Content uploaded by Md. Noor-E-Alam

Author content

All content in this area was uploaded by Md. Noor-E-Alam on Oct 21, 2016

Content may be subject to copyright.

Content uploaded by Ruilin Ouyang

Author content

All content in this area was uploaded by Ruilin Ouyang on Jul 22, 2016

Content may be subject to copyright.

Proceedings of the 2016 International Conference on Industrial Engineering and Operations Management

Detroit, Michigan, USA, September 23-25, 2016

An ILP Model for Healthcare Facility Location Problem

with Long Term Demand

Ruilin Ouyang

Department of Mathematics

Northeastern University

360 Huntington Ave, Boston, MA 02115, USA

ouyang.ru@husky.neu.edu

Tasnim Ibn Faiz and Md Noor-E-Alam

Department of Mechanical and Industrial Engineering

Northeastern University

360 Huntington Ave, Boston, MA 02115, USA

faiz.t@husky.neu.edu, mnalam@neu.edu

Abstract

Facility location decisions is one of the most crucial commitments that manufacturing and service

industries face to impacts their interaction with the end users or customers. This is a long term decision

and recourse option is very difficult once the decision has been implemented. By optimally placing the

facilities considering the probable future expansion in the market, a business entity can gain the

dominance over its competitors. In the face of rapid urbanization and increasing demand, making facility

location decision based only on existing demand is not optimal; taking into account possible realization of

future demand would result in more robust decisions. The present study focuses on developing a

mathematical model for making optimal decisions regarding healthcare facilities taking into account long

term demand. It utilizes the concept of grid-based location problems to divide the area of interest into

discrete cells. The model provides the optimal locations for facilities to be built at present time and the

potential location of facilities in the future. Finally, the model is programmed with a standard modeling

language AMPL and solved with the CPLEX solver. Results show that the model is efficient in solving

small to moderate sized problems. The developed approach can be used by government or relevant

agencies to make optimal location decisions for healthcare or other service facilities.

Keywords

Healthcare facility location, Grid-based location problem, Long term location decision.

1. Introduction

Numerous studies have addressed the facility location problem for both manufacturing and service sectors; many

algorithms have been developed for determining the optimal numbers and locations for facilities to be built. The

survey done by Brandeau el al. (1989) provides an overview of the studies focusing on location decision problems

conducted in the earlier part of the twentieth century. According to the survey, one of the most important inputs for

the location problem is the demand for the products or services that the facility will provide. But very few of these

studies have taken into account the future demand of customers or end users, in addition to current or existing

requirements. In one of these studies, done by Brancolini et al. (2006), asymptotical location problem was

considered; the authors compared long term and short-term strategies and their effects on location decisions. Their

study suggested that considering long term demand in making location decisions is beneficial. In a slightly similar

manner, Fernández at el. (2007) considered a location problem and the price setting in order to maximize profit, in

which the authors considered long term competition on prices and showed the effect of it on location decisions.

Chou (2009) proposed an integrated short-term and long term MCDM (multiple-criteria decision-making) Model for

location problems. It showed the importance of integrating the short and long term evaluation method with

examples. Kim and Kim (2010) studied long term healthcare facilities problem, which can balance the numbers of

© IEOM Society International

840

Proceedings of the 2016 International Conference on Industrial Engineering and Operations Management

Detroit, Michigan, USA, September 23-25, 2016

patients assigned to each facility; although cost was not considered in the model, which is a limitation of this study.

Marić et al. (2015) used Hybrid metaheuristic method for long term health-care facility problem that focused on

minimizing the maximum number of patients assigned to established facilities. Carlo et al. (2012) discussed and

proposed several approaches for the long term location problem with the objective of minimizing the total cost of

interacting with a set of existing facilities. Öhman and Lämås (2003) studied the effect of long term location

planning on harvest activities. In considering the long term demand, a few studies have been done that strived to

make the optimal location decisions in presence of demand uncertainty. Hosseini and MirHassani (2015) proposed a

2-stage stochastic location model for refueling station under uncertainty. But the model is too complicated to be

solved by available solvers. Albareda-Sambola et al. (2013) presented a compact binary formulation for the

deterministic equivalent model of the problem under uncertainty. Temur (2016) presented a multi-attribute decision

making approach for location decision under high uncertainty. It showed that location decision is very sensitive to

the uncertainty. Bai and Liu (2014) examined the influence of uncertain transportation costs and customers’

demands on the location decisions.

