Waste Management 113 (2020) 197–209

Contents lists available at ScienceDirect

Waste Management
journal homepage: www.elsevier.com/locate/wasman

A reverse supply chain for medical waste: A case study in Babol

healthcare sector
Saeed Kargar, Mohammad Mahdi Paydar ⇑, Abdul Sattar Safaei
Department of Industrial Engineering, Babol Noshirvani University of Technology, Babol, Iran

a r t i c l e i n f o a b s t r a c t

Article history: Medical waste generation is rapidly rising, which may cause a serious risk for both humans and environ-
Received 9 October 2019 ment if it is not properly managed. Designing an efficient and reliable medical waste reverse supply chain
Revised 30 May 2020 (MWRSC) is extremely beneficial to society. Most studies on this issue have only considered the gener-
Accepted 31 May 2020
ated waste and have not reported the uncertainty in the amount of medical waste generation and other
Available online 11 June 2020
MWRSC parameters. Sustainability criteria and environmental issues in choosing treatment technology
are rarely considered as well. In this research, a linear programming model under uncertainty is devel-
Keywords:
oped to design an MWRSC. The proposed model is multi-item and multi-period with three objective func-
Reverse Supply Chain
Medical Waste
tions. The first objective function minimizes total costs, the second objective function is relevant to the
Multi-objective best treatment technology selection and the third objective function minimizes the total medical waste
Fuzzy goal programming stored. A robust possibilistic programming approach is utilized to handle imprecise parameters in the
Robust possibilistic programming model and a fuzzy goal programming method is employed to build up a multi-objective model. A real
case study is conducted to illustrate the potential of the proposed model which involves different attri-
butes and problems, such as the location and capacity of facilities, and the type of treatment technology.
Furthermore, the transferring amount of medical waste among different nodes is calculated.
Ó 2020 Elsevier Ltd. All rights reserved.

1. Introduction (W.H.O., 2017). GMW which is not typically considered hazardous,

has the major share of medical waste in a medical center and
Access to health care has been improved along with new devel- includes plastic, paper, and office waste. GMW can be disposed
opments in health care instruments and health centers. The through a regular process and does not need any special handling
amount of generated medical waste has dramatically increased (W.H.O., 2017). IMW includes blood-soaked bandages, sharps, sur-
as well. According to the World Health Organization (WHO) report, gical waste, human or body parts, blood and body fluids, unwanted
the term ‘‘medical waste” contains all the waste generated within microbiological cultures, and swabs. There are different technolo-
health-care centers, research centers, and laboratories associated gies to treat medical waste such as incinerators, autoclaves, micro-
with medical procedures (W.H.O., 2014). Medical waste which is waves and chemical materials, applying an efficient technology is a
not appropriately handled and disposed creates a great risk of serious challenge for medical center managers.
infection or injury to the healthcare staff as well as a minor risk Proper MWRSC management may be useful in appropriate stor-
to the general public through the spread of micro-organisms from ing, collecting, treating, and disposing. Moreover, it may minimize
healthcare facilities into the environment (Windfeld and Brooks, the health and environmental impacts of IMW as well as total costs
2015). Medical waste needs a different management method com- (Korkut, 2018). In general, a reverse supply chain can play an
pared to municipal waste and should therefore be treated before important role in the product life cycle (Genovese et al., 2017).
disposal. Thus, the scholars are interested in continuing their research on
There are four main types of medical waste in terms of haz- the design of reverse supply chain networks. Therefore, the reverse
ardousness. They include non-hazardous or general medical waste supply chain network design for various products, such as elec-
(GMW), hazardous or infectious medical waste (IMW) which may tronic waste, paper, bulk waste, household, and vehicles has been
be sharps and non-sharps, and chemical & pharmaceutical waste widely studied (Doan et al., 2019). Some of the most recent and rel-
evant researches on reverse logistics are also listed here (Paydar
and Olfati, 2018); (Oyola-Cervantes and Amaya-Mier, 2019);
⇑ Corresponding author. (Trochu et al., 2019); (Dutta et al., 2020).
E-mail address: [email protected] (M.M. Paydar).

https://doi.org/10.1016/j.wasman.2020.05.052
0956-053X/Ó 2020 Elsevier Ltd. All rights reserved.
198 S. Kargar et al. / Waste Management 113 (2020) 197–209

Despite various studies regarding waste management, very few Table 1 presents some key features of the reviewed studies in
studies have been reported in medical waste management. While the field of medical waste management. The main features and
most studies on medical waste management has only focused on detected research gaps in the literature addressed by the current
the management aspect, studies on analytical models or quantities study are as follows:
techniques are rare. Some of the current studies in this field are
listed here (Lee et al., 2004); (Yong et al., 2009); (Voudrias and 1. Most current published researches in this area only reported
Graikos, 2014); (Windfeld and Brooks, 2015); (Caniato et al., the generated waste in various parts of the world. An optimiza-
2015); (Makajic-Nikolic et al., 2016); (Rolewicz-Kalińska, 2016); tion of a reverse supply chain for medical waste is rarely taken
(Rajan et al., 2018); (Dehghani et al., 2019); (Mazloomi et al., into consideration. The existence of an efficient reverse supply
2019). A review of the most recent and relevant researches on chain network can reduce costs and undesirable cases.
MWRSC which applied quantities techniques is presented below. 2. Due to the large variety of medical services and the imprecision
Nolz et al. (2014) considered the problem of designing a logis- of demand, it is difficult to determine the exact amount of gen-
tics system for handling a sufficient collection of infectious medical erated waste in this sector. In previous studies, uncertainty in
waste. This issue is developed as a collector-managed inventory the amount of medical waste generation and other MWRSC
routing problem by utilizing the radio frequency identification parameters were rarely considered. The results of the model
(RFID) technology. Social objectives, especially the satisfaction of will be more reliable after considering these uncertainties.
pharmacists and local government, in addition to the minimization 3. In the real-world situation, the researchers have considered dis-
of public health risks are considered in a real-world-motivated tance minimization in addition to total costs which are most
inventory routing problem. important in this problem. However, sustainability criteria
Budak and Ustundag (2017) presented a reverse logistics opti- and environmental issues in choosing treatment technology
mization for waste collection and disposal in Turkish health cen- are rarely considered.
ters. A multi-period and multi-product MILP model with the
minimizing of the total cost was developed to calculate the optimal The aim of this study is to design a reverse supply chain opti-
locations and number of facilities for efficient waste management. mization model for the safe disposal of medical waste. First, a
Alshraideh and Qdais (2017) developed a stochastic model of med- tri-objective mixed-integer linear programming (MILP) mathemat-
ical waste collection in Northern Jordan. A route scheduling model ical model is designed to optimize the medical waste management
is proposed to minimize the total travel distance, which ultimately system for the waste generated in medical centers. Then, a recent
minimizes the cost of transportation and reduces emissions. possibilistic programming approach called robust possibilistic pro-
Mantzaras and Voudrias (2017) presented an optimization model gramming (RPP) is utilized to control the uncertain parameters.
to minimize the total cost of a collection, haul, transfer, treatment, Finally, a fuzzy goal programming (FGP) is employed to handle
and disposal system for IMW in the region of Greece. The model imprecise goals in the goal programming (GP) approach. The pro-
optimizes the number and location of opened waste management posed model determines the optimized number, capacity, costs,
facilities and the optimal route of transportation with minimizing location of facilities, and the type of treatment center technology,
the total costs. Gergin et al. (2019) developed an artificial bee col- by considering uncertain medical waste quantities.
ony based on clustering algorithm for solving continuous multiple The paper has the following elements of novelty:
facility location problems in Istanbul Municipality. Osaba et al.
(2019) developed a multi-attribute or rich vehicle routing problem. 1. A multi-objective mixed-integer linear programming model is
They utilized a discrete and improved bat algorithm to optimize a designed for MWRSC under uncertainty.
real-world drug distribution problem with pharmacological waste 2. The model suggests the optimal location, capacity, costs, and
collection in Spain. the number of facilities.

Table 1
Recent research in medical waste.

