentioning that all of the data is derived from SIMAPRO software [44] and Transportation European Commiss ion [45]. (a) (b) (c) Figure 1 4 a) Comparative GWP of transportation alternatives, 1 4 b) Comparative HH impact of transportation alternatives, 15 c) Cost analysis for 1 ton km It is worth mentioning that the functional unit for this comparison study is 1 ton km. Figure 1 4 c) illustrates that diesel truck shipment cost is 8% and global warming potential (GWP) is almost 10% of the airplane shipmen t mode. But ocean freight cost is 27% of diesel truck and GWP of ocean freight EI is 17% of diesel truck. So it is obvious 0.12 0.02 1.23 0 0.2 0.4 0.6 0.8 1 1.2 1.4 Diesel Truck Ocean Freight Airplane GWP impact (Kg CO2e) global warming 1.87E 07 1.32E 07 4.06E 06 0.00E+00 5.00E-07 1.00E-06 1.50E-06 2.00E-06 2.50E-06 3.00E-06 3.50E-06 4.00E-06 4.50E-06 Diesel Truck Ocean Freight Airplane HH impact ( Kg PM10 e) Human Health 0.49 0.13 6.1 0 1 2 3 4 5 6 7 Diesel Truck Ocean Freight Airplane Transportation cost (USD 2014) Transportation Cost

PAGE 49

34 that ocean freight is the cheapest and most sustainable option for inter continental vaccine transportation. 3.4 Off grid refrigerato rs alternatives The goal of this section is to compare the environmental impacts of different off grid vaccine cold storage options in the context of low and middle income countries. Five different types of vaccine refrigerat ion systems with different sys tem boundaries and character istics are considered 3.5 Process d escription and s ystem b oundaries In this analysis, we compare the environmental impacts of small scale off grid solar refrigeration systems with absorption refrigerators which use fossil fuel s for vaccine storage. Different refrigeration options that are most relevant and related to existing cold storage infrastructure in low and middle income countries are considered: a) Solar Powered Supplemented Battery (SPSB) ; b) Solar Powered Battery Fr ee (SPBF); c) Kerosene Refrigerators (KR); d) LPG Refrigerators (LR); and e) Solar Absorption Refrigerators (SAR). Most of these types are commercially available brands of vaccine refrigerators. The characteristics of these refrigerators are depicted in t he Table 3

PAGE 50

35 Table 3 : Features and characters of various off grid vaccine refrigerators Technical Characters SPSB SPBF KR LR SAR Vaccine Storage net Capacity(Liter) 38.7 54.4 102 102 102 Shipping weight (kg) 91 99 118 86 118 Energy Consumption at ambient temperature of 43 C (kWh/day) 0.61 0.86 6.3 6.3 6.3 Fuel Consumption NA NA 0.9 Lit Kerosene in 24 hours 500 g LPG in 24 hours NA Autonomy 7 days 7 days 3.10 hours 3.02 hours 3.10 hours refrigerant R134 a R134 a ammonia ammonia ammonia Blowin g Agent HCFC141 b HCFC141 b HCFC141 b HCFC141 b HCFC141 b Refrigeration Cycle Vapor Compression Vapor Compression Absorption Absorption Absorption Manufacturing Location US US Sweden Sweden Sweden Life span time (years) 15 15 5 5 5 The system boundary of the two types of photovoltaic vaccine refrigerators is illustrated in Figure 3 (Chapter2) For photovoltaic refrigeration options, the unit processes included: photovoltaic production battery production (for SPSB ) refrigerant production (R134 a) blo wing agent production (HCFC141 b) other material s needed for

PAGE 51

36 refrigerator production assembly of refrigerator; transportation and disposal of refrigerators in landfill that includes the leakage of R134a and HCFC141b. The system boundaries for the two s olar options included no significant impacts during the use phase. The system boundary for fossil fuel refrigerators (Figure 1 5 ) includes ammonia (refrigerant) production, other material s for refrigerator production assembly of refrigerator, transportatio n, in use phase (See Figure 1 5 ) the kerosene and LPG (fuel) production and combustion is considered and in disposal section, leakage of R134a and HCFC141b are included. Most of the unit processes of SAR are the same with KR and LR but SAR has the producti on and manufacturing of solar thermal plate in manufacturing phase and does not have fossil fuel production and combustion during use phase.

PAGE 52

37 Figure 1 5 : Fossil fuel and SAR refrigerators system boundary 3.6 Functional u nit A functional unit of 1 liter of cold storage capacity per year was selected for this study. 3.7 Life c ycle i nventory of a bsorption r e frigeration s ystems The LCI has already been presented in Chapter 2. Therefore, in this section the LCI of absorption refrigeration systems is presented. The refrigerant for absorption refrigeration cycles is ammonia. Based on the Boudhenn et al study [46], 0.33g of ammonia is required for a KR, LR, or SAR options for the functional unit. All of the data related to

PAGE 53

38 LCI of production of ammonia were derive d from the Mendivil et al. [47] study. The amount of blowing agent (2.03g) was derived based on the work of James et al. (Ref) for 1 liter cold storage for one year. The energy needed for material process (1.95 kWh) and assembling of absorption systems (0. 48 kWh) was estimated based on the work of Boustani et al. [33]. Crude oil for the production and combustion of kerosene and LPG is assumed to be extracted from oilfields in Nigeria. The crude oil is transported to Ghana, and the refining process is assume d to take place in Ghana, and e kerosene and LPG were transported inside Ghana. In order to calculate the emissions associated with the production of a SAR unit, the size of the solar collector plate must be determined. Based on the solar radiation of Ghan a (200 w/ ) and the amount of thermal energy needed to operate absorption refrigerator (120 MJ based on functional unit), the net surface area of the required solar collector plate was estimated as 70 c All of these calculations were based on data from a solar collector company [48] and Ardente et al [49] work 3.8 Impact a ssessment The life cycle impact assessment methods are used to translate the inventory emissions data into environmental impacts through the use of characterization factors. These char acterization factors represent relative impacts of different chemicals (as in the case of global warming potential) or are a function of fate, exposure and effect of the chemicals (as in the case of human health and ecotoxicity). In this study we adopted the life cycle impact assessment methodology (LCIA) based on Tools for the Reduction and Assessment of Chemical and other Environmental Impacts (TRACI) developed by the U.S. Environmental Protection Agency. This study considered only midpoint impacts (deri ved from TRACI) as it was deemed that they provide sufficient details needed for

PAGE 54

39 our analysis and the use of endpoint damage modeling would bring in further uncertainty. The following impact categories were considered: (1) global warming; (2) acidification ; (3) eutrophication; (4) smog formation; (5) human health criteria; and (6) ozone depletion We applied classification and characterization steps to relate individual elementary flows in the inventory to the relevant impact categories and to identify char acterization factors based on the media where the emissions occur red We did not consider the normalization step as emissions at different stages in the life cycle occur red in different geographical locations. 3.9 Results and d iscussion This section summar izes the key findings comparing the different off grid vaccine refrigeration options. The total scores related to all of the 6 impact categories for the five different options are shown in Table 4. Results are based on functional unit of 1 liter vaccine co ld storage per year. The impact profiles related to SPSB, SPBF, KR, LR and SAR are shown in Table 4. In the category of Global Warming Potential, the KR has the maximum and SAR has the minimum GWP. Burning the LPG produces 45% fewer CO 2 e emissions than ker osene so LPG is a more eco friendly fuel than kerosene. Production of PV modules in SPSB also results in considerable GWP but less than KR and LR.