Healthcare facility location decision has a significant impact on the effectiveness of a facility to provide reliable and

safe services to the patients in the long run. Once such decisions have been made and implemented, it is extremely

difficult and costly to take recourse actions. Therefore, in deriving an optimal solution for the facility location

problem, considering future demand is required in addition to current demand. To the best of our knowledge, a little

effort has been made to incorporate future demand in determining optimal healthcare facility locations. Our

objective of this study is to develop a new integer linear programming (ILP) model for a long term healthcare

facility location problem. This model will consider both, the present and future demand, i.e. total number of patients

requires treatment at present and additional number of patients that would require treatment in the future. Therefore,

for both the present time and the future time horizon, two types of decision variables will be considered. The first

type of decision variables will provide the optimal locations of facilities to be built, while the second type will

determine the optimal allocation of patients to the facilities. The objective of the model is to minimize the total cost

of building and maintaining all the facilities in the considered time horizon, while meeting all the demands.

2. Problem description

In order to formulate a mathematical model for the long term healthcare location problem, a region is considered,

where the patients are known to be present. The total number of patients can be treated as the demand and their

locations as the demand location for services. The region of interest can be divided into two-dimensional grids to

consider the problem as grid-based location problems (Noor-E-Alam et al., 2012), where each cell with the same

dimension; patients located at a particular grid can be represented by the coordinates of a cell. Cell coordinates (,)

serve as the demand locations, and number of patients present at each cell determines the demand. For each cell

(,), demand is during the time interval beginning from time point to time point , and during the time

interval (beginning from time point to the end of planning time horizon), demand is . These demands are

assumed to be determined and will be considered as parameters in our model. To serve the patients, healthcare

facilities need to be built, for which fixed costs of building facilities () and maintenance cost per unit time () will

be incurred. Each facility is capacitated (maximum capacity is ), and patients from each cell can only go to the

facility that is nearest to them. For the facility locations, the same region is considered, divided into two-dimensional

grids, where cells are indicated by a different set of coordinates (,). If a facility is to be built at (,) at time

point , the binary variable associated with this decision takes a value of 1, and if a patient located at (,) goes

to facility located at (,) during the time interval beginning from time point to time point , then the binary

decision variable takes a value of 1. The corresponding binary decision variables for time point and time

interval , i.e.

and

follow the same logic. While making these decisions, capacity restrictions of facilities

and conditions for allocation of patients to the nearest facilities must be met. The problem is to decide optimal

numbers and locations for facilities to be built now and in the future, and allocate patients to their nearest facilities.

The objective is to keep the total cost of building and maintaining the facilities at the minimum level.

3. Location model

In the following subsections, the assumptions, parameters and variables included in the model are described. Figure

1 shows the time horizon considered in developing the model. Following that, objective function and the constraints

of the developed model are presented and explained in detail.

© IEOM Society International

841

Proceedings of the 2016 International Conference on Industrial Engineering and Operations Management

Detroit, Michigan, USA, September 23-25, 2016

3.1 Assumptions

We have made the following assumptions to simplify our decision problem:

• All the current patients at each cell go to one facility in the entire planning horizon.

• All the future patients estimated at each cell at time point b will also go to a single facility during the time

interval beginning from time point to the end of planning time horizon.

• All the patients will go to the nearest facility for the entire planning time horizon.

• All the facilities have the same capacity.

• Cost of maintenance for a facility will remain same in the planning time horizon.

• Costs of building a facility at time point is the same as that at time point .

Figure 1: Planning horizon to develop optimization model.