Author(s) Year Multi- Multi- Multi- Uncertainty Case Approach Software

objective period product study
(Shih and Lin, 2003) 2003 U U China MILP Lingo
Dynamic programming GIS
(Shi et al., 2009) 2009 U China MILP Matlab
Genetic algorithm
(De Almeida, 2010) 2010 U Portugal MILP GAMS
(Nolz et al., 2014) 2014 U U U France MILP CPLEX
Adaptive large neighborhood
search
(Budak and Ustundag, 2017) 2017 U U U Turkey MILP Fico Xpress IVE
(Alshraideh and Qdais, 2017) 2017 U U U Jordan MILP Matlab
Genetic algorithm
(Mantzaras and Voudrias, 2017 U Greece MILP Evolver
2017) Genetic algorithm Crystal Ball
Monte Carlo simulation
(Gergin et al., 2019) 2019 U U U Turkey MILP Microsoft Visual
Artificial bee colony C#
(Osaba et al., 2019) 2019 U U Spain MILP Matlab
Bat algorithm
Firefly algorithm
Present study 2019 U U U U Iran MILP Lingo
Fuzzy goal programming
Robust possibilistic
programming
 S. Kargar et al. / Waste Management 113 (2020) 197–209 199

3. The model determines the optimal treatment technology based best technology under different criteria. To reduce the risk of infec-
on sustainability indicators. tions in medical centers, IMW storage in the centers should be
4. A real case study of medical waste in Iran is analyzed. shortened, the third objective function minimizes the stored waste
5. The risk of spreading infections by minimizing stored medical in the storage centers. The most important decisions in the model
waste in the centers is modeled. are as follow:

2. Problem description (1) Waste flow within facilities

(2) Stored waste in each center
Medical waste generated by medical centers in the process of (3) Opening new facilities in the network
collection is divided into four main categories including GMW, (4) Selecting the appropriate treatment technology
sharps IMW, non-sharps IMW, and chemical & pharmaceutical
waste. Because of two main reasons, the fourth category is ignored:
lower volume and different disposal methods compared to other
waste generated by medical centers. However, the first group is Indices
non-infectious while the second and third ones are infectious i Medical centers (1 . . . i . . .I)
waste. IMW must be treated before entering the municipal waste j Storage centers (1 . . . j . . .J)
network because it contains highly dangerous substances. GMW c Treatment centers (1 . . . c . . .C)
enters the municipal waste network without the need for treat- n Transfer stations (1 . . . n . . .N)
ment. According to the Medical Waste Management Act in Iran, d Disposal centers (1 . . . d . . .D)
every medical center is obliged to manage both GMW and IMW. f Waste types (1- GMW, 2- Sharp IMW, 3-Non-sharp
It is important to note that the amount of GMW makes up a large IMW)
share of the medical center waste and in addition, the costs k Technology types (1 . . . k . . .K)
involved are significantly high. Therefore, due to the cost optimiza- a Technology selection criteria (1 . . . a . . .A)
tion and comprehensive waste management, it is necessary to con- t Periods (1 . . . t . . .T)
sider all the waste generated by medical centers in Iran. According Parameters
to the regulations in most countries, the following steps should be HJfj Unit storage cost for waste f in storage center j
taken for medical waste management.
HNn Unit storage cost for waste in transfer station n
Indoor Collection: IMW and GMW are segregated at the genera-
CT k Unit treatment cost by technology k
tion source. IMW is packed in yellow containers and bags, while
DC d Unit disposal cost for waste in disposal center d
GMW is collected in blue containers and black bags.
R Unit transportation cost for GMW
Temporary storage warehouse (storage center): Storage center is a
R0 Unit transportation cost for IMW
special warehouse that is built according to waste management
LC Unit labor cost for collection waste
regulations. IMW is stored in the temporary warehouse until it
FIXJj Fixed cost for opening storage center j
reaches a certain amount, which is then sent to treatment centers.
FIXC ck Fixed cost for opening treatment center c with
Treatment center: IMW is treated by one of the treatment tech-
technology k
nologies (incinerator, autoclaves, microwave, and sanitary land-
FIXNn Fixed cost for opening transfer station n
fills) and then sent to the transfer station.
FIXDd Fixed cost for opening disposal center d
Transfer station: A waste transfer station is an area near the city.
DIJij Distance from medical center i to storage center j
Treated infectious medical waste (TIMW) and GMW are trans-
ported to the transfer stations by low-capacity vehicles and later DJC jc Distance from storage center j to treatment center c
to disposal centers by higher-capacity vehicles. DINin Distance from medical center i to transfer station n
Disposal / Landfill: In these centers, waste is buried according to DCNcn Distance from treatment center c to transfer station n
the instructions. DNDnd Distance from transfer station n to disposal center d
In this research, five levels in an MWRSC are considered: The CAPJfj Maximum capacity of waste f in storage center j
medical centers, storage centers, treatment centers, transfer sta- CAPC ck Maximum capacity of treatment center c with
tions, and disposal centers. These facilities play important roles technology k
in this network. Fig. 1 illustrates the scheme of the proposed CAPNn Maximum capacity of transfer station n
MWRSC network. CAPDd Maximum capacity of disposal center d
In this study, we have considered the current situation of DEfit Amount of waste f generated in the medical center i
MWRSC and suggested some improvements. In the current situa- in period t
tion, there are a number of treatment centers that are decentral- a A coefficient of the capacity of the treatment center,
ized. Both centralized and decentralized treatment centers have which determines the startup of the equipment
their own advantages and disadvantages. According to the Medical TMka A score of treatment technology k from criteria a
Waste Management Act in Iran, treatment centers should be as WIa Importance of criteria a
decentralized as possible, and some centralized centers can serve Decision variables
the medical centers which do not have a treatment center. In gen- ZC ct 1 If treatment center c is utilized in period t; 0
eral, both approaches are utilized in the proposed model. otherwise
Bj 1 If storage center j is opened; 0 otherwise
2.1. The proposed mathematical model W ck 1 If treatment center c with technology k is opened; 0
otherwise
The proposed model of the MWRSC is an MILP model. This Pn 1 If transfer station n is opened; 0 otherwise
model has three objective functions. The first objective function DId 1 If disposal center d is opened; 0 otherwise
minimizes the total costs of the MWRSC network considering stor- Y fijt Amount of waste f transported from medical center i
age, treatment, collection, and fixed costs. Each treatment technol- to storage center j in period t
ogy is scored by experts according to different criteria. The second
(continued on next page)
objective function maximizes the sum of these points to select the
200 S. Kargar et al. / Waste Management 113 (2020) 197–209

Fig. 1. The scheme of the proposed medical waste reverse supply chain (MWRSC) network.

⇑ (continued) cost for opening storage centers, treatment centers, transfer sta-
tions, and disposal centers, respectively.
Indices ! !
i Medical centers (1 . . . i . . .I) P
3 PP PPPP
Min Z 1 ¼ HJfj  IJ fjt þ CT k  W ck  Z jct
Z jct Amount of waste transported from storage center j to f ¼2 j t j c k t
treatment center c in period t    
PP PPP
Q cnt Amount of waste transported from treatment center þ HN n  INnt þ DC d  Endt
n t n d t
c to transfer station n in period t !!
Q 0int Amount of waste transported from medical center i P
3 PPP PPP
þ R DIJ ij  Y fijt þ DJC jc  Z jct
to transfer station n in period t f ¼2 i j t j c t
Endt Amount of waste transported from transfer station n  
PPP PPP
to disposal center d in period t þ R0  DINin  Q 0int þ DCNcn  Q cnt
IJfjt Inventory level of waste f in storage center j in period i n t c n t

t PPP
INnt Inventory level of transfer station n in period t
þ DNDnd  Endt
n d t
  
PPP 0 PPP PPP
þ LC  Q int þ Q cnt þ Endt
i n t c n t n d t
2.2. Objective functions !  
P PP
þ FIXJj  Bj þ FIXC ck  W ck
In this model, the objective function (1) minimizes the total cost j c k
that is composed of the following components. The first term com-    
P P
putes the storage cost of IMW in the storage center. The second þ FIXNn  Pn þ FIXDd  DId
n d
term calculates the treatment cost of IMW in the treatment center.
The third term is the storage cost of medical waste (GMW and ð1Þ
TIMW) in the transfer station. The fourth term is the disposal cost Treatment technologies were scored under different criteria.
of medical waste in the disposal center. The fifth term is the trans- Technology with higher total scores is nominated as the best tech-
portation cost of IMW from medical centers to storage centers and nology. The objective function (2) maximizes the total technology
then to treatment centers. The sixth term is the transportation cost score of treatment centers for selecting appropriate treatment
of medical waste (GMW and TIMW) from medical centers and technology under various criteria.
treatment centers to transfer stations then disposal centers. The XXX
seventh term is the labor cost of collection and handles medical Max Z 2 ¼ WIa  TMka  W ck ð2Þ
c k a
waste between centers. The eighth to eleventh terms are the fixed
 S. Kargar et al. / Waste Management 113 (2020) 197–209 201