PAGE 55

40 Table 4 : EI of various off grid vaccine refrigerators Impact GWP (kgCO2 eq) HH Criteria kg PM10eq Acidi fication Kg H+ moles Eutrophication Kg N eq Smog air Kg O3 eq Ozone Depletion (kg CFC 11) SPS B 8.39 0. 0039 1. 38 9.28E 4 0.52 6.76E 05 SPBF 6.98 0.0022 1.07 7.12E 4 0.40 1.03E 04 KR 24.03 0.012 3.60 1.69E 3 1.28 2.45E 4 LR 13.56 0.0031 1.19 8.33E 4 0. 58 2.45E 4 SAR 5.72 0.0045 1.24 7.15E 4 0.39 2.45E 4 GWP of SPBF is less than SPSB because SPBF does not require lead acid battery production. The SAR type has the minimum GWP impact among all of the vaccine refrigerators. This system uses an absorption refrigeration cycle, similar to KR and LR. By replacing kerosene with a solar thermal collector, the CO 2 e emissions decreases by about 78% in comparison to KR. Policymakers concerned about GWP impact, should consider SARs as the best option. In the catego ry of Human Health (HH) Criteria impact, KR has the highest scores, and it is apparent that the ranking of other options is completely different to that of GWP ranking. Burning of kerosene fuel in the use phase of KR generates significant amount of PM10, S O x and NO x which results in higher HH impact score. While the SAR unit scored the lowest in GWP, it generated the second highest score in the HH impact category. This is due to the emissions caused during the production of solar collectors, water tanks a nd external supports. It is also interesting to note that LPG refrigeration is a cleaner option in terms of HH impact than SAR. The ozone depletion impact scores results entirely from the discharge of the blowing agent

PAGE 56

41 (HCFC141 b) to the atmosphere. The am ount of blowing agent in the types of KR, LR and SAR are the same, thus the ozone depletion impacts are equal for these three systems but the SPSB and SPBF systems use much less of the blowing agent resulting in a lower ozone depletion score. In the eutrop hication, acidification, and smog air categories, the trends can easily be seen from Table 4. Figure 1 6 expand s on the above results by providing the breakdown of individual contributions of different unit processes towards GWP, and HH criteria for all of the refrigerator types For the KR and LR option, it can be seen that kerosene production and combustion has by far the highest contribution towards both GWP and HH criteria impacts. For the SPSB, PV module production has the highest effect on GWP, while i n the category of HH criteria the lead acid battery production and PV module production have an equal contribution. So it is clear that HH impact of SPBF is less than SPSB. In the case of SAR, production of solar collectors has the highest contribution on both GWP and HH impact. It is obviously seen that for KR and LR types, most of these impacts take place in use phase (located in developing countries) while for the solar options most of the impacts are attributable to production which occurs in developed countries.

PAGE 57

42 (a) (b) Figure 1 6 (a): GWP breakdown for various refrigeration technologies 1 6 (b) HH breakdown for various refrigeration technologies In Figure 1 7 the total cost (capital cost + operating cost) of the four types of off grid vaccine refrige ration options are shown. The interesting point is that although the capital cost of solar options are much higher than that of the fossil fuel options, the total cost of solar options are less than fossil fuels. Therefore, the solar options are not only 0 5 10 15 20 25 30 KR LR SPSB SPBF SAR R134a Production ammonia Production R134a loss blowing agent loss LPG production and combustion kerosene production and combustion transportation solar collector production battery production PV module production assembly of refrigerator 0.00E+00 2.00E-03 4.00E-03 6.00E-03 8.00E-03 1.00E-02 1.20E-02 1.40E-02 KR LR SPSB SPBF SAR LPG production and combustion kerosene production and combustion transportation solar collector production battery production PV module production assembly of refrigerator material process of Refrigerator

PAGE 58

43 sustainable in many locations but also are more economical when operating costs are considered. Figure 1 7 : Total cost of various refrigeration systems for functional unit 3.10 Medical w aste t reatment s ystem In this section, an analysis is done around th e disposal echelon of vaccine supply chain and two alternative technologies are considered: 1 incineration, and 2 autoclaving. 3.10.1 Incineration Vaccine waste disposal is considered as an infectious medical waste. Therefore, to decrease the hazards the medical wastes are exposed to high temperature for long period of time (incineration). However, there are certain drawbacks such as ash disposal and emission of very toxic pollutants like dioxin. For quantifying the LCA emissions of incineration the system boundary shown in Figure 1 8 is considered [50].

PAGE 59

44 Figure 1 8 : Incineration process system boundary 3.10.2 Autoclaving Autoclaving is another disposal method. In this method, syringes and needles are sterilized by high temperature and pressure of steam. This method is costly but has very low pollutant emissions to the atmosphere. Disinfection of needles, syringes and open vials is guaranteed with autoclaving. The needles could sometimes carry HIV or other infectious viruses. Therefore, sterilizing them is vital. The system boundary of autoclaving process is shown in Figure 19 [50]. Ancillary Materials production (cement and etc) External energy production (el ectricity, natural gas) Hazardous waste incineration ( incineration, f lue gas cleaning, electricity production) Ash Sludge Residues landfill (stabilization, leachate treatment)

PAGE 60

45 Figure 19 : Autoclaving process system boundary In Table 5 the various environmental impacts and cost of both disposal technologies for 1 kg of immunization waste is illustrated. Table 5 : EI and Cost of alternative medical waste treatment technologies Impact GWP (kgCO2 eq) HH Criteria kg PM10eq Acidification Kg H+ moles Eutrophication Kg N eq Smog air Kg O3 eq Cost (USD 2013) Incineration 1.07 0. 0035 0.145 9.17 E 5 0.051 2.28 Autoclaving 0.35 0.00 19 0.058 6.24E 4 0. 036 6.68 Table 5 shows that global warming and human health impacts of incineration is much higher than autoclaving but the cost of incineration is lower than autoclaving. 3.11 Reference f low For quantifying the total life cycle impacts of vaccine SC, it is required to sum up all of the echelons LCA results, but the functional units for each of them is different. So, we should establish an equivalence relationship between all echelons. This re lationship is Ancillary Materials production External energy production Steam autoclave sterilization Sanitary landfill (leachate treatment residues landfill, electricity production ) Waste water treatment

PAGE 61

46 called reference flow. The reference flow for pneumococcal vaccine SC is displayed in Figure 20 Figure 2 0 : Vaccine SC reference flow Manufacturing of 1 dose of vaccine 0.984 tkm transportation 0.5 mL vaccine cold storage capacity 0.11 0 Medical waste treatments Equivalent to Equivalent to Equivalent to