3.2 Parameters

total time interval under consideration (from time point to the end of planning time

horizon)

time interval from time point to the end of planning time horizon

current demand of cell (,) (demand during time interval beginning from time point

to time point )

future demand of cell (,) during time interval , beginning from time point to the

end of planning time horizon

fixed cost of building a facility

cost of maintain one facility per unit time

distance between points (,) and (,)

maximum capacity of one facility

large cost parameter

3.3 Variables

if patients from (,) go to facility (,) during time interval beginning from time point

to time point

, then

= 1

, else

= 0

if a facility is built at (,) at time point , then = 1, else = 0

if patients from (,) go to facility (,) during time interval ,then

= 1, else

=

0

if a facility is built at (,) at time point , then

= 1, else

= 0

3.4 ILP model

With the above assumptions, we have proposed the following ILP model to solve our long term healthcare facility

location problems:

(+)

+

(+)

(1)

© IEOM Society International

842

Proceedings of the 2016 International Conference on Industrial Engineering and Operations Management

Detroit, Michigan, USA, September 23-25, 2016

(2)

+(1 )

(3)

(4)

= 1

(5)

+

1

(6)

+

(7)

+(1

)

(8)

(9)

= 1

(10)

The objective function (1) is a cost minimization function that aims to minimize the total cost of building and

maintaining the facilities. It is assumed that the building cost is the same at time points and time . The term

indicates the maintenance cost of one facilities built at time point for time length , whereas the term

indicates the maintenance cost of one facilities setting at time point for time length . Constraint (2) describe the

relation between the variables and , that is a patient from (,) can go to (,) during time interval

beginning from time point to time point for service only if a facility is located at (,). Constraint (7) specify

the same relation during time point to the end of planning horizon between the variables

and

. Constraint

(3) restricts the patients to go only to the nearest facility built at time point . As Figure 2 indicates if there is a

facility built at (,) at time point , then for any other point (,), =1 if and only if distance .

Constraint (8) represents the same relation for facilities and patients for time point . Constraint (4) restricts the total

number of patients that can go to a facility to the maximum facility capacity for the current time period and

constraint (9) does the same for future time period. Constraints (5) and (10) restrict the number of facilities that

patients at each grid cell can visit to one. Constraint (6) specify that at any location there can be at most one facility,

i.e. if a facility is built at (,) at time point , then no facility can be built at the same location at time point .

Figure 2: Geometry of distance for three points.

4. Computational results

The model is solved on a computer with Intel(R) Core(TM) i5 CPU running at 3.19 GHz with 4 GB memory and a

64-bit operating system. For implementation, AMPL programming language is used, and CPLEX 12.6.3 solver is

utilized to solve. Three different instances are solved using the model, e.g. 5 × 5 Grid (Shown in Fig. 3), 5 × 8 Grid

(,)

(,)

(,)

© IEOM Society International

843

Proceedings of the 2016 International Conference on Industrial Engineering and Operations Management

Detroit, Michigan, USA, September 23-25, 2016

(Shown in Fig. 4) and 10 × 10 (Shown in Fig. 5) Grid. In these instances, other parameter are assumed as follows:

==10, =20, =10, =10 and =( )+ ( ). For , values were randomly generated

ranging from 0 to 5 and for , random values range from 0 to 8. In Figures 3-5, the numbers in each cell before the

sign ‘/’ specify the demand at time , and the number after the sign ‘/’ indicate the demand at time . The grey grid

cells represent facilities are built at those locations at time , while the black grid cells represent building of

facilities at time . The white gird cells indicate that no facility is required at those locations.

Figure 3: Demand distribution and result for 5×5 grid.

Figure 4: Demand distribution and result for 5×8 grid.

4/6

0/6

1/7

1/4

3/4

5/6

1/0

5/6

2/3

0/4

1/3

3/2

5/3

5/6

4/5

5/5

5/0

2/8

4/6

5/6

3/1

5/1

4/0

5/1

3/6

4/1

1/4

3/1

4/3

2/7

1/3

4/6

4/2

5/6

5/1

0/8

0/7

2/5

3/0

1/3

1/3

1/2

5/2

4/0

1/1

4/3

3/7

3/2

4/7

5/4

2/7

5/6

0/2

4/4

1/8

3/3

1/0

3/1

4/1

3/7

0/1

1/3

1/4

4/1

5/0

© IEOM Society International

844

Proceedings of the 2016 International Conference on Industrial Engineering and Operations Management

Detroit, Michigan, USA, September 23-25, 2016

Figure 5: Demand distribution and result for 10×10 grid.

Table 1: Runtime statistics.