!
The objective function (3) minimizes the total IMW stored in X X X
storage centers. Q cnt þ Q 0int þ INnðt1Þ ¼ INnt þ Endt 8n; t ð16Þ
XXX c i d
MinZ 3 ¼ IJfjt ð3Þ
Constraint (17) is the disposal centers capacity. Constraints (18)
f j t
and (19) enforce the range of variables.
X
2.3. Constraints Endt 6 CAPDd  DId 8d; t ð17Þ
n

Constraint (4) ensures that the total amount of IMW generated Bj ; W ck ; Pn ; DId ; ZC ct 2 f0; 1g 8i; j; s; k; n; q ð18Þ
in medical centers is equal to the total amount of IMW stored in
storage centers. Constraint (5) ensures that the total amount of
Y fijt ; Z jct ; Q cnt ; Q 0int ; Endt ; INnt ; IJfjt P 0 8i; j; s; k; n; q ð19Þ
GMW generated in medical centers is equal to the total amount
of GMW stored in transfer stations. Constraint (6) represents a bal-
ance constraint for IMW flow between storage centers and treat- 2.4. Linearization
ment centers.
X The proposed model includes two types of nonlinear terms. The
Y fijt ¼ DEfit 8f ¼ 2; 3; i; t ð4Þ
first type is the multiplication of a binary variable in a continuous
j
variable, and the second one is the multiplication of two binary
X variables. The second term of objective function (1) is the first type
Q 0int ¼ DEfit 8f ¼ 1; i; t ð5Þ
n
of non-linear one that is transformed into a linear function using a
new positive variable, KK jckt ¼ W ck  Z jct and the original non-
X
3 X
3 X X X
3 linear objective function is replaced with the constraints presented
IJfjðt1Þ þ Y fijt ¼ Z jct þ IJfjt 8j; t ð6Þ in Eqs. (20)–(24) (Arabsheybani et al., 2018):
f ¼2 f ¼2 i c f ¼2 XXXX
CT k  KK jckt ð20Þ
Constraints (7) and (8) are related to storage centers. Constraint j c k t
(7) is a capacity constraint. Constraint (8) ensures that the IMW is
transported to storage centers when it is opened. KK jckt 6 M  W ck 8j; c; k; t ð21Þ
IJfjt 6 CAPJ fj  Bj 8f ; j; t ð7Þ
KK jckt P Z jct  M  ð1  W ck Þ 8j; c; k; t ð22Þ
XXX
Y fijt 6 M  Bj 8j ð8Þ
f i t
KK jckt 6 Z jct þ M  ð1  W ck Þ 8j; c; k; t ð23Þ

Constraints (9)–(12) are related to treatment centers. Constraint KK jckt P 0 and integer 8j; c; k; t ð24Þ
(9) is a capacity constraint. Constraint (10) ensures that the
amount of transported IMW more than a specified amount by In constraints of treatment centers, constraints (9) and (10) are
the owners. This amount is multiplied by a coefficient called alpha the second type of non-linearization terms that may be trans-
(a) in the treatment center’s capacity. Constraint (11) ensures that formed into a linear function using a new binary vari-
each treatment center utilizes one treatment technology at most. ableWZ ckt ¼ W ck  Z ct . The original non-linear constraints are
Constraint (12) represents a balance constraint for waste flow replaced with the constraints presented in Eqs. (25)–(29) (Paydar
between treatment centers and transfer stations. and Saidi-Mehrabad, 2015).
X X X X
Z jct 6 ðCAPC ck  W ck  ZC ct Þ 8c; t ð9Þ Z jct 6 ðCAPC ckt  WZ ckt Þ 8c; t ð25Þ
j k
j k
! !
X X X X
Z jct P a  CAPC ck  W ck  ZC ct 8c; t ð10Þ Z jct P a  CAPC ckt  WZ ckt 8c; t ð26Þ
j k j k

X
W ck 6 1 8c ð11Þ WZ ckt  W ck  ZC ct þ 1:5 P 0 8c; k; t ð27Þ
k

X X 1:5  WZ ckt  W ck  ZC ct 6 0 8c; k; t ð28Þ

Q cnt ¼ Z jct 8c; t ð12Þ
n j WZ ckt 2 f0; 1g 8c; k; t ð29Þ
Constraints (13)–(16) are related to transfer stations. Con-
straint (13) is the transfer station capacity. Constraints (14)
3. Robust possibilistic optimization
and (15) ensure that the waste flow is transported to transfer
stations when it is opened. Constraint (16) represents a balance
3.1. Uncertainty in MWRSC
constraint for waste flow between treatment centers and trans-
fer stations.
One of the main concerns of the MWRSC is uncertainty in the
INnt 6 CAPN n  Pn 8n; t ð13Þ type and amount of waste generated. Some medical centers can
XX be a trauma center where most of the patients suffering from
Q cnt 6 M  Pn 8n ð14Þ major traumatic injuries such as falls, motor vehicle collisions, or
c t gunshot wounds are sent to these centers. Due to these accidents
XX are unpredictable, accurate estimation of the amount of waste
Q 0int 6 M  Pn 8n ð15Þ can be associated with high uncertainty. According to the climate
i t and geophysical conditions of Iran, the possibility of unpredictable
202 S. Kargar et al. / Waste Management 113 (2020) 197–209

 

natural disasters such as floods and earthquakes are very likely.
Nec Ax ¼ d P h ð37Þ
Therefore, the demand for medical services may be associated with
high uncertainty. Thus, the amount of medical waste generated
will be very uncertain. On the other hand, the inaccurate estima- Bx ¼ 0 ð38Þ
tion of the parameters leads to inadequate and unreliable results.
n  o
These results can lead to making incorrect decisions by medical Nec Sx 6 N y P b ð39Þ
centers managers which may cause very severe health threats.
Thus, the correct estimation of MWRSC parameters is vital.
Tx 6 1 ð40Þ
3.2. Possibilistic optimization framework
y 2 f0; 1g; x P 0 ð41Þ
Supply chain models may confront two types of uncertainty. b and h represent the confidence level of chance constraints.
The first type of uncertainty is the deficiency of knowledge about According to the fuzzy number ranking method, (Jiménez et al.,
    
the accurate values of the model parameters (imprecise/ dubious 2007), if g is a real number and d ¼ ðd1 ; d2 ; d3 ; d4 Þis a trapezoidal
parameters). The second type of uncertainty is the flexible target number then it can be stated as Eqs. (42)–(44):
value of goals and constraints (ambiguous/unsharp boundaries).
 
