Ahmadi-Javid, Seyedi, Syam - 2017 - A Survey of Healthcare Facility Location

Computers & Operations Research 79 (2017) 223–263
Contents lists available at ScienceDirect
Computers & Operations Research

journal homepage: www.elsevier.com/locate/caor
A survey of healthcare facility location

Amir Ahmadi-Javid a,n, Pardis Seyedi a, Siddhartha S. Syam b
a
Department of Industrial Engineering, Amirkabir University of Technology, Tehran, Iran
b
Department of Management, Marquette University, Milwaukee, WI, USA
art ic l e i nf o a b s t r a c t
Article history: Healthcare facility (HCF) location has attracted considerable attention from the operations research
Received 12 March 2015 community over nearly four decades as one of the most important strategic issues in healthcare systems,
Received in revised form disaster management, and humanitarian logistics. However, the lack of a comprehensive review in the
2 April 2016
last decade is a serious shortcoming in the literature of HCF location. This survey presents a framework to
Accepted 30 May 2016
classify different types of non-emergency and emergency HCFs in terms of location management, and
Available online 15 October 2016
reviews the literature based on the framework. The papers on HCF location problems are classified in
Keywords: detailed tables along ten descriptive dimensions, which are consideration of uncertainty, multi-period
Healthcare facility location setting, particular input/setting, objective function, decision variable, constraint, basic discrete location
Health systems
problem, mathematical modeling approach, solution method, and case study inclusion. For each HCF
Emergency medical services
type, research gaps and possible future directions are identified. Moreover, the literature and future
Optimization methods
Operations research research possibilities are analyzed in terms of modeling approach and solution method.
Discrete location problems & 2016 Elsevier Ltd. All rights reserved.
Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
2. An overview of discrete location problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
2.1. Covering-based problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
2.1.1. Set covering location problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
2.1.2. Maximal covering location problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
2.1.3. p-center location problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
2.2. Median-based problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
2.2.1. p-median location problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
2.2.2. Fixed charge facility location problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
2.3. Other problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
3. Scope of literature survey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
3.1. A framework for classification of HCFs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
3.2. Descriptive dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
4. Non-emergency HCF location . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
4.1. Primary care facilities (hospitals, clinics, off-site public access devices, etc.) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
4.2. Blood banks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
4.3. Specialized services facilities. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
4.3.1. Organ transplant centers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
4.3.2. Detection and prevention centers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
4.3.3. Other specialized services facilities. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
4.4. Medical laboratories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
4.5. Mobile healthcare units. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
4.6. Home healthcare centers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
4.7. Rehabilitation centers, doctors' offices, and drugstores . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
4.8. Long-term nursing care centers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
n
Corresponding author.
E-mail address: [email protected] (A. Ahmadi-Javid).
http://dx.doi.org/10.1016/j.cor.2016.05.018
0305-0548/& 2016 Elsevier Ltd. All rights reserved.
224 A. Ahmadi-Javid et al. / Computers & Operations Research 79 (2017) 223–263
4.9. Combinations of several types of HCFs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242

5. Emergency HCF location . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
5.1. Permanent emergency HCFs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
5.1.1. Emergency off-site public access devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
5.1.2. Emergency centers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
5.1.3. Trauma centers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
5.1.4. Ambulance stations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
5.2. Temporary emergency HCFs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
5.2.1. Temporary medical centers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
5.2.2. Points of dispensing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
6. Analyses of literature from different perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
6.1. Basic discrete location problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
6.2. Modeling approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
6.3. Solution methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
6.4. Case studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
6.5. Categorization of literature with respect to all descriptive dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
7. Future research directions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
7.1. Future research directions from computational perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
7.1.1. Future research directions: modeling approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
7.1.2. Future research directions: solution methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
7.2. Future research directions in terms of HCF type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
7.3. General future research directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
8. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
Appendix A. (Supplementary details on HCF types) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
A.1. Non-emergency facilities. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
A.1.1. Primary care facilities (hospitals, clinics, off-site public access devices, etc.) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
A.1.2. Blood banks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
A.1.3. Specialized services facilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
A.1.4. Medical laboratories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
A.1.5. Mobile healthcare units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
A.1.6. Home healthcare centers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
A.1.7. Rehabilitation centers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
A.1.8. Doctors' offices. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
A.1.9. Drugstores . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
A.1.10. Long-term nursing care centers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
A.2. Emergency facilities. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
A.2.1. Permanent emergency facilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
A.2.2. Temporary emergency facilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260
1. Introduction keen interest from the operations research (OR) community.

Perhaps the earliest location-allocation study in the field of
Facility location decisions play a critical role in the strategic healthcare facilities (HCFs) was presented by Gould and Lienbach
design of systems for a wide range of private and public organi- [2]. In this study, the problem of locating hospitals and de-
zations (e.g., retail facilities, warehouses, airline hubs, police sta- termining their capacities was considered as a p-median problem
tions, hospitals, etc.). This is because poorly located facilities or an in the western part of Guatemala. The transportation algorithm
improper number of facilities can greatly increase capital and in- was used to solve the problem. From 2000 onwards, researchers
ventory costs and degrade customer services. The first theoretical reviewed different parts of the literature of HCF location from
study on the location of facilities began in 1909 when Alfred various perspectives (see [3–11]). The scopes of these review pa-
Weber [1] introduced a warehouse location problem to minimize pers are summarized in Table 1 and explained in more detail
the total distance between a warehouse and a set of customers. subsequently.
Thereafter, location theory and its applications were developed in The use of location-allocation models in development planning
different research areas along with a variety of models. of health services in developing nations was reviewed by Rahman
In healthcare, incorrect facility location decisions have a serious and Smith [3]. They classified studies into four categories:
impact on the community beyond simple cost and service metrics; (i) finding a set of optimal sites, (ii) locating optimal sites in a new
for instance, hard-to-access healthcare facilities are likely to be area, (iii) measuring the effectiveness of past location decisions,
associated with increased morbidity and mortality. From this and (iv) improving existing location patterns. Brotcorne et al. [4]
perspective, facility location modeling for healthcare is more cri- investigated the evolution of ambulance location and relocation
tical than similar modeling for other areas. In addition, because of models over the course of thirty years. For this purpose, they
globally pervasive trends, such as decreasing birth rates, higher classified models into deterministic and probabilistic categories,
longevity and associated growth in elderly population, and in- and summarized the literature in two corresponding tables. Fur-
creasing environmental problems (e.g., sound and air pollution), thermore, they noted the introduction of dynamic models in am-
healthcare and the associated healthcare facility (HCF) location bulance location.
problems have become noticeably more critical and important to Daskin and Dean [5] classified the location models used in
society. Due to this, HCF location modeling continues to attract healthcare literature into three categories: accessibility,
A. Ahmadi-Javid et al. / Computers & Operations Research 79 (2017) 223–263 225
Table 1 papers that consider a specific type of HCFs, not a generic type of
Review papers related to HCF location from 2000 onwards. service facility. Actually, there are papers that study typical (mo-
bile or immobile) service facilities with specific properties, which
Reference Year The scope of review
are not reviewed in this paper unless they provided a case study
[3] 2000 Location-allocation models for health service development on a healthcare location problem. For a review of such papers, the
[4] 2003 Ambulance location and relocation models reader may refer to the recent survey [12] and references therein.
[5] 2004 Location of HCFs from modeling perspective The models developed in such papers can be potentially adapted
[6] 2011 Emergency response facility location
[7] 2011 An overview of several applications of OR in healthcare
for different types of HCFs, depending on the assumptions un-
[8] 2012 Methodological advancements in healthcare accessibility derlying each model.
[9] 2013 Home healthcare logistics One should pay careful attention to the point that HCFs are
[10] 2013 An overview on planning and management of EMSs widely considered in many different interrelated research fields.
[11] 2015 An introduction and a short review of three types of HCFs
These fields along with related survey papers are listed as follows:
healthcare operations management ([13,14]), healthcare services
supply chains ([15,16]), services supply chains ([17]), pharmaceu-
adaptability, and availability models. In their view, accessibility tical supply chains ([18–20]), healthcare waste management ([21]),
models are extensions of location models whose goals were pre- disaster operations management ([22–24]), emergency logistics
dominantly to maximize coverage or to minimize average dis- ([25]), relief distribution ([26]), humanitarian logistics ([25,27–
tance. Adaptability models attempt to find solutions that perform 29]), homeland security ([30,31]), emergency response ([5,32,33]),
properly across a range of possible scenarios and conditions. emergency services stations ([34]), emergency services vehicles
Availability models are divided into deterministic, queuing-based ([35]), and supply chain with disruptions ([36]). This indicates that
and probabilistic models. These models address very short-term HCFs have various types and widespread usages in different fields,
changes that result from facilities being busy. which made our survey process more challenging.
The literature of covering models and optimization techniques The remainder of the paper is organized as follows: Section 2
for locating emergency response facilities was studied by Li et al. presents an overview of discrete location problems. Section 3
[6]. Rais and Vianna [7] briefly surveyed several applications of provides a framework for the classification of HCFs from a location
operations research in healthcare planning (e.g., demand fore- analysis perspective and describes the structure of this review
casting, location selection, and capacity planning), healthcare paper. Section 4 and Section 5 are devoted to review and scrutiny
of non-emergency and emergency HCF location papers, respec-
management and logistics (e.g., resource and staff scheduling), and
tively, based on the framework proposed in Section 3. Section 6
other applications (e.g., disease diagnosis and treatment planning).
analyzes the literature from different perspectives. Section 7 pre-
Wang [8] presented a literature review regarding three issues re-
sents directions for possible future research. Section 8 concludes
lated to inequality in healthcare accessibility: measurement, op-
the review. Appendix A provides definitions and details on dif-
timization, and impact, with emphasis on methodological ad-
ferent types of HCFs.
vancements and implications for public policy.
Gutiérrez and Vidal [9] reviewed the literature of home
healthcare logistics in terms of a three-dimension framework. In
2. An overview of discrete location problems
the first dimension, home healthcare planning levels were dis-
tinguished according to the time horizon, namely strategic, tac-
Facility location theory refers to the modeling, formulation, and
tical, and operational levels. In the second dimension, logistics
solution methods of a class of problems that deal with locating
management decisions were divided into four groups: network
facilities in some given space. Since facility location is a critical
design, transportation management, staff management, and in-
subject at the strategic planning level, location theory and its ap-
ventory management. In the third dimension, service processes
plications have received increasing attention from the OR com-
were defined as the set of steps performed in the delivery of home munity. The study of facility location models has its roots in
healthcare services. These service processes include medical pre- the pioneering work of Weber in 1909. Thereafter, numerous pa-
scription, patient admission, appointment scheduling, visiting pers and books have dealt with facility location problems (see, e.g.,
patients, and medical discharge. Ingolfsson [10] briefly reviewed [37–39] and references therein).
research on the planning and management of emergency medical Over the years, the broad spectrum of location problems has
services (EMSs) with emphasis on four topics: performance mea- been divided in several ways. For instance, Revelle et al. [40]
surement; location of ambulance stations; allocation of ambu- proposed a taxonomy for location problems based on the space in
lances to stations; and forecasting of demand, response times, and which the problems are modeled. They divided these location
workload. Recently, Gunes and Nickel [11] provided an overview of problems into four broad basic classes: analytic, continuous, net-
facility location problems in health systems with a focus on three work, and discrete. Among these classes, discrete location pro-
main areas in the healthcare context: public facility location, am- blems have been used in numerous practical contexts including, in
bulance planning, and hospital layout. particular, health systems.
By considering the scope of the recent review papers (given in Facility location problems can be extended by specializing them
Table 1), one finds that each paper covers a part of the healthcare in various ways; for example, stochastic location problems, hier-
services. Thus, the field of OR continues to lack a comprehensive archical location problems, multi-criteria location problems, hub
review of facility location in healthcare. In this regard, we decided to location problems, dynamic and online location problems, com-
provide a thorough classification of HCF location models and survey petitive location problems, etc. (see, the recent book [41] for a
the literature on HCF location in the last decade. The review con- review of different types of location problems). In addition, the
siders 18 types of facilities in three main categories (see Section 3.1) integration of location decisions with other important logistical
along 10 descriptive dimensions (see Section 3.2). For this purpose, decisions or other related decisions became an increasingly im-
we identified approximately 150 articles that have been published portant topic in the literature. Instances of this integration include
since 2004. Note that almost all earlier papers on HCF location location-inventory problems (see, e.g., [42,43]), location-routing
published before 2004 have been reviewed in surveys [3–11]. problems (see, e.g., [44,45]), location-routing-inventory problems
Furthermore, the scope of this paper is to review only those (see, e.g., [46,47]), location-pricing problems (see, e.g., [48–50]),
and location-inventory-pricing problems (see, e.g., [51,52]). problems find the number and location of facilities such that all
In general, location problems may be either continuous (in demand points are within a specified travel distance (or time) of
which facilities may be located anywhere in the feasible region) or the facilities that serve them. In formulating a basic set covering
discrete (in which they can be established only at candidate lo- location problem, the following notation is used:
cations that can include the demand points) [38]. We focus on
discrete location problems since they comprise one of the best Sets:
known categories of location problems and frequently arise in
healthcare settings. Daskin [53] classified discrete location pro-
I The set of demand points.
blems into three broad categories: covering-based problems,
J The set of candidate locations.
median-based problems, and other problems. This classification,
with slight changes, is shown in Fig. 1. Ni The set of all candidate locations which can cover demand
Discrete location problems comprise an important set of ap- point i∈I , Ni={ j∈J: dij≤Di }.
plications of location modeling and theory. Thus, it is not sur-
prising that many extensions to basic discrete problems have been
proposed and studied in the literature of health system problems. Input parameters:
Given our focus on healthcare facility location, we will consider
and discuss many of these extensions in the rest of this paper.
dij The travel distance (or time) from demand point i∈I to
On the one hand, covering-based and median-based problems
candidate location j∈J .
are well-known facility location problems for modeling real-world
situations. On the other hand, important problems, such as p- f j The fixed cost of locating at candidate location j∈J .
center and p-median location problems on a general network or in Di The maximum acceptable travel distance or time from de-
the plane are NP-hard for p>1 [53–55]. These are problems with mand point i∈I (the cover distance or time).
no known polynomial-time exact solution algorithms. It means
that the time required to exactly solve an instance of these pro-
blems may increase very rapidly as the size of the problem in- Decision variables:
stance grows, often well beyond any reasonable time frame. As a
consequence, a variety of algorithms have been developed in the xj 1, if a facility is established (located or opened) at candidate
literature to solve these problems both optimally and heuristically. location j ∈ J ; 0 otherwise.
In the rest of this section, we study the three broad categories
given in Fig. 1 and mathematical programming models of basic
Formulation:
discrete location problems in these categories. The analysis of HFC
literature based on different types of discrete location problems min ∑ f j xj
will be presented in Section 6.1. j∈J (1)
subject to
2.1. Covering-based problems
∑ xj ≥1, i∈I
Covering-based problems assume that demand locations need j ∈ Ni (2)
to be within a specific coverage distance (or time) from facilities
which service them, in order to be covered by the service, or sa- xj ∈{ 0, 1}, j∈J . (3)
tisfactorily served. This class of problems includes three basic
types: set covering problems, maximal covering problems, and p- In this model, the objective function (1) minimizes the location
center problems. In particular, we note that covering-based pro- cost of the facilities which are needed to cover all demand points.
blems are typically appropriate for determining the location of Constraints (2) ensure that each demand point must be covered
emergency service facilities. and Constraints (3) are integrality constraints.
2.1.1. Set covering location problems 2.1.2. Maximal covering location problems
In a set covering location problem, the goal is minimizing the Maximal covering location problems (MCLPs) determine the
number of established facilities or the total location cost, given a location of p facilities in order to maximize the demand covered
specified level of demand coverage which must be achieved. These within a pre-specified maximum coverage distance. These
Fig. 1. A classification of discrete location problems.

problems differentiate between points with large and small de- Sets:
mand, by taking into account the level of demand at each point. A
basic MCLP can be formulated with the following notation: I The set of demand points.
Sets:
Ni The set of all candidate locations which can cover demand
point i∈I , Ni={ j∈J: dij≤Di }.
Ni The set of all candidate locations which can cover demand Input parameters:
point i∈I , Ni={ j∈J: dij≤Di }.
Input parameters:
wi The demand at point i∈I .
Di The maximum acceptable travel distance or time from de-
dij The travel distance (or time) from demand point i∈I to mand point i∈I (the cover distance or time).
candidate location j∈J . p The number of candidate locations to be established.
Di The maximum acceptable travel distance or time from de-
mand point i∈I (the cover distance or time). Decision variables:
p The number of candidate locations to be established.
xj 1, if a facility is established at candidate location j∈J ;
0 otherwise.
Decision variables:
yij 1, if demand point i is assigned to a facility at candidate
location j∈Ni ; 0 otherwise.
0 otherwise.
zi 1, if demand point i∈I is covered; 0 otherwise. Formulation:
min L (9)
Formulation: subject to
max ∑ wi zi ∑ yij =1, i∈I
i∈I (4) j ∈ Ni (10)
subject to
∑ xj =p
∑ xj =p j∈J (11)
j∈J (5)
∑ dij yij ≤L, i∈I
z i≤ ∑ xj , i∈I j ∈ Ni (12)
j ∈ Ni (6)
yij ≤xj , i∈I , j∈Ni (13)
zi∈{ 0, 1}, i∈I (7)
yij ∈{ 0, 1}, i∈I , j∈Ni (14)
xj ∈{ 0, 1}, j∈J . (8)
xj ∈{ 0, 1}, j∈J (15)
The objective (4) maximizes the total covered demand. Con-
straint (5) states that p facilities are to be located. Constraints (6)
require that demand points are only covered by open facilities. L ≥ 0. (16)
Constraints (7) and (8) are integrality constraints. The objective (9) minimizes the maximum demand-weighted dis-
tance (or time) between a demand point and the (nearest) facility al-
2.1.3. p-center location problems located to it. Constraints (10) guarantee that each demand point is
p-center location problems (PCLPs) are the third classical type covered by only one facility. Constraint (11) specifies the total number
of covering-based problems, which minimize the maximum travel of facilities to be established. Constraints (12) determine the maximum
distance (or time) among all demand points and the allocated demand-weighted distance (or time). Note that L is an auxiliary vari-
facilities, considering that every demand point is covered. When able (not a decision variable) that is used to compute the maximum
the facilities are uncapacitated, the demand points are assigned to distance. Constraints (13) show that demand points are only covered
the closet open facilities. These covering-based location problems by open facilities. Finally, Constraints (14)-(16) are domain constraints.
are a type of minmax problem and may be also referred to as lo-
cation-allocation problems since they require simultaneous facility 2.2. Median-based problems
location and allocation of the demand points to the open facilities.
To formulate the canonical form of a basic PCLP, the following Median-based problems locate facilities at candidate points so
notation is used: as to minimize the weighted average distance costs between de-
mand points and the facilities to which they are assigned. These
locations are the medians of the network. This class of problems attempt to minimize the total cost of facility opening and travel-
may be referred to as location-allocation problems as they de- ing. The notation and formulation of a basic uncapacitated FCLP
termine both location and allocation decisions. p-median and the are described below:
fixed charge location problems are important problems in this
class. Sets:
2.2.1. p-median location problems I The set of demand points.

p-median location problems (PMLPs) are among the most J The set of candidate locations.
popular problems in facility location. These problems aim to locate
p facilities in a network. The sets, parameters, and decision vari-
ables used in the formulation of a basic PMLP are as follows: Input parameters:
Sets:
f j The fixed charge of establishing a facility at candidate lo-
cation j .
v The transportation cost per item per distance unit (the
Input parameters:
variable transportation cost).
candidate location j∈J . Decision variables:
p The number of candidate locations to be established. xj 1, if a facility is established at candidate location j∈J ;
0 otherwise.
yij 1, if demand point i∈I is assigned to a facility at candidate
Decision variables:
location j∈J; 0 otherwise.

0 otherwise. Formulation:
yij 1, if demand point i∈I is assigned to a facility at candidate
location j∈J ; 0 otherwise.
min ∑ f j x j +v ∑ ∑ wi dij yij
j∈J i∈I j∈J (23)
subject to
Formulation:
∑ yij =1, i∈I
min ∑∑ wi dij yij j∈J (24)
i∈I j∈J (17)
yij ≤xj , i∈I , j∈J (25)
subject to
∑ yij =1, i∈I yij ∈{ 0, 1}, i∈I , j∈J (26)

j∈J (18)
xj ∈{ 0, 1}, j∈J . (27)
∑ xj =p
j∈J (19) The objective (23) minimizes the total cost which includes the
facility-opening and transportation costs. Constraints (24) ensure
yij ≤xj , i∈I , j∈J (20) that each demand node is assigned to an open facility, while
Constraints (25) restrict assignments to open facilities. Constraints
(26) and (27) are integrality constraints.
yij ∈{ 0, 1}, i∈I , j∈J (21) In a capacitated FCLP, a new parameter Uj is defined as the
maximum capacity of each facility j . The formulation of a basic
xj ∈{ 0, 1}, j∈J . (22) capacitated FCLP is similar to the one given above with the fol-
lowing additional capacity constraints:
In this model, the objective (17) minimizes the total demand-
weighted travel distance (or time). Constraints (18) show that each ∑ wi yij ≤Uj, j∈J .
demand point is assigned to only one facility. Constraint (19) i∈I
specifies the total number of facilities to be established. Con-

straints (20) limit assignments to open facilities. Constraints (21) 2.3. Other problems
and (22) are integrality constraints.
Finally, the problems which are not in either of the above two
2.2.2. Fixed charge facility location problems broad categories (covering-based and median-based problems)
Fixed charge facility location problems (FCLPs) are closely re- belong to the last category. For example, in the p-dispersion pro-
lated to PMLPs. While PMLPs disregard the differences in the fa- blem, p facilities are located on a network in order to maximize the
cility establishing costs at different candidate locations, FCLPs minimum separation distance between any pair of open facilities,
which can be applied to facilities with a threat to each other, or to deterministic stockpile location model which includes disaster
systems of retail or service franchises [56]. The maxisum disper- specific casualty characteristics, such as the severity and type of
sion problem is a similar problem, which maximizes the average medical condition, and the unique nature of each type of disaster).
separation distance between open facilities [56]. Another example The main reason perhaps is that these areas are covered by the
is the maximum-number-of-sustainable-facilities (MNSF) location extensive literature of supply chain location management (see
problem that maximizes the number of bases for healthcare [60,61] and references therein). Despite the fact that producers are
workers (HCW), which can work sustainably in an area. An open keenly interested in improving products and ensuring quality and
HCW basis works sustainably if a sufficient number of self-help safety, they often have insufficient expertize in logistics and dis-
groups (SHGs), within a certain travel distance, are allocated to it, tribution. Moreover, reducing those expenses is not a priority for
where SHGs are located on a given network [57]. One may add producers or distributors because in most cases the transportation
other problems to this category. cost is ultimately passed on to the healthcare providers. Therefore,
logistics and distribution are promising future areas for cost-cut-
ting in health systems [62].
3. Scope of literature survey Medical waste comprises a wide range of waste materials
generated by the broad variety of facilities related to health sys-
This section includes two subsections. The first presents the tems, including infectious and non-infectious waste; anatomical
framework that is used to classify the different types of HCFs, and and pathological waste; pharmaceutical waste; genotoxic waste;
the second introduces the descriptive dimensions considered to chemical waste; heavy-metal waste; hazardous and non-ha-
analyze each research paper. zardous materials, and used medical devices. Nowadays, the tre-
mendous rise in the amount of medical waste poses grave chal-
3.1. A framework for classification of HCFs lenges to all of the facilities related to health systems [21]. Com-
ponents of an HCWM system are responsible for collection, dis-
In most societies, the healthcare industry has grown over time, tribution, recycling, and disposal of medical waste (see [21,63] for
leading to high levels of aggregation and integration of various further study of HCWM). In this regard, determining the best lo-
sectors in order to provide healthcare products and services effi- cation and optimal number of the related facilities is a pivotal
ciently and effectively to the society members. The five major strategic decision in health systems in order to avoid the trans-
players of the healthcare industry are producers, distributors, mission of infections as well as toxic effects, injuries, and pollution
providers, waste management actors, and fiscal intermediaries of the environment. Given the importance of strategic planning in
(Fig. 2). Producers include all health suppliers, such as pharma- HCWM, a corresponding investigation of this area from an op-
ceutical companies, medical-surgical product companies, and erations research perspective has not been seen in the literature.
medical equipment manufacturers. Distributors who comprise Nevertheless, we believe that proper location of the facilities for
links between producers and providers include different types of collection, distribution, recycling, and disposal of medical waste
wholesalers and distributors in the field of health products. The will receive considerable attention from government and society
main body of a healthcare system is made up of providers that are in the future, and that the OR community can significantly con-
responsible for provision of a variety of healthcare services to the tribute to this emerging important need (see [64] for a review of
people of a country or an area. The last, but not least, components OR models for solid waste management).
in a health supply chain are healthcare waste management Healthcare providers, by virtue of forming the main body of
(HCWM) facilities which are responsible for the collection, dis- health systems, are the key players of these systems. These facil-
tribution, recycling, and disposal of medical waste and used ities provide a variety of healthcare services to the people of so-
medical equipment. Finally, the health systems also involve the ciety. Accordingly, developing location models for healthcare
participation of fiscal intermediaries, such as insurance agencies providers has received attention from the OR community in gen-
and health maintenance organizations. The objective of this paper eral and specialists in location management in particular. The rest
is to present a survey of the location literature that applies to of this paper is dedicated to the literature on location of healthcare
health systems from an OR perspective. providers. Hence, healthcare facilities (HCFs) refer to healthcare
Our review of the literature indicated that on the one hand provider facilities in the remainder of this paper.
supply chains for health services have been addressed by many As was mentioned, there are a wide range of services related to
researchers (see survey papers [15,16]); on the other hand, de- the improvement of human health. Nowadays, many types of HCFs
termination of the optimal location of medical producers, dis- are required to perform these services with each providing dif-
tributors, and fiscal intermediaries has not received much atten- ferent features and applications. High or optimal performance of a
tion from the OR community even though that is crucially im- health system depends on determining the optimal number of
portant in the timely delivery of medical products and services. facilities, their optimal locations, and a system of communication
Actually, we could find only two papers studying the location of between these HCFs. Considering the wide range of healthcare and
medical distributors in disaster situations [58,59] (Mete and Za- medical activities, a framework must be developed based on a
binsky [58] modeled the location of stockpiles, i.e., medical sup- systematic way in order to classify all types of HCFs based on their
plies warehouses, in disaster situations using a two-stage sto- locational properties. Using such a framework, the healthcare lo-
chastic program; and Paul and Hariharan [59] developed a cation literature can be comprehensively reviewed, and gaps and
Waste management
Producers Distributors Providers
actors
Fiscal intermediaries
Fig. 2. The major players in health systems.