PAGE 62

47 CHAPTER 4 MATHEMATICAL PROGRAMMING MODEL FOR OPTIMAL DESIGN OF SUSTAINABLE VACCINE SUPPLY CHAIN 4.1 Motivation This chapter addresses the optimal design and planning of vaccine SC with environmental and economic objectives. The environmental objective is measured by the LCA methodology; the economic objective is measured by the total inv estment cost of supply chain. A multi objective continuous linear programing model is developed that accounts for major characteristics of vaccine supply chains, including geographical and industrial diversity of suppliers, various shipment met hods, multip le refrigeration technolog ies different disposal treatment option s, demand distribution, cold chain and transportation capacity. The resulting Pareto optimal curves display the tradeoff between the environmental and economic dimension s of vaccine supply c hain. 4.2 Introduction Vaccine supply chain (SC) which can be defined as the network necessary for d istributing vaccine from manufacturers to the final consumers includes various sections such as production and packaging transportation cold chain and di sposal. Vaccine demand fluctuations with location and time and an unfixed production rate of manufacturers are challenging points in this SC In the low and middle income countries due to higher birth rate s vaccine demand is continuously increasing Equit y and efficiency are two parameters that should be considered in vaccine Supply Chain Management (SCM) It means that vaccine should be provided uniformly for each recipient in the system while the infectious diseases are averted as much as possible

PAGE 63

48 Vacci ne SC consists of s everal echelons all of which have significant cost and environmental burdens. Furthermore, there are several alternative options available in each section of SC Therefore, design and planning of a sustainable vaccine SC could offer sign ificant benefits to immunization programs. In this study we propose a methodology to minimizing the environmental impact (EI) and cost of the vaccine SC by developing a generic mathematical model. The C ontinues L inear Progr amming problem is optimized usin g stochastic multi objective function method called Non Dominated Sorting Genetic Algorithm (NSGA ) [51] For quantifying the EI of the SC LCA methodology is utilized [52, 53, 54] Finally, a chieving long range planning and designing the most sustain able SC are the goals of this study. 4.3 Problem f ormulation The economic and environmentally conscious alternative selection problem for the planning and design of vaccine SC can be expressed as follow. A set of clinics and their demands for various type s of vaccine over a given long term future (planning horizon) A set of candidate medical waste treatments technologies A set of candidate vaccine refrigeration options in clinics A set of potential geographical sites for the various types of manufacturing locations. A set of production and packaging options for the desired vaccines, The target is the following: Design the supply chain network of vaccine that would meet the demand over the entire planning horizon. Such that both the (1) The overall capital and operating investment cost at the end of the planning horizon is minimized (2) EI of the entire SC is minimized.

PAGE 64

49 As illustrated in Figure 22, the model suggested here to solve the problem is based upon a network structure consisting of a set of h type of vacc ines with a set of J manufacturing facilities (sites) with X types of packaging options. Regarding the locations and availability of UNICEF walk in cold rooms, K UNICEF refrigeration warehouses are taken into account. I different shipment modes from UNICEF warehouses to national cold storage facilities is considered. From there vaccines are shipped to M existing regional cold storage facilities. From there the vaccines move towards NN clinics at the next level. We consider SS refrigeration types in clinics and O locations for vaccine vial and needle disposal with FF medical waste treatment options. Optimal decisions include selection of the optimum combination of manufacturing and packaging methods and locations from available options. In addition, the opt imal structure of transportation links between the chosen sites and existing clinics needs to be designed. All these choices are made within a finite number of T time periods. Unlike traditional methodologies, which just consider economic objective functi ons, the methodology developed here tries to reach the optimum superstructure configuration and schedule a planning strategy that minimizes both environmental burdens and the cost of the SC. The solutions of this multi objective continues linear programmi ng problem lead to trade offs between conflicting economic and environmental objectives.

PAGE 65

50 Figure 2 1 : Schematic graph of vaccine network superstructure For quantifying the total investment of the SC, the procurement cost of var ious vaccine types transportation cost, cold box cost, energy consumption cost, disposal cost capital and operating cost of the clinic refrigerators are all used as part of the economic objective function. On the other hand, the environmental objective f unction is based upon the EI Vaccine Supplier with packaging technology 1 UNICEF Store National Level Storag e u hlmt Regional Level Storage v hmnnsst Clinic with Refrigeration Technology 1 Clinic wi th Refrigeration Technology 2 Clinic with Refrigeration Technology SS w hnnofft Disposal Method Technology 1 Disposal Method Technology 2 Disposal Method Technology FF Vaccine Supplier with packaging technology 2 Techno logy 1 Vaccine Supplier with packaging technology X Technology 1 y hjkxt z hklit Shipment Mode 1 Mode 1 Shipment Mode 2 Mode 1 Shipment Mode I Mode 1

PAGE 66

51 outcomes from the operation of the entire SC throughout the entire time horizon. This is evaluated through the concept of LCA and using IMPACT2002+ methodology [55] 4.4 Mathematical model 4.4.1 Production, refrigeration and t ransportation capacities and material balances Equation (1) implies that the volume of vaccines stored in the UNICEF cold room k in time period t should be less than that of cold room capacity in time t (1) (2) E quation (2) ensures that vac cines received by UNICEF store k during time interval t is less than its transportation capacity. (3) E quation (3) ensures that vacci ne h produced in manufacturing location j with technology x should be less than its production capacity during time interval t (4) (5) Equation (6) ensures the mass balance between input and output vaccines for UNICEF cold room k (6) Constraint (7) ensures that the history of product quantities shipped by each production plant to the UNICEF centers from beginning until the time period t does not exceed the product outward flow rate.

PAGE 67

52 (7) Constraint (8) ensures that the difference between the outgoing and incoming quantities, from and to the UNICEF warehouses, in each period is con fined to the holding capacity of that warehouse. (8) Constraint (9) ensures that the product amounts which are receiv ed by n ational warehouse l do not exceed the transportation capacity of it in time interval t (9) Equation (10) sets up a balance betw een the total incoming and outgoing items, to and from the national warehouse cold room l respectively. (10) Equation (11) assures that the history of product amounts sent by each production plant to the national warehouse center l from the beginning until the time period t does not exceed the product outgoing flow rate. (11) Constraint (12) assures that the difference between the outgoing and incoming amounts, from and to the regional warehouses m respectively, in each period is confined to the h olding capacity of that warehouse. (12) Constraint (13) assures that the vaccine amounts which are received by r egional cold room warehouse m do not exceed the transportation capacity of it. (13)

PAGE 68

53 Equation (14) sets up a balance between the total incomi ng and outgoing items, to and from the regional warehouse cold room m (14) Constraint (15) assures that the history of v accine amounts sent by each production plant to the regional warehouse center m from beginning until the time period t does not exceed the vaccine outgoing flow rate. (15) Equation (16) represents that the vaccine provided for clinic nn is 1.1 of the total demand of it during the planning horizon. 10 percent vaccine wastage through SC is considered in this study. (16) h Constraint (17) assures that the vaccine amounts which are received by clinic nn should not exceed the transportation capacity of the vehicle wh ich delivers vaccine to clinic nn (17) Constraint (18) represents that the amount of vaccine inventory in clinic nn should be les s than total various type s of refrigerating capacity of clinic nn (18) Equation (19) displays the vaccine volume balance for vaccine volume inputs to clinics and empty vaccine vials volume outgoing to disposal locations for each clinic nn (19) Constraint (20) represents that the vaccine wastage amounts in each vaccine disposal location o have not been allowed to violate medical wastage treatment device capacity.