Possibilistic programming and robust optimization are two meth-
ods utilized to cope with the first type of uncertainty. Robust pos-
Nec d 6 g P h $ g P ð1  hÞd3 þ hd4 ð42Þ
sibilistic optimization is a hybrid approach to handle uncertainty.
Some reasons for the combination of possibilistic programming  
and robust optimization approaches are as follows. Firstly, in Nec d P g P h $ g 6 hd1 þ ð1  hÞd2 ð43Þ
possibilistic programming, the confidence level of constraint’s sat-
isfaction is a parameter that is determined by the decision-maker.   ( 
Therefore, the determined level of confidence cannot be an opti- g 6 2h d3 þ 1  2h d4
Nec d ¼ g P h $  ð44Þ
mal value, but in a robust optimization model, the confidence g P 2h d2 þ 1  2h d1
level is optimized because the confidence level is determined by
the model. Secondly, in the possibilistic programming model, So, the model of possibilistic programming in Eqs. (36)–(41) can
there is low attention to the feasible robustness and optimally be reformulated as Eqs. (45)–(51):
robustness however in robust optimization, the final solution  
c þ c þ c þ c 
has feasibly robustness and optimally robustness. Third, the little fi1 þ fi2 þ fi3 þ fi4 1 2 3 4
MinE½Z  ¼ yþ x ð45Þ
attention to the deviations of the objective function due to the 4 4
uncertainty of the parameters can cause considerable and s.t.
irreparable costs for owners and organizations, while in possi-  
bilistic programming do not handle this issue sufficiently. The h h
Ax 6 d3 þ 1  d4 ð46Þ
compressed form of a deterministic supply chain model can be 2 2
stated as Eqs. (30)–(35):
 
MinZ ¼ fi  y þ c  x ð30Þ h h
Ax P d2 þ 1  d1 ð47Þ
2 2
s.t.

Ax ¼ d ð31Þ Bx ¼ 0 ð48Þ

Bx ¼ 0 ð32Þ Sx 6 ½ð1  bÞN2 þ bN1 y ð49Þ

Sx 6 Ny ð33Þ
Tx 6 1 ð50Þ
Tx 6 1 ð34Þ
y 2 f0; 1g; x P 0 ð51Þ
y 2 f0; 1g; x P 0 ð35Þ
The vectors fi, c, and d respectively relate to fixed opening costs,
3.3. The robust possibilistic optimization model
transportation costs and production, and demand levels. Matrix A,
B, N, S, and T are the coefficients matrix of constraints. In addition, y
For a possibilistic programming model in Eqs. (45)–(51),
and  vectors are binary variables. We assume that the values of c according to (Pishvaee et al., 2012), this model can be turned into
and fi are imprecise parameters in the objective function, and d and
an RPP model as Eqs. (52)–(58):
coefficient matrix N are imprecise parameters in the constraints of
the compact formulation. It should be noted that a trapezoidal
MinE½Z  þ cðZ max  E½Z Þ þ p½bN 1 þ ð1  bÞN2  N1 y
function is utilized to formulate imprecise parameters. The trape-      
 h1 h1
zoidal possibility distribution imprecise parameterdis shown in þ d1 d4   d2  1   d1
2 2
Fig. S1.     
h1 h1
The compact formulation form of a possibilistic programming þ d2  d3 þ 1   d4  d1 ð52Þ
model can be stated as Eqs. (36)–(41): 2 2
hi   hi s.t.
Min E Z ¼ E f i y þ E c x ð36Þ  
h h
Ax 6 d3 þ 1  d4 ð53Þ
s.t. 2 2
 S. Kargar et al. / Waste Management 113 (2020) 197–209 203

 
h h MinZ 0 1 ¼ E½Z 1  þ cðZ 1max  E½Z 1 Þþ
Ax P d2 þ 1  d1 ð54Þ
2 2 " #
3 P P
P h 
d1 DEfit4  1
 DEfit2  1  h21  DEfit1
Bx ¼ 0 ð55Þ f ¼2 i t
2

Sx 6 ½bN 1 þ ð1  bÞN2 y ð56Þ " #

3 P P 
P 
d1 h1
2
 DEfit3 þ 1  h1
2
 DEfit4  DEfit1 þ
Tx 6 1 ð57Þ f ¼2 i t ð72Þ
" #
y 2 f0; 1g; x P 0; 0:5 6 h; b 6 1 ð58Þ P P P h 
d2 DEfit4  2
2
 DEfit2  1  h22  DEfit1 þ
The coefficients c, p, d1 and d2 represent the weight (impor- f ¼1 i t

tance) of each term against other terms in the objective function.

" #
Z max is calculated as Eq. (59): P P P h2 
d2 2
 DEfit3 þ 1  h22  DEfit4  DEfit1
Z max ¼ fi4 y þ c4 x ð59Þ f ¼1 i t

The above model is a nonlinear model. To transform it into a lin- XXX

ear model, letv be an auxiliary variable that is defined as Eq. (60): MaxZ 2 ¼ WIa  TMka  W ck ð73Þ
c k a
v ¼by ð60Þ
XXX
Then the nonlinear model is transformed into a linear model as MinZ 3 ¼ IJ fjt ð74Þ
f j t
Eqs. (61) – (71):
s.t.
MinE½Z  þ cðZ max  E½Z Þ þ p½bN1 þ ð1  bÞN2  N1 yþ ð61Þ Eqs. (6)–(8), (11)–(19), (21)–(24), (25)–(29) and
      X    
h1 h1
d1 d4 
h1
 d2  1 
h1
 d1 Y fijt P  DEfit2 þ 1   DEfit1 8f ¼ 2; 3; i; t ð75Þ
2 2 j
2 2
    
h1 h1
X    
þ d2  d3 þ 1   d4  d1 ð62Þ h1 h1
2 2 Y fijt 6  DEfit3 þ 1   DEfit4 8f ¼ 2; 3; i; t ð76Þ
j
2 2
s.t.
  X    
h2 h2
h
Ax 6 d3 þ 1 
h
d4 ð63Þ Q 0int P  DEfit2 þ 1   DEfit1 8f ¼ 1; i; t ð77Þ
2 2 n
2 2

  X    
h2 h2
h
Ax P d2 þ 1 
h
d1 ð64Þ Q 0int 6  DEfit3 þ 1   DEfit4 8f ¼ 1; i; t ð78Þ
2 2 n
2 2

0:5 6 h; b 6 1 ð79Þ
Bx ¼ 0 ð65Þ

Sx 6 v N1 þ ðy  v ÞN2 ð66Þ 4. The fuzzy goal programming method

v 6 My ð67Þ As discussed in Section 3.2, one of the uncertainties that the

supply chain face is the flexible target value of goals and con-
straints. This uncertainty is modeled by a flexible programming
v P M ð y  1Þ þ b ð68Þ
method. In flexible programming methods, the membership func-
tions of the constraints and fuzzy goals are determined by the
v 6b ð69Þ decision-maker. One of the tools for considering uncertainty is
fuzzy logic that is based on expert knowledge and experience. Dif-
Tx 6 1 ð70Þ ferent methods are introduced for FGP. In this method, the problem
is defined as Eqs. (80) – (83):
y 2 f0; 1g; x; v P 0; 0:5 6 h; b 6 1 ð71Þ find X ð80Þ
s.t.
3.4. The proposed robust possibilistic optimization model
f i ðXÞ  bi i ¼ 1; 2; :::; n ð81Þ

In the MWRSC the value of R, R0 and CT k respectively, unit
transportation cost for GMW, unit transportation cost for IMW hk ðXÞ ¼ ð6 or P Þ k ¼ 1; 2; :::; q ð82Þ
and unit treatment cost by technology k are imprecise parame-
ters in the first objective function. In addition, the value of XP0 ð83Þ
DEfit is equal to the amount of waste f generated in a medical
where f i ðxÞ is the fuzzy objective i, bi is the value of objective i,
center i in period t which is an imprecise parameter in the
hk ðxÞ ¼ ð6 or PÞ is the set of system constraints and X is an n-
model constraints. Considering the previous description and
dimensional vector of decision variables. If the fuzzy objective is
referring to the (Pishvaee et al., 2012) model, the proposed
for i ¼ 1; 2; :::; io asf i ðxÞ < bi then the membership function will be
robust possibilistic optimization model of MWRSC is defined as 
Eqs. (72) – (79): as Eq. (84) and presented in Fig. S2.
204 S. Kargar et al. / Waste Management 113 (2020) 197–209