deficiencies can be identified. papers using these methods to visualize or validate input and/or
To develop the framework used in this paper, all activities as- output data are considered in this paper (e.g., [174]). Furthermore,
sociated with human health were first considered based on the HCF layout problems, as a special class of HCF planning problems,
last edition of the international standard industrial classification are not considered in this review. An overview of these problems
(ISIC), introduced in 2008. Then, all types of HCFs performing can be found in Section 21.4 of [11].
these activities were collected through two sources: Iran's minis- The following sections elaborate on optimization methods used
try of health and medical education and the north American in- in each HCF category of Fig. 3. In order to analyze the literature, we
dustry classification system (NAICS). In Fig. 3, these HCFs are propose a review that is structured using ten main descriptive di-
classified with regard to the location management purposes. This mensions. These are consideration of uncertainty, multi-period
framework is quite comprehensive in the scene that it considers all setting, particular input/setting, objective function, decision vari-
types of known HCFs for all types of known heath activities. This able, constraint, basic location problem, mathematical modeling
will also serve to make researchers aware that the “story” is sig- approach, solution method, and case study inclusion. These di-
nificantly different for each type of HCF found in the literature. mensions, which are described in Tables 3 and 4, investigate the
As shown in Fig. 3, the HCFs are classified into two main ca- papers from two different perspectives: location theory and com-
tegories: (1) non-emergency facilities and (2) emergency facilities. putation, respectively. The last three columns in Table 3 indicate the
Emergency facilities are also divided into permanent and tem- general HCF types for which each row-item is applicable.
porary groups. In addition, each category includes several sub- Note that a decision problem can be modeled using various
categories which are described in detail in Appendix A. approaches, and each resulting model can be solved by different
Almost all of the published literature over the last decade is solution methods (see Table 4). Moreover, the classification re-
classified in Table 2 according to the proposed framework. Table 2 garding solution methods in Table 4 could be presented in more
aids in both the concise representation of non-emergency and detail, but this level of detail is sufficient for our survey. Some of
emergency HCFs considered in the literature in an accessible the subclasses may have overlaps and interconnections which are
fashion, and in the identification of gaps in the literature. not detailed here. It should also be noted that solution methods
It is also possible to draw some conclusions based on Table 2. are here divided into two main classes: A and B. A method in Class
Almost 50% of the surveyed papers refer to non-emergency HCF A finds, in a given time, either an optimal (exact) solution or a
location. Among these, about half of the papers (52%) address the perturbed (or near-optimal) solution with a known deterministic
location of primary care facilities (hospitals, clinics, etc.). Emer- error bound on the (relative or absolute) optimality gap of the
gency HCF location makes up almost 50% of the papers. Further, resulting solution. However, a method in Class B does not provide
about 16% of these examined the HCF location in disaster situa- an error bound and basically cannot determine the quality of its
tions (temporary emergency facilities). Fig. 4 shows the share of resulting solution. Borrowing from terminology used in the lit-
each HCF type in the literature of location management. erature of metrology, methods in Classes A and B may be simply
referred to as accurate and inaccurate, respectively [181].
3.2. Descriptive dimensions Any non-exact method is sometimes called heuristic. This de-
finition is much broader than the definition of heuristic methods
Various methods are used in the search for solutions of HCF in this paper. Based on our definition, heuristic methods are non-
location problems; however, optimization methods are widely exact methods in Class B which do not fall in the other two sub-
used. Other methods include GIS (e.g., see [178]) and simulation classes: metaheuristic and approximate stochastic optimization.
(e.g., see [179,180]) are sparsely considered in the literature, and One should also note that polynomial time algorithms with
therefore not discussed in this review paper. However, those bounded relative errors are referred to as approximation
Fig. 3. The proposed framework for the classification of healthcare provider facilities.
Table 2
The breakdown of the HCF location literature into HCF types given in the framework (Fig. 3).
Non-emergency facilities Primary care facilities (hospitals, clinics, off-site public access devices, etc.) [65–91]
Blood banks [92–97]
Organ transplant centers [98–101]
Detection and prevention centers [102–110]
Other specialized services facilities [111–114]
Medical laboratories [115]
Mobile healthcare units [116]
Long-term nursing care centers [117,118]
Emergency facilities Permanent emergency facilities Emergency off-site public access devices [119–124]
Emergency centers [125]
Trauma centers [126–131]
Ambulance stations [132–165]
Temporary emergency facilities Temporary medical centers [57,166–170]
Points of dispensing [171–174]
Combination of several types of HCFs Combination of 1.1 and 1.3.3* [175,176]

Combination of 1.1 and 1.6 * [177]
*
See Fig. 3 for HCF type codes.
Fig. 4. The frequencies of papers on HCF types in the HCF location literature (see Fig. 3 for HCF type codes).
algorithms in the combinatorial optimization literature. These are effective treatment. Primary care is usually performed by a general
rarely used in the location literature because most bounded-error practitioner and has great potential for referral to specialty ser-
methods used in the location literature, such as branch-and-bound vices, which are secondary, tertiary, or quaternary care; or
and Lagrangian relaxation, do not result in polynomial-time so- non-medical services ([182]). Primary care, which is also referred
lution algorithms. Moreover, under specific conditions, it is pos- to as primary medical care ([183]), differs from a broader concept
sible to present some convergence analysis or probabilistic error of primary health care that includes primary care services, health
analysis for methods in Class B (which are randomized in nature). promotion and disease prevention, and population-level public
However, these are not elaborated here for the sake of brevity. health functions ([184,185]).
Ease of access for all sections of society is an important goal for
primary care facilities (PCFs). Almost all of these facilities are open
4. Non-emergency HCF location 24 hours a day and clients refer to the nearest one. Attention must
be paid to PCFs because the optimal location of these facilities is
Following the classification of HCFs illustrated in Fig. 3, content currently, as in the past, the subject of ongoing debate among
within the non-emergency literature can be classified into several researchers, as reflected in the fact that about half of this section’s
categories. In the following, the reviewed papers in each category papers are dated from 2012 and onward.
are analyzed according to Tables 3 and 4. It should be mentioned that hospitals and most clinics provide,
beside primary care, specialty services, which are secondary, ter-
4.1. Primary care facilities (hospitals, clinics, off-site public access tiary, or quaternary care. Therefore, if the main task of HCFs is the
devices, etc.) provision of primary care to the public, they will be considered as
PCFs which are always available for all sections of society. How-
All the facilities in this class of HCFs provide primary care, i.e., ever, if the main task of a hospital, such as specialty, super-spe-
first-contact care, which is early diagnosis, and timely and cialty, or multi-specialty hospitals, is to provide specialty care
Table 3
Survey descriptive dimensions from location-theory perspective: consideration of uncertainty, multi-period setting, particular input/setting, objective function, decision
variable, constraint, and basic location problem.
Survey dimension Code Descriptive dimension Non-emergency Permanent emer- Temporary emer-
facilities gency facilities gency facilities
Consideration of Y Considering uncertainties ✓ ✓ ✓

uncertainty N Not considering uncertainties ✓ ✓ ✓
Multi-period setting S Static ✓ ✓ ✓
D-1 Multi-period short-term decisions (e.g., am- ✓ ✓ ✓

bulance deployment or shift resource
Dynamic allocation)
D-2 Multi-period long-term decisions (e.g., ✓ ✓ ✓
location)
Particular input/ P1 Demand ✓ ✓ ✓

setting P2 Travel time ✓ ✓ ✓
P3 Travel distance ✓ ✓ ✓
P4 Facility capacity ✓ ✓ ✓
P5-1 Fixed cost ✓ ✓ ✓

P5-2 Cost Variable cost ✓ ✓ ✓
P5-3 Penalty for lost demand ✓ ✓ ✓
P6 Waiting time ✓ ✓ ✓
P7 Multiple servers: Considering several servers at each facility ✓ ✓ ✓
P8 Multiple services/Multi-type demand ✓ ✓ ✓
P9 Elastic demand: Demand depends on distance, waiting time, etc. ✓ ✓
P10 Busy fraction: Probability of an ambulance being busy ✓
P11 Hierarchical system ✓ ✓ ✓
P12 Other items, e.g., number of periods and different coefficients ✓ ✓ ✓
Objective function O1 Minimize total number of facilities ✓ ✓ ✓

O2 Minimize total number of ambulances ✓
O3 Minimize total travel distance (or time) ✓ ✓ ✓
O4 Minimize sum of costs ✓ ✓ ✓
O5 Minimize maximum travel distance (or time) ✓ ✓ ✓
O6 Maximize participation ✓
O7 Maximize demand coverage ✓ ✓ ✓
O8 Maximize multiple coverage ✓
O9 Minimize number of uncovered demand ✓ ✓
O10 Other objectives, e.g., maximize number of voluntary facilities, mini- ✓ ✓ ✓
mize number of ambulance relocations, and minimize maximum
transfer time between stations
Decision variable D1 Location of facilities ✓ ✓ ✓

D2 Allocation of resources ✓ ✓
D3 Deployment (location or relocation) of ambulances in stations ✓
D4 Allocation of HCFs to demand points ✓ ✓ ✓
D5-1 Once ✓ ✓ ✓
Demand coverage
D5-2 More than once ✓ ✓
D6 Dispatch (assignment) of ambulances to demand points ✓
D7 Number of required resources ✓ ✓
D8 Other items, e.g., demand flow and number of required facilities ✓ ✓ ✓
Constraint C1 Full coverage ✓ ✓ ✓

C2 Partial coverage ✓
C3 Multiple coverage ✓
C4 Maximum number of required facilities ✓ ✓ ✓
C5 Maximum travel distance (or time) ✓ ✓ ✓
C6 Ambulance reliability/service level (probabilistic coverage) ✓ ✓ ✓
C7 Maximum number of ambulances at each station ✓
C8 Maximum available resources ✓ ✓ ✓
C9-1 Maximum capacity for ✓ ✓ ✓

demand response
Service capacity
C9-2 Minimum capacity for de- ✓
mand response
C10 Budget ✓ ✓ ✓
C11 Other items: e.g., no-vacant and flow constraints ✓ ✓ ✓
Table 3 (continued )
Survey dimension Code Descriptive dimension Non-emergency Permanent emer- Temporary emer-
facilities gency facilities gency facilities
Basic location SCL Set covering location problem ✓ ✓ ✓

problem MCL Maximal covering location problem ✓ ✓ ✓
PCL p-center location problem ✓ ✓ ✓
PML p-median location problem ✓ ✓ ✓
FCL Fixed charge facility location problem ✓ ✓ ✓
O Other items, e.g., p-dispersion, maxisum dispersion, and MNSF location ✓ ✓ ✓
problems
rather than primary care, it should not be classified as a pure PCF. Formulation:
For such facilities one may use multi-objective optimization
models to compromise between different location criteria (e.g., see min ∑ ∑ dik uik + ∑ ∑ dkj vkj
i∈I k∈K k∈K j∈J (28)
[75,84]). Finally, it is worth noting that off-site public access de-
vices, which are a new type of HCF (see Section 5.1.1 and Appendix subject to
A), can be possibly used as PCFs in the near future.
PCFs can be classified into different levels. For such facilities, ∑ uik =wi, i∈I
k∈K (29)
hierarchical extensions of PMLPs can be useful (see [186] for a re-
view of hierarchical location problems). In this regard, we formulate
a p-median single-flow hierarchical problem for locating a set of ∑ vkj=θ k ∑ uik , k∈K
PCFs with two levels, in which the total travel distance (or time) for j∈J i∈I (30)
patients is minimized. The sets, parameters and decision variables
used in the formulation of this problem are as follows: ∑ uik ≤Ck1xk1, k∈K
i∈I (31)
Sets:
∑ vkj≤C2j x2j , j∈J
k∈K (32)
K The set of candidate locations for a level-1 PCF (e.g., clinics).
∑ xk1=p
J The set of candidate locations for a level-2 PCF (e.g., k∈K (33)
hospitals).
∑ x2j =q
j∈J (34)
Input parameters:
uik ≥0, i∈I , k∈K (35)
dik The travel distance (or time) between demand point i∈I
and a level-1 PCF in candidate location k∈K . vkj≥0, k∈K , j∈J (36)
dkj The travel distance (or time) between a level-1 PCF in
candidate location k∈K and a level-2 PCF in candidate
xk1∈{ 0, 1}, k∈K (37)
location j∈J .
wi The population size at demand point i∈I .
x2j ∈{ 0, 1}, j∈J . (38)
C1k The capacity of a level-1 PCF in candidate location k∈K .
C2j The capacity of a level-2 PCF in candidate location j∈J . In this model, the objective (28) minimizes the total demand-
p The number of alevel-1 PCFs to be established. weighted travel distance (or time). Constraints (29) show that the
q The number of level-2 PCFs to be established. entire population of patients at each demand point must be assigned
θk The proportion of patients in a level-1 PCF at candidate to level-1 PCFs. Constraints (30) stipulate that θk proportion of patients
location k∈K referred to a level-2 PCF. in a level-1 PCF are referred to open level-2 PCFs. Constraints (31) and
(32) control the capacities of open level-1 and level-2 PCFs. Constraints
Decision variables: (33) and (34) specify the total number of level-1 and level-2 PCFs to be
established. Constraints (35)–(38) are domain constraints.
Table 5 provides a breakdown of the optimization studies on
xk1 1, if a level-1 PCF is established at candidate location k∈K ;
the location of PCFs. It should be noted that Burkey et al. [65] and
0 otherwise.
Fo and Mota [66] also addressed this class of HCFs, but their work
x 2j 1, if a level-2 PCF is established at candidate location j∈J ; has not been summarized in Table 5 since they limited themselves
0 otherwise.
to comparing the performance of different existing discrete loca-
uik The flow of patients between demand point i∈I and a level-
tion models in health systems.
1 PCF at candidate location k∈K . It can be seen from "Consideration of uncertainty" column of
vkj The flow of patients referred form a level-1 PCF at candi- Table 5 that almost all of the papers deal with deterministic models.
date location k∈K to a level-2 PCF at candidate location j∈J . The exceptions are the papers [69,70,83,84,88]. Oliveira and Bevan
[70] developed a utilization-based model which used behavioral
Table 4
The survey descriptive dimensions from computational perspective: modeling approach, solution method, and case study inclusion.
information generated by gravity models in order to improve geo- refers to the large number of static (i.e., single-period) models
graphic equity. Mitropoulos et al. [79] presented a probabilistic ex- when compared with dynamic (i.e., multi-period) models (ap-
tension of the p-median model which combines data envelopment proximately 83% against 17%). A short-term dynamic model for
analysis (DEA) and location analysis. The uncertainty in the model seasonally moving populations was introduced in [71], and long-
was associated with the number of the treatment population which term dynamic location-allocation models were considered in
they estimated in exponential form using SPSS. In [83], a scenario- [81,84,85].
based location-allocation model was presented, which aimed to The papers on location of PCFs emphasized the need for pri-
balance hospitals’ usage, minimize congestion at the hospitals, and mary care of various sectors of society. While different sectors of
increase accessibility to the hospitals. Mestre et al. [84] developed society are multi-type demand and require different services, a
two location models to address the planning of hospital networks few papers consider multiple services (see [72,75]) and multi-type
over a planning horizon under uncertainty. They did this using a set demand (see [80,89]). In addition, health systems are generally
of discrete scenarios. Shishebori and Babadi [88] proposed a robust hierarchical in nature leading to several types of services which
and reliable model which simultaneously takes uncertainty in de- may differ in cost and complexity. In this context, facilities tend to
mand and transfer cost, and system disruptions into account. specialize in the sophistication of services they provide – consider,
Another important conclusion that can be drawn from Table 5 for example, services provided by hospitals compared to those
Table 5
Non-emergency healthcare facilities: primary care facilities (hospitals, clinics, off-site public access devices, etc.).
Reference Consideration Multi-period Particular input/setting Objective Decision Constraint Basic location Modeling Solution Case
(year) of uncertainty setting function variable model approach method study
[67] (2000) N S P1, P12 O10 D1, D2 C1, C4, C9-1, C9-2, C11 SCL MILP, GP, MCDM SL Y
[90] (2001) N S P1, P3, P5-1, P5-2 O7 D1, D4 C4, C5, C10, C11 MCL ILP SC, H Y
[68] (2002) N S P1, P3, P8, P11, P12 O3 D1, D8 C1, C4, C9-1, C10 PML MILP SC, LR Y
[89] (2004) Y S P1, P3, P5-1, P5-2, P8, P12 O7 D1, D4 C4, C5, C10, C11 MCL ILP, O(QT) SC N
[85] (2004) N D-2 P3, P4, P5-2, P7, P12 O3, O4, O9, O10 D7, D8 C9-1 FCL MILP, MCDM MH-TS Y
[69] (2006) N S P1, P3, P4 O3, O5 D1, D4 C1, C4, C5, C8, C9-1 PML MILP, MCDM SX Y
[70] (2006) Y S P1, P3, P4 O10 D4, D8 C4, C9-1, C9-2, C11 O PSP, MINLP SO Y
[71] (2008) N D-1 P1, P5-1, P5-2 O4 D1, D4, D8 C1, C4, C9-2 FCL ILP SL Y
[72] (2008) N S P1, P4, P5-1, P5-2, P8 O7 D1, D4, D8 C5, C9-1, C10, C11 MCL MILP – Y
[73] (2009) N S P1, P3, P5-1, P5-2, P11 O9 D1, D4, D8 C4, C5, C10 MCL MILP SO Y
[74] (2009) N S P1, P3, P11, P12 O10 D1, D4 C4, C5, C11 O ILP SX Y
A. Ahmadi-Javid et al. / Computers & Operations Research 79 (2017) 223–263

[75] (2012) N S P1, P2, P4, P8, P11, P12 O3 D1, D4, D7, D8 C1, C5, C9-1, C9-2, C11 PML MILP SG Y
[76] (2012) N S P1, P3, P4 O7 D1, D4 C1, C4, C5, C9-1 MCL ILP SC, MH-GA Y
[77] (2012) N S P1, P3, P4, P7, P9 O3/O6/07 D1, D4, D7 C1, C4, C8, C9-1, C9-2, MCL ILP – Y
C11
[78] (2012) N S P1, P3, P5-1, P5-2 O3 D1, D4, D8 C1, C10, C11 FCL MILP SC Y
[86] (2012) N S P1, P3, P8, P11, P12 O3, O10 D1, D4 C1, C4, C5, C11 PML ILP, MCDM SX Y
N S P1, P3, P8, P11, P12 O7, O10 D1, D4, D5-1 C4, C5, C11 MCL MILP, GP, MCDM SX Y
[79] (2013) Y S P1, P3, P8, P12 O3, O10 D1, D4, D8 C1, C5, C9-2 PML MILP, PSP, MCDM SX Y
[80] (2013) N S P1, P3, P5-1, P8, P12 O7 D1, D4 C5, C9-1, C10, C11 MCL ILP SC, LR Y
[81] (2013) N D-2 P1, P3, P5-1, P5-2, P12 O4 D1, D4, D8 C1, C10, C11 FCL MINLP SC, MH-SA, H Y
[82] (2014) N S P1, P3, P5-1, P12 O3/04/O10 D1, D4, D8 C1, C4, C9-1, C9-2 PML MILP, MCDM MH-GA Y
[83] (2014) Y S P1, P2, P4, P5-1, P5-2 O4, O10 D1, D4 C1, C4 FCL INLP, 2-SSP, SG, MH-O Y
MCDM
[87] (2014) N S P1, P3, P5-1, P5-2, P8, P12 O3, O4, O10 D1, D4 C1, C5, C9-2 FCL MILP SC Y
[84] (2015) Y D-2 P1, P2, P5-1, P5-2, P5-3, P8, P11, O3, O4, O9 D1, D7, D8 C4, C5, C9-1, C9-2 FCL, MCL MILP, 2-SSP, SG Y
P12 MCDM
[88] (2015) Y S P1, P3, P4, P5-1, P5-2, P12 O4 D1, D4, D8 C4, C9-1, C10, C11 FCL MILP, RO SG Y
[91] (2016) N S P1, P2, P12 O10 D1, D8 C4, C11 O MILP O Y
235
provided by health centers. Furthermore, there are many links developed a multi-objective model to determine the size and location
between the different levels of PCFs which makes it impossible to of departments in facilities within a given network of hospitals.
solve the location problems for each level independently [187]. Thus, given the detailed analysis provided above, future re-
Nevertheless, the small number of P11s in "Particular input/set- search could be conducted in the following fields:
ting" column of Table 5 (see [68,73–75,84,86]) shows that hier-
archical location models have not received much attention. As Proposing dynamic PCF location models (i.e., multi-period lo-
seen from the lack of P6 in the "Particular input/setting" column cation models) that take into account changes in the problem
and O(QT) in the "Modeling approach" column, waiting time and setting over time, such as population migration, significant
queuing theory considerations are rarely included in the literature. changes in management objectives, transportation and facility
Efficiency and effectiveness, equity, and demand coverage are capacities, patient population, etc.
three major criteria to evaluate accessibility to non-emergency Designing a hospital network with different types of PCFs
HCFs. Cost minimization is another major objective that is used in (hospitals, clinics, ambulatory healthcare centers, and off-site
HCF location problems. Table 5 depicts the types of objective public access devices).
functions that measure the locational performance of PCFs. A Developing location models by considering PCFs with different
major policy issue for a national health system in particular is the payment systems (e.g., with or without insurance).
efficiency and effectiveness of the public HCFs with respect to their Incorporating transportation modes (e.g., air, truck, and rail)
locations in local communities [69]. Distance (or time) minimiza- and routing decisions into PCF location models.
tion is a key factor in enhancing the efficiency and effectiveness of Developing statistical methods to estimate the input parameters
these facilities (see [68,69,75,77–79,82,84–87]). The most frequent of the existing models.
service goal which is used to improve the equity and fairness of Incorporating logistics and distribution considerations into the
services is minimizing the maximum distance to the nearest fa- existing PCF location models.
cility (see [69]). However, Beheshtifar and Alimohammadi [82] Integrating PCF location with related healthcare planning
proposed a new definition for equity by minimizing the variability decisions.
of access distance to healthcare services. It should be noted that Extending the existing PCF location models in a competitive
cost minimization has recently gained particular attention for environment for private primary care providers (see, e.g., [89]).
these PCFs (see [71,81,83–85,87,88]). As seen in Table 5, more than Developing models for centralizing locational decisions for a set
half of the papers (approximately 53%) utilized median-based lo- of (dependent or independent) primary care providers to im-
cation problems and about 38% used covering-based location prove the service quality and the utilization of the common
problems. resources.
The considerable number of D1s and D4s in the "Decision
variable" column of Table 5 clearly indicates that location deci- 4.2. Blood banks
sions are frequently combined with allocation of demand (see [69–
77,79,80,82,86–90]) and rarely with allocation of resources (see The use of blood and blood products on a daily basis is ex-
[67,75]). However, incorporating transportation modes (e.g., air, tensive worldwide for accident victims, cancer patients, and other
truck, and rail) and routing decisions have not attracted attention patients undergoing various surgeries, organ and marrow trans-
from researchers. plants, inherited blood disorders, etc. The literature on blood
Another important conclusion from the "Modeling approach" stockpile management focuses on the complexity of effective and
column of Table 5 is the large number of ILP and MILP models (see efficient inventory management of blood [92]. For instance, Pras-
[68,69,71–80,82,84,89–91]), compared to the number of stochastic tacos [188] reviewed the studies that incorporates OR techniques
(see [70,79,83,84]), dynamic, fuzzy, nonlinear (see [81,83]), and into blood inventory management theory and practice, and Beliën
goal programming (see [67]) models. and Forcé [189] surveyed the literature on inventory and supply
In addition, Table 5 gives an overview of the types of solution chain management of blood products.
methods that have been used to solve location problems in this class Extreme shortage of blood occurs in over 80% of the countries
of HCFs. As is evident from the table, these problems have been solved in the world [93]. In addition, the world health organization has
frequently by general-purpose software packages (see [67,69–71,73– stated that many patients requiring transfusion do not have timely
75,78,79,84,86–89]), Lagrangian relaxation method (see [68,80]), access to safe blood. The median blood donation rate in high-in-
heuristic methods (see [81,90]), and genetic algorithms (see [76,82]). It come countries is 36.8 donations per 1000 population, 11.7 in
can be concluded that approximately 72% of the papers used Class A middle-income countries, and 3.9 in low-income countries (note
methods and 28% used Class B methods to solve their models. that the blood donation rate is an indicator of the general avail-
The papers that deal with PCFs can be divided into sub- ability of blood in a country). Consequently, the lack of availability
categories, such as hospitals, clinics, ambulatory healthcare cen- and accessibility of blood can be attributed to the inefficient al-
ters, and off-site public access devices (see Appendix A). Never- location of resources (i.e., blood collection methods) that may be
theless, we deliberately chose to avoid more detailed classification the result of poor geographic location of the blood supply sources.
since papers do not determine any specific subcategory (it seems Hence, access to different types of blood banks (e.g., blood
there is no clear difference among these subcategories from a lo- transfusion providers, blood centers, blood stations and mobile
cation analysis perspective). Special types of PCFs are, however, units) is highly important in health systems. A major problem for
studied in some papers. For instance, Ndiaye and Alfares [71] blood banks is that human blood is a perishable, scarce, and va-
considered primary health units that are seasonally operated for luable product with a life time of only 21 days. Moreover, both
nomadic population groups. Griffin et al. [72] determined the best demand and supply of blood are stochastic. The extension of
location and number of new community health centers (CHCs) as covering-based problems for blood stations and mobile units as
well as the services each CHC should offer using publicly available well as median-based problems for the blood centers and trans-
data with the goal of maximizing the coverage of the weighted fusion organization can be considered as an effective location
demand given budget and capacity constraints. method in order to increase people’s participation in blood do-
The combination of departments’ layout and hospitals’ location is nation and timely access to safe blood. In addition, applying a
less considered in the literature. For instance, Stummer et al. [85] hierarchical approach is also suggested (see [92]).
Given the importance of blood bank locations in a health sys-
Case study
tem, only a few papers [92–97] studied the location of blood
study
Case
banks. Şahin et al. [92] considered a blood bank location problem
N
Y
Y
Y
Y
Y
Y
Y
in a hierarchical system in which the regional blood centers (RBCs)
SG, MH-SA, MH-O