PAGE 69

54 (20) Constraint (21) assures that the vaccine vials and needle amounts which are received by disposal devices do not exceed the transportation capacity of the delivery vehicle in each time interval t (21) 4.4.2 EI q uantification m ethodology For measuring the EI with LCA approach, the definition of system boundary and functional unit are essential For the model implemented here, the sy stem boundary of SC network (Fig ure 1) cover s the 12 set s of life cycle stages. The functional unit for the overall system is defined as the delivery of various types of vaccine s to the clinics over the entire planning horizon. The impacts due to operation of each of these stages were calculated separately. The LCA method categorizes the emission inventories of each life cycle stage in terms of appropriate impact category. The IMPACT 2002+ indicator estimates the environmental damages based upon three main c ategories: Global Warming (GW), Human Health (HH) and Ecosystem Quality (EQ). Equation (22) illustrates the EI measuring process with LCA method [53]. (22) is the reference flow required for life cycle stage during time t (such as a liter of vaccine in the manufacturing echelon or the weight and distance of vaccine transported in transportation echelon etc.) specified for each life cyc le stage. Secondly, which is the emissions factor per unit of reference flow of life cycle stage is specified from various resources (Such as SIMAPRO and EIO [42]). It is worth mentioning that is the

PAGE 70

55 set of environmental burdens. At the third step, which is the characterization factor of substance b contributing to the impact indicator, e is derived from IMPACT2002+ indicator. ( ).The quan tity are represent the human health impacts with the unit of Disability Adjusted Life Years (DALY), or PDF m 2 yr for the ecosystem quality category which represents the potentially disappeared fraction per square meter per year, or Kg CO 2 e for the global warming potential category for the life cycle stage pe during time interval t. For the vaccine production and packaging stage, Equation (22 ) is rewritten as Equation (23) (23) At the next level, the EI resulting from transporting vaccines from manufacturing locations to UNICEF cold rooms needs to be assessed. T he functio nal unit for transportation is based upon both the distance and weight of vaccine carried. Therefore, knowing the distance, between the location of the vaccine supplier at site j and UNICEF warehouse k and using the transportation emissions inventory, the impact asso ciated with transportation can be assessed through Equation 24 (24) EI resulting from the ice production step for ice needed for cold boxes during transportation is calculated based on Equation 25. The functional unit for this stage is based on the amount of ice needed to carry 1 lite r of YF vaccine. (Equation 25)

PAGE 71

56 (25) EI resulting from the refrigeration process of v accines in walk in cold rooms at UNICEF warehouse is derived from equation 26. The functional unit for this stage is refrigeration of 1 liter of vaccine per year. is the emissions inventory of the refrigeration process based on electricity profile of the UNICEF warehouse location. The EI score is proportional to the amount of vaccine inventory. (26) Equations (27 33) are used for determining EI scores for refrigeration, transportation and ice production in various echelons of the supply chain. (27) (28) (29) (30) (31) (32)

PAGE 72

57 (33) Equation 34 illustrates how the impact score of the waste treatment stage is calculated. The f unctional unit is disposal of 1 liter vaccine glass vial and is emission inventory of waste treatment with ff technology met hod. (34) By using the normalization factors ( ) of eac h category, the set of impact indicators for the various life cycle stages are normalized as follows (Equation 35): (35) To clarify: At the final step the scalar IMPACT 2002+ score can be achieved by using the weighting factor of relative importance to the normalized categories (which are represented as )[ 53 ]. This factor in this study is considered 0.2, 0.6 and 0.2 for hu man health, global warming and ecosystem quality respectively. (36) 4.4.3 Cost o bjective f unction The economic performance of the SC network is relates to the total cost of investment at the end of the pl anning horizon. Similar to the e nvironmental objective function each stage of vaccine SC has a specified investment cost that is determined individually by Eq uations (3 7 4 8 ). (37) Production Stage

PAGE 73

58 (38) Transportation from Supplier to UNICEF Store (39) Ice Production for Transformation stage 2 (40) Refrigeration in UNICEF Store (41) Transportation from UNICEF Store to National Store (42) Ice Production for Transformation stage 5 (43) Refrigeration in National Store (44) Transportation from National Store to Regional Store (45) Ice Production for Transformation stage 8 (46) Refrigeratio n in Regional Store (47) Refrigeration in Clinic (48) Disposal Stage In all of the above equations, C represents the unit cost at each SC stage. Unlike the environmental function there is another stage in cost function that represents t he cost of Open Vial Waste of vaccines that is measurable by e q uation ( 49 ).

PAGE 74

59 (49) Cost of Open Vial Waste represents the percentage of open vial waste of vaccine h at clinic with refrigeration technology during time interval t with vaccine production technology is determined by equation (50). It is worth mentioning, that is vaccine h demand at clinic nn which is kept in refrigerator type ss in time interva l t. (50) Percentage of open vial vaccine waste In the resulting section it will be shown that can be considered as a separate objective fu nction for the optimizing of this SC [ 56 ] The overall cost function is shown in Equation (51). (51) 4.5 Computational e xperiments We have coded this model in MATLAB and NSGA algorithm within MATLAB was used to solve the model. The algorithm was tested with real data of a vaccine supply chain. To run each condition it took 8 hours of 1 CPU to reach a solution with a 0% total gap on a 3.4 GHz Intel Core i7 with 16 GB physical memory. 4.6 Parameter s etting This algorithm was run with different value of generation number with an initial population size of 50 and with the following parameters for NSGA The probability of crossover was set as 0.7 and the probability of mutation was set at 0.02. In addition the genetic operations such as the Pareto dominance ranking procedure and the elitist selections were used.

PAGE 75

60 4.7 Vaccine s upply c hain in Ghana 4.7. 1 Network s uperstructure This section will be used to illustrate the capability of the supply chain model which consists of different types of vaccine needed for children under the age of five and for pregnant women in an immunization program for Ghana (th e list of vaccines and their dosage are available in Table 6) where the vaccines are demanded at twenty province clinics Vaccines are produced in 3 UNICEF approved locations Russia, France and Senegal with 3 different product formulations 5 doses, 10 doses and 20 doses per vial [43]. All of these vaccines are first gathered at UNICEF walk in cold ro om in Copenhagen Then these products are transported to national cold room in Accra, A total of three potential transportation modes are considered, Diesel Truck, Ocean Freight and Airplane. Next the vaccines are shipped to 10 regio nal cold rooms [57], at Ashanti, Brong Ahafo, Central, Eastern, Greater Accra, Northern, Upper East, Upper West, Volta and Western Regions From they are transported to twenty province clinics There are four potential refrigeration options in the clinics namely, Electrical Ice lined Refrigerators, Solar Refrigerators Supplemented with Battery, Solar Refrigerators Battery Free and Kerosene [58]. The final echelon in this supply chain is vial, needle and syringe waste treatment. Twenty disposal locations with two technologies, Incineration and Autoclaving [19], are considered in this study. The geographic locations of the SC echelons are shown in Figure 2 2 Table 6 : Ghana i mmunization program for under five ages childr en and pregnant women