8
>
< 1 if f i ðXÞ 6 bi ðAXÞi þ di 6 bi
 l
i ¼ ko þ 1; :::; k ð100Þ
U i f i ðXÞ
li ¼ if bi 6 f i ðXÞ 6 U i ð84Þ
> Ui bi
:
0 if f i ðXÞ P U i li ; di ; dþi P 0 i ¼ 1; :::; k ð101Þ

where U i an upper tolerance is a limit for the fuzzy objective f i ðxÞ X 2 Cs ð102Þ
and Di ¼ U i  bi is the tolerance range. If the fuzzy objective for
i ¼ io þ 1; :::; n isf i ðxÞ > bi , then the membership function will be as In this model, Eqs. (93) and (94) are related to minimization
 problems, Eqs. (95) and (96) are related to the maximization prob-
Eq. (85) and is shown in Fig. S3. lems, Eqs. (97) and (98) handle problems with triangular member-
8 ship function, Eqs. (99) and (100) are for problems with a
>
< 1 if f i ðXÞ P bi
f i ðXÞLi trapezoidal membership function (in the triangular and trape-
li ¼ if Li 6 f i ðXÞ 6 bi ð85Þ
> bi Li
: zoidal membership function, the deviations must be minimized
0 if f i ðXÞ 6 Li on both sides), and Eq. (102) are related to the main constraints
of the problem.
Which Li is a lower tolerance limit for the fuzzy objectivef i ðxÞ
Their method considers the absolute value of deviations and the
and Di ¼ bi  Li is the tolerance range. To minimize the member-
degree of membership functions in optimal solutions where devi-
ship function, the objective function is as Eq. (86):
 ation values can be employed by decision-makers for further anal-
Max Min li ð86Þ ysis. In this model, one can consider different weights for positive
X i
and negative deviations. Moreover, the model minimizes the
If we assumek ¼ Mini ðli Þ, then the definitive model will be as changes required for sensitivity analysis.
Eqs. (87)–(91): FGP is utilized when a decision-maker is not able to determine
Max k ð87Þ the exact levels of the aspirations; therefore, changes in tolerable
deviations and levels of aspirations may often occur. In previous
s.t. models, one needs more changes for sensitivity analysis, while in
U i  f i ðXÞ the utilized model, changing the coefficients of the deviation vari-
k6 i ¼ 1; 2; :::; io ð88Þ þ 
ables (di anddi ) or the values of the uncertain aspirations (Bi ) is
U i  bi
sufficient. Thus, the proposed FGP optimization model of MWRSC
f i ðXÞ  Li is presented as Eqs. (103)–(110):
k6 i ¼ io þ 1; :::; n ð89Þ
bi  Li d1
þ 
d2 d3
þ
MinFGP ¼ w1 R
þ w2 þ w3 ð103Þ
hk ðXÞ ¼ ð6 or P Þ k ¼ 1; 2; :::; q ð90Þ D1 DL2 DR3
s.t.
XP0 ð91Þ Eqs. (6)–(8), (11)–(19), (20)–(24), (25)–(29), (75)–(79) and
Several weighted additive models are introduced to solve þ
Z 01  d1 6 B1 ð104Þ
FGP problems. These models are based on both GP and fuzzy
programming techniques. However, some of these models are þ
d1
not able to solve all FGP problems and the main weakness is l1 þ ¼1 ð105Þ
the lack of precision in their formulations. (Yaghoobi et al., DR1
2008) presented a new weighted additive model to solve the 
FGP problems in 2008 by improving in the WGP models and Z 2 þ d2 P B2 ð106Þ
combining it with fuzzy programming techniques as presented 
in Eqs. (92)–(102). d2
l2 þ ¼1 ð107Þ
DLi
!
X
io
di
þ X
jo 
di X
K 
di
þ
di
Min wi R
þ wi þ wi þ ð92Þ þ
Z 3  d3 6 B3 ð108Þ
i¼1 Di i¼io þ1 DLi i¼jo þ1 DLi DRi
þ
s.t. d3
l3 þ ¼1 ð109Þ
ðAXÞi  di 6 bi
þ
i ¼ 1; :::; io ð93Þ DR3

þ 
di
þ di ; di ; DRi ; DLi ; Bi P 0 8i ð110Þ
li þ ¼ 1 i ¼ 1; :::; io ð94Þ
DRi

5. Case study
ðAXÞi þ di P bi i ¼ io þ 1; :::; jo ð95Þ
 5.1. Case description
di
li þ ¼ 1 i ¼ io þ 1; :::; jo ð96Þ
DLi Iran is one of the developing countries in the Middle East. In
recent years, the number of medical centers in Iran have increased
 þ
ðAXÞi þ di  di ¼ bi i ¼ jo þ 1; :::; ko ð97Þ significantly. According to the Iranian Ministry of Health and Med-
ical Education, the total number of active beds in hospitals were
 þ
di di about 120,612 in 2018. In addition to 2,632 rural health clinics,
li þ þ ¼ 1 i ¼ jo þ 1; :::; ko ð98Þ 2,783 urban health centers provide health care services. On the
DLi DRi
other hand, Iran’s population has doubled between the years
þ u 1980 and 2017. Population growth and increasing access to health
ðAXÞi  di 6 bi i ¼ ko þ 1; :::; k ð99Þ
services led to an increase in medical waste generation. The
 S. Kargar et al. / Waste Management 113 (2020) 197–209 205

environmental organization and the Ministry of Health and Medi- The capacity of the transfer station and the disposal center is 15
cal Education are responsible for monitoring the management of and 200 ton, respectively. The minimum amount of waste that the
waste generated by medical centers. Segregated medical waste in treatment equipment starts operating is different in each center.
one of Iran’s medical centers is shown in Fig. S4. Usually, this parameter is between 20 and 80 percent of the treat-
The northern part of Iran has dense forests, unique vegetation ment center’s capacity per day. In the current study, this parameter
and a high population density compare to other parts of the coun- is about 40 percent. There are three potential treatment centers
try. Because of this, medical waste management in northern Iran is that each of them should be equipped with one treatment technol-
much more important than other parts. Thus, Babol city with a ogy. Information about potential treatment technologies was
250,126 population which is located in Mazandaran province cho- received from reputable companies in Iran in 2019 and shown in
sen as a case study. Table S2. Moreover, there is a potential storage facility near each
Medical centers with a larger share in providing medical ser- treatment center, the parameters for these centers are shown in
vices (generating more medical waste) to the public are chosen Table S3.
in this case. Twenty high demanding centers are considered. The To evaluate the fixed cost parameters ‘‘Equivalent Uniform
list includes eight hospitals and twelve clinics. Seven medical cen- Annual Cost” (EUAC), one of the ‘‘engineering economics analysis”
ters have storage centers and treatment centers while other cen- method, is employed. All costs are transmitted to the annual cost.
ters do not have. Three types of waste, a seven-day period, a According to this method as presented in Eq. (118), cost parame-
disposal center (landfill) which is located in the suburb of Babol, ters are composed of asset initial cost (p), asset salvage value end
and a transfer station in the southern part of this city are consid- of life (SV), lifetime (n), annual operating and maintenance costs
ered. Three potential locations for the treatment centers in eastern, (M), and rate of interest (i). Then by dividing the EUAC by the days
southern, and western parts of the city are identified by experts. in a year, equivalent uniform period time cost (seven days) is
Generally, storage centers for infectious waste storage are set up derived. The cost parameter’s unit is Iranian million Tomans
alongside the treatment center. Thus, each potential treatment (Approximately 100 US dollars).
center is considered with one potential storage center. To illustrate
more clearly the location of medical centers, treatment centers and EUAC ¼ PðA=P; i%; nÞ  SVðA=F; i%nÞ þ M ð118Þ
storage centers on the existing and potential network is shown in The average maintenance cost of sharps and non-sharps IMW in
Fig. S5. the storage centers are 0.12 and 0.10 per ton/day, respectively.
Treatment cost is an uncertain parameter. The trapezoidal fuzzy
5.2. Input parameters number of this parameter per ton by various technologies is shown
in Table S4. The storage cost of waste in the transfer station is 0.01
In the studied network five main levels, including twenty med- per ton. Moreover, transportation cost is an uncertain parameter.
ical centers, seven storage centers, seven treatment centers, one The transportation cost of IMW is estimated as a trapezoidal fuzzy
transfer station, and one disposal center are identified. Currently, number of (0.45, 0.48, 0.5, 0.53) per ton/km. Similarly, this cost for
there are treatment centers and storage centers only located in GMW is (0.009, 0.010, 0.011, 0.012) per ton/km. Labor costs for col-
medical centers. The value of the parameters for the medical cen- lecting waste is 0.13 per ton. Disposal cost which is composed of
ters, including the amount of generated waste, storage center’s waste processing cost and landfill cost, in the Anjilsi disposal cen-
capacity, the technology type and capacity of the treatment center ter, is equal to 0.03 cost unit/per ton.
are shown in Table 2. The volume of waste generated by medical There are different treatment technologies for medical waste.
centers is an uncertain parameter. The assignable and actual value Choosing the best technology is an important decision, because
for this parameter is shown in Table S1 which is modeled as a incorrect choices may cause health problems, economic losses,
trapezoidal fuzzy number. For example, medical waste generated and even social discontent. In this study, the selection is according
within the hospital (A) is illustrated in Table S1. to the methods used in the most recent research. In choosing