Solution method
are defined as the upper level; and blood centers, blood stations,
and mobile units are defined as the lower level. They decomposed
SG, DP, O
Solution
the problem into three sub-problems. The first sub-problem solves

method
LR, SG
a hierarchical PMLP, called the pq-median location problem, to

SO
SG
SC
SL
SG
SC
SL
determine the locations of q RBCs (regional blood centers) which
MILP, 2-SSP, MCDM
provide services to p blood centers located at the demand points.
MINLP, FP, PSP, MCDM, O(QT)

MINLP, PSP, RO, FP
In the second sub-problem, location of blood stations–as sup-

porting facilities–are determined to improve the accessibility at
MINLP, GP
Modeling
approach
Modeling approach
MILP, RO
ILP/MILP
the demand points by set covering models. Finally, the third sub-
MINLP, PSP, FP, RO

MINLP
problem is formulated as an IP model to reduce inefficiency in the
MILP,MCDM
use of mobile units and ensure a homogenous distribution of
mobile units among the service regions. Cetin and Sarul [93] de-
MILP
veloped a model which determines a set of independent blood
Basic location
banks that should be established and the assignment of hospitals

Basic location
to the open blood banks. In their model, the locations of blood

PML/SCL
model
banks are determined based on the center of gravity method.

FCL, MCL
PML
FCL
FCL
FCL
FCL
model
The location of fixed and temporary blood facilities, and the

PML
PML
FCL
C1, C4, C5, C8, C9-1, C11
C2, C4, C5, C8, C9-1, C11
design of distribution networks over a multi-period horizon with

uncertainty in the input parameters were addressed by Zahiri et al.
C1, C5, C6, C8, C9-1, C11
C1, C4, C5, C8, C10, C11
C2, C4, C5, C6, C8, C11

C5, C8, C9-1, C11
C1, C4, C9-1, C11
[95]. Elalouf et al. [94] focused on the determination of centrifuge

C1, C4, C8, C11
centers in a three-echelon blood sample collection chain com-

Constraint
posed of clinics, centrifuge centers, and testing laboratories under

Decision variable Constraint
C1, C4, C5
the assumption that centrifugation services can be performed in

C10
clinics and centrifuge centers.

Jabbarzadeh et al. [97] and Fahimnia et al. [96] addressed the
D4, D8
D4, D8
D7, D8
location of permanent and/or temporary blood facilities in dis-

D4
asters. Jabbarzadeh et al. [97] proposed an FCLP-based robust

D4, D8
D4, D8
D7, D8
Decision
variable
D2,
D2,
D2,
D2,
D4
D4
model to determine the number and location of permanent and

D8
D1,
D1,
D1,
D1,
D1,
D1,
temporary blood facilities as well as to periodically determine

D2,
D2,
D2,
D2,
blood inventory levels by coordinating supply and demand during

D1,
D1,
D1,
D1,
and after disasters. Fahimnia et al. [96] also presented a two-stage

bi-objective model to determine location and relocation planning
Objective
function
O3, O10
O1/O3/O7
Objective
function
O3, O4
of mobile blood facilities; and to specify blood collection and

O3, O4
O3
O4
transportation quantities, and inventory levels for efficient and

O4
O4
O4
O4
timely supply of blood in disasters. In addition to the brief de-

P5-1, P5-2, P5-3, P8, P12
P5-1, P5-2, P7, P8, P12
P1, P2, P3, P4, P5-1, P5-2, P12
scription provided above, the papers are listed in Table 6.

P1, P3, P4, P5-1, P5-2, P12
P1, P3, P4, P5-1, P5-2, P12
In this regard, some possible future directions are as follows:

Particular input/setting
P1, P3, P4, P5-1, P12
P5-1, P8, P12
Extending location models for handling both independent blood

P1, P5-1, P5-2, P12
P1, P3, P11, P12
banks and dependent blood bank units (inside hospitals or

input/setting
clinics).
Particular
Non-emergency healthcare facilities: organ transplant centers.
P3
Developing models that consider stochastic and dynamic con-

P2,
P2,
P2,
P2,
ditions (see, e.g., [95]).

P1,
P1,
P1,
P1,
Considering new settings, such as budget constraints or multi-

Non-emergency healthcare facilities: blood banks.
ple-server blood banks.

Multi-period
Multi-period
Taking into account the different levels of the blood supply’s life
time(s) when locating blood banks; in particular, considering
setting
setting
the fact that some specialized treatments require the use of only
D-1
D-1
D-1
D-1
D-1
S
S
fresh blood.
S
S
Developing a model for optimal location/relocation of compo-

of uncertainty
of uncertainty
Consideration
Consideration
nents of a blood supply chain.

Proposing online location models for mobile blood units in or-
der to effectively increase voluntary blood donation rate and
N
N
enrich blood banks (see, e.g., [92,97]).

Y
Y
Y
N
Y
Y
Y
Incorporating the average age, sex, and blood groups of people

[100] (2014)
[101] (2014)
(2009)
(2007)
[98] (2006)
(2015)
(2015)
(2015)
(2014)
[99] (2013)
who can donate blood in each region in blood station location

Reference
Reference
models.
(year)
(year)
Table 6
Table 7

[92]
[93]
[96]
[95]
[94]
[97]
Addressing emergency conditions arising at times of disasters,

or other similar events, in locating blood banks (see, e.g., [96]).
4.3. Specialized services facilities Formulation:
There are two important considerations in the location of min ∑∑ aij xij + ∑ ∑ wk tkj ykj +E
j∈J i∈I k∈K j ∈ Tk (39)
specialized services facilities. On the one hand, proper location is
related to availability of specialized and experienced resources. On subject to
the other hand, their role in enhancing the quality of healthcare,
satisfaction of demand, and social welfare, is important to gov- ∑ xij =1, i∈I
j∈J (40)
ernments, the private sector, and individuals alike. The literature
on the location of specialized services facilities is examined in the
following three subsections. ∑ ykj =1, k∈K
j ∈ Tk (41)
4.3.1. Organ transplant centers
Organ transplant centers (OTCs) are the main components of ∑ zj=p
organ transplantation programs in healthcare systems, which have j∈J (42)
three distinctive features that should be considered in their
location: ykj ≤zj , k∈K , j∈J (43)
1. As the demand for organs continues to exceed the supply, organ xij ≤zj , i∈I , j∈J (44)
transplants suffer from long waiting lists.
2. The time that elapses between donor notification and trans-
E≥ ∑ wk ykj , j∈J
plantation is vital and very important in the process of organ k∈K (45)
donation.
3. Organ transplantation involves both a donor (a person who ykj ∈{ 0, 1}, k∈K , j∈J (46)
donates an organ intended for transplant) and a recipient (a
person who receives an organ).
xij ∈{ 0, 1}, i∈I , j∈J (47)
Hence, the strategic planning of the organ transplantation
programs requires a different location decision approach. For this zj∈{ 0, 1}, j∈J (48)
purpose, we present a PMLP-based model which takes into con-
sideration the above three features. The following MILP model is a E≥0. (49)
generalization of the one given in [98], with the notation:
In this model, the objective (39) minimizes the sum of the total
Sets: demand-weighted travel time (or distance) from donor hospitals
and organ recipient points to active OTCs, and the maximum size
of waiting list among all active OTCs. Constraints (40) and (41)
I The set of donor hospitals.
stipulate that each recipient point and donor hospital is only as-
J The set of candidate OTC locations.
signed to one active OTC. Constraint (42) specifies the total
K The set of potential organ recipient points. number of OTCs to be established. Constraints (43) and (44) limit
Tk The set of all candidate OTCs which are within an acceptable assignments to active OTCs. Constraints (45) determine the max-
travel time (or distance) of organ recipient point k∈K , imum size of the waiting lists of active OTCs. Note that E is an
Tk={ j: tkj≤T}. auxiliary variable (not a decision variable), used to compute the
maximum size. Constraints (46)-(49) are domain constraints.
Both Bruni et al. [98] and Belien et al. [99] formulated an OTC
Input parameters: location problem based on a PMLP to minimize the time components
of transplant, with the difference that the latter work considers both
the donors and the recipients in the problem. Furthermore, the for-
tkj The travel time (or distance) for transferring patients from
mer work considers three situations that all have different paths and
organ recipient point k∈K to an OTC at candidate location
travel times. Zahiri et al. [100] also investigated an OTC location
j∈J . problem, but slightly differently from the previous papers. They
aij The travel time (or distance) for transferring organs from presented a fuzzy programming model for a long-term dynamic lo-
donor hospital i∈I to candidate OTC j∈J . cation-allocation problem to minimize the total costs including fixed,
wk The demand size at organ recipient point k∈K . variable, and unsatisfied demand costs. Thereafter, this model was
T The maximum acceptable travel time (or distance) from an extended by Zahiri et al. [101]. They considered alternative trans-
organ recipient point (the cover distance or time). portation mode as well as uncertainty in demand and supply of or-
p The number of candidate OTCs to be established. gans in a multi-objective location-queuing model.
Although most research studies on OTC planning focused primarily
on topics, such as transplant waiting list and allocation policies (see,
Decision variables: e.g., [190–193]), it can be seen from Table 7 that only a few papers
addressed the location of OTCs. Indeed, proper location of OTCs plays a
zj 1, if an OTC is active at candidate location j∈J ; 0 otherwise. vital role in successful transplants (in terms of saving time and opti-
mizing links with other required units).
ykj 1, if recipient point k∈K is served by an OTC at candidate
Based on Table 7, locating OTCs in terms of both modeling and
solution methods, appears to have many opportunities for im-
xij 1, if donor hospital i∈I serves an OTC at candidate location
provement, such as
j∈J ; 0 otherwise.
Integrating location of OTCs with other related components
(hospitals and emergency departments) in a health system. Formulation:

Developing multi-stage stochastic programming models to in-
corporate other relevant OTC planning aspects in the determi- max ∑∑ σij yij
j∈J i∈I (50)
nation of OTC locations.
Centralizing decision making on locating/relocating OTCs. subject to
Extending the existing models to take dynamic and online as-
pects of transplant procedures into account. ∑ xj =p
Considering more realistic transportation features (e.g., sto- j∈J (51)
chastic travel times and ambulance busy fractions) in OTC lo-
cation problems. ∑ yij ≤1, i∈I
j∈J (52)
4.3.2. Detection and prevention centers
Detection and prevention centers (DPCs) provide healthcare ∑ σij yij ≥Wmin xj , j∈J
services which are different from other healthcare services (e.g., i∈I (53)
healthcare services for acute diseases). These services are usually
defined based on (local or national) detection and prevention yij ≤xj , i∈I , j∈J (54)
programs. DPCs have the following two key characteristics:
yij ∈{ 0, 1}, i∈I , j∈J (55)
1. Each center needs to have a minimum number of clients in
order to retain accreditation.
2. People have more flexibility as to when or where to receive xj ∈{ 0, 1}, j∈J . (56)
services, and might not refer to the closest DPC; therefore,
The objective (50) maximizes the total expected number of
participation in detection and prevention programs depends on
clients who take services from open DPCs. This is a relevant
the accessibly as well as quality of services provided by DPCs.
measure for participation in detection and prevention programs
operated by open DPCs. Constraint (51) states that p DPCs are to be
As a result, DPCs require a different location decision approach
established. Constraints (52) specify that each client point is
to incorporate these characteristics. The first characteristic can be
served by at most one open DPC. Constraints (53) ensure that each
satisfied by adding constraints in which each center has at least a
open DPC has the minimum number of clients. Constraints (54)
minimum number of clients. The second can be managed by
show that demand points are only covered by open DPCs. Con-
considering distance or waiting-time elasticity of demands. In this
straints (55) and (56) are integrality constraints.
regard, by using the model proposed in [102], a DPC location
Table 8 represents the papers dealing with DPC location pro-
problem can be formulated as follows:
blems. Because one of the purposes of detection and prevention
programs is to maximize participation, the majority of the papers
Sets:
in this category considered participation maximization objectives
(see [102–105,107,108,110]). Moreover, some papers (see
I The set of client points. [106,109]) used travel distance (or time) minimization objectives,
J The set of candidate DPC locations. which implicitly results in higher accessibility, and thus, in in-
creased participation.
Almost all of the papers in this section considered uncertainty
Input parameters: in the problem (see, [103,104,106–110]). In contrast, dynamic
conditions have not received attention from researchers. Another
dij The travel distance (or time) from client point i∈I to conclusion from Table 8 is that the basic location problem for
candidate DPC location j∈J . approximately all of the papers, excepting [106,109], is the MCLP.
wi The population size at client point i . From the O(QT)s in the "Modeling approach" column, it is also
seen that papers [103,104,106,108–110] used queuing theory to
σij The expected number of clients at client point i∈I who
consider service quality, which depends on the expected total time
take services from a DPC at candidate location j∈J , which
in the system, to increase participation in detection and preven-
is a function of wi and dij , e.g., σij=wi f (min {dij , Di }/Di )
tion programs. Moreover, all of the papers presented case studies
with f a decreasing function. demonstrating the importance of this approach in the real-world.
Di The maximum acceptable travel distance or time (the Breaking down the literature into the smaller parts identified in
cover distance or time) from client point i∈I . Table 8 leads us to propose specific future research suggestions.
p The number of DPCs to be established. These include the following:
Wmin The minimum required clients at an open DPC.
Developing dynamic DPC location models.
Applying other unconsidered settings such as multiple services.
Decision variables: Considering various types of costs related to the health system
and clients, and budget constraints.
Incorporating non-spatial factors, such as demographic and
xj 1, if a DPC is established at candidate location j∈J ;
socioeconomic variables into the existing models to increase
0 otherwise.
participation in detection and participation programs.
yij 1, if client point i∈I is assigned to an open DPC at candidate
location j∈J ; 0 otherwise. 4.3.3. Other specialized services facilities
Other specialized services facilities, which are discussed in this
section, were not included in the above two sections. These pro-
study vide services, such as exercise stress test, radiation therapy, EEG,
Case
ECG, etc., which are not related to detection and prevention pro-
Y
Y
Y
Y
Y
Y
grams or organ transplants. These are typically units with two
characteristics:
SC, H, MH-GA
MH-TS, H
H, MH-O
Solution
method
1. They can be located either as independent units, or embedded
S-SBO
O, BB
O, SC
in hospitals and other healthcare centers.
O, H
CP
H
2. If they are embedded, they should support their host facilities

while also providing service to their own clients. In this case,
Modeling approach
two types of demand are considered: flexible demand (coming

INLP, GP, PSP, FP,
MILP, PSP, O(QT)
from their own clients) and non-flexible demand (coming from

MINLP, O(QT)
MINLP, O(QT)
MINLP, O(QT)
MINLP, O(QT)
INLP, O(QT)
ILP, MCDM
MCDM the hosting facility).
As it can be seen from Table 9, the location of these units has

ILP
received considerable attention from OR/MS researchers in recent

years, after 2011 (see [111–114]). The two types of demand were
Basic location
addressed by [111,114]. Moreover, Syam and Côté [113] defined

several acuity levels for patients. In addition, the majority of pa-
pers located specialized services facilities within hospital net-
model
MCL
MCL
MCL
MCL
MCL
MCL
MCL
PML
works as dependent units (see [111–113]).

FCL
Although establishment of these facilities enhance the quality

C1, C5, C9-2, C11
C5, C8, C9-2, C11
of healthcare, satisfaction of demand, and social welfare; they are

C9-1, C9-2, C11
C9-1, C9-2, C11

C8, C9-1, C9-2,
C1, C9-2, C10,

C4, C5, C9-2
generally very expensive. Hence, the main objective sought in the

C1, C10, C11
Constraint
literature is related to minimization of the various costs involved

C5, C9-2
as evident in Table 9 (see [111–114]). For instance, Guo et al. [114]

C11
C11
embraced a multi-objective approach by considering optimizing

both cost and efficiency. They extended the basic FCLP and used
the non-inferior set estimation method (see [194]) to evaluate the
D4, D7
D4, D8
D1, D4, D8
D1, D4, D7
tradeoffs between cost and service. Furthermore, all papers in this

Decision
variable
D2, D4
D4, D7
D4
D4
section have capacity constraints in their models. In contrast, none

D1,
D1,
D1,
D1,
D1
of them takes budget constraints, multiple servers, or multiple

services into account. Moreover, only one paper considered un-
certainty or dynamic setting. All studies used the MCDM modeling
approach and applied software packages to solve the models.
Objective
function
O6, O10
O3, O10
O6, O10
Future studies on locating other specialized services facilities

can be conducted to cover the following gaps:
O6
O6
O6
O6
P1, P3, P5-1, P5-2, P6, P7, P9, O4
O6
Extending the existing location models for other specialized

services facilities to handle uncertainties.
Studying the location of other specialized services facilities with
P7, P9, P12
P1, P3, P5-1, P5-2, P9
multiple servers, budget constraints, or multiple services.

P6, P9
P6, P9
P1, P2, P4, P6, P9
Developing dynamic location models for other specialized ser-

P1, P2, P5-1, P6
input/setting
P9
P9
vices facilities.
Particular
P4,
P4,
P4,
P4,
P6,
Using accurate methods to solve large-sized instances of the

P3,
P2,
P2,
P3,
P2,
Non-emergency healthcare facilities: detection and prevention centers.
resulting location models for other specialized services facilities.

P12
P1,
P1,
P1,
P1,
P1,
Integrating location of other specialized services facilities with

other related HCFs with both flexible demand and non-flexible
demand.
Multi-period
4.4. Medical laboratories

setting
S
S
S
S
S
Medical laboratories provide services in testing and diagnosing

diseases (especially infection and contagious diseases). Though the
location of medical laboratories is critical for disease management,
infection control, and public health; research on location of these
Consideration of
HCFs is lacking. Considering various types of laboratories (see

uncertainty
Appendix A), a different location decision approach seems to be

needed. In this regard, only Shemshaki et al. [115] presented a
PMLP-based model to design a network of tuberculosis (TB) testing
laboratories where there are two possibilities: opening new la-
N
N
Y
Y
boratories or equipping existing ones. Their model minimizes the

[102] (2002)
[103] (2009)
[109] (2012)
[104] (2010)
[106] (2014)
[105] (2010)
[108] (2015)
[107] (2014)
[110] (2012)
total time required for transportation of samples that are collected

Reference
daily in new or existing laboratories by considering laboratory

(year)
Table 8
capacities, budgets, and maximum transportation times (between

origins and laboratories) as constraints. They applied their model
to the TB reference laboratory in the Canadian province of British
study
study
Case
Case
Columbia.
N
N
Y
Y
Y
SL, SO, SG
4.5. Mobile healthcare units
Solution
method
Solution
method
SC, BB
Mobile healthcare units often provide primary care services in
SC
SC
SL
SG
areas where there is no fixed primary HCF (i.e., hospital or clinic).
MINLP, PSP, MCDM

While this situation tends to be relatively rare in developed
countries, it can still be found in some developing countries. The
MILP, 2-SSP
MILP, MCDM
INLP, MCDM
Modeling
approach
ILP, MCDM
main task of the mobile healthcare units is to enable governments
Modeling
approach
to provide essential public health programs. This includes pre-
MILP
venting the spread of dangerous diseases (e.g., polio, diphtheria,
tetanus, and hepatitis) often through specific vaccination program
tion model
Basic loca-
tion model
for children and adults. The dynamic nature of location/relocation
Basic loca-
MCL
MCL
MCL
MCL
decisions is an essential aspect of location modeling of mobile
FCL,
FCL,
FCL,
FCL,
healthcare units.
PCL
FCL
Doerner et al. [116] proposed a model for locating mobile HCFs.
C1, C4, C5, C6, C8, C9-1, C9-2, C11

They considered a closed tour with selected stops and formulated
C4, C5, C6, C9-1, C9-2, C11

a multi-objective location-routing model to determine the optimal
number and location of these stops. As a result, they evaluated
tours according to three criteria, namely efficiency, average dis-
tances, and coverage.
C5, C8, C9-1, C9-2

C1, C6, C9-1, C9-2
C4, C5, C9-1, C11
Constraint
C1, C4, C5
4.6. Home healthcare centers
Constraint
Home healthcare (HHC) services emerged around 1950 as a
way to reduce costs of health systems and improve patients’
D1, D2, D4, D7, D8

quality of life [9]. In recent years, the number of HHC service
Decision variable
Decision variable
providers and demand for these services have grown rapidly. For
instance, the number of HHC companies in France increased 137%
D4, D7
D2, D4
D2, D4
D2 D4
from 2005 to 2010 (in just five years). Hence, the OR/MS literature
D1, D4
investigated different challenges of the HHC services, such as
D1,
D1,
D1,
D1,
routing, scheduling nurses and patients, and various resource al-
Objective function
location issues. However, our literature review reveals that no

paper studies the location of HHC centers. This shows that more
Objective
function
emphasis is required on location management and the strategic
O4, O7, O10
planning level of these HCFs (see [9]).
O10
O4
O4, O7
O4, O7
O4,O9
4.7. Rehabilitation centers, doctors' offices, and drugstores
P1, P2, P4, P5-1, P5-2, P7, P8, P12

P4, P5-1, P5-2, P5-3, P7, P8, P9
Rehabilitation centers, doctors’ offices, and drugstores are

P4, P5-1, P5-2, P5-3, P12
identified in Appendix A as constituting crucial components of

health systems. Nevertheless, research on location management of
P4, P5-1, P5-2, P5-3
these HCFs is lacking in the literature. It should be noted that most

Non-emergency healthcare facilities: other specialized services facilities.
HCFs which are discussed in this section are private and easy to
input/setting
Non-emergency healthcare facilities: long-term nursing care centers.
construct. Since they also make up the conduit for widely used
P5-1, P5-2
Particular
P1, P3, P4
services to reach patients, determination of the optimal number

input/setting
and proper location of them is very effective in increasing system

Particular
profitability (by cost reduction and revenue increase). Thus, given

P3,
P3,
P3,
P3,
that this shortage is identified, we hope that research on the lo-

P1,
P1,
P1,
P1,
Multi-peri-
od setting
cation of these HCFs will be included in the OR/MS research lit-

erature in the future.
Multi-peri-
od setting
D-2
S
S/D-1
4.8. Long-term nursing care centers

S
S
S
of uncertainty
Consideration
Due to decrease in both birth rates and mortality rates, many

of uncertainty
Consideration
developed and developing countries are experiencing significant

aging of the population. Aging societies require specific social and
medical services that have to be provided in healthcare centers,
N
Y
N
N
N
such as long-term nursing care centers. Undoubtedly, govern-