PAGE 76

61 Figure 2 2 [59] : Vaccine SC map for Ghana 4.7.2 Results Multi (two) objective optimization of this SC model leads to a set of solutions (Pareto Front) as presented in Figure 2 3 From the Pareto Front so lutions, it is obvious that a conflict exists between a structure design that leads to a minimum EI and a minimum total cost. The figure shows that reduction of EI is only possible if policy makers compromise on the cost. The extremes of the Pareto front r each the optimum structural Vaccines Doses per person Volume per one dose (ml) BCG 1 dose 1.3 DTP HepB Hib 3 doses 13.1 Yellow Fever 1 dose 3.6 Oral polio Virus (OPV) 4 doses 1 Tetanus Toxoid (TT) 5 doses 2.6 Measles 2 doses 5.2 Pneumococcal (PCV) 3 doses 21 Rotavirus 2 doses 156

PAGE 77

62 design that leads to minimum EI and minimum total cost. The point which is located at the middle of the Pareto front represents the multi objective optimum network superstructure which is shown in Figure 2 3 Furthermore, it is w orth mentioning that the discontinuity in Pareto front (Figure 2 3 ) comes from the switching of one structure to another. This switch provides a chance to achieve a significant improvement in environmental performance at a marginal increase in cost by rebui lding the network and implementing various sustainable policies. For instance, around the cost value of USD, a compromise in the total cost of only 0.21% can reduce the EI by 2%. This is achieved by rebuilding the network structure from configu ration B to A. Figure 2 3 : Efficient set of Pareto solutions to the vaccine SC in Ghana Fig 2 4 shows a configuration located in the middle of the Pareto space where the combination of 5 doses per vial packaging technology, located in Russia and 20 d oses per vial packaging technology located in Senegal may result in an optimum alter native. The cost and EI of 20 do ses per vial packaging technology are less than other options so the major portion of vaccine flow rate was passed through this manufactu ring location O n the other hand, the cost due to open vial waste of 5 doses per vial is less tha n the 4.80E+08 4.85E+08 4.90E+08 4.95E+08 5.00E+08 5.05E+08 5.10E+08 5.15E+08 4.50E+07 4.60E+07 4.70E+07 4.80E+07 4.90E+07 Environmental Impact Score Cost (USD) Configuration A Configuration B Configuration C Feasible Solutions Region Multi Objective Optimum Point

PAGE 78

63 others and hence the remaining portion of this vaccine flow rate passes through 5 doses per vial manufacturer in Russia. For intercontinental transportat ion the ocean freight is the optimum option due to less cost and EI. The combination of Electrical Ice Lined refrigerator and Solar direct derive refrigerator is the optimum option in terms of EI and total cost functions. Autoclaving medical waste treatmen t is a more sustainable method for disposal section because this technology does not have haz ardous emissions such as Dioxin [50], but incineration is much cheaper. Therefore, a combination of 23% of autoclaving and 77% of incineration from total vaccine w astage flow rate were obtained from multi objective optimization. As it can be observed, the multi objective optimization resul ts in a set of Pareto solutions; among these set s of non dominated solutions the policy makers should select one. The stakeholder be dependent on the weights that they give to each of the objectives. UNICEF Store Vaccine Supplier with Technology of 5 doses Vaccine Supplier with Technology of 10 doses per Vaccine Supplier with Technology of 10 doses National Level Store Regional Level Store Shipment with Truck Shipment with airplane Shipment with Ship

PAGE 79

64 Figure 2 4 : Network Configuration for multi objective optimization 4.7.3 Other i mportant o bjective f unction s for v accine s upply chain 4. 7.3.1 Open Vial Waste (OVW) Open vial wastage occurs due to the fact that when multi dose vials are opened during a vaccine session and if the vial is not completely used then the remaining doses have to be discarded and cannot be used for subsequent sessi ons. So for efficient vaccine SC, minimizing the OVW is very important. This fact illustrates that although, using multi dose vials is sometimes an economic option it also increases the risk of unaccounted cost due to OVW. The definition of OVW is brought in equation (50).

PAGE 80

65 4.7. 4 Percentage of v accine m et d emand Larger vial sizes can improve the percentage of vaccine met demand. If the clinics are equipped with large vial sizes, they can easily provide sufficient amount of vaccine for unpredicted demands (mo re than estimated demand). For instance, if our estimation is 14 doses of a vaccine and there are: 14 one dose or three 5 dose or two 10 dose packaging options available, then with a more than estimated arrival rate large vial sizes could meet the demand better. So maximizing this objective function could be worthy in vaccine SC optimization. Therefore, optimization based on opposite objective functions (like vaccine met demand and OVW) is necessary to balance the trade offs. Therefore, multi objective opt imization is completely relevant in this case. With curve fitting process, equation (51) is derived from Rajgopal et al. study [60]. This correlation displays the percentage of vaccine met demand based on vial sizes. If the percentage of open vial waste of vaccine (OVW) which is defined in equation (50) is considered as a separate objective function, a 3D Pareto front will be achieved (Figure 2 5 ) that not only tries to provide the minimum cost and EI supply chain network, but also attempts to reduce the ope n vial waste. Figure 2 6 displays the 3D Pareto front when percentage of vaccine met demand is replaced with OVW. The points that the surface passes through are optimum solutions of vaccine SC with each of them being a detailed design and scheduling solutio n that leads to the Pareto fronts. In the next section the optimum scheduling pattern is discussed. (51)

PAGE 81

66 Figure 2 5 : 3D Pareto front for multi objective optimization (OVW as a third objective function)

PAGE 82

67 Figure 2 6 : 3D Pareto front for multi objective optimization (Percentage of vaccine met demand as a thi rd objective function) 4.7. 5 Contribution of each echelon in cost and EI of vaccine SC Table 7 shows the contribution of each supply chain echelon to global warming and investment cost. The vaccine production and packaging has a significant role in CO 2e e missions (96.7%); the second echelon that generates the most CO 2e is vaccine intercontinental transportation, but it is far less than the first echelon (2.6%). This suggests that the activities for CO 2e reductions should be focused on the vaccine productio n echelon.