Grid

size

Number of

facilities at

time

Number of

facilities at

time

CPU time

(seconds)

Simplex

iterations

Branch and

bound nodes

MIP

Gap

5×5

9

4

5.834

38360

1388

0.001

5×8

11

5

281.005

865350

6091

0.001

10×10

27

16

>36338.577

21005880

27202

0.05

Table 1 summarizes the runtimes and other measures that were obtained through solving for these three instances

using the developed model. From our preliminary experiment we see that our proposed model is capable to identify

the optimum healthcare facility locations and assigned facilities for the patients located in each grid for a long-term

planning horizon within a reasonable runtime at least for 5x5 and 5X8 grids. However, when we try to solve larger

grids, it takes very long time to reach the optimality (after running a 10x10 grid for more than 10 hours, we saw that

the mipgap was 5%). Therefore, we future plan is to reformulate the problem such that it will be scalable and

capable to solve large-scale instances.

5. Conclusion

In this paper, a grid-based integer linear programming model is developed for identifying optimal location of

healthcare facilities to be built now and in the future. It takes into consideration the long term, i.e. future demand

realizations in addition to the existing demand. Although this study focuses on the healthcare facilities, this model

can also be used for a variety of service sectors, for example parks and recreational centers, as well as in placing

retail stores. The model is scalable up to a degree, but for solving the large instances it takes longer time than we

expect. To rectify this limitation, future extensions would consider relaxation of some constraints and addition of

some logical constraints. Furthermore, another extension of the model will consider multiple periods and demands

as functions of time periods, which will make the model more robust.

4/1

1/4

3/1

4/3

2/7

1/3

4/6

4/2

2/4

0/7

5/6

5/1

0/8

0/7

2/5

3/0

1/3

1/3

4/2

0/6

1/2

5/2

4/0

1/1

4/3

3/7

3/2

4/7

3/6

3/4

5/4

2/7

5/6

0/2

4/4

1/8

3/3

1/0

3/2

4/3

3/1

4/1

3/7

0/1

1/3

1/4

4/1

5/0

5/5

5/4

0/5

1/7

4/7

4/1

2/1

2/4

5/1

2/1

1/1

1/2

1/2

2/4

4/1

3/7

2/2

5/3

3/8

1/5

4/3

3/4

3/5

5/8

2/3

2/5

3/1

2/7

1/8

1/6

4/5

2/4

5/6

5/1

1/2

0/4

4/1

1/3

1/5

2/5

3/6

0/7

5/6

5/4

1/6

0/1

4/2

1/1

1/0

2/4

3/1

2/6

© IEOM Society International

845

Proceedings of the 2016 International Conference on Industrial Engineering and Operations Management

Detroit, Michigan, USA, September 23-25, 2016

References

Abdelmegid, M. A., Shawki, K. M., and Abdel-Khalek, H., GA optimization model for solving tower crane location

problem in construction sites, Alexandria Engineering Journal, vol. 54, no. 3, pp. 519-526, 2015.

Albareda-Sambola, M., Alonso-Ayuso, A., Escudero, L. F., Fernández, E., and Pizarro, C., Fix-and-relax-

coordination for a multi-period location–allocation problem under uncertainty, Computers & operations

research, vol. 40, no. 12, pp. 2878-2892, 2013.

Bai, X., and Liu, Y., Minimum risk facility location-allocation problem with type-2 fuzzy variables, The Scientific

World Journal, vol. 2014, 2014.

Berman, O., Larson, R. C., and Chiu, S. S., Optimal server location on a network operating as an M/G/1 queue,

Operations research, vol. 33, no. 4, pp. 746-771, 1985.

Brancolini, A., Buttazzo, G., Santambrogio, F., and Stepanov, E., Long-term planning versus short-term planning in

the asymptotical location problem, ESAIM: Control, Optimisation and Calculus of Variations, vol. 15, no. 03,

pp. 509-524, 2009.

Brandeau, M. L., and Chiu, S. S., An overview of representative problems in location research, Management science,

vol. 35, no. 6, pp. 645-674, 1989.

Carlo, H. J., Aldarondo, F., Saavedra, P. M., and Torres, S. N., Capacitated continuous facility location problem

with unknown number of facilities, Engineering Management Journal, vol. 24, no. 3, pp. 24-31, 2012.

Chou, C. C., Integrated short-term and long-term MCDM model for solving location selection problems, Journal of

Transportation Engineering, vol. 135, no. 11, pp. 880-893, 2009.