Y
ments (and private providers) need to plan for the ideal location of
[112] (2012)
[113] (2012)
[114] (2013)
[111] (2011)
[118](2015)
[117](2010)
these centers in order to provide the best possible services to aged

Reference
Reference
people. According to Appendix A, long-term nursing centers have

Table 10
(year)
(year)
Table 9
some features that should be considered in terms of location

modeling:
1. Long-term nursing centers provide medical (including nursing the other hand, temporary emergency HCFs include temporary
care) and social services to inpatients, and therefore the con- medical centers and points of dispensing, which deal with
current determination of optimal capacity levels (in terms of the healthcare services in disaster situations.
number of beds), inventory levels, and locations is essential. It should be recalled that emergency HCFs, especially tempor-
2. The clients of long-term care often do not require emergency ary emergency HCFs, are considered by many different fields, for
services and can wait until beds become available. example: disaster operations management, emergency logistics,
relief distribution, humanitarian logistics, homeland security,
The location of this category of HCFs was studied by Kim and emergency response, emergency departments, and emergency
Kim [117] and Cardoso et al. [118]. Kim and Kim [117] formulated service stations and vehicles. In the introduction, we have pro-
this problem as a p-center location model to minimize the max- vided appropriate survey papers on each topic for those readers
imum load of open facilities for load balancing. Moreover, they who are interested in more exploration.
suggested a branch and bound algorithm for the location problem
as well as a heuristic method to find an initial feasible solution. 5.1. Permanent emergency HCFs
Cardoso et al. [118] developed an FCLP-based model by taking into
consideration demand uncertainty, multiple services, and various Permanent emergency facilities provide wide range of
forms of equity (access, utilization, socioeconomic, and geo- emergency healthcare services. According to Fig. 3, these facil-
graphical equities). Table 10 describes these papers in detail. ities can be classified into four main categories in order to
According to what was mentioned above, we believe that the identify research gaps. These categories are studied in the se-
long-term nursing centers will receive considerable attention from quel. Since the nature of emergency and non-emergency HCFs
government and society in the future. There is a great need to are very different, descriptive fields, such as particular input/
address the proper location of these centers, and the OR com- setting, objective function, decision variable, and constraint are
munity could greatly contribute to this area; and thereby improve examined from a somewhat different perspective than those
the quality of healthcare and social welfare in general, and the considered in Section 4 (see Table 3).
quality of life of the elderly in particular. In this regard, future
research could be directed towards: 5.1.1. Emergency off-site public access devices
Off-site public access devices (OPADs) are non-interactive and
Improving the current models by taking other related HCFs into interactive facilities for providing a variety of healthcare services
account. in out-of-HCF environments (see Appendix A). OPADs have some
Extending the models by considering multiple services and features that should be considered in terms of location modeling:
service quality.
Clustering demand points based on non-spatial factors, such as 1. OPADs are public access devices, which mean that they em-
various emergency categories, social classes, age, race, etc. power individuals or bystanders to receive healthcare services
Developing competitive location models to represent situations without the presence of any trained medical personnel.
where private long-term nursing care centers compete for 2. Priority locations for OPADs placement are typically public and
clients. non-residential buildings (e.g., schools, transportation build-
Incorporating logistics considerations into locating long-term ings, commercial, civic and industrial sites, and recreational
nursing care centers. areas), and spaces containing high foot-traffic.
4.9. Combinations of several types of HCFs Hence, the strategic planning of OPAD placement programs
requires a location decision approach with regard to maximizing
In this section, the papers that considered the combination of coverage and ensuring timely access for the public. For this pur-
several types of HCFs are reviewed. These papers are listed in Ta- pose, extensions of covering-based problems seem appropriate. In
ble 11. In the literature, Galvao et al. [176], as well as Baray and addition, applying the MCDM modeling approach is also suggested
Cliquet [175], studied the combination of primary care facilities in order to improve the quality of service, reduce response time,
(hospitals, clinics, off-site public access devices, etc.) and other and consider both emergency and non-emergency situations.
specialized services facilities. Since the paper by Baray and Cliquet Recently, the investigation into locating automated external
[175] used the models of the basic MCLP and PMLP without any defibrillators (AEDs) as non-interactive emergency OPADs has
extension to locate maternity hospitals, it is not included in Table 11. been widely considered in the literature with different approaches
Galvao et al. [176] presented a p-median capacitated three-level (see, e.g., [195–201]). For instance, Folke et al. [198] used simple
hierarchical model to assign the prenatal HCFs to three levels in a geospatial techniques to evaluate and compare the effectiveness of
hierarchy and developed a Lagrangian heuristic to solve it. Kim et al. alternative AED placement strategies. Lerner et al. [199] and
[177] studied the problem of determining locations of public HCFs Warden et al. [200] used GIS to locate AEDs. Dao et al. [196] de-
which provide both hospital services and homecare services. Though veloped stochastic multi-time window MCLP and PMLP, and also
there are many interrelations among different types of non-emer- presented visualization techniques for 3D AEDs layout in a multi-
gency or emergency HCFs, the above review shows that a few stu- story academic building. Brooks et al. [201] presented a way to
dies have addressed the combination of multiple HCF types. This quantify the demand in AED location models which may be im-
represents a promising future research opportunity area. portant when a funding agency evaluates the deployment of AEDs
in practice. For this purpose, they sought to identify types of lo-
cations (e.g., race track/casino, jail, hotel/motel, hostel/shelter, and
5. Emergency HCF location rail station) with higher per-site risk for cardiac arrest.
Since 2009, several papers have studied the emerging problem
We classify the emergency HCFs according to whether HCFs of locating AEDs in different environments (schools, university, ur-
perform under permanent or temporary emergency situations. ban, etc.) with regard to the mentioned features [119–124]. Myers
Permanent emergency HCFs include emergency centers, emer- and Mohite [122] and Chan et al. [121] located AEDs optimally
gency off-site public access devices, trauma centers, and ambu- based on the criterion that a person should be covered by an AED
lance stations which provide emergency services all the time. On which is located no further than a particular travel time threshold
to ensure effective service delivery. Siddiq et al. [123] developed
Case study
this problem further by considering an effective range for each AED,
study
Case
which is affected by storage in a low-visibility and limited-access
N
Y
Y
Y
location, and lack of registration with local EMS authorities as po-
tential barriers to AED usage. Multi-responder and gradual coverage
Solution method
by a probabilistic extension of the basic MCLP were studied by Chan
Solution
et al. [120]. Bonnet et al. [119] also presented a multi-objective
MH-GA, SO
method
SC, LR
SC, H
optimization along with simulation to test the optimization output
O, SO
for the AEDs performance in terms of time-to-retrieve. Further-
more, they displayed the results in an interactive decision-making
web-based user interface to visualize potential deployment con-
Modeling approach
MILP, MCDM
figurations. These papers are listed in Table 12. Since papers [121–
MILP, MINLP, PSP

Modeling
approach
124] used the basic MCLP without any extension to locate AEDs,
ILP, MCDM
they are not included in Table 12.
ILP
Based on the discussion above, we believe that OPADs, parti-

cularly interactive and web-based devices, will receive consider-
able attention by governments, private sectors, and societies in the
Basic location
Basic location model
future due to their countless benefits in ways similar to what

happened with automated teller machine (ATMs) in financial
systems. There is a great need to address the appropriate OPAD
model
PML
placement locations. In this regard, future research could be di-

C1, C5, C9-1, C9- SCL
MCL, SCL
rected towards:
MCL
Extending models to determine locations for placing OPADs by

C1, C4, C9-1
Constraint
considering more realistic factors, such as uncertainty in de-

C2, C4, C11
Constraint
mands; the weight of each building (e.g., based on its popula-

tion); traffic patterns; and building accessibility influenced by
2
C11
locked doors, multiple floors, hours of operation, etc.

Developing location models for OPADs, especially for interactive
Decision variable
and web-based ones, which can be covered by other related

Decision
variable
D1, D4
D1, D4
HCFs, such as emergency centers and trauma centers if required.

D1, D5-1
D1, D5-1
Proposing integrated models for locating/relocating OPADs and

other related HCFs.
5.1.2. Emergency centers

Objective function
Objective
Emergency centers (or emergency departments) are permanent

function
emergency facilities which provide medical care to both walk-ins

O1, O7, O10
P1, P3, P4, P8, P11, P12 O3
O1
(unscheduled patients) and emergency patients transported by an

ambulance, who require immediate and urgent medical treatment.
In the OR literature, locating these facilities has not received much
O7
P1, P3, P4, P5-1, P9,
attention, despite the fact that many papers studied location

input/setting
problems on other related emergency facilities, such as ambulance

Particular
stations (see Section 5.1.4). In this regard, only Silva and Serra
[125] studied the location of emergency centers. They presented a
P12
priority queuing-based covering location problem to consider

P1, P3, P7, P9
Emergency healthcare facilities: emergency off-site public access devices.
P1, P3, P5-1
different service priority levels and applied a greedy randomized

adaptive search procedure to solve randomly-generated problem
instances. Thus, location of emergency centers, particularly those
Multi-period
associated with hospitals and clinics or ambulance stations is a gap

Multi-period setting
in the literature and needs to receive attention.

setting
S
S
5.1.3. Trauma centers

Trauma centers are hospitals that provide specialized medical
and nursing care to patients with traumatic injuries, and typically
have helicopter platforms to transfer patients. Although trauma
Combinations of several types of HCFs.
S
S
Consideration of
patients require immediate transfer to trauma centers, the estab-

of uncertainty
Consideration
lishment of helicopter platforms and using air transport are

uncertainty
expensive.
Extensions of the maximal backup coverage problem (BACOP)
can be deployed to address the modeling of location problems for
N
N
N
Y
trauma centers. The BACOP, which was introduced by Hogan and

Revelle [202], is based on an MCLP. In this regard, an extension of
[120] (2015)
[176] (2006)
[177] (2012)
[119] (2015)
the BACOP is presented to model a joint location problem of

Reference
Reference
trauma centers and helicopters with a budget constraint. In this

Table 12
Table 11
(year)
(year)
problem, we assume that the helicopters are not initially sent to

any demand point, and that they only provide backup coverage for
ambulances that are dispatched to crash locations. To formulate

study this problem, the following notation is used:
Case
N
Y
Y
Y
Y
Y
Y
Sets:
SL, MH-SA
SC, BD, H
I The set of demand points (trauma patients' locations).
Solution
method
J The set of candidate locations for trauma centers.

SC, H
SC, H
SC, H
SC
SL
K The set of candidate locations for helicopter platforms.

Ni The set of all candidate locations which can cover demand
{
point i by ambulances, Ni= j: tijA≤Ti . }
MILP, MCDM
ILP, MCDM
Si The set of all pairs of candidate locations of trauma centers

Modeling
approach
MINLP
and helicopter platforms, which can provide backup cov-

MILP
MILP
ILP
ILP
erage for demand point i by helicopters,

{ }
Si= (j, k ): (tijH +tkiH )≤Ti .
Basic location
FCL, MCL
FCL, MCL
Input parameters:
model
MCL
MCL
MCL
C4, C5, C7, C9-1, C11 MCL

SCL
wi The population size at demand point i∈I .

C1, C5, C6, C9-1, C11
C1, C5, C9-1, C9-2,
tijA The driving time from demand point i∈I to a trauma

C3, C5, C10, C11
C1, C3, C5, C11
center at candidate location j∈J .

The flying time from demand point i∈I to a trauma center
Constraint
tijH
at candidate location j∈J .
C4, C5
C4, C5
The flying time from a helicopter platform at candidate

C11
tkiH
location k∈K to demand point i∈I .
D1, D3, D5-1, D5-2,
pTC The number of trauma centers to be established.

Decision variable
D1, D3, D5-1, D8
p HP The number of helicopter platforms to be established.

The fixed cost to establish a trauma center at candidate
D4, D8
D1, D3, D4
cTC
D5-1
D5-1
j
location j∈J .
D4
ckHP The fixed cost to establish a helicopter platform at candi-

D1,
D1,
D1,
D1,
D8
date location k∈K .

α1 The weight parameter, between 0 and 1, representing the
importance of the total population with primary coverage.
O1, O2, O7
Objective
function
α2 The weight parameter, between 0 and 1, representing the

O4, O9
O4, O9
O7, O8
importance of the total population with backup coverage

O7
O7
O7
( α2 = 1−α1).
Ti The maximum acceptable travel time (the cover time)
P4, P5-1, P5-2, P5-3, P9,
P4, P5-2, P5-3, P8, P9
from demand point i∈I .

B The maximum budget that can be allocated for establish-
ing trauma centers and helicopter platforms.

P5-1, P7, P12
P5-1, P7, P12
P1, P3, P4, P7, P10

P12
P12
Decision variables:
P1, P2,
P1, P2,
P1, P3,
P1, P3,
P1, P2,
P1, P2,
P12
xTC
j
1, if a trauma center is established at candidate location
j∈J ; 0 otherwise.
xkHP 1, if a helicopter platform is established at candidate lo-
Multi-period
Emergency healthcare facilities: trauma centers.
cation k∈K ; 0 otherwise.

ykj 1, if a helicopter platform is established at candidate lo-
setting
cation k∈K and a trauma center is established at candidate

S
S
S
S
S
S
ziA 1, if demand point i∈I is covered by ambulance;

of uncertainty
Consideration
0 otherwise.
ziH 1, if demand point i∈I is covered by helicopter;
0 otherwise.
N
N
N
N
N
N
N
(2000)
(2007)
(2001)
(2010)
[130] (2010)
[131] (2013)
Reference
Formulation:
Table 13
(year)
[126]
[128]
[129]
[127]
max α1 ∑ wi ziA+α2 ∑ wi ziH

i∈I i∈I (57)
subject to in almost all the papers cited in Table 13, and other objectives have
not received attention.
∑ c TC TC
j xj + ∑ ckHP xkHP ≤B
Branas et al. [126] and Cho et al. [131] studied integrated lo-
j∈J k∈K (58)
cation of trauma centers, and associated helicopter platforms and
helicopter depots. Erdemir et al. [130] considered the possibility of
∑ xTC
j =p
TC
using hybrid transportation modes in which both ambulances and
j∈J (59)
helicopters (i.e. air ambulances) are used to transfer trauma pa-
tients to trauma centers when the scene of an incident does not
∑ xkHP =p HP have a suitable nearby area where a helicopter can safely land.
k∈K (60)
Future studies can address combinations of ambulance stations
and trauma centers with hybrid transportation modes.
ziA−ziH ≥0, i∈I (61) Concluding the above discussion, suggestions for future re-
search directions are as follows:
∑ xTC A
j −zi ≥0, i∈I
j ∈ Ni (62) Extending the existing models to capture more realistic as-
sumptions, such as uncertainty of demand, multi-type demand,
∑ ykj −ziH ≥0, i∈I and multiple server settings.
(j, k)∈ Si (63) Developing location models for trauma centers in multi-period
settings.
ykj ≤0. 5 (xTC HP
j +xk ), j∈J , k∈K (64)
Presenting hierarchical facility location models for trauma
centers.
Combining ambulance stations, helicopter depots, and trauma
xTC HP
j +xk −1 ≤ ykj , j∈J , k∈K (65) centers with hybrid transportation modes.
Developing integrated models for simultaneously locating/re-
ykj ∈{ 0, 1}, k∈K , j∈J (66) locating trauma centers, and other related permanent and
temporary emergency HCFs, such as emergency departments,
xTC OPADs, temporary medical centers, etc.
j ∈{ 0, 1} , j∈J (67)
Incorporating scenarios that may occur in disasters into the
location of trauma centers.
xkHP ∈{ 0, 1}, k∈K (68)
5.1.4. Ambulance stations
ziA, ziH ∈{ 0, 1}, i∈I . (69) EMSs (emergency medical services) play a pivotal role in health
systems, which are generally concerned with providing out-of-
The objective function (57) maximizes the weighted combina- hospital acute medical care and transferring patients to emergence
tion of primary and backup coverage given to demand points. centers, emergency departments within hospitals, or trauma
Constraint (58) is the budget constraint to establish and equip the centers for definitive care [10]. Each EMS is typically a service
selected trauma centers and helicopter platforms. Constraints (59) process including the following four main steps: (i) receiving an
and (60) specify the total number of facilities to be established. emergency call and evaluating the situation, (ii) dispatching an
Constraints (61) ensure that backup coverage is credited in the ambulance(s) to the scene, if required, (iii) serving on-scene
proper order. Constraints (62) assure that demand points are only emergency services, and (iv) transferring a patient(s) to a related
covered by open trauma centers, and Constraints (63) show that HCF, if required, and coming back to a station or other emergency
demand point i is covered by helicopters when at least one pair of sites.
trauma center and helicopter platform within the set Si services it. Ingolfsson [10] surveyed EMS planning and management from
Constraints (64) and (65) provide the logical links between the four perspectives: forecasting of demand, response times, and
binary variables. Finally, Constraints (66)-(69) are integrality workload; performance measurement; choosing station locations;
constraints. and allocation of ambulances to stations. Moreover, Chaiken [203]
Papers that studied the location of trauma centers have been provided key lessons for implementing OR studies in EMS cases. Li
listed in Table 13. Examination of this table allows us to draw some et al. [6] reviewed the covering models and optimization methods
conclusions. Despite random variation (especially in demand) for emergency response facility location and planning. In this pa-
being the main characteristic of these centers, none of the papers per, several effective research directions are presented at the
has incorporated uncertainty. In addition, dynamic models (either conclusion. Other related review papers on emergency response
long-term or short-term) have not been applied to problems of are [33,204,205].
trauma center location. Ambulances are a major resource for EMSs. Ambulances must be
Another deficiency observed from Table 13 is the lack of hier- located at appropriate points in order to provide adequate coverage
archical facility location models when injuries can be classified at and minimize the response time. Hence, a great deal of attention has
different therapeutic levels depending on predetermined factors been paid to the location of ambulance stations, the deployment
(e.g., the intensity and time of the event, resource availability at each (location or relocation) of ambulances in the stations, and the dis-
level of hierarchy, and the type of injury). These different therapeutic patch (assignment) of ambulances to the demand points (emer-
classes could subsequently be treated at different levels of a facility gency sites). One should also note that there are two types of loca-
hierarchy. Moreover, the inclusion of multiple servers and/or busy tion decisions related to ambulances: locating ambulance stations
fractions of ambulances can assist in the proper handling of trau- and locating ambulances in stations (also known as ambulance de-
matic events, but this is not seen in the literature. ployment or ambulance relocation). When ambulances are homo-
Furthermore, since equipping these centers is expensive, the in- geneous, the ambulance deployment is equivalent to determine the
clusion of cost factors in objective functions and/or constraints is number of ambulances at each station.
needed. In general, the objective is either cost minimization (see Brotcorne et al. [4] provided a detailed review of the literature
[128,129]) or demand coverage maximization (see [126,127,130,131]) with further focus on relocation and dispatching, up to 2003, and
Jeffrey and Goldberg [35] surveyed the research done on emer- subject to
gency vehicles, including ambulances, up to 2004. Location of
emergency services stations before 2012 are reviewed in [34].
∑ uj, t ≤Mt , t∈T
j∈J (71)
One should note that ambulance stations differ from emer-
gency centers (or emergency departments). Emergency centers are
units equipped and staffed to provide immediate and urgent ∑ yik, t ≤ ∑ aij, t uj, t , i∈I , t∈T
k∈K j∈J (72)
medical care to unscheduled patients, who show up or are trans-
ported by an ambulance. However, ambulance stations are re-
yik, t ∈{ 0, 1}, i∈I , k∈K , t∈T (73)
sponsible for dispatching ambulances which provide EMSs on
scene or during transport to an emergency center, an emergency
department, a trauma center, etc.; and return to a predetermined uj, t ≥0 & integer , j∈J , t∈T . (74)
station to await another call.
The objective function (70) maximizes the expected coverage of
We assert that extensions of the basic MCLP can be useful for
total demand by the ambulances over multiple periods. Con-
locating ambulance stations or deploying ambulances. An example
straints (71) state that in each time interval up to Mt ambulances
of such extensions is the maximum expected coverage location
can be deployed. Constraints (72) show that demand points in
problem (MEXCLP), developed by Daskin [206]. This formulation
each time interval are only covered by the stationed ambulances.
seeks to maximize the expected covered demand. It models each
Constraints (73) and (74) are integrality constraints.
ambulance as being busy with probability p and operating in-
One can alternatively minimize the number of stationed am-
dependently from other ambulances. In the sequel, we formulate a
bulances:
multi-period version of MEXCLP in order to consider the re-
deployment of ambulances as well as changes in the quantity of min ∑ ∑ uj , t
available ambulances in the candidate stations. The dynamic t∈T j∈J
MEXCLP can be formulated with the following notation:

by assuring that the expected coverage of total demand is larger
than a predetermined constant ct in each time period,
Sets:
∑
i∈I
wi, t (∑ k∈K )
(1 − pt ) ptk − 1yik, t ≥ct , t∈T .
J The set of candidate locations for ambulance stations. Batta et al. [207] introduced the adjusted MEXCLP (AMEXCLP)
K The set of ambulances. in order to relax the assumption that ambulances are independent
T The set of time intervals. by using the hypercube queuing model, which is a spatial queuing-
theoretic model that tracks the status of mobile servers in order to
compute performance measures, such as the mean user response
Input parameters: times, fraction of time that the system is in each state, fraction of
dispatches of each server to each region, server workloads, etc.
[204,208–210].
Di The maximum acceptable travel distance or time (the
This section covers location models that addressed ambulance
cover distance or time) from demand point i ∈ I .
stations and deployment from the past decade and breaks them
aij, t 1, if an ambulance at candidate location j∈J covers de- down into categories. Results are given in Table 14. In fact, around
mand point i∈I in cover distance Di at time interval t ∈T ; 31% of the total number of papers which are reviewed in this
0 otherwise. paper and also approximately 61% of the papers that discussed
wi, t The population size at demand point i∈I at time interval emergency HCF location are included in Table 14, highlighting the
t ∈T . importance of these location models.
Mt The maximum number of ambulances to be stationed at It can be seen from Table 14 that about 75% of the surveyed
time interval t ∈T . papers deal with stochastic problems. This is because location
pt The probability that an ambulance is not working at time management of ambulance stations involves the inherent un-
interval t ∈T . certainty of EMS systems. The busy fraction of ambulances and
ct The minimum expected coverage requirement at time in- uncertainty in service requests (i.e., the number of emergency
terval t ∈T . calls) are considered in [134,141,147,154], which can be managed
in different ways including queuing theory, chance constraints
([134,147]), two-stage stochastic programming ([141]), and robust
Decision variables: optimization ([154]). C6s in the “Constraint” column represent that
ambulance reliability (also known as service level) are frequently
considered by many papers using chance constraints or queuing
uj, t The number of ambulances stationed at candidate location theory ([133, 134, 136–141, 143, 145, 147, 151, 152, 159, 164, 165]).
j∈J at time interval t ∈T . An EMS system is called reliable if there is at least one ambulance
yik, t 1, if demand point i∈I is covered by at least k∈K available to provide service whenever a service demand arises.
ambulances at time interval t∈T; 0 otherwise. Unfortunately, based on the “Solution method” column in Table 14,
it can be seen that there are a few papers that developed exact or
bounded-error algorithms.
Formulation: Another important conclusion that can be drawn from Table 14
refers to dynamic (i.e., multi-period) models which have emerged
⎛ ⎞ in the literature as a significant development. Several papers
max ∑ ∑ wi, t ⎜⎜ ∑ (1 − pt ) ptk − 1yik, t ⎟⎟ considered short-term dynamic settings (see [132,138,140,144,
t∈T i∈I ⎝ k∈K ⎠ (70)
146,150,152,158,159]). This is sometimes motivated by the
Table 14
Emergency healthcare facilities: ambulance stations.
Reference (year) Consideration of Multi-period Particular input/ Objective Decision Constraint Basic location Modeling Solution Case
uncertainty setting setting function variable model approach method study
[132] (2001) N D-1 P1, P2, P4, P5-2, P7, P12 O8, O10 D3, D5-2, D6 C1, C3, C7, C11 MCL ILP SC, MH-TS Y
[133] (2002) Y S P1, P2, P5-2, P10 O4, O8 D1, D5-2, D6 C4, C6 MCL MILP, MCDM, O SL, SBO Y
(QT)
[134] (2004) Y S P1, P3, P4, P5-1, P5-2, O4 D1, D3 C1, C5, C6, C7 FCL ILP, PSP SC N
P7