PAGE 83

68 Table 7 : Contribution of life cycle CO 2e emission and cost of various echelons of vaccine SC vaccine echelon CO 2e Contribution investment Cost Contribution Vaccine Production and Packaging 96.7% 84.31% Transportation from Manufacturing Lo cation to UNICEF Store in Copenhagen(1) 0.17% 0.36% Ice Production for Cooling Vaccines During Transportation (1) 0.03% 0.02% Refrigeration Process in UNICEF Store in Copenhagen 0.005% 0.001% Transportation from UNICEF Store in Copenhagen to National St ore in ACCRA(2) 2.60% 0.17% Ice Production for Cooling Vaccines During Transportation (2) 0.29% 0.02% Refrigeration Process in National Store in ACCRA 0.006% 0.001% Transportation from National Store in ACCRA to Regional Stores(3) 0.009% 0.02% Ice Prod uction for Cooling Vaccines During Transportation (3) 0.03% 0.02% Refrigeration Process in Regional Stores 0.02% 0.005% Refrigeration Process in Clinics 0.10% 0.08% Transportation from Regional Stores to Clinics(4) 0.008% 0.02% Ice Production for Cool ing Vaccines During Transportation (4) 0.06% 0.57% Vaccination Waste Treatment 0.18% 0.18% Open Vial Waste (OVW) 14.36% Similar to CO 2e emissions, vaccine production stage represents 84.12% of total vaccine supply chain investment cost, but the cost associated with Open Vial Waste is next with

PAGE 84

69 14.36% of total cost. Thus, solutions that decrease the rate of wastage could result in significant cost savings. As mentioned before, the cost and environmental emissions per dose of large vial sizes is the low est. On the other hand, the higher dosage per vial will result in an increase in rate of OVW and consequently the OVW cost will increase. Multi objective optimization will find a trade off solution (optimum mixture of vaccine production/packaging technolog y). 4.8 Time s cheduling for v accine t ransportation: Table 8 illustrates the current vaccine transportation plan and Optimum results from NSGA code for Ghana Currently the vaccine transportation from national level to regional level takes place 4 times per year (every 3 months) [57]. I t is obvious that this schedule is not effi cient since it requires high capacity for cold chain infrastructu re B ut in our optimum scenario the transportation frequency is monthly and this frequency prevent s the high vaccine accumulation in cold rooms and reduce s the inventory cost.

PAGE 85

70 Table 8: Comparison of current and opt imum vaccine shipment fre quency Optimum Shipment Schedule Current Shipment Schedule in Ghana

PAGE 86

71 4. 8 1 Optimization r esults based on different objective functions: Fig ure 28 shows the life cycle CO2e emissions ( i ndirect + direct emission) through vaccine SC in Ghana, the current SC in Ghana produce s CO2e. As it is shown in F ig 28 EI optimization causes the highest reduction for CO2e emission (24%). This reduction is equal to carbon emissions of 1004 passenger vehicle during one year [61].It is depicted in Fig ure 29 that optimization base d on total investment cost cause s a $ 13, 158,000 reduction in vaccine supply chain in Ghana (22% redu ction). When this cost is averted policy makers will be able to invest more towards new infrastructure Figure 2 7 : Effect of different objective functions on CO 2e Figure 2 8 : Effect of dif ferent objective functions on t otal Cost 0.00E+00 5.00E+06 1.00E+07 1.50E+07 2.00E+07 2.50E+07 Cost Optimization Environmental Impact Optimization Multi-Objective Optimization Current Situation Kg CO2e 0.00E+00 1.00E+07 2.00E+07 3.00E+07 4.00E+07 5.00E+07 6.00E+07 7.00E+07 Cost Optimization Environmental Impact Optimization Multi-Objective Optimization Current Situation Total Investment Cost of Vaccine Supply Chain in Ghana (USD 2014)

PAGE 87

72 4.9 Conclusion: In this work, a method to design environmental ly friendly and profitable vaccine SC is presented. The model consisted of multi period Continues linear programming that accounts for multi objective op timization of economics and environmental considerations The model consider s the long term strategic decisions (e.g. select ion suppliers, shipment modes, r efrigeration technologies) with mid term time schedule planning for SCs. T he vaccine SC was modeled as a multi objective optimization model and a stochastic methodology (NSGA ) was adopted as solution methodology Design results indicate that a combination of 20 and 5 dose vaccine in production /packaging section, ocean freight in inter continental trans portation combination of electrical ice lined and solar direct drive vaccine refrigerators in cold storage, and combination of incineration and autoclaving for medical waste disposal results in optimum configuration network based on economic and EI object ive s. The solution to the scheduling problem suggests that although increasing the shipment frequency between different echelons leads to increased shipment cost and EI it results in significant reductions to the inventory cost and EI associated with cold room warehouses Therefore, high shipment frequenc y leads to optimum results. By implementing both the designing and time scheduling optimization scenario s, 24% carbon emission and 22% total investment cost reduction s can be achieved

PAGE 88

73 R EFERENCES 1 WHO w orld health statistics 2012: Home Page http://www.who.int/gho/publications/world_health_statistics/2012/en/ (Accessed June 17, 2013) 2 WHO world health statistics 2012: Ho me Page http://www.who.int/gho/publications/world_health_statistics/2012/en/ (Accessed June 17, 2013) 3 Gavi Alliance H ome P age http://gavialliance.org/support/nvs/pneumococcal/ (accessed June 26 2013) 4 Atherly, D. E., Lewis, K. D., Tate, J., Parashar, U. D., & Rheingans, R. D. (2012). Projected health and economic impact of rotavirus vaccination in GAVI eligible countries: 2011 2030. Vaccine, 30 A7 A14. 5 Saxenian, H., Cornejo, S., Thorien, K., Hecht, R., & Schwalbe, N. (2011). An analysis of how the GAVI alliance and low and middle income countries can share costs of new vaccines. Health affairs, 30 (6), 1122 1133. 6 Chen, Sheng I. Modeling the WHO EPI vaccine supply chain in low and middle income countries PhD Dis sertation University of Pittsburgh, 2012. 7 Lee, Bruce Y., et al. "Impact of introducing the pneumococcal and rotavirus vaccines into the routin e immunization program in Niger." American journal of public health 102 .2 (2012): 269 276. 8 Assi, T. M., Brown, S. T., Kone, S., Norman, B. A., Djibo, A., Connor, D. L., ... & Lee, B. Y. (2013). Removing the regional level from the Niger vaccine supply cha in. Vaccine, 31 (26), 2828 2834. 9 Ferrari, Matthew J., Ottar N. Bjrnstad, and Andrew P. Dobson. "Estimation and inference of R 0 of an infectious pathogen by a removal method." Mathematical biosciences 198. 1 (2005): 14 26. 10 Bojarski, A. D., Lanez, J. M., E spua, A., & Puigjaner, L. (2009). Incorporating environmental impacts and regulations in a holistic supply chains modeling: An LCA approach. Computers & Chemical Engineering, 33 (10), 1747 1759.