Dinler, D., Tural, M. K., and Iyigun, C., Heuristics for a continuous multi-facility location problem with demand

regions, Computers & Operations Research, vol. 62, pp. 237-256, 2015.

Erbıyık, H., Özcan, S., and Karaboğa, K., Retail store location selection problem with multiple analytical hierarchy

process of decision making an application in Turkey, Procedia-Social and Behavioral Sciences, vol. 58, pp.

1405-1414, 2012.

Fernández, P., Pelegrin, B., Pérez, M. D. G., and Peeters, P. H., A discrete long-term location–price problem under

the assumption of discriminatory pricing: Formulations and parametric analysis, European journal of

operational research, vol. 179, no. 3, pp. 1050-1062, 2007.

Hosseini, M., and MirHassani, S. A., Refueling-station location problem under uncertainty, Transportation

Research Part E: Logistics and Transportation Review, vol. 84, pp. 101-116, 2015.

Kim, D. G., and Kim, Y. D., A branch and bound algorithm for determining locations of long-term care facilities,

European Journal of Operational Research, vol. 206, no. 1, pp. 168-177, 2010.

Marić, M., Stanimirović, Z., and Božović, S. Hybrid metaheuristic method for determining locations for long-term

health care facilities, Annals of Operations Research, vol. 227, no.3, pp.3-23, 2015.

Noor-E-Alam, M., Mah, A., and Doucette, J., Integer linear programming models for grid-based light post location

problem, European Journal of Operational Research, vol. 222, no. 1, pp. 17-30, 2012.

Öhman, K., and Lämås, T., Clustering of harvest activities in multi-objective long-term forest planning, Forest

ecology and management, vol. 176, no. 1, pp. 161-171, 2003.

Serrano, A., Faulin, J., Astiz, P., Sánchez, M., and Belloso, J., Locating and Designing a Biorefinery Supply Chain

under Uncertainty in Navarre: A Stochastic Facility Location Problem Case, Transportation Research Procedia,

vol. 10, pp. 704-713, 2015.

Sonmez, A. D., and Lim, G. J. (2012)., A decomposition approach for facility location and relocation problem with

uncertain number of future facilities, European Journal of Operational Research, vol. 218, no. 2, pp. 327-338,

2012.

Syam, S. S., and Côté, M. J., A location–allocation model for service providers with application to not-for-profit

health care organizations, Omega, vol. 38, no. 3, pp. 157-166, 2010.

Temur, G. T., A novel multi attribute decision making approach for location decision under high uncertainty,

Applied Soft Computing, vol. 40, pp. 674-682, 2016.

© IEOM Society International

846

Proceedings of the 2016 International Conference on Industrial Engineering and Operations Management

Detroit, Michigan, USA, September 23-25, 2016

Biography

Ruilin Ouyang is a Master Student in the Department of Mathematics at Northeastern University, Massachusetts,

USA. He earned a Bachelor of Science degree in Mathematics from Zhejiang University, Hangzhou, China. He is

currently working under Professor Muhammad Noor E Alam.

Tasnim Ibn Faiz is pursuing his PhD in Industrial Engineering at Northeastern University. He earned his B.S. in

Industrial Engineering from Bangladesh University of Engineering and Technology, Bangladesh, and Masters in

Industrial Engineering from Louisiana State University, USA. He has published journal and conference papers,

worked in academic research projects as well as in industrial sector. His research interests include optimization,

scheduling, healthcare and renewable energy. He is a member of IIE and INFORMS.

Md Noor-E-Alam is an Assistant Professor in the Department of Mechanical & Industrial Engineering at the

Northeastern University. Prior to his current role, he was working as a Postdoctoral Research Fellow at

Massachusetts Institute of Technology. His current research interests lie in the intersection of operations research

and data analytics, particularly as applied to healthcare, manufacturing systems and supply chain. He has completed

his PhD in Engineering Management in the Department of Mechanical Engineering at the University of Alberta

(UofA) in 2013. Before coming to the UofA, he served as a faculty member (first as a Lecturer and then as an

Assistant Professor) in the Department of Industrial and Production Engineering at Bangladesh University of

Engineering & Technology (BUET). He also previously received a B.Sc. and M.Sc. in Industrial and Production

Engineering from BUET.

© IEOM Society International

847