[135] (2004) Y S P1, P3, P4, P7, P10 O9 D1, D3, D5-2 C2, C4, C7 MCL ILP – Y
[136] (2005) N S P1, P2, P4, P7, P12 O8, O9, O10 D3, D5-2, D6 C3, C5, C6, C7 MCL ILP MH-TS, MH-AC Y
[137] (2006) Y S P1, P2, P4, P7, P10 O2, O7 D1, D6 C2, C5, C6, C7 MCL ILP, GP, MCDM, O O Y
(QT)
[162] (2006) Y S P1, P2, P10, P12 O7 D3, D6, D8 C11 MCL ILP SC Y
[138] (2007) N D-1 P1, P2, P7, P8 O5 D3, D6 C5, C6, C11 PCL MILP H N
[139] (2008) Y S P1, P2, P7, P8, P10, P12 O7 D3, D8 C5, C6 MCL INLP, PSP, O(QT) SO, H Y
[140] (2008) Y D-1 P1, P3, P7, P10 O2 D3, D5-2 C2, C5, C6, C11 SCL INLP, PSP, O(QT) MH-TS N
[141] (2009) Y S P1, P3, P4, P5-1, P5-2, O4 D1, D3, D6 C2, C5, C6, C7, C11 FCL ILP, PSP, 2-SSP SC, BB, H N
P7
[142] (2009) Y S P1, P2, P4, P7, P8, P10 O8 D3, D5-2 C5, C7 MCL ILP, O(QT) SC Y
[143] (2009) Y S P1, P2, P10 O3 D3, D6 C1, C2, C5, C6 PML INLP, PSP, O(QT) MH-TS Y
[144] (2010) N D-2 P1, P2, P5-1, P5-2, P7, O4 D1, D3, D5-2, D6 C1, C3, C5, C9-1, C9- FCL ILP SL Y
P8, P12 2, C11
[145] (2010) Y S P1, P2, P10, P12 O8 D3, D5-2 C5, C6 MCL ILP SC Y
[146] (2010) N D-1 P1, P2, P4, P5-2, P7, P12 O7, O8, O10 D3, D5-2 C3, C5, C7, C9-1, C11 MCL ILP, MCDM MH-O Y
[147] (2010) Y S P1, P3, P4, P5-1, P5-2, O4 D1, D3 C5, C6, C7, C11 FCL MILP, PSP, 2-SSP H N
P7
[148] (2011) N S P1, P2, P12 O7 D3, D5-2 C2, C5, C9-1 MCL ILP S-SBO N
[149] (2011) Y S P1, P3, P4, P7, P10, P12 O5, O7, O9 D3, D5-2 C5, C7, C11 MCL ILP, MCDM, O(QT) SO Y
[159] (2011) Y D-1 P1, P2, P10, P12 O2 D3, D6 C6, C11 SCL ILP/INLP, O(QT) H, MH-TS Y
[150] (2012) Y D-1 P1, P2, P7, P10, P12 O3 D3, D6 C5, C11 PML SDP S–O Y
[151] (2012) Y S P1, P2, P5-1, P5-2, P7, O1, O2/O3 D1, D3, D8, D5-2 C2, C3, C5, C6, C11 SCL ILP SG Y
P10, P12
[152] (2013) Y D-1 P1, P2, P10 O2, O10 D3, D5-2 C2, C5, C6, C11 SCL MILP, MCDM, O MH-TS, H Y
(QT)
[153] (2013) Y S P1, P2, P10, P12 O3 D3, D6 C11 PML MINLP, O(QT) MH-GA Y
[163] (2013) Y D-1 P1, P2, P4, P5-2, P5-3, O9, O10 D3, D6 C2, C7, C11 MCL ILP, 2-SSP SC Y
P8, P12
[154] (2014) Y S P1, P3, P4, P5-1, P5-2, O4 D1, D3, D6 C1, C7, C11 FCL MILP, MCDM, RO SC Y
P5-3
[158] (2015) Y D-1 P1, P2, P5-1, P10, P12 O1, O7, O10 D1, D3, D5-2 C11 MCL, SCL MILP, MCDM SC Y
[165] (2015) N D-1 P1, P2, P4, P5-2, P5-3, O4, O10 D3, D6 C6, C7, C11 PML MILP, MCDM SC N
P12
[164] (2015) Y S P1, P3, P5-1, P5-2 O2, O4 D1, D3, D6 C1, C6, C9-1 FCL MINLP, PSP SC Y
[160] (2016) N D-1 P1, P2, P4, P8 O2 D3, D6 C1, C2, C5, C9-1, C11 SCL MILP S-SBO Y
247
application of these location models to update ambulance posi- concerned with the modification of ambulance location in parti-
tions during a specific time period. Furthermore, Table 14 shows cular stations and the determination of where to send ambulances
that stochastic settings are only used in half of dynamic models when their missions are complete. Dispatching decisions focus on
(see [140,150,152,158,159]). This is because combining a stochastic proper procedures for assigning ambulances to demand points
approach with a multi-period setting can result in very complex which are assumed to be based on a priority list, which is a list of
models that are difficult to solve. As a consequence, the papers ambulances sorted with regard to their priority of dispatch based
developing stochastic multi-period models presented metaheur- on call severity (see, e.g., [142]). In this section, D3, D6 and D5-2 in
istic algorithms (often Tabu Search) to solve their models (see the “Decision variable” column represent the ambulance deploy-
[140,159]). ment, dispatching, and multiple coverage decisions, respectively,
From the "Objective function" column in Table 14 one can which can be integrated with the ambulance station location de-
conclude that most of the papers aim for lower cost (see O4) and cision (D1). Note that in Table 14 there are papers which take
better responsiveness, and often deal with the tradeoff between ambulance stations as given, and subsequently optimize both de-
these objectives. In the literature, better responsiveness is eval- ployment and dispatching decisions (see, e.g., [138,150,163,165])
uated through the use of objectives, such as minimizing travel or a combination of other decisions including deployment deci-
distance (or time) of ambulances, minimizing unmet demand, and sions (see, e.g., [132,162]).
maximizing demand coverage (single and multiple) (see O9, O7, In conclusion, based on Table 14, we can suggest some future
and O8). Some objectives, such as minimizing the relocation and research directions:
coverage costs (included in O10) can also implicitly reduces the
response time (see, e.g., [152,158]). Furthermore, minimizing the Considering more realistic assumptions in locating ambulances
maximum response time and the number of ambulances are rarely and their stations, such as ambulance capacity, interruptions,
considered as objectives (see O2 and O10). dynamic setting, real-time setting, and general travel and ser-
In addition, according to the "Modeling approach" column in vice probability distributions.
Table 14, ILP and MILP are the basis of almost all the models for Extending the hierarchical location models for ambulance sta-
locating ambulances and their stations while nonlinear program- tions and other related HCFs, such as trauma centers, OPADs,
ming (see [153,159]), goal programming (see [137]), robust opti- emergency centers, etc.
mization (see [154]), and dynamic programming (see [150]) are Integrating ambulance station and deployment decisions with
rarely found in the literature. Further, fuzzy programming, hier- other EMS strategic and tactical decisions, such as fleet size,
archical programming, and multi-stage stochastic programming staff number, crew planning, standby sites, etc.
have not been considered in the literature at all. Extending location models for ambulance stations with con-
There are a few papers that could not be listed in Table 14 ([155– sidering their role in disasters.
157, 161]). Paper [155] compared the performance of existing discrete Integrating ambulance station decisions with most related op-
location models for locating ambulances. The method proposed in erational decisions, i.e., deployment and dispatching decisions.
[156] is based on embedding a hypercube model into a hybrid genetic Incorporating real-time (online) relocating or dispatching stra-
algorithm to find the optimal location and coverage areas of ambu- tegies into locating ambulance stations ([214]).
lances in order to minimize the response time. Paper [161] presented Developing multi-stage stochastic programming to more accu-
two greedy algorithms using genetic algorithm which can be em- rately determine ambulance station locations under a set of
bedded with either an exact or approximate hypercube model. stochastic scenarios.
Erkut et al. [157] introduced a new covering location problem for Proposing exact or bounded-error algorithms for solving exist-
EMS stations which is called the maximal survival location problem ing ambulance location models.
(MSLP) where the objective is to maximize the expected number of
patients who survive. For this purpose, they incorporated a de- 5.2. Temporary emergency HCFs
creasing function of the response time into existing covering models.
Their empirical comparison of the MSLP with the corresponding Temporary emergency events occur suddenly and infrequently,
MCLP and PMLP showed that the MSLP was more appropriate for but lead to great demand for a wide range of emergency services. As
the EMS location. a result, a variety of medical, social, and relief services are needed.
Bélanger et al. [211] empirically investigated the location and Temporary emergency HCFs are crucially important to rescue large
relocation strategies in ambulance fleet management. First, they number of people facing a catastrophic disaster or major emergency
briefly reviewed the literature, and then proposed four management situations. The optimal location of these HCFs in pre-disaster plan-
strategies, ranging from simple to sophisticated strategies. Finally, ning leads to risk mitigation and reduction in response time.
they designed a simulation tool to analyze each of these strategies According to Fig. 3, temporary medical centers and points of
with extensive simulation experiments. Aringhieri et al. [160] also dispensing are two major types of temporary emergency HCFs,
combined optimization and simulation for ambulance location. which are studied in the following subsections.
It should be mentioned that there are other techniques that can
be used to improve optimization modeling in the ambulance lo- 5.2.1. Temporary medical centers
cation literature. For instance, Alanis et al. [212] proposed a sto- Temporary medical centers (TMCs) are providers of healthcare
chastic queuing model that could be used to evaluate the perfor- services to people who are affected by disasters or large-scale
mance of a repositioning plan for ambulances. This model can, in emergencies, which may be of a catastrophic nature. This class
principle, be incorporated into an optimization model for ambu- contains field hospitals, Red Crescent and/or Red Cross tents, ca-
lance repositioning. Moreover, Budge et al. [213] provided a sta- sualty collection points (CCPs), and any existing hospitals and
tistical model that can be used to estimate input parameters as- clinics, that are pre-planned to play short-term roles in disasters.
sociated with travel times for ambulance location models. TMCs have some distinctive features that should be considered in
Review of the literature indicates that many ambulance loca- terms of location modeling:
tion problems have been addressed by integrated models which
incorporated deployment strategies and/or dispatching decisions 1. Pre-planned TMCs may be completely disrupted or their service
into ambulance station location (see, e.g., [141,154,164]). Deploy- capacities may be significantly reduced in disasters, and thus
ment decisions, which are indispensable for fleet management, are they cannot be operationalized in some disaster scenarios.
2. In disaster events, the medical services provided by oper-

ationalized TMCs may be cut off from emergency sites because
study
study
Case
Case
of the destruction of the area’s civil infrastructure (freeways,
N
Y
Y
Y
Y
roads, communications, emergency medical services, etc.).
3. The demand for emergency cares stochastically varies under
different scenarios associated with disasters.
SC, MH-SA
SC, MH-SA
SC, MH-GA
H, MH-TS
SC, DP, O
Solution
method
Solution
4. It is possible to use helicopters for air transportation of medical
method
MH-GA
supplies or of people in need of further care to other HCFs inside
SG
H
or outside the affected area.
5. Pre-planned TMCs are not required to be operational in all
MINLP, PSP, O(QT)

disaster scenarios.
Modeling
approach
Modeling
approach
CCPs are mass collection TMCs which are used for provision of first
MCDM
ILP, DP
ILP, PSP
MILP
MILP
MILP
aid in some public or private facilities, such as colleges, schools, and
ILP
public parks, which are large enough to accommodate a large number
of people, and are relatively free from falling debris ([168–170]). CCPs
are basically pre-determined units that are operationalized and staffed
Basic location
Basic location
after a disaster occurrence by medical teams who have been preas-

MCL/PML
signed. People in need of medical attention have to get to these CCPs

model
C1, C4, C5, C9- SCL/PML

model
on their own by foot or off-road vehicles, and subsequently transferred

PML
PML
PCL
MCL
C1, C4, C8, C11 PML
O
to more equipped HCFs or shelters if needed.

Below, a two-stage stochastic programming model is proposed
C1, C4, C8, C9-1, C9-2,
based on an MCLP. In this model, a set of candidate TMCs is de-

C4, C8, C9-1
Constraint
C1, C4, C9-1, C11
termined in the pre-disaster planning phase, considered as the first

C1, C9-1, C11
stage. All pre-planned TCMs will be established when a disaster

Constraint
C1, C5, C11
happens. The assignment of open TMCs to demand points is decided

1
C4, C5
in the post-disaster planning phase, considered as the second stage.

C11
D1, D4, D7, D8
The capacities of TCMs and the populations seeking urgent care at

D1, D2, D7
demand points vary with the disaster scenarios.

Decision
D2, D3, D4,
variable
D1, D4
Sets:
Decision
variable
D1, D4
D4
D4
D4
D1,
D1,
D1,
D1,
D8

Objective
function
K The set of possible disaster (large-scale emergency) scenarios.

O1/O3
Objective
function
O7
O3
O7/O3
O10
Input parameters:
O3
O4
O5
P1, P2, P3, P4, P6, P7, P8,

P1, P2, P4, P7, P8, P9,
P1, P3, P4, P9, P12
pk The probability of occurring disaster scenario k∈K .

Particular input/
wik The population at demand point i∈I who need emergency

P1, P3, P4, P5
care under disaster scenario k∈K .

P1, P3, P4
P1, P3, P12
P1, P3, P4
cjk The service capacity of a TMC at candidate location j∈J

setting
P1, P3
P12
under disaster scenario k∈K .

Temporary emergency healthcare facilities: temporary medical centers.
P12
p The maximum number of candidate TMCs to be

established.
Temporary emergency healthcare facilities: points of dispensing.
Multi-period
Multi-period
Decision variables:
setting
setting
D-1
S
S
S
S
S
S
xj 1, if a TMC is established at candidate location j ∈ J ;

0 otherwise (first-stage decision variable).
yijk The fraction of the demand of demand point i ∈ I served by
a TMC at candidate location j ∈ J under disaster scenario
Consideration of
Consideration of
k ∈ K ; 0 otherwise (second-stage decision variable).

uncertainty
uncertainty
zik 1, if demand point i ∈ I is covered under disaster scenario

k ∈ K ; 0 otherwise (second-stage decision variable).
N
N
N
N
N
N
Y
Y
[171] (2009)
[167] (2008)
[170] (2006)
[172] (2012)
Formulation:
[168] (2014)
[174] (2015)
[57] (2010)
Reference
Reference
⎛ ⎞
Table 16
Table 15
(year)
(year)
max ∑ pk ⎜⎜ ∑ wik zik ⎟⎟

k∈K ⎝ i∈I ⎠ (75)
subject to decade by several authors (see the survey paper by Synder et al.
[36]). Moreover, incorporating priorities can be another feature of
∑ xj ≤p
such models. For example, Oran et al. [215] developed a location-
j∈J (76)
routing problem where demand points have different priorities in
getting service in emergency response planning.
∑ yijk =zik , i∈I , k∈K The above observations can be used to suggest studies to cover
j∈J (77)
the following research gaps:
∑ wik yijk ≤cjk xj , j∈J , k∈K Incorporating uncertainty of demand and service capacities into
i∈I (78)
existing TMC location models.
Developing multi-stage stochastic programming models to
xj ∈{ 0, 1}, j∈J (79) adequately model disaster operations management under dif-
ferent scenarios.
0≤yijk ≤1, i∈I , j∈J , k∈K (80) Presenting models for simultaneously locating different types of
TMCs, such as field hospitals and CCPs simultaneously.
zik∈{ 0, 1}, i∈I , k∈K . (81)
Integrating the location of helipads with TMCs.
Adapting existing location models with disruptions to locate
The objective (75) maximizes the expected coverage by open TMCs, whose service capacities fluctuate in disasters.
TMCs. Constraint (76) specifies the maximum number of TMCs to Considering concerns of disaster management and humanitar-
be established. Constraints (77) specify which TMCs meet the ian logistics in the location of non-emergency HCFs which can
demand of a covered demand point. Constraints (78) specify the play a temporary role in large-scale emergencies.
maximum capacities of open TMCs. Finally, Constraints (79)–(81) Developing location models for TMCs with disruptions im-
are domain constraints. pacting both TMCs and links, under different disaster scenarios
Papers that studied the location of TMCs are listed in Table 15. (see e.g., [88,167]).
Paper [169] on CCP location and paper [166] on location of medical
services facilities for large-scale emergencies are not included in 5.2.2. Points of dispensing
this table since they compared different existing location models. A point of dispensing (POD) is a mass medication dispensing
As mentioned above, important characteristics of such situa- site that is capable of providing medicine and medical supplies
tions are the possibility of facilities being destroyed, changes in the (e.g., vaccines, drugs, and therapeutics) to protect the general
capacities of facilities, changes in the capacities of roads, and un- population in infectious disease disasters (e.g., epidemics, pan-
certainty in the size and location of demand. Nevertheless, we are demics, or an outbreak of an emerging infectious disease). PODs
surprised to see in Table 15 that all papers investigate problems are one type of the temporary emergency HCFs used in disaster
under deterministic conditions. situations, but differ from HCFs given in the previous subsection.
Moreover, another main characteristic of these facilities, CCPs They are more similar to typical public service facilities where
in particular, is the possibility of accommodating helicopter plat- reduction of congestion costs is critical and where all people
forms for air transportation, which has not received attention in should be able to access them with no trouble.
the literature. In addition, based on Table 15, we believe that more In this regard, the OR community has concentrated on bio-
attention should be paid to hierarchical models, required re- terrorist attacks as a kind of infectious disease disaster (see [216])
sources, and multiple services in order to make the models more with reference to this HCF type ([171–174,217–219]) and has ad-
realistic. dressed the various challenges in mass dispensing: medical supply
As indicated by Larson [205], traditional facility location pro- distribution, locations of dispensing facilities, optimal facility
blems may be highly inappropriate in disasters or major emer- staffing and resource allocation, routing of the population, and
gencies where one must consider the destruction of facilities and dispensing methods. For this purpose, interactive tools have re-
infrastructure. In such cases, to increase the probability of survi- cently received attention to help decision makers in mapping out
vability of the drug and supplies distribution system, one may real-time dynamic optimization, and analyzing the economic and
want to position more than the usual number of facilities, each potential benefits. Moreover, the integration of simulation tools is
containing fewer medications and supplies than usual, resulting in growing, since simulation is typically a much more realistic eva-
a problem that has some similarity to the p-dispersion location luator of system performance and is useful for validating the re-
problem (see Section 2.3). One may use the location models with sults returned from optimization models. For instance, Lee et al.
disruptions, which have been extensively studied over the last [171] focused on the problem of selecting an adequate number of
Fig. 5. The frequencies of basic discrete location problems used in the HCF location literature.
Temporary emergency facilities
medical centers
Temporary
[167,168]
[168]
[57]
–
–
dispensing
Points of
[171,174]
[172]
[171]
–
137,139,142,145,146,148,149,157,158,162,163] Fig. 6. The frequencies of modeling approaches used in the HCF location literature.
well-placed PODs where established PODs are dynamically de-

signed and staffed in a stochastic environment. For this purpose,
they integrated simulation and optimization into a decision sup-
port system (Real opt-POD) that can perform optimization of
strategic and operational planning over each simulation iteration.
[134,141,144,147,154,164]
The papers ([171,172,174]) which proposed location models for

[140,151,152,158–160]
PODs are analyzed in Table 16.

Ambulance stations
The following directions can be suggested for future studies on

[143,150,153,165]
POD location:
[132,133,135–
Integrating related humanitarian logistics decisions with POD

[138]
location decisions.
Applying the MCDM approach to locate PODs.
Developing accurate methods for solving the resulting POD lo-
[126–131]
[128,129]
Trauma
centers
cation models.
Studying POD location when different disasters (e.g., bioterrorist
[130]
Permanent emergency facilities
attacks during an earthquake) can happen simultaneously.

Emergency
centers
6. Analyses of literature from different perspectives

[125]
Emergency HCFs
–
–
We complete our review of the literature on facility location

Emergency
models in health systems by presenting additional information

[119–124]
regarding the surveyed papers. For this purpose, we analyze the

OPADs
[119]
types of basic discrete location problems corresponding to these

–
–
–
papers, specify the modeling approaches and solution methods

used to solve the problems, and determine whether a case study is
105,107,108,110–113]
[92,94,106,115,116]
[93,95–97,109,111–
considered or not. Finally, we break down the papers with respect

The basic corresponding discrete location problems which are used in the literature.
to the subcategories of the survey’s descriptive dimensions.

B-NE-HCF
[72,73,76,77,80,84,86,89,90,100] [92,102–
6.1. Basic discrete location problems

114,118]
[117]
[92]
Fig. 5 and Table 17 provide an overview of the basic discrete

location problems that have been used in modeling HCF location
[71,78,81,83–85,87,88,100,101]
[68,69,75,79,82,86,98,99,176]
problems (i.e., the strategic planning level of health systems). As

can be seen from Fig. 5, the basic MCLP (41.9%), PMLP (19.4%), and
Non-emergency HCFs
FCLP (22.6%) are three popular basic location problems used to

study healthcare location problems. Furthermore, approximately
54% of the basic location problems are covering-based problems;
about 43% are median-based problems, and about 4% of the pro-
A-NE-HCF
[67,177]
blems are not in either of the above two broad classes.

Table 17 indicates the basic discrete location problems which
–
are used in the surveyed papers. In this table, the non-emergency

discrete location problems
HCFs have been divided into two main classes, namely, A-NE-HCF
↓The basic corresponding
PMLP
MCLP
PCLP
FCLP
SCLP
and B-NE-HCF. The HCFs in Class A-NE-HCF provide mainly

→The Types of HCFs
healthcare services that directly affect clients’ experience of health

treatment while those in Class B-NE-HCF mainly provide suppor-
Covering-based
Median-based
tive services that are used in other healthcare services, such as

problems
problems
radiology, CT scan, blood provision, etc. Class A-NE-HCF can in-

clude hospitals and clinics (general or specialized), organ trans-
Table 17
plant centers, and home healthcare centers.

The median-based location problems find the median locations
Fig. 7. The frequencies of solution methods used in the HCF location literature.
Fig. 8. The classification of papers on HCF location problems based on whether or not they present a case study (NEF, PEF and TEF stand for non-emergency, permanent
emergency, temporary emergency facilities, respectively).
among the candidate locations of facilities in order to minimize made in different stages. It seems that the HCF literature has
the demand-weighted average travel distance (or time) while the mostly tended to use a modeling approach that results in simple
covering-based location problems emphasize the demand cover- models, which can be optimally solved using existing optimization
age within a specified travel distance. As a consequence, the pro- solvers within reasonable times (see also the next subsection), but
posed non-emergency HCF location problems were generally sacrifices or dilutes the validity of the models. This shows that
median-based problems and the emergency HCF location pro- there is much room for OR experts to use more advanced modeling
blems were generally covering-based, as can be seen from Fig. 5 approaches.
and Table 17. However, studies on temporary emergency HCF lo-
cation problems tended to apply both median-based and covering- 6.3. Solution methods
based problems.
Fig. 7 summarizes the different types of solution methods that
6.2. Modeling approaches have been used to solve HCF location problems. From this figure,
one can see that a variety of solution methods in both Classes A
Fig. 6 depicts the frequencies of the modeling approaches used and B (accurate and inaccurate) have been developed to solve
to formulate the HCF location problems. Over 50% of the papers these problems. Approximately half of the papers used general-
used integer programming, including ILP, INLP, MILP and MINLP, purpose optimization software (43%); and, among these, non-
with ILP and MLIP as the most popular modeling approaches. emergency problems comprise the majority.
MCDM is also widely used to deal with HCF location problems Moreover, approximately 45% of the papers solve the problems
which naturally involve several performance measures. Un- heuristically using methods in Class B. Among metaheuristic
fortunately, only a few studies incorporated an uncertain optimi- methods, Tabu Search (32%) and Genetic (32%) methods are more
zation approach, such as fuzzy programming, stochastic pro- popular compared to Simulated Annealing (14%) and Ant Colony
gramming, stochastic dynamic programming, and robust optimi- (4%) methods. In addition, among the methods in Class A, branch
zation, despite the fact that uncertainty is an important modeling and bound and Lagrangian relaxation methods have been used
factor which should not be simplified. Moreover, multi-stage sto- more than dynamic programming, decomposition, and cutting
chastic programming and stochastic dynamic programming are plane methods (see Table 18). Note that the branch and cut, branch
rarely used though many decisions in HCF location problems are and price, and column generation methods have not been applied
Table 18
The categorization of the literature with respect to the survey descriptive dimensions.
Dimension Codes Papers
Consideration of uncertainty Y [58,70,79,83,84,88,89,95–97,99–101,103,104,106–111,118,120,121,133–135,137,139–143,145,147,149–154,158,159,162–164,172,174]

N [57,67–69,71–78,80–82,85–87,90–94,98,102,105,112–114,117,119,125–132,136,138,144,146,148,160,165,167,168,170,171,176,177]
Multi-period setting S [57,58,67–70,72–80,82,83,86–94,98,99,102–114,117,119–121,125–131,133–137,139,141–143,145,147–149,151,153,154,162,164,167,170–172,174,176,177]

D-1 [71,95–97,100,101,112,132,138,140,146,150,152,158–160,163,165,168]
D-2 [81,84,85,118,144]
Particular input/setting P1 [57,58,67–84,86–114,117–121,125–154,158–160,162–165,167,168,170–172,174,176,177]

P2 [58,75,83,84,91,96,98–101,103,104,106,108,110,118,121,126,127,130,132,133,136–139,142–146,148,150–153,158–160,162,163,165,168,174]
P3 [57,58,68–70,73,74,76–82,85–90,92,93,95–98,102,105,107,109,111–114,117,119–121,125,128,129,131,134,135,140,141,147,149,154,164,167,170–
172,174,176,177]
P4 [58,69,70,72,75–77,83,85,88,93,95–97,102–105,108,111–113,117,118,128,129,131,132,134–137,141,142,146,147,149,160,163,165,167,168,171,172,174,176,177]
P5-1 [71–73,78,80–84,87–90,93–97,99–101,106,107,109,111–114,118,119,129,130,134,141,144,147,151,154,164,177]