PAGE 89

74 11 Zaffran, M., Vandelaer, J., Kristensen, D., Melgaard, B., Yad av, P., Antwi Agyei, K. O., & Lasher, H. (2013). The imperative for stronger vaccine supply and logistics systems. Vaccine, 31 B73 B80. 12 Introduction of rotavirus vaccines into national immunization programs, Immunization, Vaccines and Biologicals, WHO H ome P age http://whqlibdoc.who.int/hq/2009/WHO_IVB_09.09_eng.pdf ( accessed June 17, 2013) 13 Burton, A. (2007). Solar Thrill: Using the sun to cool vaccines. Environmental health perspect ives, 115 (4), A208. 14 WHO, H ome P age http://who.int/immunization_standards/vaccine_quality/pqs_e03_fridges_freez ers/en/ (accessed 17 June 2013) 15 RFVB134a Sol ar Vaccine Refrigerator, Sunfrost Company. Arcata, California, US. Home page http://www.sunfrost.com/vaccine_refrigerators.html (accessed 17 June 2013) 16 BFRV55, Solar Battery Free Vaccine Refrigerator. Sundanzer company, Texas, US. Home P age http:// sundanzer.com (accessed 26 June 2013) 17 BLF100, Electricity Powered Vaccine Refrigerator, Surechill Company, UK Home Page http://surechill.com (accessed 26 June 2013) 18 Solar Air Conditioning a nd Refrigeration; Sayigh, A.A.M, MCViegh, J.C, Pergamon Press, 1992 19 Innovation Performance Savings. ENERGY STAR Report, Refrigerators 2007 Partner Resource Guide. Home Page www.howards .com/eStar/2007Refrigerator_prg.pdf (accessed 1 December 2013) 20 US Life Cycle Inventory: National Research Laboratory, (NREL), US H ome P age http://www.nrel.gov/lci/ (accessed 26 June 2013) 21 Life Cycle Inventories of Energy Systems, Results for Current Systems in Switzerland and other UCTE Countries. Ecoinvent Report NO 5 December 2007. Home page http://www.ecolo.org/ documents/documents_in_english/Life cycle analysis PSI 05.pdf

PAGE 90

75 22 Refrigerant Charge Calculator (Version 1.1). Guidance on Impact of the Fluorinated Gas Regulations for Users of Stationary Refrigeration, Air Conditioning and Heat Pump January 2007 23 McCull och, A., & Lindley, A. A. (2003). From mine to refrigeration: a life cycle inventory analysis of the production of HFC 134a. International Journal of Refrigeration, 26 (8), 865 872. 24 Global Comparative analysis of HFC and alternative Technologies Refrigerat ion, Air Conditioning, Foam, Solvent, Aerosol Propellant and Fire Protection Applications ; Arthur D. Little, Inc. Cambridge Massachusetts, 2002 25 Agarwal, R; Anderson, J; Bivens, D; Colbourne D; Hundy, G; Knig, H; Lundqvist, P; McInerney, E; Neks, P. I PCC/TEAP Special Report: Safeguarding the Ozone Layer and the Global Climate System (Chapter 4: Refrigeration). Home page http://www.ipcc.ch/pdf/special reports/sroc/sroc04.pdf (access ed 26 June 2013). 26 Sand, J. R., Fischer, S. K., & Baxter, V. D. (1997). Energy and global warming impacts of HFC refrigerants and emerging technologies (pp. 5 60). Oak Ridge, TN, USA: Oak Ridge National Laboratory 27 Boxwell, M. Solar Electricity Handbook. 2013 Edition Home page http:// solarelectricityhandbook.com (accessed 26 June 2013) 28 Sunvia Company. OPTXXX 60 4 1B0 (60 CELL MODULE). Home page http:// sunvia.com (accessed 17 June 2013) 29 Fthenakis, V. M., Kim, H. C., & Alsema, E. (2008). Emissions from photovoltaic life cycles. Environmental science & technology, 42 (6), 2168 2174. 30 Trojan Company. Deep Cycle Battery. US Home Page http:// trojanbattery.com (accessed 26 June 2013) 31 Sullivan, J. L., & Gaines, L. (2010 ). A review of battery life cycle analysis: state of knowledge and critical needs (No. ANL/ESD/10 7). Argonne National Laboratory (ANL). 32 Horie, Y. A. Life Cycle Opt imization of Household Refrigerator Freezer Replacement. MS Dissertation ( Report No. CSS04 13) University of Michigan, August 2004.

PAGE 91

76 33 Boustani, A; Gutowski, T; Graves, S. Appliance Remanufacturing and Energy Savings. January 2010. Home Page http://web.mit.edu/ebm/www/Publications/MITEI 1 a 2010.pdf (accessed 26 June 2013) 34 Guide to Electric Power in Ghana, University of Ghana. 2005 Home Page http://www.beg.utexas.edu/energyecon/IDA/USAID/RC/Guide_to_Electric%2 0Power_in_Ghana.pdf (accessed 26 June 2013) 35 Riberio, F. M; Da Silva, G. A. Life Cycle Inventory for Hydroelectric Generat ion: A Brazilian Case Study. J. Cleaner Production 2010 18(1), 44 54. 36 U.S. Environmental Protection Agency (EPA), Tools for the Reduction and Assessment of Chemical and other Environmental Impacts (TRACI) H ome P age http://epa.gov/nrmrl/std/traci/traci.ht m l (accessed June 26 2013) 37 Ryberg, M., Vieira, M. D., Zgola, M., Bare, J., & Rosenbaum, R. K. (2014). Updated US and Canadian normalization factors for TRACI 2.1. Clean Technologies and Environmental Policy, 16 (2), 329 339. 38 Southern Africa Grid Map, Esko m Holdings Limited Annual report 2009 H ome P age http://www.financialresults.co.za/eskom_ar2009/ar_2009/dow nloads/01_profile.pdf (accessed June 26 2013) 39 Eckelman, M.J,Mauter, M.S, Isaacs,J.A, Elimelech, M. New Perspectives on NanoMaterial Aquatic Ecotoxicity : Production Impacts Exceed Direct Exposure Impacts for Carbon Nanotoubes. Journal Environmental Science and Technology 2012, 46, 2902 2910 40 US Energy Information Administration (EIA), Home Page: http://www.eia.go v/countrie ( Accessed 30 November 2013) 41 Electrical Power in Africa, Home Page: http://mbendi.com/indy/power/af/p0005.htm#country profiles (accessed 24 June 2013). 42 Hendrickson, C., Ho rvath, A., Joshi, S., & Lave, L. (1998). Peer reviewed: economic input output models for environmental life cycle assessment. Environmental science & technology, 32 (7), 184A 191A. 43 UNICEF vaccine price data, home page, http://www.unicef.org/supply/index_57476.html (accessed January 2015) 44 SIMAPRO 8 LCA software. Home page: ( http://simapro.com/ )