Table 2
Information about medical centers.

Medical center Storage center’s capacity of IMW Treatment center’s Treatment center’s GMW (kg) IMW
Capacity (kg) Technology (kg)
Sharps (kg) Non-sharps (kg) Sharps (kg) Non-sharps (kg)
Hospital A 200 2000 320 Autoclave 220 20 120
Hospital B 200 2000 160 Autoclave 300 15 120
Hospital C 200 2000 640 Autoclave 470 40 570
Hospital D 200 2000 400 Autoclave 700 40 360
Hospital E 50 800 150 Autoclave 180 20 100
Hospital F 50 500 25 Autoclave 25 5 17
Hospital G 80 1000 75 Autoclave 40 13 32
Hospital H – – – – 30 4 18
Clinic I – – – – 10 4 19
Clinic J – – – – 14 5 30
Clinic K – – – – 11 6 35
Clinic L – – – – 14 7 34
Clinic M – – – – 12 4 20
Clinic N – – – – 6 12 20
Clinic O – – – – 12 8 34
Clinic P – – – – 14 5 31
Clinic Q – – – – 6 12 22
Clinic R – – – – 8 8 20
Clinic S – – – – 22 10 39
Clinic T – – – – 13 14 31

IMW: infectious medical waste.

GMW: general medical waste.
206 S. Kargar et al. / Waste Management 113 (2020) 197–209

medical waste treatment technology various criteria such as eco- TIMW, and labor costs which equal 5.051 million Tomans and
nomic, environmental, technical and social criteria are considered. 4.866 million Tomans, respectively. All costs of the proposed sup-
The criteria for treatment technology selection are borrowed from ply chain network are shown in Table S9. The comparison of these
Xiao (2018). costs is shown in Fig. S6. The optimal MWRSC network is shown in
The criteria for selecting treatment technologies include eco- Fig. S7. In the optimal solution, two new treatment centers, with
nomic, social, technical and environmental criteria. These criteria two storage ones, in the western and eastern part of the city sug-
have a series of sub-criteria which are shown in Fig. 2. According gested to be opened. These treatment centers are equipped with
to Fig. 2, the economic criterion has a sub-criterion called net cost autoclave technology and minimal capacity. The results of the
per ton as well as some risk levels associated with each technology opening new centers are illustrated in Table 3.
are described respectively. ‘‘Waste residuals” means the amount of The IMW, including sharps and non-sharps, are sent from med-
waste that remains after the treatment process. Due to the reduc- ical centers to the storage ones and then to the treatment center.
tion in waste volume, incinerator and autoclave are the best tech- According to the model optimal solution, the amount of this waste
nology and sanitary landfill is the worst ones. The ‘‘Noise” is the is shown in Table S10. In addition, GMW generated in medical cen-
amount of noise generated during the treatment process. Auto- ters is separately collected from IMW. Like other municipal waste,
clave and sanitary landfill generate less noise than an incinerator. this waste will be sent directly to the transfer station without any
‘‘Release health effects” is the impact of treatment technology on special handling. The amount of GMW which is sent from medical
the various health indicators. Many hazardous microorganisms centers to the transfer station number 1 is shown in Table S11. The
remain in sanitary landfills, causing environmental and human amount of waste sent from the transfer station to the disposal cen-
pollution. ‘‘Reliability” means the quality of treatment being trust- ter in each period is shown in Table S12.
worthy in various conditions. In this criterion, autoclave and incin-
erator are the best technology and sanitary landfill is the worst 5.4. Sensitivity analyses and discussion
ones. ‘‘Treatment effectiveness” is the degree in which the treat-
ment process is successful and producing a desired result. The Sensitivity analysis of the key parameters is important that
results of the sanitary landfills method are not highly effective. attempts to provide a measure of the sensitivity of either parame-
‘‘Occupational hazards” are the dangers and injuries that laborers ter, or forcing functions, or sub models to the state variables of
may suffer, like infectious illnesses. Correspondingly, in terms of greatest interest in the model. In this section, four sensitivity anal-
occupational hazards, sanitary landfills are the worst technology. yses are performed. In addition, managerial insights are provided
‘‘Public acceptance” is the acceptance of treatment technology at the end of the section.
equipment by the people around the treatment center areas. Due
to the emission of hazardous gases, incinerators are one of the 5.4.1. Objective functions analysis
worst technology in public acceptance. In Xiao’s (2018) study, The proposed model has tri-objective functions. In this section,
average weights are reported according to some decision-maker’s each objective function is optimized individually. The optimizing
comments. The treatment technologies score under various criteria results of the single and multiple objectives are presented in
are shown in Table S5. The weight of each criterion is shown in Table S13. As shown in Table S13, when only the first objective
Table S6. function is considered, the amount of total costs is minimized,
and the new treatment center is not opened so existing ones are
5.3. Results used only. In these conditions, 2.7 tons of IMW overall per week
is stored in the storage center which may cause a serious risk.
The proposed tri-objective model with the real case study is When only the second objective function is considered, we expect
implemented by a personal computer with IntelÒ CoreTM i5 2410 the technology total score of treatment centers that utilize for
CPU 4.00 GHz, 6.00 GB RAM and Lingo 17 software package. The selecting appropriate treatment technology under various criteria
optimal value of the first objective function is obtained 22.914 mil- is maximized. Therefore, regardless of the minimizing costs, three
lion Tomans. The value of all objective functions is presented in new treatment centers are opened also the highest value for the
Table S7. In the FGP objective function, the deviations values of second objective function is obtained.
the second and third functions are as low as possible, which is
excellent, and the first objective function with 1.825 million 5.4.2. Coefficient of starting treatment equipment (a) analysis
Tomans in deviation has an acceptable solution. It means 12.5% A minimum required volume of IMW to start the treatment
deviation increase in costs can minimize waste stored in storage center equipment is defined as a coefficient that called alpha mul-
facilities that reduce the risk of spreading the infection. The result tiplied by the treatment center capacity. The alpha coefficient is
of the FGP model presented in Table S8. The highest cost of the determined by the manager of each treatment center according
MWRSC is treatment cost which is equal to 7.976 million Tomans. to the different criteria. The second sensitivity analysis is related
Besides, the next ranks are the transportation cost of GMW and to the alpha coefficient. The effect of alpha coefficient variations

Fig. 2. Treatment technology selection criteria.

 S. Kargar et al. / Waste Management 113 (2020) 197–209 207

Table 3
Result of Treatment center opening.