P5-2 [58,71–73,78,81,83–85,87–90,94–97,100,101,107,109,111–114,118,128,129,132–134,141,144,146,147,151,154,158,163–165]
P5-3 [58,84,100,111–113,128,129,163,165]
P6 [103,104,106,108–110,174]
P7 [77,85,101,109,113,118,120,130–132,134–142,144,146,147,149–151,168,174]
P8 [58,68,72,75,79,80,84,86,87,89,99–101,113,118,128,138,139,142,144,160,163,168,174]
P9 [77,102–105,107–110,113,120,128,129,168,172,177]
P10 [131,133,135,137,139,140,142,143,145,149–153,158,159,162]
P11 [68,73–75,84,86,92,176]
P12 [67,68,74,75,79–82,84–89,91–97,99–101,109,110,112,118,126,127,129,130,132,136,139,144–146,148–151,153,158,159,162,163,165,168,170,172,174,176,177]
Objective function O1 [92,119,130,151,158,171,177]

O2 [130,137,140,151,152,159,160,164]
O3 [58,68,69,75,77–79,82,84–87,92,96,98,99,101,106,143,150,151,153,167,168,171,174,176]
O4 [58,71,81–85,87,88,93–97,100,101,109,111–114,118,129,133,134,141,144,147,154,164,165,167]
O5 [57,69,138,149]
O6 [77,102–105,107,108,110]
O7 [72,76,77,80,86,89,90,92,112–114,119–121,125–127,131,137,139,146,148,149,158,162,168,172]
O8 [130,132,133,136,142,145,146]
O9 [73,84,85,111,128,129,135,136,149,163]
O10 [58,67,70,74,79,82,83,85–87,91,98,105–107,112,117,119,132,136,146,152,158,163,165,170]
Decision variable D1 [57,58,67–69,71–84,86–105,108–114,117–121,125–131,133–135,137,141,144,147,151,154,158,164,167,168,170–172,174,176,177]

D2 [58,67,75,92,95–97,99–101,106,111,113,114,118,168,172]
D3 [130–132,134–136,138–154,158–160,162–165,168]
D4 [57,69–83,86–90,92–96,98,99,102–104,106–114,117,118,121,125,128,129,131,167,168,170,171,174,176,177]
D5-1 [86,119,120,126,127,130]
D5-2 [130,132,133,135,136,140,142,144–146,148,149,151,152,158]
D6 [132,133,136–138,141,143,144,150,153,154,159,160,162–165]
D7 [58,75,77,84,97,100,104,107,108,112,118,172,174]
D8 [58,68,70–73,75,78,79,81,82,84,85,88,91,95–101,109,110,118,121,128,130,139,151,158,162,168,174]
Constraint C1 [57,58,67–69,71,75–79,81–83,86,92,93,96,98–100,103,106,107,111,113,117,129,132,134,143,144,154,160,164,167,168,171,174,176,177]
C2 [95,101,120,135,137,140,141,143,148,151,152,160,163]
C3 [130,132,136,144,146,151]
C4 [67–71,73,74,76,77,82–84,86,88–93,95,96,98,99,101,105,112,113,117,118,120,121,125–127,131,133,135,167,168,170–172,174,176]
C5 [57,69,72–76,79,80,84,86,87,89,90,95–105,112–114,117,118,121,125–131,134,136–143,145–152,160,170,171,177]
C6 [100,101,111,113,128,133,134,136–141,143,145,147,151,152,159,164,165]
C7 [132,134–137,141,142,146,147,149,154,163,165]
C8 [58,95–97,100,101,172,174]
C9-1 [58,67–70,72,75–77,80,82–85,88,93,95–97,100,106,108–114,118,128,129,131,144,146,148,149,160,164,167,168,171,172,176,177]
C9-2 [67,70,71,75,77,79,82–84,87,102–106,108–111,113,114,118,129,144,168,177]
C10 [68,72,73,78,80,81,88–90,94,99,106,107,130,176]
C11 [57,58,67,70,72,74,75,77,78,80,81,86–93,95–97,99–101,103,104,106–110,112,113,118–121,128,129,131,132,138,139,141,144,146,149–154,
158–160,162,163,165,167,168,174]
253
254
Dimension Codes Papers
Basic location problem SCL [67,92,93,119,130,140,151,152,158–160,171,177]

MCL [72,73,76,77,80,84,86,89,90,92,100,102–105,107,108,110–113,119–121,124–133,135–137,139,142,145,146,148,149,158,162,163,168,172]
PCL [57,117,138]
PML [68,69,75,79,82,86,92,94,96,98,99,106,115,143,150,153,165,167,168,171,174,176]
FCL [58,71,78,81,83–85,87,88,93,95–97,100,101,109,111–114,118,128,129,134,141,144,147,154,164]
O [70,74,91,170]
Modeling approach ILP [57,71,74,76,77,80,86,89,90,92,102,105,112,119,125,129,130,132,134–137,141,142,144–146,148,149,151,159,162,163,171,172,177]

INLP [83,103,107,113,139,140,143,159]
MILP [58,67–69,72,73,75,78,79,82,84–88,91,92,96–99,110,114,117,118,120,121,126–128,133,138,147,152,154,158,160,165,167,168,176]
MINLP [70,81,93–95,100,101,104,106,108,109,111,120,131,153,164,174]
GP [67,86,93,107,137]
FP [95,100,101,107]

PSP [70,79,95,100,101,107,111,120,134,139–141,143,147,164,172,174]
2-SSP [58,83,84,96,118,141,147,163]
DP [57]
SDP [150]
RO [88,95,97,100,154]
MCDM [58,67,69,77,79,82–86,96,98,101,105,107,111–114,119,128,129,133,137,146,149,152,154,158,165,170,176]
O [89,101,103,104,106,108–110,121,133,137,139,140,142,143,149,152,153,156,159,161,174]
NLP, 1-SSP, M-SSP, SP-O, CP, MPDM -
Solution method SL [67,71,97,98,111,112,128,129,133,144]

SC [57,68,76,78,80,81,87,89,90,92,99,108,110,113,114,117,125–127,130–132,134,141,142,145,154,158,162–165,167,171,176,177]
SX [69,74,79,86]
SG [58,75,83,84,88,94–96,100,101,111,118,151,168]
SO [70,73,93,111,119,120,139,149]
BB [102,117,141]
CP [106]
LR [68,80,96,176]
BD [131]
DP [57,94]
O [57,91,94,102,108,109,120,137]
BC, BP, BCP -
H [81,90,103–105,109,110,127,130,131,138,139,141,147,152,159,170,172,177]
MH-TS [85,104,132,136,140,143,152,159,170]
MH-GA [76,82,110,119,153,156,161,171,174]
MH-SA [81,101,129,167]
MH-AC [136]
MH-O [83,101,103,121,146]
S-SBO [107,133,148,160]
S-O [150]
S-SA, S-SAA, S-SO -
Case study inclusion Y [58,67–88,90–95,97–110,112,114,115,118–120,125–128,130–133,135–137,139,142–146,149–154,156–164,167,168,171,172,174,176]

N [57,89,93,96,113,121,129,134,138,140,141,147,148,165,170,177]
to solve the HCF location problems. Modeling complex HCF location problems using constraint
programming which simplifies the statement of constraints.
6.4. Case studies Developing models with equilibrium constraints or multi-level
programming models to address HCF location problems where
Generally, researchers provide evidence of the applicability of health system actors can decide independently after learning
their research through the process of validation. Some research the decisions made by upper-level decision makers.
studies apply their results to a case study, which refers to a real- Using the simulation approach for modeling HCF location pro-
life example using historical data or implementation in practice to blems which cannot be mathematically modeled or their
demonstrate the importance of their results in the real world. mathematical models cannot be solved efficiently.
Fig. 8 classifies the surveyed literature according to whether they
used a case study or not. As shown in Fig. 6, 80% of the papers 7.1.2. Future research directions: solution methods
presented case studies while the remaining 20% only tested their As shown in Fig. 6, approximately half of the papers used
results using test problems, which are randomly generated. general-purpose optimization software packages (43%), corre-
It may be noted that data collection for testing HCF location sponding to the first subclass of Class A methods (i.e., accurate
models in disaster situations is very difficult. Indeed, in such si- methods: exact or bounded-error methods). There is relatively
tuations data may not be available or may not be easy to com- little work on the other subclasses of Class A methods in the lit-
municate. However, we are pleased to have found that 77% of the erature. Furthermore, about 45% of the papers used Class B
papers dealing with temporary emergency HCF location problems methods (i.e., inaccurate methods: heuristic, metaheuristic, and
tested their results using real-life case examples. approximate stochastic optimization methods) for solving their
models. Thus, based on detailed analyses of various tables pro-
6.5. Categorization of literature with respect to all descriptive vided previously in this review, other conclusions in terms of so-
dimensions lution methods are summarized below:
The papers related to each type of HCFs were analyzed in detail Using accurate solution methods (exact or bounded-error
in the previous sections. Nevertheless, if the readers are interested methods) to solve the existing models which are basically
in reviewing the literature corresponding to each subcategory of solved by software packages and cannot be solved in large scale
the descriptive dimensions, they have to check each section. by the commercial solvers within reasonable times.
Therefore, we presented an inverse categorization in Table 18 Applying advanced IP methods (e.g., branch and cut, Lagrangian
which helps the readers to find the papers belonging to each relaxation, and Benders decomposition) to solve existing ILP or
subcategory of these dimensions. MILP models.
Solving those HCF problems tackled by inaccurate solution
methods (e.g., heuristics or meta-heuristics) using accurate so-
7. Future research directions lution methods.
Applying simulation-based methods to solve location problems
In general, we have provided most of our suggestions for future in complex health systems, for which other solution methods
research in each section. However, in this section we summarize cannot be used.
related discussions and potential future research directions that Developing software packages or web-based programs which
are drawn from the overall review. We do this with respect to can be used for free by all health systems worldwide.
(i) the computational perspective (mathematical modeling ap- Organizing a data base which systematically collects benchmark
proach and solution method) and (ii) different types of location test problems for HCF location problems.
problems in health systems.
7.2. Future research directions in terms of HCF type
7.1. Future research directions from computational perspective
The health systems require various types of HCFs to perform a
We divided the computational future research directions into wide range of services related to human health. Each type of HCF has
two separate subsections in terms of modeling approach and so- different characteristics and applications. Naturally, these character-
lution method. istics and applications interact with the optimal locations of HCFs.
Therefore, relevant characteristics should be considered in location
7.1.1. Future research directions: modeling approaches modeling in order to make location models more efficient and closer
Considering the analysis presented in Section 6.2, we can to reality. With respect to this aspect, the literature can be scrutinized
suggest the following future guidelines: to identify the distinctive features of different types of HCFs that
should be considered in location modeling of health systems.
Incorporating stochastic or robust optimization into the HCF Thus, there are many research directions which are useful for
location models in static and dynamic settings. addressing more realistic HCF location problems. With regard to
Using multi-stage uncertain programming, such as multi-stage this, Table 19 summarizes some suggestions for identifying po-
stochastic programming, stochastic dynamic programming, or tentially fertile areas in real-world HCF location modeling.
adjustable robust optimization approaches to address more
realistic real-world applications in an uncertain dynamic 7.3. General future research directions
setting.
Applying game theory to model HCF location problems in In addition to the categorized suggestions in Table 19, some
competitive environments. future research directions could also be suggested for all HCFs:
Taking advantage of queuing theory when developing HCF lo-
cation models in order to capture congestion and related service Combining related types of HCFs in location models.
quality metrics. (e.g., papers [106, 174] incorporated congestion Integrating location decisions with other strategic, tactical, or
terms into their objective functions and papers [103, 104, 108– operational decisions in HCF location models.
110] considered constraints on waiting times). Extending the existing HCF location models to multi-period
Table 19
The future research directions in terms of HCF type.
The types of HCFs The future research directions in terms of the HCF type
Primary care facilities (hospitals, clinics, off-site public access de- Proposing dynamic PCF location models (i.e., multi-period location models) that take into
vices, etc.) account changes in the problem setting over time, such as population migration, significant
changes in management objectives, transportation and facility capacities, patient population,
etc.
Designing a hospital network with different types of PCFs (hospitals, clinics, ambulatory
healthcare centers, and off-site public access devices).
Developing location models by considering PCFs with different payment systems (e.g., with or
without insurance).
Incorporating transportation modes (e.g., air, truck, and rail) and routing decisions into PCF
location models.
Developing statistical methods to estimate the input parameters of the existing models.
Incorporating logistics and distribution considerations into the existing PCF location models.
Integrating PCF location with related healthcare planning decisions.
Extending the existing PCF location models in a competitive environment for private primary
care providers.
Developing models for centralizing locational decisions for a set of (dependent or in-
dependent) primary care providers to improve the service quality and the utilization of the
common resources.
Blood banks Extending location models for handling both independent blood banks and dependent blood
bank units (inside hospitals or clinics).
Developing models that consider stochastic and dynamic conditions.
Considering new settings, such as budget constraints or multiple-server blood banks.
Taking into account the different levels of the blood supply’s life time(s) when locating blood
banks; in particular, considering the fact that some specialized treatments require the use of
only fresh blood.
Developing a model for optimal location/relocation of components of a blood supply chain.
Proposing online location models for mobile blood units in order to effectively increase vo-
luntary blood donation rate and enrich blood banks.
Incorporating the average age, sex and blood groups of people who can donate blood in each
region in blood station location models.
Addressing emergency conditions arising at times of disasters, or other similar events, in lo-
cating blood banks.
Organ transplant centers Integrating location of OTCs with other related components (hospitals and emergency de-
partments) in a health system.
Developing multi-stage stochastic programming models to incorporate other relevant OTC
planning aspects in the determination of OTC locations.
Centralizing decision making on locating/relocating OTCs.
Extending the existing models to take dynamic and online aspects of transplant procedures
into account.
Considering more realistic transportation features (e.g., stochastic travel times and ambulance
busy fractions) in OTC location problems.
Detection and prevention centers Developing dynamic DPC location models.

Applying other unconsidered settings, such as multiple servers and multiple services.
Considering various types of costs related to the health system and clients, and budget
constraints.
Incorporating non-spatial factors, such as demographic and socioeconomic variables into the
existing models to increase participation in detection and participation programs.
Other specialized services facilities Extending the existing location models for other specialized services facilities to handle
uncertainties.
Studying the location of other specialized services facilities with multiple servers, budget
constraints, or multiple services.
Developing dynamic location models for other specialized services facilities.
Using accurate methods to solve large-sized instances of the resulting location models for
other specialized services facilities.
Integrating location of other specialized services facilities with other related HCFs with both
flexible demand and non-flexible demand.
Long-term nursing care centers Improving the current models by taking other related HCFs into account.
Extending the models by considering multiple services and service quality.
Clustering demand points based on non-spatial factors such as various emergency categories,
social classes, age, race, etc.
Developing competitive location models to represent situations where private long-term
nursing care centers compete for clients.
Incorporating logistics considerations into locating long-term nursing care centers.
Off-site public access devices Extending models to determine locations for placing OPADs by considering more realistic
factors, such as uncertainty in demands; the weight of each building (e.g., based on its po-
pulation); traffic patterns; and building accessibility influenced by locked doors, multiple
floors, hours of operation, etc.
Developing location models for OPADs, especially for interactive and web-based ones, which
can be covered by other related HCFs, such as emergency centers and trauma centers if
required.
Proposing integrated models for locating/relocating OPADs and other related HCFs.
The types of HCFs The future research directions in terms of the HCF type
Trauma centers Extending the existing models to capture more realistic assumptions, such as uncertainty of
demand, multi-type demand and multiple server setting.
Developing location models for trauma centers in multi-period settings.
Presenting hierarchical facility location models for trauma centers.
Combining ambulance stations, helicopter depots, and trauma centers with hybrid transpor-
tation modes.
Developing integrated models for simultaneously locating/relocating trauma centers and
other related permanent and temporary emergency HCFs, such as emergency departments,
OPADs, temporary medical centers, etc.
Incorporating scenarios that may occur in disasters into the location of trauma centers.
Ambulance stations Considering more realistic assumptions in locating ambulances and their stations, such as
ambulance capacity, interruptions, dynamic setting, real-time setting, and general travel and
service probability distributions.
Extending the hierarchical location models for ambulance stations and other related HCFs,
such as trauma centers, OPADs, emergency centers, etc.
Integrating ambulance station and deployment decisions with other EMS strategic and tactical
decisions, such as fleet size, staff number, crew planning, standby sites, etc.
Extending location models for ambulance stations with considering their role in disasters.
Integrating ambulance station decisions with most related operational decisions, i.e., reloca-
tion and dispatching decisions.
Incorporating real-time (online) deployment or dispatching strategies into locating ambulance
stations.
Developing multi-stage stochastic programming to more accurately determine ambulance
station locations under a set of stochastic scenarios.
Proposing exact or bounded-error algorithms for solving existing ambulance location models.
Temporary medical centers Incorporating uncertainty of demand and service capacities into existing TMC location models.
Developing multi-stage stochastic programming models to adequately model disaster op-
erations management under different scenarios.
Presenting models for simultaneously locating different types of TMCs, such as field hospitals
and CCPs simultaneously.
Integrating the location of helipads with TMCs.
Adapting existing location models with disruptions to locate TMCs whose service capacities
fluctuate in disasters.
Considering concerns of disaster management and humanitarian logistics in the location of
non-emergency HCFs which can play a temporary role in large-scale emergencies.
Developing location models for TMCs with disruptions impacting both TMCs and links, under
different disaster scenarios.
Points of dispensing Integrating related humanitarian logistics decisions with POD location decisions.
Applying the MCDM approach to locate PODs.
Developing accurate methods for solving the resulting POD location models.
Studying POD location when different disasters (e.g., bioterrorist attacks during an earth-
quake) can happen simultaneously.
Other types: medical Laboratories, mobile healthcare units, home Using the basic location models to solve location problems of these facilities, given that little
healthcare centers, rehabilitation centers, doctors' offices, drugstores, study has been done on locating these facilities.
emergency centers, and other facilities given in Fig. 2, which are not Investigating real-world case studies on location problems of these facilities.
healthcare providers Developing specialized location models for these facilities.
Studying the location of these facilities in competitive environments.
Incorporating the location of these facilities into the existing location models developed for
other HCFs.
settings. theoretic models (see, e.g., [89]).

Considering medical waste in proper location of HCFs. Applying integrated HCF location models, such as location-
Studying network design problems involving healthcare provi- routing, location-inventory, and location-pricing to HCF
ders in medical supply chains, which include the various players location.
in health systems (see Fig. 1). Using continuous location models in HCF applications.
Incorporating disaster management (humanitarian logistics) Applying the existing location models with disruption risk to
considerations into determination of non-emergency and per- address HCF location problems with unreliable HCFs or to in-
manent emergency HCF locations, which can provide special corporate disaster management considerations.
services in disaster-level events. Incorporating engineering economic models into the design
Integrating location of permanent and temporary emergency phase of multi-period models, which leads to more realistic
facilities with other facilities in the disaster management and estimation of costs and revenues (see, e.g., [144]).
humanitarian logistic networks. Studying HCF location problems with multiple servers at var-
Considering reliability and resilience considerations in HCF lo- ious levels.
cation models under possibility of interruptions and disrup- Extending hierarchical HCF location models to consider given
tions, and other uncertainties. priorities in health treatment procedures.
Studying the HFC facility location problems in competitive en- Introducing environmental and sustainability concerns into the
vironments using cooperative and non-cooperative game- HCF location models.
Studying HCFs for animal visitation, animal-assisted activities, circumstances. Moreover, except for clients for home healthcare
animal-therapy, and any health-related programs for animals facilities, clients in this group refer to physicians, surgeons or
considering infection prevention and control concerns [220]. paramedics for prevention, examination, diagnosis, treatment
Considering the concept of medical tourism and globalization (both surgical and non-surgical), or checkup either on an in-pa-
issues in location management of HCFs. tient or on an out-patient basis.
Developing a holistic model to optimally design (redesign) a health
system’s physical network for a new (existing) city or district. A.1.1. Primary care facilities (hospitals, clinics, off-site public access
Studying location of IT-based and web-based (online) HCFs, e.g., devices, etc.)
interactive OPADs, mobile HCFs, or temporary ambulance sta- This class of HCFs provides the most basic and the broadest
tions, possibly by using radio-frequency identification (RFID), scope of healthcare services for clients of all ages across all social,
internet of things (IoT), and Big Data. economic, and geographic categories. Primary care facilities (PCFs)
comprise a wide range of primary care providers, such as public
and specialized hospitals and clinics, polyclinics, ambulatory
8. Conclusions healthcare centers, and off-site public access devices. Due to the
essential services that these facilities provide, ease of access for
In this paper, we have reviewed almost the entire emergency clients is crucial. In keeping with the need for proximity, primary
and non-emergency healthcare literature on facility location ana- care clients usually refer to the nearest PCF.
lysis over the last decade (2004-2016). By analyzing the existing It should be noted that ambulatory care centers are PCFs that
surveys, we show that the lack of a comprehensive review of HCF
provide diagnosis, observation, consultation, treatment, and in-
location is a significant shortcoming in the healthcare literature.
tervention services directly to out-patients who do not require in-
Therefore, we introduced a comprehensive framework to classify
patient services. These PCFs offer a mix of telemedicine, imaging,
HCFs in terms of location management. The optimization models
short-term observation care, and surgery. Technology used in
in each classification of HCFs are analyzed in a detailed table with
these centers allows patients to avoid being kept overnight for
ten descriptive dimensions (consideration of uncertainty, multi-
monitoring. Many routine checks can be done through remote
period setting, particular input/setting, objective function, decision
digital technologies [221].
variable, constraint, basic discrete location problem, mathematical
In many countries, there can be found PCFs that are operated as
modeling approach, solution method, and case study inclusion) so
an independent, non-profit entity, governed by a volunteer board
as to identify gaps in research and provide essential future re-
of directors. These PCFs provide accessible, affordable, quality
search directions.
Throughout this literature review, we identify and highlight primary medical, dental, and mental healthcare services to ev-
several research gaps in every section. Beyond the future research eryone, regardless of ability to pay.
possibilities identified in each section, we summarize overall po-
tential research directions in terms of (i) a computational per- A.1.2. Blood banks
spective (modeling and solution methodology) and (ii) different Blood banks are centers or caches that collect, process, store,
types of location problems in health systems. and distribute blood and blood products. Blood banks include
In conclusion, this review indicates that there is still a lot of several types of centers, such as blood transfusion providers, blood
room for the study of more realistic HCF location problems, and centers, blood stations, and mobile units.
the development of both new optimization models and solution
methods in HCF location planning. Moreover, the existing HCF A.1.3. Specialized services facilities
location problems can be adapted for other service industries or Specialized services units provide exceptional services, such as
may be extended to new general problems in location theory. We radiography, CT scan, MRI, radiation therapy, electro-en-
hope that the challenges presented in this article arouse interest in cephalogram (EEG), electro-cardiogram (ECG), etc., which are lo-
readers and encourage them to conduct research in healthcare cated either as independent units, or within hospitals or other
location modeling, an area that is indispensable from both in- health centers. For a more detailed review, we divided this class of
dustry and societal perspectives. HCFs into three subsections: organ transplant centers, detection
and prevention centers, and other specialized services units.
Acknowledgments A.1.3.1. Organ transplant centers. Organ transplant centers are

hospitals or other HCFs that have an established transplant pro-
The authors sincerely acknowledge the oversight and assis- gram including transplants for end-stage organ disease, such as
tance provided during the review process by the three anonymous marrow or cord blood transplants. These facilities, in cooperation
referees as well as Editor-in-Chief Prof. Stefan Nickel. Their gui- with blood centers, test patients, examine the possibility of
dance has helped us to significantly improve this survey paper. transplantation, and perform organ transplantation procedures.
They would also like to thank those who made comments on an
earlier version of this paper that was presented at 2015 MSOM A.1.3.2. Detection and prevention centers. Detection and prevention
annual conference. centers are established to save lives and improve the quality of life
by early diagnosis of serious medical conditions and prevention of
disease. They implement public vaccination programs or perform
Appendix A. (Supplementary details on HCF types) tests to detect potential cancerous cells. These facilities often
contain units for radiology, radiography, CT scan, MRI, etc.
A.1. Non-emergency facilities
A.1.3.3. Other specialized services facilities. Other specialized ser-
The class of non-emergency facilities contains a large number vices facilities are not in either of the above two classes of HCFs.
of HCFs which are divided into 10 subclasses as indicated in Fig. 3. These facilities provide some services, such as exercise stress test,
Clients for this group of facilities are not under emergency radiation therapy, EEG, ECG, etc.
A.1.4. Medical laboratories A.1.10. Long-term nursing care centers