PAGE 92

77 45 Transportation European Commission, Home Page : http://ec.europa.eu/transport/index_en.htm 46 Boudhenn, F., Demasles, H., Wyttenbach, J., Jobard, X., Chze, D., & Papillon, P. (2012). Development of a 5 kW cooling capacity ammonia water absorption chiller for s olar cooling applications. Energy Procedia, 30 35 43. 47 Mendivil, R., Fischer, U., Hirao, M., & Hungerbhler, K. (2006). A New LCA Methodology of Technology Evolution (TE LCA) and its Application to the Production of Ammonia (1950 2000)(8 pp). The Internat ional Journal of Life Cycle Assessment, 11 (2), 98 105. 48 Solar Collector Inc. Home Page: ( http://www.solarcollectorinc.com ) 49 Ardente, F., Beccali, G., Cellura, M., & Brano, V. L. (2005). Life cycle assessmen t of a solar thermal collector Renewable Energy, 30 (7), 1031 1054. 50 Zhao, W., van der Voet, E., Huppes, G., & Zhang, Y. (2009). Comparative life cycle assessments of incineration and non incineration treatments for medical waste. The International Journal of Life Cycle Assessment, 14 (2), 114 121 51 Kalyanmoy, et al. "A fast and elitist multi objective genetic algorithm: NSGA II." Evolutionary Computation, IEEE Transactions on 6 .2 (2002): 182 197. 52 Baumann, H., & Tillman, A. M. (2004). The Hitch Hiker's Guid e to LCA. An orientation in life cycle assessment methodology and application External organization 53 Hugo, A., & Pistikopoulos, E. N. (2005). Environmentally conscious long range planning and design of supply chain networks. Journal of Cleaner Production, 13 (15), 1471 1491. 54 You, F., Tao, L., Graziano, D. J., & Snyder, S. W. (2012). Optimal design of sustainable cellulosic biofuel supply chains: multiobjective optimization coupled with life cycle assessment and input output analysis. AIChE Journal, 58 (4), 1157 1180. 55 Jolliet, O., Margni, M., Charles, R., Humbert, S., Payet, J., Rebitzer, G., & Rosenbaum, R. (2003). IMPACT 2002+: a new life cycle impact assessment

PAGE 93

78 methodology The International Journal of Life Cycle Assessment, 8 (6), 324 330. 56 Assi, T. M., B rown, S. T., Djibo, A., Norman, B. A., Rajgopal, J., Welling, J. S., ... & Lee, B. Y. (2011). Impact of changing the measles vaccine vial size on Niger's vaccine supply chain: a computational model. BMC Public Health,11 (1), 425. 57 Ministry of Health Ghana, Immunization Program Comprehensive Multi year Plan (2010 2014) Home Page, file:///C:/Users/Bahador/Downloads/Ghan a Comprehensive%20multi year%20plan%20for%202010 2014%20 %20Year%20Unknown%20(2).pdf (accessed January 2015) 58 Kilfoyle, D., & Ventre, G. G. (1988). Test and evaluation of vaccine refrigeration systems. In Photovoltaic Specialists Conference, 1988., Confe rence Record of the Twentieth IEEE (pp. 1200 1205). IEEE. 59 Google Map, Home Page, https://www.google.com/maps (accessed in August 2105) 60 Rajgopal, J., Connor, D. L., Assi, T. M., Norman, B. A., Chen, S. I., Baile y, R. R., ... & Lee, B. Y. (2011). The optimal number of routine vaccines to order at health clinics in low or middle income countries. Vaccine, 29 (33), 5512 5518. Deb, 61 EPA, Office of Transportation and Air Quality, Home Page, http://www.epa.gov/otaq/consumer/420f08024.pdf (accessed in January 2015).

PAGE 94

79 APPENDIX Unit Processes Refrigerant Production PV Panel Production Battery Production Material Processing Assembly Transportation Re frigerant Leakage Blowing agent Leakage Electricity input (KWh) 2.1E 4 2.51 0.53 0.32 0.17 Heat input (KWh) 1.33E 3 7.54 1.58 0.96 0.15 Emissions to air (g) Arsenic 3.72E 8 0.000316 0.007855 5.94E 5 3.1E 5 Benzene 1.38E 5 0.0 94388 0.007209 0.0123 0.003322 Cadmium 3.13E 9 9.46E 5 0.007855 4.99E 6 2.6E 6 Carbon dioxide 0.49 3348.137 3749.91 489.507 162.702 915.204 261.816 441.653 Carbon monoxide 7.41E 5 0.660 0.2397 0.086373 0.035 1.595 Chromium 7.74E 8 0.000 474 0.007855 0.000124 6.45E 5 -Hydrogen chloride 2.99E 5 0.363 0.076192 0.04768 0.025 These unit processes happen in US Transportation Ghan a a

PAGE 95

80 Lead 5.1E 8 0.000592 0.00024 8.14E 5 4.25E 5 Mercury 3.45E 8 0.000158 6.25E 6 3.89E 6 2.03E 6 Methane 9.82E 4 7.225 0.9192 1.029 0.29456 0.031 57 Hydrogen fluoride 3.72E 6 0.0452 0.0095 0.005938 0.0031 NO x 0.00596 6.606 0.6993 0.7349 0.315 12.472 PM10 1.67E 5 0.097 1.139 0.012617 0.002157 0.2898 PM 2.5 SO x 0.000559 7.98 1.198 0.890644 0.4647 0.180274 NM VOC 0.000517 0.699 0.599 0.091 0.021 0.57536 Calcium Sulfate 0.0052 N 2 O 8.5E 6 0.10336 0.025424 0.012113 7.056E 3 0.0201376 Nickel 8.25E 8 0.001471 0.000247 0.000132 6.87E 5 Emissions to water Arsenic 0.00646 Cadmium 0.0060 Chromium 0.0050 Lead 0.000959 Mercury 6.89E 9 BOD 3.3E 6 0.021 0.001307 0.002788 0.000672 COD 8.24E 6 0.069 0.009275 0.00906 0.003435 Table A1 : LCI data of d ifferent unit process for the case of SPSB manufactured in US and used in Ghana for functional unit of 1 liter cold storage per year

PAGE 96

81 Unit Processes Refrigeran t Production Material Processing Assembly Transportation Electricity consumption in Ghana Electricity consumption in South Africa Refrigeran t Leakage Blowing Agent Leakage Electricity input (KWh) 2.1E 4 0.32 0.17 9.21 9.21 Heat input (KWh) 1.33E 3 0.96 0.15 Emissions to air (g) Arsenic 3.72E 8 2.91E 7 0.000474 Benzene 1.38E 5 9.08E 5 Cadmium 3.13E 9 Carbon dioxide 0.49 360.2621 118.205 141.8524 1485.08 14363.99 233.764 853.41 Carbon monoxide 7.41E 5 0.2863 0.04507 0.2036 4.7541 Chromium 7.74E 8 1.04E 7 Hydro gen chloride 2.99E 5 Lead 5.1E 8 Mercury 3.45E 8 4.58E 14 UK Transportation Ghana South Africa Ghana or South Africa

PAGE 97

82 Methane 9.82E 4 0.02624 0.00413 0.00342 4.3850 3.4651 Hydrogen fluoride 3.72E 6 NO x 0.0059 0.4122 0.18091 1.0104 1.2345 28.57161 P M10 1.67E 5 0.003991 0.00062 0.01951 0.137863 2.4893 PM 2.5 0.03027 0.01625 0.40602 3.8387 SO x 0.0005 0.40268 0.20260 0.031249 0.5744 134.1568 NMVOC 0.0005 0.0642 0.02065 0.069801 0.6664 0.450654 Calcium Sulfate 0.0052 N 2 O 8.5E 6 0.01945 0.00030 0.005129 3.26E 6 0.25537 Nickel 8.25E 8 1.32E 7 Emissions to water Arsenic Cadmium Chromium Lead Mercury 6.89E 9 BOD 3.3E 6 0.08868 0.04760 0.014764 COD 8.24E 6 1.36E 6 0.38341 Table A2 : LCI data of different unit processes of EP manufactured in UK and used in Ghana or South Africa for functional unit of 1 liter of cold storage per year