Opening new centers Treatment center Storage center

Technology type Capacity (kg) IMW (kg) GMW (kg)
East potential center Autoclave 160 200 2000
West potential center Autoclave 160 200 2000
South potential center – – – –

on objective functions and the opened treatment centers are It means the maximum amount of waste generated should be
shown in Table 4. As shown in Table 4, two new treatment centers considered for IMW and NIMW thus lead to all of the new treat-
are opened when the alpha coefficient values are 0.2 or 0.4. How- ment centers are opened. Considering the second case, the IMW
ever, when the alpha value is 0.6 or 0.8, only one new treatment is considered to be the least possible, therefore the number of
center is opened. A small alpha coefficient cause frequent starting opened centers is the minimum. This case has a higher risk,
and depreciation of treatment equipment. On the other hand, high because the amount of medical waste may not commonly be the
values of alpha cause IMW is stored in the storage center longer lowest (best) case. This is an optimistic forecast and cannot be con-
time, which may cause serious risks. For these reasons, the alpha sidered the basis of this research. Thus, the first case is more
value of 0.4 is suggested in this study. realistic.

5.4.4. FGP model coefficients analysis

5.4.3. RPP model coefficients analysis The third sensitivity analysis is based on the importance of
In this section, sensitivity analysis is performed on the coeffi- deviations from the goals of the objective functions in the FGP
cients of the RPP model function terms. The results of this sensitiv- model. The coefficients of each term in the FGP model objective
ity analysis are presented in Table 5. A higher penalty coefficients function represent their importance. The results of this sensitivity
means the maximum amount (worth case) is considered for the analysis are shown in Table 6. According to Table 6, the increase in
generation of medical waste. According to Table 5, the coefficients each coefficient causes the value of the objective functions closest
of the second and third terms are higher than the first term. to the aspiration level. One of the main goals of this study was to

Table 4
The effect of alpha coefficient variations on the objective functions and opening new centers.

Objective Alpha coefficient (a)

0.2 0.4 0.6 0.8

Z 01 (millionTomans) 22.915 22.915 22.974 23.134

Z2 14.10 14.10 8.6 8.6
Z 3 (kg) 0 0 0 0
Number of opened 2 2 1 1
treatment centers
Treatment centers Autoclaves in the eastern and Autoclaves in the eastern and Autoclave in the eastern part of the Autoclave in the eastern
feature western part of the city western part of the city city (capacity:160 kg) part of the city
(capacity:160 kg) (capacity:160 kg) (capacity:160 kg)

One Million Tomans: Approximately 100 US dollars.

Table 5
Sensitivity analysis of robust possibilistic programming (RPP) model coefficients.

Cases Weight (importance) Objective (value) RPP Number of opened treatment centers
c d1 d2 Z 01 (million Tomans) Z2 Z 3 (kg)

1 5 10 10 22.915 14.10 0 95.917 2

2 10 5 10 20.500 8.6 0 94.925 1
3 10 10 5 21.450 14.10 10 96.854 2
4 10 10 10 23.250 14.10 0 99.545 2

One Million Tomans: Approximately 100 US dollars.

Table 6
Sensitivity analysis of fuzzy goal programming (FGP) model coefficients.

Cases FGP Coefficient Deviation Number of opened treatment center future

þ  þ
W1 W2 W3 d1 (million Tomans) d2 d3 (kg)

1 0.333 0.333 0.333 4.5 0 0 Two autoclaves

2 0.6 0.2 0.2 3 2.4 5.9 One autoclaves
3 0.2 0.2 0.6 6.8 2.4 0 One autoclaves
4 0.2 0.6 0.2 7.5 0 3.8 Three autoclaves

One Million Tomans: Approximately 100 US dollars.

208 S. Kargar et al. / Waste Management 113 (2020) 197–209

minimize the amount of IMW stored in storage centers. For this centers among three potential centers, in the west and east of
reason, the third objective function is more important. The coeffi- Babol are suggested to be opened. In addition, to evaluate the
cients of the first case in Table 6 are selected as the optimal coef- results of the model, four types of sensitivity analyses are per-
ficients. In this case, deviations of the third objective function are formed on the key parameters. Some suggestions for future work
the minimum possible value, and fewer deviations in the other on this problem are as follows:
objective functions are seen in comparison with the third case.
1. Considering the routing problem for MWRSC from the medical
5.4.5. Managerial insights centers to the disposal centers
The purpose of this study was to properly manage medical 2. Considering other objective functions, such as the emissions or
waste and optimize the costs and minimizing risks of medical revenues from the reverse supply chain network.
waste. In addition, the proposed model can assist the medical cen- 3. Considering the amount of microbial contamination of different
ter’s managers in making strategic decisions. According to con- waste and minimizing the amount of damage to laborers and
ducted research in the field of medical waste in Iran, it can be people associated with medical waste.
understood that despite many costs paid in this area and several 4. Focusing on the generation of medical waste and creating a con-
facilities opened Iran has not reached its proper place in this field dition that minimizes the amount of medical waste.
yet. In fact, if Iran can optimize costs and efforts in this area and 5. Developing and implementing a meta-heuristic algorithm for
have a proper MWRSC network, it is hoped that in recent years, large-scale problems.
it can improve its situation and be able to even provide a model
for developing countries.
In the following, some of the recommendations for health sys- Declaration of Competing Interest
tem managers are presented to improve the MWRSC in Iran:
The authors declare that they have no known competing finan-
1. Assigning the budget and proofing the coordination needed to cial interests or personal relationships that could have appeared
integrated medical waste management in each city. The non- to influence the work reported in this paper.
integrated management increases the risks of the treatment
process therefore the quality of the process may be reduced
Appendix A. Supplementary material
in some centers, and the optimization of the cost will be very
difficult as well.
Supplementary data to this article can be found online at
2. The amount of IMW is lower than NIMW. If GMW waste is com-
https://doi.org/10.1016/j.wasman.2020.05.052.
bined with IMW, all of them will become infectious. Therefore,
if these two types of waste are combined, the volume of IMW
References
will be greatly increased that will create high risks and costs.
As far as possible, special attention should be paid to the waste Alshraideh, H., Qdais, H.A., 2017. Stochastic modeling and optimization of medical
segregation in the source of generation and these processes waste collection in Northern Jordan. J. Mater. Cycles Waste Manage. 19, 743–
should be carried out properly. 753. https://doi.org/10.1007/s10163-016-0474-3.
Arabsheybani, A., Paydar, M.M., Safaei, A.S., 2018. An integrated fuzzy MOORA
3. The deficiency of proper training plans for employees to learn method and FMEA technique for sustainable supplier selection considering
the risks of medical waste and methods to handle these risks quantity discounts and supplier’s risk. Journal of Cleaner Production 190, 577–
can cause irreparable damage to their health. In order to evade 591.
Budak, A., Ustundag, A., 2017. Reverse logistics optimisation for waste collection
potential risks and proper segregation, it is recommended to and disposal in health institutions: the case of Turkey. Int. J. Logist. Res. Appl.
provide continuing training to laborers, and the effectiveness 20, 322–341. https://doi.org/10.1080/13675567.2016.1234595.
of training should be evaluated. Caniato, M., Tudor, T., Vaccari, M., 2015. International governance structures for
health-care waste management: A systematic review of scientific literature. J.
4. Reducing as much as possible the effect of MWRSC on the envi- Environ. Manage. 153, 93–107. https://doi.org/10.1016/j.jenvman.2015.01.039.
ronment, health, and social issues. The selection of sites for a set De Almeida, J.M.N., 2010. A Cost Optimization Model for Hazardous Medical Waste
of potential new centers should be done by the experts. Management in Portugal.
Dehghani, M.H., Ahrami, H.D., Nabizadeh, R., Heidarinejad, Z., Zarei, A., 2019.
Medical waste generation and management in medical clinics in South of Iran.
6. Conclusions MethodsX 6, 727–733. https://doi.org/10.1016/j.mex.2019.03.029.
Doan, L.T.T., Amer, Y., Lee, S.-H., Phuc, P.N.K., Dat, L.Q., 2019. A Comprehensive
In this research, a tri-objective MILP model is designed to opti- Reverse Supply Chain Model using an Interactive Fuzzy Approach-a Case Study
on the Vietnamese Electronics Industry. Appl. Math. Model. https://doi.org/
mize the medical waste management system. The objective func- 10.1016/j.apm.2019.06.003.
tions of this study include minimizing the total costs, Dutta, P., Mishra, A., Khandelwal, S., Katthawala, I., 2020. A multiobjective
maximizing appropriate treatment technology selection under sus- optimization model for sustainable reverse logistics in Indian E-commerce
market. J. Cleaner Prod. 249,. https://doi.org/10.1016/j.jclepro.2019.119348
tainability criteria, and minimizing total IMW that stored in stor- 119348.
age centers. To handle uncertain parameters, a novel possibilistic Genovese, A., Acquaye, A.A., Figueroa, A., Koh, S.L., 2017. Sustainable supply chain
programming approach, called RPP, is employed. In addition, to management and the transition towards a circular economy: Evidence and
some applications. Omega 66, 344–357. https://doi.org/10.1016/j.
handle imprecise goals, the fuzzy goal programming is applied. omega.2015.05.015.
In this context, for the uncertain quantities of medical waste, the Gergin, Z., Tunçbilek, N., Esnaf, Sß., 2019. Clustering Approach Using Artificial Bee
optimized number and location of facilities and the types of tech- Colony Algorithm for Healthcare Waste Disposal Facility Location Problem. Int.
J. Operat. Res. Info. Syst. (IJORIS) 10, 56–75. https://doi.org/10.4018/
nologies in treatment centers are determined. In this research, a IJORIS.2019010104.
real case study of the medical waste network in one of the northern Jiménez, M., Arenas, M., Bilbao, A., Rodrı, M.V., 2007. Linear programming with
cities of Iran, Babol, was conducted to ensure the model applicabil- fuzzy parameters: an interactive method resolution. Eur. J. Oper. Res. 177,
1599–1609. https://doi.org/10.1016/j.ejor.2005.10.002.
ity. The case includes twenty medical centers, seven storage cen-
Korkut, E.N., 2018. Estimations and analysis of medical waste amounts in the city of
ters, seven treatment centers, one transfer station, and one Istanbul and proposing a new approach for the estimation of future medical
disposal center. Furthermore, the model was implemented by the waste amounts. Waste Manage. 81, 168–176. https://doi.org/10.1016/j.
Lingo 17 software package and solved by the FGP method. Accord- wasman.2018.10.004.
Lee, B.-K., Ellenbecker, M.J., Moure-Ersaso, R., 2004. Alternatives for treatment and
ing to the optimal solution, two new treatment centers with auto- disposal cost reduction of regulated medical wastes. Waste Manage. 24, 143–
clave technology and specific capacity, along with two storage 151. https://doi.org/10.1016/j.wasman.2003.10.008.
 S. Kargar et al. / Waste Management 113 (2020) 197–209 209