Medical laboratories are units where various tests are done on This class of HFCs contains facilities, such as nursing homes,
clinical specimens in order to get information about the state of long-term rehabilitation centers, hospices, etc. According to ICHA
clients’ health for the diagnosis, treatment, and prevention of dis- 2000, long-term healthcare comprises ongoing health and nursing
ease. These are classified into two main types: hospital-based or care given to in-patients who need assistance because of chronic
stand-alone private (or public) laboratories. They can be clinical impairments, and a reduced degree of independence and activities
pathology, clinical microbiology, or clinical biochemistry of daily living. Long-term care is typically a mix of medical care
laboratories. (including nursing care) and social services. Long-term care in
institutions is of three types: inpatient long-term nursing care, day
A.1.5. Mobile healthcare units cases of long-term nursing care, and long-term home care.
Mobile healthcare units are often clinics on wheels that provide
medical, dental, and behavioral healthcare to patients who have A.2. Emergency facilities
difficulty reaching the other HCFs. Since developing countries
frequently face the dilemma of very restrictive budget limitations A.2.1. Permanent emergency facilities
for healthcare expenditures and a growing population [116], mo- Occasionally, due to illness, injury, or urgent medical condi-
bile healthcare units are potentially very cost-effective vehicles for tions, patients require immediate treatment. Such patients should
delivery of primary care and some other health services. Mobile be immediately transmitted to one of the emergency facilities by
units cannot provide all of the services available in a hospital, and ambulance (air or land). Indeed, emergency facilities location di-
they should be seen as supplemental in nature, providing services, rectly affects mortality and injury rates of an area’s inhabitants.
such as outpatient surgeries, dental services, and urgent primary Hence, determining the best location of emergency facilities is
care. crucial. We divided these facilities into four subcategories: emer-
gency off-site public access devices, emergency centers, trauma
A.1.6. Home healthcare centers centers, and ambulance stations.
Home healthcare centers are facilities and institutions that
dispatch nurses and paramedical technicians to provide home A.2.1.1. Emergency off-site public access devices. Off-site public ac-
healthcare services. According to the international classification cess devices (OPADs) are non-interactive or interactive medical
for health accounts (ICHA) 2000, home healthcare comprises devices designed to provide healthcare services in out-of-HCF
medical and paramedical services that are delivered to patients at environments. OPADs are able to receive individuals’ health status
home. Included in this category are obstetric services, home dia- information by connecting the sensors to the body and/or asking
lysis, home visit by a general practitioner, telematics services, etc. some queries, and provide appropriate diagnosis and prescription.
Furthermore, these facilities can be equipped with signature cap-
A.1.7. Rehabilitation centers ture devices in order to present documentation to the individual,
Rehabilitation centers are outpatient or residential HCFs pro- collect their signature, and empower them to check-in for their
viding rehabilitative care. According to ICHA 2000, rehabilitative scheduled appointments. In cases where individuals must make
care comprises medical and paramedical services delivered to co-payments, OPADs can also collect payment. In the provision of
patients during rehabilitation. Rehabilitative care includes services healthcare, OPAD placement programs could facilitate maximiza-
tion of coverage, improvement of services quality, and reduction in
where the emphasis lies on improving the functional levels of
response times and costs.
patients where the functional limitations are either due to a recent
A known subclass of non-interactive emergency OPADs are
illness or injury, or of a recurrent nature (regression or progres-
automated external defibrillators (AEDs), which are portable de-
sion). It requires frequent (daily to weekly) patient assessment,
vices to check the heart rhythm and send electric shocks to the
and review of the clinical course and treatment plan for a limited
heart in order to try to restore a normal rhythm. AEDs are used to
time period (several days to several months) until the condition is
treat sudden cardiac arrest ([222]).
stabilized or a pre-determined treatment course is completed.
OPADs could be of either non-emergency or emergency type.
Rehabilitation centers including centers for optometry, audio-
However, due to the co-characteristics of these two types, greater
metry, physiotherapy, occupational therapy, speech therapy, etc.
importance of emergency devices in determining optimal locations,
and that OPADs generally need to have the ability to provide emer-
A.1.8. Doctors' offices
gency services (e.g., to give medical guidelines, to contact with am-
Doctors’ offices including specialists’ and general practitioners’
bulances, or emergency centers, etc.), the emergency type of these
offices, dental and psychology offices, etc., are solely private.
HCFs could be more important in terms of location management.
Hence, services that doctors offer in their offices are often primary
care described in Section 3.1. However, for greater generality, A.2.1.2. Emergency centers. Generally, emergency centers are in-
doctors’ offices have been categorized into a separate section. In dependent centers, or embedded in hospitals or other healthcare
both urban and rural areas of developed countries as well as urban centers, but in a separate section from other units, which are also
areas of developing countries, the distance between patients and known as emergency departments, casualty rooms, or emergency
HCFs is a factor influencing a patient’s decision regarding where to rooms. These facilities provide medical care to unscheduled pa-
get medical services (e.g., a doctor’s office or a clinic). In contrast, tients requiring immediate and urgent medical attention, who
in rural areas of developing countries, due to lack of efficient show up or are brought by an ambulance. Due to the uncertain
transportation infrastructure, low incomes, or high travel costs, nature of patients’ arrivals, these centers have to provide a wide
distance is the decisive factor in deciding whether or not to use range of initial treatment to their clients. Emergency centers are
medical services [116]. open 24 hours a day and their staffing levels vary depending on
patient volume.
A.1.9. Drugstores
Drugstores are sometimes considered as a variety of HCFs that, A.2.1.3. Trauma centers. Trauma centers are hospitals equipped
after obtaining accreditation, provide medicines, dietary supple- and staffed to provide comprehensive specialized medical and
ments, cosmetics products, and medical devices for clients. nursing services to patients suffering from traumatic injuries.
Trauma centers typically have helicopter platforms to transfer Sainfort F, Pierskalla WP, editors. Operations Research and Health Care. New
patients, and require specialized and experienced multi-dis- York: Springer; 2004. p. 43–76.
[6] Li X, Zhao Z, Zhu X, Wyatt T. Covering models and optimization techniques
ciplinary treatment and specialized resources. When the number for emergency response facility location and planning: a review. Math
of trauma centers is limited, helicopters play an important role in Methods Oper Res 2011;74(3):281–310.
effective treatment of injuries. [7] Rais A, Viana A. Operations research in healthcare: a survey. Int Trans Oper
Res 2011;18(1):1–31.
[8] Wang F. Measurement, optimization, and impact of health care accessibility:
A.2.1.4. Ambulance stations. Ambulance stations are structures for a methodological review. Ann Assoc Am Geogr 2012;102(5):1104–12.
storage of ambulance vehicles and other medical supplies, and [9] Gutiérrez EV, Vidal CJ. Home health care logistics management problems: a
critical review of models and methods. Rev Fac Ing 2013;68:160–75.
they are responsible for ambulance dispatch (air or land). Ambu-
[10] Ingolfsson A. EMS planning and management. In: Zaric GS, editor. Operations
lances are equipped with highly trained medical personnel and Research and Health Care Policy. New York: Springer; 2013. p. 105–28.
equipment to revive patients which can take either of two forms: [11] Güneş ED, Nickel S. Location problems in healthcare. In: Laporte G, Nickel S,
Saldanha da Gama F, editors. Location science. Cham: Springer; 2015.
(i) providing care to patients outside of HCFs or (ii) transferring
p. 555–79.
sick or injured patients to or between HCFs. Since the response [12] Berman O, Krass D. Stochastic location models with congestion. In: Laporte
time is a critical factor in the reduction of mortality and morbidity G, Nickel S, Saldanha da Gama F, editors. Location science. Cham: Springer;
rates, determining optimal ambulance station locations, as well as 2015. p. 443–86.
[13] Hulshof PJ, Kortbeek N, Boucherie RJ, Hans EW, Bakker PJ. Taxonomic clas-
the optimal deployment (location or relocation) of ambulances in sification of planning decisions in health care: a structured review of the
the stations, is pivotal. state of the art in OR/MS. Health Syst 2012;1(2):129–75.
[14] Langabeer II JR, Helton J. Health care operations management: a systems
perspective. 2nd ed. Burlington: Jones & Bartlett Publishers; 2015.
A.2.2. Temporary emergency facilities [15] de Vries J, Huijsman R, de Vries J, Huijsman R. Supply chain management in
Temporary emergencies (natural disasters, military attacks, health services: an overview. Supply Chain Manag: Int J 2011;16(3):159–65.
infectious disease disasters, etc.) that rarely occur, lead to a big [16] Dobrzykowski D, Deilami VS, Hong P, Kim S-C. A structured analysis of op-
erations and supply chain management research in healthcare (1982–2011).
surge in the demand for medical supplies. In such situations, lo-
Int J Prod Econ 2014;147(2):514–30.
cation of HCFs directly affects mortality and injury rates, and thus [17] Wang Y, Wallace SW, Shen B, Choi T-M. Service supply chain management: a
the optimal locations of these HCFs play a vital role in risk miti- review of operational models. Eur J Oper Res 2015;247(3):685–98.
[18] Lemmens S, Decouttere C, Vandaele N, Bernuzzi M. A review of integrated
gation, coverage maximization, rapid distribution of resources,
supply chain network design models: key issues for vaccine supply chains.
waste avoidance, and reduction in response time. These facilities Chem Eng Res Des 2016;109:366–84.
are divided into two subcategories: temporary medical centers [19] Narayana SA, Pati RK, Vrat P. Research on management issues in the phar-
and points of dispensing. maceutical industry: a literature review. Int J Pharm Healthc Mark 2012;6
(4):351–75.
[20] Shah N. Pharmaceutical supply chains: key issues and strategies for opti-
A.2.2.1. Temporary medical centers. This class of HFCs contains fa- misation. Comput Chem Eng 2004;28(6):929–41.
cilities, such as hospitals, clinics, field hospitals, Red Crescent and/ [21] Thakur V, Ramesh A. Healthcare waste management research: a structured
analysis and review (2005–2014). Waste Manag Res 2015;33(10):855–70.
or Red Cross tents, etc. as providers of healthcare services to [22] Altay N, Green III WG. OR/MS research in disaster operations management.
people affected by disasters, which could be of a catastrophic Eur J Oper Res 2006;175(1):475–93.
nature. Determining the best locations of these HCFs plays a pi- [23] Galindo G, Batta R. Review of recent developments in OR/MS research in
disaster operations management. Eur J Oper Res 2013;230(2):201–11.
votal role in mitigating the number of serious injuries and [24] Hoyos MC, Morales RS, Akhavan-Tabatabaei R. OR models with stochastic
mortalities. components in disaster operations management: a literature survey. Comput
Casualty collection points (CCPs) are temporary centers, which Ind Eng 2015;82:183–97.
[25] Caunhye AM, Nie X, Pokharel S. Optimization models in emergency logistics:
are used for provision of first aid to accommodate a large number of
a literature review. Socio-Econ Plan Sci 2012;46(1):4–13.
people coming to them on foot, who are subsequently transferred [26] Anaya-Arenas AM, Renaud J, Ruiz A. Relief distribution networks: a sys-
to hospitals or shelters, if required. In pre-disaster planning phase, tematic review. Ann Oper Res 2014;223(1):53–79.
CCPs are initially established in public or private facilities, such as [27] Çelik M, Ergun Ö, Johnson B, Keskinocak P, Lorca Á, Pekgün P, Swann JL.
Humanitarian logistics. In: Mirchandani P, editor. New directions in infor-
colleges, schools, and public parks which are open, large enough, matics, optimization, logistics, and production. Maryland: INFORMS Tutorials
and relatively free from falling debris. Then, in catastrophic disaster in Operations Research; 2012. p. 18–49.
situations, they may be operationalized when existing centers are [28] Gutjahr WJ, Nolz PC. Multicriteria optimization in humanitarian aid. Eur J
Oper Res 2016;252(2):351–66.
destroyed or cannot satisfy the demand ([169]). [29] Özdamar L, Ertem MA. Models, solutions and enabling technologies in hu-
manitarian logistics. Eur J Oper Res 2015;244(1):55–65.
A.2.2.2. Points of dispensing. A point of dispensing (POD) is a mass [30] McLay LA. Discrete optimization models for homeland security and disaster
management. In: Aleman D, Thiele A, editors. The operations research re-
medication dispensing site for provision of medicine and medical volution. Maryland: INFORMS Tutorials in Operations Research; 2015.
supplies (i.e., vaccines, drugs, and therapeutics) to protect the p. 111–32.
general population from infectious disease disasters, which are [31] Wright PD, Liberatore MJ, Nydick RL. A survey of operations research models
and applications in homeland security. Interfaces 2006;36(6):514–29.
events in which a biological agent/disease involves a large portion
[32] Larson RC. Decision models for emergency response planning. In: Kamien D,
of people, such as bioterrorist attacks, pandemics, or an outbreak editor. The McGraw-Hill handbook of homeland security. Columbus: Ohio:
of an emerging infectious disease (further study of infectious McGraw-Hill; 2005. p. 911–27.
disease disasters can be found in [216]). [33] Simpson N, Hancock P. Fifty years of operational research and emergency
response. J Oper Res Soc 2009;60(1):S126–39.
[34] Başar A, Çatay B, Ünlüyurt T. A taxonomy for emergency service station lo-
cation problem. Optim Lett 2012;6(6):1147–60.
References [35] Goldberg JB. Operations research models for the deployment of emergency
services vehicles. EMS Manag J 2004;1(1):20–39.
[36] Snyder LV, Atan Z, Peng P, Rong Y, Schmitt AJ, Sinsoysal B. OR/MS models for
[1] Weber A. Theory of the location of industries [translated by CJ Friedrich from supply chain disruptions: a review. IIE Trans 2016;48(2):89–109.
Weber’s 1909 book]. Chicago: The University of Chicago Press; 1929. [37] Drezner Z, Hamacher HW, editors. Facility location: applications and theory.
[2] Gould PR, Leinbach TR. Approach to the geographic assignment of hospital Berlin: Springer; 2002.
services. Tijdschr voor Econ En Soc Geogr 1966;57(5):203–6. [38] Eiselt HA, Marianov V, editors. Foundations of location analysis. New York:
[3] Rahman S-u, Smith DK. Use of location-allocation models in health service Springer; 2011.
development planning in developing nations. Eur J Oper Res 2000;123 [39] Daskin MS. Network and discrete location: models, algorithms, and appli-
(3):437–52. cations. 2nd ed. New York: John Wiley & Sons; 2013.
[4] Brotcorne L, Laporte G, Semet F. Ambulance location and relocation models. [40] Revelle CS, Eiselt HA, Daskin MS. A bibliography for some fundamental
Eur J Oper Res 2003;147(3):451–63. problem categories in discrete location science. Eur J Oper Res 2008;184
[5] Daskin MS, Dean LK. Location of health care facilities. In: Brandeau ML, (3):817–48.
[41] Laporte G, Nickel S, Saldanha da Gama F, editors. Location science Cham, 2009;193(3):768–77.
Springer; 2015. [75] Mestre AM, Oliveira MD, Barbosa-Póvoa A. Organizing hospitals into net-
[42] Daskin MS, Coullard CR, Shen Z-JM. An inventory-location model: formula- works: a hierarchical and multiservice model to define location, supply and
tion, solution algorithm and computational results. Ann Oper Res 2002;110 referrals in planned hospital systems. OR Spectr 2012;34(2):319–48.
(1-4):83–106. [76] Shariff S, Moin NH, Omar M. Location allocation modeling for healthcare
[43] Farahani RZ, Rashidi Bajgan H, Fahimnia B, Kaviani M. Location-inventory facility planning in Malaysia. Comput Ind Eng 2012;62(4):1000–10.
problem in supply chains: a modelling review. Int J Prod Res 2015;53 [77] Güneş E, Yaman H, Çekyay B, Verter V. Matching patient and physician
(12):3769–88. preferences in designing a primary care facility network. J Oper Res Soc
[44] Prodhon C, Prins C. A survey of recent research on location-routing pro- 2012;65(4):483–96.
blems. Eur J Oper Res 2014;238(1):1–17. [78] Cocking C, Flessa S, Reinelt G. Improving access to health facilities in Nouna
[45] Ahmadi-Javid A, Seddighi AH. A location-routing problem with disruption district, Burkina Faso. Socio-Econ Plan Sci 2012;46(2):164–72.
risk. Transp Res Part E: Logist Transp Rev 2013;53:63–82. [79] Mitropoulos P, Mitropoulos I, Giannikos I. Combining DEA with location
[46] Ahmadi Javid A, Azad N. Incorporating location, routing and inventory de- analysis for the effective consolidation of services in the health sector.
cisions in supply chain network design. Transp Res Part E: Logist Transp Rev Comput Oper Res 2013;40(9):2241–50.
2010;46(5):582–97. [80] Kim D-G, Kim Y-D. A Lagrangian heuristic algorithm for a public healthcare
[47] Ahmadi-Javid A, Seddighi AH. A location-routing-inventory model for de- facility location problem. Ann Oper Res 2013;206(1):221–40.
signing multisource distribution networks. Eng Optim 2012;44(6):637–56. [81] Ghaderi A, Jabalameli MS. Modeling the budget-constrained dynamic un-
[48] Ahmadi-Javid A, Ghandali R. An efficient optimization procedure for de- capacitated facility location–network design problem and solving it via two
signing a capacitated distribution network with price-sensitive demand. efficient heuristics: a case study of health care. Math Comput Model 2013;57
Optim Eng 2014;15(3):801–17. (3):382–400.
[49] Fattahi M, Mahootchi M, Govindan K, Husseini SMM. Dynamic supply chain [82] Beheshtifar S, Alimoahmmadi A. A multiobjective optimization approach for
network design with capacity planning and multi-period pricing. Transp Res location‐allocation of clinics. Int Trans Oper Res 2015;22(2):313–28.
Part E: Logist Transp Rev 2015;81:169–202. [83] Rahmaniani R, Rahmaniani G, Jabbarzadeh A. Variable neighborhood search
[50] Plastria F, Vanhaverbeke L. Maximal covering location problem with price based evolutionary algorithm and several approximations for balanced lo-
decision for revenue maximization in a competitive environment. OR Spectr cation–allocation design problem. Int J Adv Manuf Technol 2014;72(1-
2009;31(3):555–71. 4):145–59.
[51] Ahmadi-Javid A, Hoseinpour P. A location-inventory-pricing model in a [84] Mestre AM, Oliveira MD, Barbosa-Póvoa AP. Location–allocation approaches
supply chain distribution network with price-sensitive demands and infor hospital network planning under uncertainty. Eur J Oper Res 2015;240
ventory-capacity constraints. Transp Res Part E: Logist Transp Rev (3):791–806.
2015;82:238–55. [85] Stummer C, Doerner K, Focke A, Heidenberger K. Determining location and
[52] Ahmadi-Javid A, Hoseinpour P. Incorporating location, inventory and price size of medical departments in a hospital network: a multiobjective decision
decisions into a supply chain distribution network design problem. Comput support approach. Health Care Manag Sci 2004;7(1):63–71.
Oper Res 2015;56:110–9. [86] Smith HK, Harper PR, Potts CN. Bicriteria efficiency/equity hierarchical lo-
[53] Daskin MS. What you should know about location modeling. Nav Res Logist cation models for public service application. J Oper Res Soc 2012;64(4):500–
2008;55(4):283–94. 12.
[54] Kariv O, Hakimi SL. An algorithmic approach to network location problems. [87] Graber-Naidich A, Carter MW, Verter V. Primary care network development:
II: the p-medians. SIAM J Appl Math 1979;37(3):539–60. the regulator's perspective. J Oper Res Soc 2015;66(9):1519–32.
[55] Megiddo N, Supowit KJ. On the complexity of some common geometric lo- [88] Shishebori D, Babadi AY. Robust and reliable medical services network de-
cation problems. SIAM J Comput 1984;13(1):182–96. sign under uncertain environment and system disruptions. Transp Res Part
[56] Kuby MJ. Programming models for facility dispersion: the p‐dispersion and E: Logist Transp Rev 2015;77:268–88.
maxisum dispersion problems. Geogr Anal 1987;19(4):315–29. [89] Marianov V, Ríos M, Taborga P. Finding locations for public service centres
[57] Huang R, Kim S, Menezes MB. Facility location for large-scale emergencies. that compete with private centres: effects of congestion. Pap Reg Sci 2004;83
Ann Oper Res 2010;181(1):271–86. (4):631–48.
[58] Mete HO, Zabinsky ZB. Stochastic optimization of medical supply location [90] Marianov V, Taborga P. Optimal location of public health centres which
and distribution in disaster management. Int J Prod Econ 2010;126(1):76–84. provide free and paid services. J Oper Res Soc 2001;52(4):391–400.
[59] Paul JA, Hariharan G. Location-allocation planning of stockpiles for effective [91] Ares JN, De Vries H, Huisman D. A column generation approach for locating
disaster mitigation. Ann Oper Res 2012;196(1):469–90. roadside clinics in Africa based on effectiveness and equity. Eur J Oper Res
[60] Daskin MS, Snyder LV, Berger RT. Facility location in supply chain design. In: 2016;254(3):1002–16.
Langevin A, Riopel D, editors. Logistics systems: design and optimization. [92] Şahin G, Süral H, Meral S. Locational analysis for regionalization of Turkish
New York: Springer; 2005. p. 39–65. Red Crescent blood services. Comput Oper Res 2007;34(3):692–704.
[61] Melo MT, Nickel S, Saldanha-da-Gama F. Facility location and supply chain [93] Cetin E, Sarul LS. A blood bank location model: a multiobjective approach.
management–a review. Eur J Oper Res 2009;196(2):401–12. Eur J Pure Appl Math 2009;2(1):112–24.
[62] Rubenfire A. Growing hospital systems stretch supply chains. Mod Healthc [94] Elalouf A, Hovav S, Tsadikovich D, Yedidsion L. Minimizing operational costs
2016;46(5):20–2. by restructuring the blood sample collection chain. Oper Res Health Care
[63] Chartier Y, Emmanuel J, Pieper U, Prüss A, Rushbrook P, Stringer R, Townend 2015;7:81–93.
W, Wilburn S, Zghondi R, editors. Safe management of wastes from health- [95] Zahiri B, Torabi S, Mousazadeh M, Mansouri S. Blood collection management:
care activities. 2nd ed. Geneva: World Health Organization; 2014. methodology and application. Appl Math Model 2015;39(23-24):7680–96.
[64] Ghiani G, Laganà D, Manni E, Musmanno R, Vigo D. Operations research in [96] Fahimnia B, Jabbarzadeh A, Ghavamifar A, Bell M. Supply chain design for
solid waste management: a survey of strategic and tactical issues. Comput efficient and effective blood supply in disasters. Int J Prod Econ 2015.
Oper Res 2014;44:22–32. [97] Jabbarzadeh A, Fahimnia B, Seuring S. Dynamic supply chain network design
[65] Burkey ML, Bhadury J, Eiselt HA. A location-based comparison of health care for the supply of blood in disasters: a robust model with real world appli-
services in four US states with efficiency and equity. Socio-Econ Plan Sci cation. Transp Res Part E: Logist Transp Rev 2014;70(1):225–44.
2012;46(2):157–63. [98] Bruni ME, Conforti D, Sicilia N, Trotta S. A new organ transplantation loca-
[66] Fo A, Mota I. Optimization models in the location of healthcare facilities: a tion–allocation policy: a case study of Italy. Health Care Manag Sci 2006;9
real case in Brazil. J Appl Oper Res 2012;4(1):37–50. (2):125–42.
[67] Chu SC, Chu L. A modeling framework for hospital location and service al- [99] Beliën J, De Boeck L, Colpaert J, Devesse S, Van den Bossche F. Optimizing the
location. Int Trans Oper Res 2000;7(6):539–68. facility location design of organ transplant centers. Decis Support Syst
[68] Galvao RD, Acosta Espejo LG, Boffey B. A hierarchical model for the location 2013;54(4):1568–79.
of perinatal facilities in the municipality of Rio de Janeiro. Eur J Oper Res [100] Zahiri B, Tavakkoli-Moghaddam R, Pishvaee MS. A robust possibilistic pro-
2002;138(3):495–517. gramming approach to multi-period location-allocation of organ transplant
[69] Mitropoulos P, Mitropoulos I, Giannikos I, Sissouras A. A biobjective model centers under uncertainty. Comput Ind Eng 2014;74(1):139–48.
for the locational planning of hospitals and health centers. Health Care [101] Zahiri B, Tavakkoli-Moghaddam R, Mohammadi M, Jula P. Multi-objective
Manag Sci 2006;9(2):171–9. design of an organ transplant network under uncertainty. Transp Res Part E:
[70] Oliveira MD, Bevan G. Modelling the redistribution of hospital supply to Logist Transp Rev 2014;72:101–24.
achieve equity taking account of patient's behaviour. Health Care Manag Sci [102] Verter V, Lapierre SD. Location of preventive health care facilities. Ann Oper
2006;9(1):19–30. Res 2002;110(1-4):123–32.
[71] Ndiaye M, Alfares H. Modeling health care facility location for moving po- [103] Zhang Y, Berman O, Verter V. Incorporating congestion in preventive
pulation groups. Comput Oper Res 2008;35(7):2154–61. healthcare facility network design. Eur J Oper Res 2009;198(3):922–35.
[72] Griffin PM, Scherrer CR, Swann JL. Optimization of community health center [104] Zhang Y, Berman O, Marcotte P, Verter V. A bilevel model for preventive
locations and service offerings with statistical need estimation. IIE Trans healthcare facility network design with congestion. IIE Trans 2010;42
2008;40(9):880–92. (12):865–80.
[73] Ratick SJ, Osleeb JP, Hozumi D. Application and extension of the Moore and [105] Gu W, Wang X, McGregor SE. Optimization of preventive health care facility
ReVelle hierarchical maximal covering model. Socio-Econ Plan Sci 2009;43 locations. Int J Health Geogr 2010;9(1):9–17.
(2):92–101. [106] Vidyarthi N, Kuzgunkaya O. The impact of directed choice on the design of
[74] Smith HK, Harper PR, Potts CN, Thyle A. Planning sustainable community preventive healthcare facility network under congestion. Health Care Manag
health schemes in rural areas of developing countries. Eur J Oper Res Sci 2015;18(4):459–74.
[107] Davari S, Kilic K, Ertek G. Fuzzy bi-objective preventive health care network servers. IIE Trans 2009;41(8):730–41.
design. Health Care Manag Sci 2015;18(3):303–17. [143] Rajagopalan HK, Saydam C. A minimum expected response model: for-
[108] Aboolian R, Berman O, Verter V. Maximal accessibility network design in the mulation, heuristic solution, and application. Socio-Econ Plan Sci 2009;43
public sector. Transp Sci 2015;50(1):336–47. (4):253–62.
[109] Aboolian R, Berman O, Krass D. Profit maximizing distributed service system [144] Coskun N, Erol R. An optimization model for locating and sizing emergency
design with congestion and elastic demand. Transp Sci 2012;46(2):247–61. medical service stations. J Med Syst 2010;34(1):43–9.
[110] Zhang Y, Berman O, Verter V. The impact of client choice on preventive [145] Sorensen P, Church R. Integrating expected coverage and local reliability for
healthcare facility network design. OR Spectr 2012;34(2):349–70. emergency medical services location problems. Socio-Econ Plan Sci 2010;44
[111] Mahar S, Bretthauer KM, Salzarulo PA. Locating specialized service capacity in (1):8–18.
a multi-hospital network. Eur J Oper Res 2011;212(3):596–605. [146] Schmid V, Doerner KF. Ambulance location and relocation problems with
[112] Benneyan JC, Musdal H, Ceyhan ME, Shiner B, Watts BV. Specialty care single time-dependent travel times. Eur J Oper Res 2010;207(3):1293–303.
and multi-period location–allocation models within the Veterans Health [147] Noyan N. Alternate risk measures for emergency medical service system
Administration. Socio-Econ Plan Sci 2012;46(2):136–48. design. Ann Oper Res 2010;181(1):559–89.
[113] Syam SS, Côté MJ. A comprehensive location-allocation method for specia- [148] Lim CS, Mamat R, Braunl T. Impact of ambulance dispatch policies on per-
lized healthcare services. Oper Res Health Care 2012;1(4):73–83. formance of emergency medical services. IEEE Trans Intell Transp Syst
[114] Guo M, Li B, Zhang Z, Wu S, Song J. Efficiency evaluation for allocating 2011;12(2):624–32.
community-based health services. Comput Ind Eng 2013;65(3):395–401. [149] Chanta S, Mayorga ME, McLay LA. Improving emergency service in rural
[115] Saveh-Shemshaki F, Shechter S, Tang P, Isaac-Renton J. Optimizing locations areas: a bi-objective covering location model for EMS systems. Ann Oper Res
and capacities of new tuberculosis testing laboratories. IIE Trans Healthc Syst 2011;221(1):133–59.
Eng 2012;2(4):248–58. [150] Schmid V. Solving the dynamic ambulance relocation and dispatching pro-
[116] Doerner K, Focke A, Gutjahr WJ. Multicriteria tour planning for mobile blem using approximate dynamic programming. Eur J Oper Res 2012;219
healthcare facilities in a developing country. Eur J Oper Res 2007;179 (3):611–21.
(3):1078–96. [151] Shariat-Mohaymany A, Babaei M, Moadi S, Amiripour SM. Linear upper-
[117] Kim D-G, Kim Y-D. A branch and bound algorithm for determining locations bound unavailability set covering models for locating ambulances: applica-
of long-term care facilities. Eur J Oper Res 2010;206(1):168–77. tion to Tehran rural roads. Eur J Oper Res 2012;221(1):263–72.
[118] Cardoso T, Oliveira MD, Barbosa-Póvoa A, Nickel S. An integrated approach [152] Saydam C, Rajagopalan HK, Sharer E, Lawrimore-Belanger K. The dynamic
for planning a long-term care network with uncertainty, strategic policy and redeployment coverage location model. Health Syst 2013;2(2):103–19.
equity considerations. Eur J Oper Res 2015;247(1):321–34. [153] Toro-Díaz H, Mayorga ME, Chanta S, McLay LA. Joint location and dispatching
[119] Bonnet B, Dessavre DG, Kraus K, Ramirez-Marquez JE. Optimal placement of decisions for Emergency Medical Services. Comput Ind Eng 2013;64(4):917–
public-access AEDs in urban environments. Comput Ind Eng 2015;90:269–80. 28.
[120] Chan TC, Demirta D, Kwon R. Optimizing the deployment of public access [154] Zhang Z-H, Jiang H. A robust counterpart approach to the bi-objective
defibrillators. Manag Sci 2016. http://dx.doi.org/10.1287/mnsc.2015.2312. emergency medical service design problem. Appl Math Model 2014;38
[121] Chan TC, Li H, Lebovic G, Tang SK, Chan JY, Cheng HC, Morrison LJ, Brooks SC. (3):1033–40.
Identifying locations for public access defibrillators using mathematical op- [155] Erkut E, Ingolfsson A, Sim T, Erdoğan G. Computational comparison of five
timization. Circulation. 2013;127(17):1801–9. maximal covering models for locating ambulances. Geogr Anal 2009;41
[122] Myers D, Mohite M. Locating automated external defibrillators in a university (1):43–65.
community. J Oper Res Soc 2009;60(6):869–72. [156] Iannoni AP, Morabito R, Saydam C. An optimization approach for ambulance
[123] Siddiq AA, Brooks SC, Chan TC. Modeling the impact of public access defi- location and the districting of the response segments on highways. Eur J
brillator range on public location cardiac arrest coverage. Resuscitation. Oper Res 2009;195(2):528–42.
2013;84(7):904–9. [157] Erkut E, Ingolfsson A, Erdoğan G. Ambulance location for maximum survival.
[124] Sun CL, Demirta D, Brooks SC, Morrison LJ, Chan TC. Overcoming spatial and Nav Res Logist 2008;55(1):42–58.
temporal barriers to public access defibrillators via optimization. J Am Coll [158] van den Berg PL, Aardal K. Time-dependent MEXCLP with start-up and re-
Cardiol 2016;68(8):836–45. location cost. Eur J Oper Res 2015;242(2):383–9.
[125] Silva F, Serra D. Locating emergency services with different priorities: the [159] Rajagopalan HK, Saydam C, Setzler H, Sharer E. Ambulance deployment and
priority queuing covering location problem. J Oper Res Soc 2008;59(9):1229– shift scheduling: an integrated approach. J Serv Sci Manag 2011;4(1):66–78.
38. [160] Aringhieri R, Carello G, Morale D. Supporting decision making to improve the
[126] Branas CC, MacKenzie EJ, ReVelle CS. A trauma resource allocation model for performance of an Italian emergency medical service. Ann Oper Res
ambulances and hospitals. Health Serv Res 2000;35(2):489. 2016;236(1):131–48.
[127] Branas CC, Revelle CS. An iterative switching heuristic to locate hospitals and [161] Iannoni AP, Morabito R, Saydam C. Optimizing large-scale emergency med-
helicopters. Socio-Econ Plan Sci 2001;35(1):11–30. ical system operations on highways using the hypercube queuing model.
[128] Côté MJ, Syam SS, Vogel WB, Cowper DC. A mixed integer programming Socio-Econ Plan Sci 2011;45(3):105–17.
model to locate traumatic brain injury treatment units in the Department of [162] Gendreau M, Laporte G, Semet F. The maximal expected coverage relocation
Veterans Affairs: a case study. Health Care Manag Sci 2007;10(3):253–67. problem for emergency vehicles. J Oper Res Soc 2006;57(1):22–8.
[129] Syam SS, Côté MJ. A location–allocation model for service providers with [163] Naoum-Sawaya J, Elhedhli S. A stochastic optimization model for real-time
application to not-for-profit health care organizations. Omega 2010;38 ambulance redeployment. Comput Oper Res 2013;40(8):1972–8.
(3):157–66. [164] Zhang Z-H, Li K. A novel probabilistic formulation for locating and sizing
[130] Erdemir ET, Batta R, Rogerson PA, Blatt A, Flanigan M. Joint ground and air emergency medical service stations. Ann Oper Res 2015;229(1):813–35.
emergency medical services coverage models: a greedy heuristic solution [165] Zhi J, Keskin BB, Melouk SH. A multi-period dynamic location planning
approach. Eur J Oper Res 2010;207(2):736–49. model for emergency response. IIE Trans Healthc Syst Eng 2015;5(4):211–24.
[131] Cho S-H, Jang H, Lee T, Turner J. Simultaneous location of trauma centers and [166] Jia H, Ordóñez F, Dessouky M. A modeling framework for facility location of
helicopters for emergency medical service planning. Oper Res 2013;62 medical services for large-scale emergencies. IIE Trans 2007;39(1):41–55.
(4):751–71. [167] Paul JA, Batta R. Models for hospital location and capacity allocation for an
[132] Gendreau M, Laporte G, Semet F. A dynamic model and parallel tabu search area prone to natural disasters. Int J Oper Res 2008;3(5):473–96.
heuristic for real-time ambulance relocation. Parallel Comput 2001;27 [168] Apte A, Heidtke C, Salmerón J. Casualty collection points optimization: a
(12):1641–53. study for the District of Columbia. Interfaces 2014;45(2):149–65.
[133] Harewood S. Emergency ambulance deployment in Barbados: a multi-ob- [169] Drezner T. Location of casualty collection points. Environ Plan C: Gov Policy
jective approach. J Oper Res Soc 2002;53(2):185–92. 2004;22(6):899–912.
[134] Beraldi P, Bruni ME, Conforti D. Designing robust emergency medical service [170] Drezner T, Drezner Z, Salhi S. A multi-objective heuristic approach for the
via stochastic programming. Eur J Oper Res 2004;158(1):183–93. casualty collection points location problem. J Oper Res Soc 2006;57(6):727–
[135] Tavakoli A, Lightner C. Implementing a mathematical model for locating EMS 34.
vehicles in Fayetteville, NC. Comput Oper Res 2004;31(9):1549–63. [171] Lee EK, Chen C-H, Pietz F, Benecke B. Modeling and optimizing the public-
[136] Doerner KF, Gutjahr WJ, Hartl RF, Karall M, Reimann M. Heuristic solution of health infrastructure for emergency response. Interfaces 2009;39(5):476–90.
an extended double-coverage ambulance location problem for Austria. Cent [172] Murali P, Ordóñez F, Dessouky MM. Facility location under demand un-
Eur J Oper Res 2005;13(4):325–40. certainty: response to a large-scale bio-terror attack. Socio-Econ Plann Sci
[137] Alsalloum OI, Rand GK. Extensions to emergency vehicle location models. 2012;46(1):78–87.
Comput Oper Res 2006;33(9):2725–43. [173] Lee EK, Pietz F, Benecke B, Mason J, Burel G. Advancing public health and
[138] Andersson T, Värbrand P. Decision support tools for ambulance dispatch and medical preparedness with operations research. Interfaces 2013;43(1):79–98.
relocation. J Oper Res Soc 2007;58(2):195–201. [174] Ramirez‐Nafarrate A, Lyon JD, Fowler JW, Araz OM. Point‐of‐dispensing lo-
[139] Ingolfsson A, Budge S, Erkut E. Optimal ambulance location with random cation and capacity optimization via a decision support system. Prod Oper
delays and travel times. Health Care Manag Sci 2008;11(3):262–74. Manag 2015;24(8):1311–28.
[140] Rajagopalan HK, Saydam C, Xiao J. A multiperiod set covering location model [175] Baray J, Cliquet G. Optimizing locations through a maximum covering/p-
for dynamic redeployment of ambulances. Comput Oper Res 2008;35 median hierarchical model: Maternity hospitals in France. J Bus Res 2013;66
(3):814–26. (1):127–32.
[141] Beraldi P, Bruni ME. A probabilistic model applied to emergency service [176] Galvão RD, Acosta Espejo LG, Boffey B, Yates D. Load balancing and capacity
vehicle location. Eur J Oper Res 2009;196(1):323–31. constraints in a hierarchical location model. Eur J Oper Res 2006;172(2):631–
[142] McLay LA. A maximum expected covering location model with two types of 46.
[177] Kim D-G, Kim Y-D, Lee T. Heuristics for locating two types of public health- [200] Warden CR, Daya M, LeGrady LA. Using geographic information systems to
care facilities. Ind Eng Manag Syst 2012;11(2):202–14. evaluate cardiac arrest survival. Prehosp Emerg Care 2007;11(1):19–24.
[178] Cromley EK, McLafferty SL. GIS and public health. 2nd ed. New York: Guilford [201] Brooks SC, Hsu JH, Tang SK, Jeyakumar R, Chan TC. Determining risk for out-
Press; 2012. of-hospital cardiac arrest by location type in a Canadian urban setting to
[179] Aboueljinane L, Sahin E, Jemai Z. A review on simulation models applied to guide future public access defibrillator placement. Ann Emerg Med 2013;61
emergency medical service operations. Comput Ind Eng 2013;66(4):734–50. (5) 530-8. e2.
[180] Harper PR, Shahani A, Gallagher J, Bowie C. Planning health services with [202] Hogan K, Revelle C. Concepts and applications of backup coverage. Manag Sci
explicit geographical considerations: a stochastic location–allocation ap- 1986;32(11):1434–44.
proach. Omega 2005;33(2):141–52. [203] Chaiken JM. Transfer of emergency service deployment models to operating
[181] Menditto A, Patriarca M, Magnusson B. Understanding the meaning of ac- agencies. Manag Sci 1978;24(7):719–31.
curacy, trueness and precision. Accrédit Qual Assur 2007;12(1):45–7. [204] Chiyoshi F, Iannoni AP, Morabito R. A tutorial on hypercube queueing models
[182] Starfield B. Primary care: balancing health needs, services, and technology. and some practical applications in emergency service systems. Pesqui Oper
New York: Oxford University Press; 1998. 2011;31(2):271–99.
[183] Walters DJ, Toombs M, Rabuka LA. Strengthening the foundation: the role of [205] Larson RC. Decision models for emergency response planning. In: Herrmann
the physician in primary health care in Canada. Canadian Medical Association J, editor. The McGraw-Hill handbook of homeland security. Columbus:
Journal 1994;150(6):839–47. McGraw-Hill; 2005. p. 911–27.
[184] Society, the Individual, and Medicine (SIM). Primary care: definitions and [206] Daskin MS. A maximum expected covering location model: formulation,
historical developments. Ottawa: University of Ottawa Medical School, Re- properties and heuristic solution. Transp Sci 1983;17(1):48–70.
trieved form 〈http://wwwmeduottawaca/sim/data/Primary_Carehtm〉 2016, [207] Batta R, Dolan JM, Krishnamurthy NN. The maximal expected covering lo-
June. cation problem: revisited. Transp Sci 1989;23(4):277–87.
[185] World Health Organization (WHO). Declaration of Alma-Ata. International [208] Galvao RD, Morabito R. Emergency service systems: the use of the hypercube
Conference on Primary Health Care, Alma-Ata, September, 1978. Retrieved queueing model in the solution of probabilistic location problems. Int Trans
form 〈http://www.euro.who.int/en/publications/policy-documents/declara Oper Res 2008;15(5):525–49.
tion-of-alma-ata,-1978〉 2016, June. [209] Larson RC. Approximating the performance of urban emergency service
[186] Şahin G, Süral H. A review of hierarchical facility location models. Comput systems. Oper Res 1975;23(5):845–68.
Oper Res 2007;34(8):2310–31. [210] Larson RC. Hypercube queueing model. In: Gass S,I, Fu M,C, editors. En-
[187] Hodgson M. An hierarchical location-allocation model for primary health cyclopedia of operations research and management science. New York:
care delivery in a developing area. Soc. Sci Med 1988;26(1):153–61. Springer; 2013.
[188] Prastacos GP. Blood inventory management: an overview of theory and [211] Bélanger V, Kergosien Y, Ruiz A, Soriano P. An empirical comparison of re-
practice. Manag Sci 1984;30(7):777–800. location strategies in real-time ambulance fleet management. Comput Ind
[189] Beliën J, Forcé H. Supply chain management of blood products: a literature Eng 2016;94(1):216–29.
review. Eur J Oper Res 2012;217(1):1–16. [212] Alanis R, Ingolfsson A, Kolfal B. A Markov chain model for an EMS system
[190] Kim WR, Biggins SW, Kremers WK, Wiesner RH, Kamath PS, Benson JT, Ed- with repositioning. Prod Oper Manag 2013;22(1):216–31.
wards E, Therneau TM. Hyponatremia and mortality among patients on the [213] Budge S, Ingolfsson A, Zerom D. Empirical analysis of ambulance travel times:
liver-transplant waiting list. N. Engl J Med 2008;359(10):1018–26. the case of Calgary emergency medical services. Manag Sci 2010;56(4):716–
[191] Zenios SA, Chertow GM, Wein LM. Dynamic allocation of kidneys to candi- 23.
dates on the transplant waiting list. Oper Res 2000;48(4):549–69. [214] Haghani A, Yang S. Real-time emergency response fleet deployment: Con-
[192] Alagoz O, Maillart LM, Schaefer AJ, Roberts MS. Determining the acceptance cepts, systems, simulation & case studies. In: Zeimpekis V, Tarantilis CD,
of cadaveric livers using an implicit model of the waiting list. Oper Res Giaglis GM, Minis I, editors. Dynamic fleet management. New York: Springer;
2007;55(1):24–36. 2007. p. 133–62.
[193] Lucey MR, Brown KA, Everson GT, Fung JJ, Gish R, Keeffe EB, Kneteman NM, [215] Oran A, Tan KC, Ooi BH, Sim M, Jaillet P. Location and routing models for
Lake JR, Martin P, McDiarmid SV. Minimal criteria for placement of adults on emergency response plans with priorities. In: Aschenbruck N, Martini P,
the liver transplant waiting list: a report of a national conference organized Meier M, Tölle J, editors. Future security. Berlin: Springer; 2012. p. 129–40.
by the American Society of transplant physicians and the american associa- [216] Rebmann T. Infectious disease disasters: bioterrorism, emerging infections,
tion for the study of liver diseases. Liver Transplant Surg 1997;3(6):628–37. and pandemics. In: Grota P, editor. APIC text of infection control and epi-
[194] Cohon JL. Multiobjective programming and planning. New York: Dover demiology. 4th ed. Arlington: The Association for Professionals in Infection
Publications; 2004. Control and Epidemiology (APIC); 2014 p. 120-1-22.
[195] Caffrey SL, Willoughby PJ, Pepe PE, Becker LB. Public use of automated ex- [217] Lee EK, Maheshwary S, Mason J, Glisson W. Decision support system for mass
ternal defibrillators. New Engl J Med 2002;347(16):1242–7. dispensing of medications for infectious disease outbreaks and bioterrorist
[196] Dao THD, Zhou Y, Thill J-C, Delmelle E. Spatio-temporal location modeling in attacks. Ann Oper Res 2006;148(1):25–53.
a 3D indoor environment: the case of AEDs as emergency medical devices. [218] Herrmann JW. Disseminating emergency preparedness planning models as
Int J Geogr Inf Sci 2012;26(3):469–94. automatically generated custom spreadsheets. Interfaces 2008;38(4):263–70.
[197] Folke F, Gislason GH, Lippert FK, Nielsen SL, Weeke P, Hansen ML, Fosbøl EL, [219] Richter A, Khan S. Pilot model: judging alternate modes of dispensing pro-
Andersen SS, Rasmussen S, Schramm TK. Differences between out-of-hos- phylaxis in Los Angeles county. Interfaces 2009;39(3):228–40.
pital cardiac arrest in residential and public locations and implications for [220] Darling KT. Animals visiting healthcare facilities. In: Grota P, editor. APIC text
public-access defibrillation. Circulation 2010;122(6):623–30. of infection control and epidemiology. 4th ed. Arlington: The Association for
[198] Folke F, Lippert FK, Nielsen SL, Gislason GH, Hansen ML, Schramm TK, Professionals in Infection Control and Epidemiology (APIC); 2014 p. 122-1-8.
Sørensen R, Fosbøl EL, Andersen SS, Rasmussen S. Location of cardiac arrest [221] Royse D. Who needs beds? New ambulatory centers offer everything except
in a city center strategic placement of automated external defibrillators in inpatient care. Mod Healthc 2015;45(37):20–2.
public locations. Circulation 2009;120(6):510–7. [222] National Heart, Lung, and Blood Institute (NHLBI). What is an automated
[199] Lerner EB, Fairbanks RJ, Shah MN. Identification of out‐of‐hospital cardiac external defibrillator? Maryland: National Institutes of Health; 2011. Re-
arrest clusters using a geographic information system. Acad Emerg Med trieved from 〈https://www.nhlbi.nih.gov/health/health-topics/topics/aed〉
2005;12(1):81–4. 2016, June.