Makajic-Nikolic, D., Petrovic, N., Belic, A., Rokvic, M., Radakovic, J.A., Tubic, V., 2016. Rolewicz-Kalińska, A., 2016. Logistic constraints as a part of a sustainable medical
The fault tree analysis of infectious medical waste management. J. Cleaner Prod. waste management system. Transp. Res. Procedia 16, 473–482 https://core.ac.
113, 365–373. https://doi.org/10.1016/j.jclepro.2015.11.022. uk/download/pdf/82613820.pdf.
Mantzaras, G., Voudrias, E.A., 2017. An optimization model for collection, haul, Shi, L., Fan, H., Gao, P., Zhang, H., 2009. Network model and optimization of medical
transfer, treatment and disposal of infectious medical waste: Application to a waste reverse logistics by improved genetic algorithm, International
Greek region. Waste Manage. 69, 518–534. https://doi.org/10.1016/j. Symposium on Intelligence Computation and Applications. Springer, 40–52
wasman.2017.08.037. https://link.springer.com/chapter/10.1007/978-3-642-04843-2_6.
Mazloomi, S., Zarei, A., Alasvand, S., Farhadi, A., Nourmoradi, H., Bonyadi, Z., 2019. Shih, L.-H., Lin, Y.-T., 2003. Multicriteria optimization for infectious medical waste
Analysis of quality and quantity of health-care wastes in clinical laboratories: a collection system planning. Practice Periodical Hazard. Toxic Radioactive Waste
case study of Ilam city. Environ. Monit. Assess. 191, 207. https://doi.org/ Manage. 7, 78–85. https://doi.org/10.1061/(ASCE)1090-025X(2003)7:2(78).
10.1007/s10661-019-7345-z. Trochu, J., Chaabane, A., Ouhimmou, M., 2019. A two-stage stochastic optimization
Nolz, P.C., Absi, N., Feillet, D., 2014. A stochastic inventory routing problem for model for reverse logistics network design under dynamic suppliers’ locations.
infectious medical waste collection. Networks 63, 82–95. https://doi.org/ Waste Manage. 95, 569–583. https://doi.org/10.1016/j.wasman.2019.06.012.
10.1002/net.21523. Voudrias, E., Graikos, A., 2014. Infectious medical waste management system at the
Osaba, E., Yang, X.-S., Fister Jr, I., Del Ser, J., Lopez-Garcia, P., Vazquez-Pardavila, A.J., regional level. J. Hazardou. Toxic Radioactive Waste 18, 04014020. https://doi.
2019. A discrete and improved bat algorithm for solving a medical goods org/10.1061/(ASCE)HZ.2153-5515.0000225.
distribution problem with pharmacological waste collection. Swarm Evol. W.H.O., 2014. Safe management of wastes from health-care activities. World Health
Comput. 44, 273–286. https://doi.org/10.1016/j.swevo.2018.04.001. Organization. https://www.who.int/water_sanitation_health/publications/
Oyola-Cervantes, J., Amaya-Mier, R., 2019. Reverse logistics network design for large wastemanag/en/.
off-the-road scrap tires from mining sites with a single shredding resource W.H.O., 2017. Safe management of wastes from health-care activities: a summary.
scheduling application. Waste Manage. 100, 219–229. https://doi.org/10.1016/j. World Health Organization. https://www.who.int/water_sanitation_health/
wasman.2019.09.023. publications/safe-management-of-waste summary/en/.
Paydar, M.M., Olfati, M., 2018. Designing and solving a reverse logistics network for Windfeld, E.S., Brooks, M.S.-L., 2015. Medical waste management–A review. J.
polyethylene terephthalate bottles. J. Cleaner Prod. 195, 605–617. https://doi. Environ. Manage. 163, 98–108. https://doi.org/10.1016/j.jenvman.2015.08.013.
org/10.1016/j.jclepro.2018.05.218. Xiao, F., 2018. A novel multi-criteria decision making method for assessing health-
Paydar, M.M., Saidi-Mehrabad, M., 2015. Revised multi-choice goal programming care waste treatment technologies based on D numbers. Eng. Appl. Artif. Intell.
for integrated supply chain design and dynamic virtual cell formation with 71, 216–225. https://doi.org/10.1016/j.engappai.2018.03.002.
fuzzy parameters. Int. J. Comput. Integr. Manuf. 28, 251–265. https://doi.org/ Yaghoobi, M., Jones, D.F., Tamiz, M., 2008. Weighted additive models for solving
10.1080/0951192X.2013.874596. fuzzy goal programming problems. Asia-pacific journal of operational research
Pishvaee, M.S., Razmi, J., Torabi, S.A., 2012. Robust possibilistic programming for 25, 715–733. https://doi.org/10.1142/S0217595908001973.
socially responsible supply chain network design: A new approach. Fuzzy Sets Yong, Z., Gang, X., Guanxing, W., Tao, Z., Dawei, J., 2009. Medical waste management
Syst. 206, 1–20. https://doi.org/10.1016/j.fss.2012.04.010. in China: a case study of Nanjing. Waste Manage. 29, 1376–1382. https://doi.
Rajan, R., Robin, D.T., Vandanarani, M., 2018. biomedical waste management in org/10.1016/j.wasman.2008.10.023.
Ayurveda hospitals–current practices & future prospectives. Journal of
Ayurveda and integrative medicine. https://doi.org/10.1016/j.jaim.2017.07.011.