Ahmadi-Javid, Seyedi, Syam - 2017 - A Survey of Healthcare Facility Location

Uploaded by

Copyright:

Available Formats

Ahmadi-Javid, Seyedi, Syam - 2017 - A Survey of Healthcare Facility Location

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ahmadi-Javid, Seyedi, Syam - 2017 - A Survey of Healthcare Facility Location

Uploaded by

Copyright:

Available Formats

What types of healthcare facilities are discussed in the document?

What types of healthcare facilities are discussed in the document?

What framework is used to classify the different healthcare facilities?

What framework is used to classify the different healthcare facilities?

Computers & Operations Research 79 (2017) 223–263

Contents lists available at ScienceDirect

Computers & Operations Research

A survey of healthcare facility location

4.9. Combinations of several types of HCFs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242

1. Introduction keen interest from the operations research (OR) community.

Fig. 1. A classiﬁcation of discrete location problems.

2.2.1. p-median location problems I The set of demand points.

xj 1, if a facility is established at candidate location j∈J ;

∑ yij =1, i∈I yij ∈{ 0, 1}, i∈I , j∈J (26)

speciﬁes the total number of facilities to be established. Con-

Fig. 2. The major players in health systems.

Combination of several types of HCFs Combination of 1.1 and 1.3.3* [175,176]

Consideration of Y Considering uncertainties ✓ ✓ ✓

Multi-period setting S Static ✓ ✓ ✓

D-1 Multi-period short-term decisions (e.g., am- ✓ ✓ ✓

Particular input/ P1 Demand ✓ ✓ ✓

P5-1 Fixed cost ✓ ✓ ✓

Objective function O1 Minimize total number of facilities ✓ ✓ ✓

Decision variable D1 Location of facilities ✓ ✓ ✓

Constraint C1 Full coverage ✓ ✓ ✓

C9-1 Maximum capacity for ✓ ✓ ✓

Basic location SCL Set covering location problem ✓ ✓ ✓

A. Ahmadi-Javid et al. / Computers & Operations Research 79 (2017) 223–263

Given the importance of blood bank locations in a health sys-

SG, MH-SA, MH-O

the problem into three sub-problems. The ﬁrst sub-problem solves

a hierarchical PMLP, called the pq-median location problem, to

provide services to p blood centers located at the demand points.

MINLP, FP, PSP, MCDM, O(QT)

In the second sub-problem, location of blood stations–as sup-

MINLP, PSP, FP, RO

problem is formulated as an IP model to reduce inefﬁciency in the

banks that should be established and the assignment of hospitals

to the open blood banks. In their model, the locations of blood

banks are determined based on the center of gravity method.

The location of ﬁxed and temporary blood facilities, and the

design of distribution networks over a multi-period horizon with

C2, C4, C5, C6, C8, C11

[95]. Elalouf et al. [94] focused on the determination of centrifuge

centers in a three-echelon blood sample collection chain com-

posed of clinics, centrifuge centers, and testing laboratories under

the assumption that centrifugation services can be performed in

clinics and centrifuge centers.

location of permanent and/or temporary blood facilities in dis-

asters. Jabbarzadeh et al. [97] proposed an FCLP-based robust

model to determine the number and location of permanent and

temporary blood facilities as well as to periodically determine

blood inventory levels by coordinating supply and demand during

and after disasters. Fahimnia et al. [96] also presented a two-stage

of mobile blood facilities; and to specify blood collection and

transportation quantities, and inventory levels for efﬁcient and

timely supply of blood in disasters. In addition to the brief de-

scription provided above, the papers are listed in Table 6.

P1, P3, P4, P5-1, P5-2, P12

Extending location models for handling both independent blood

Developing models that consider stochastic and dynamic con-

Considering new settings, such as budget constraints or multi-

Developing a model for optimal location/relocation of compo-

Incorporating the average age, sex, and blood groups of people

Extending the existing location models for other specialized

Developing dynamic location models for other specialized ser-

Using accurate methods to solve large-sized instances of the

Integrating location of other specialized services facilities with