Abstract
Pursuing the overarching goal of saving both lives and healthcare costs, we introduce an approach to increase the expected participation in a preventive healthcare program, e.g., breast cancer screening. In contrast to sick people who need urgent medical attention, the clients in preventive healthcare decide whether to go to a specific facility (if this maximizes their utility) or not to take part in the program. We consider clients’ utility functions to include decision variables denoting the waiting time for an appointment and the quality of care. Both variables are defined as functions of a facility’s utilization. We employ a segmentation approach to formulate a mixed-integer linear program. Applying GAMS/CPLEX, we optimally solved instances with up to 400 demand nodes and 15 candidate locations based on both artificial data as well as in the context of a case study based on empirical data within one hour. We found that using a Benders decomposition of our problem decreases computational effort by more than 50%. We observe a nonlinear relationship between participation and the number of established facilities. The sensitivity analysis of the utility weights provides evidence on the optimal participation given a specific application (data set, empirical findings).
Similar content being viewed by others
Change history
12 March 2024
A Correction to this paper has been published: https://doi.org/10.1007/s00291-024-00758-y
References
Anand KS, Paç MF, Veeraraghavan S (2011) Quality-speed conundrum: trade-offs in customer-intensive services. Manag Sci 57(1):40–56
Aros-Vera F, Marianov V, Mitchell JE (2013) p-Hub approach for the optimal park-and-ride facility location problem. Eur J Oper Res 226(2):277–285
Australian Bureau of Statistics (2011a) Greater Sydney (Statistical Area Level 1), Basic Community Profile, Excel spreadsheet 2011 Census B04 A+B New South Wales SA1, AGE BY SEX. Viewed May 16, 2018. http://www.abs.gov.au/websitedbs/censushome.nsf/home/census?opendocument&navpos=10
Australian Bureau of Statistics (2011b) Statistical Area Level 1 (SA1) ASGS Ed 2011 Digital Boundaries in ESRI Shapefile Format. Viewed May 16, 2018. http://www.abs.gov.au/AUSSTATS/abs@.nsf/DetailsPage/1270.0.55.001July
Ayer T, Alagoz O, Stout NK, Burnside ES (2016) Heterogeneity in women’s adherence and its role in optimal breast cancer screening policies. Manag Sci 62(5):1339–1362
Baron O, Berman O, Krass D (2008) Facility location with stochastic demand and constraints on waiting time. Manuf Serv Oper Manag 10(3):484–505
Batt RJ, Terwiesch C (2015) Waiting patiently: an empirical study of queue abandonment in an emergency department. Manag Sci 61(1):39–59
Belloni A, Freund R, Selove M, Simester D (2008) Optimizing product line designs: efficient methods and comparisons. Manag Sci 54(9):1544–1552
Benati S, Hansen P (2002) The maximum capture problem with random utilities: problem formulation and algorithms. Eur J Oper Res 143(3):518–530
Besbes O, Sauré D (2016) Product assortment and price competition under multinomial logit demand. Prod Oper Manag 25(1):114–127
Blankart CR (2012) Does healthcare infrastructure have an impact on delay in diagnosis and survival? Health Policy 105(2–3):128–137
Blum K, Offermanns M (2004) Umverteilungswirkungen der Mindestmengenregelung. das Krankenhaus 96(10):787–790
BreastScreen Australia (2014) BreastScreen and you—information about mammography screening. Viewed May 16, 2018. http://cancerscreening.gov.au/internet/screening/publishing.nsf/Content/breastscreen-and-you
BreastScreen Australia (2015) National Accreditation Handbook. Viewed May 16, 2018. http://www.cancerscreening.gov.au/internet/screening/publishing.nsf/Content/acc-hb-2015
BreastScreen New South Wales (2015) Viewed August 18, 2015. http://www.bsnsw.org.au/
Castillo I, Ingolfsson A, Sim T (2009) Social optimal location of facilities with fixed servers, stochastic demand, and congestion. Prod Oper Manag 18(6):721–736
Chen H, Qian Q, Zhang A (2015) Would allowing privately funded health care reduce public waiting time? Theory and empirical evidence from Canadian joint replacement surgery data. Prod Oper Manag 24(4):605–618
Codato G, Fischetti M (2006) Combinatorial Benders’ cuts for mixed-integer linear programming. Oper Res 54(4):756–766
Cordeau J-F, Furini F, Ljubic I (2019) Benders decomposition for very large scale partial set covering and maximal covering location problems. Eur J Oper Res 275(3):882–896
Dan T, Marcotte P (2019) Competitive facility location with selfish users and queues. Oper Res 67:479–497
de Bekker-Grob EW, Ryan M, Gerard K (2012) Discrete choice experiments in health economics: a review of the literature. Health Econ 21(2):145–172
Elhedhli S (2006) Service system design with immobile servers, stochastic demand, and congestion. Manuf Serv Oper Manag 8(1):92–97
Espinoza Garcia JC, Alfandari L (2018) Robust location of new housing developments using a choice model. Ann Oper Res 271:527–550
Feldman J, Liu N, Topaloglu H, Ziya S (2014) Appointment scheduling under patient preference and no-show behavior. Oper Res 62(4):794–811
Freire AS, Moreno E, Yushimito WF (2016) A branch-and-bound algorithm for the maximum capture problem with random utilities. Eur J Oper Res 252(1):204–212
Gallego G, Topaloglu H (2014) Constrained assortment optimization for the nested logit model. Manag Sci 60(10):2583–2601
GAMS Development Corporation (2017) General Algebraic Modeling System (GAMS) Release 24.8.4. GAMS Development Corporation, Washington, DC, USA
Gerard K, Shanahan M, Louviere J (2003) Using stated preference discrete choice modelling to inform health care decision-making: a pilot study of breast screening participation. Appl Econ 35(9):1073–1085
Gerard K, Shanahan M, Louviere J (2008) Using discrete choice modelling to investigate breast screening participation. In: Ryan M, Gerard K, Amaya-Amaya M (eds) Using discrete choice experiments to value health and health care, volume 11 of the economics of non-market goods and resources. Springer, Dordrecht, pp 117–137
Güneş ED, Chick SE, Akşin OZ (2004) Breast cancer screening services: trade-offs in quality, capacity, outreach, and centralization. Health Care Manag Sci 7(4):291–303
Güneş ED, Örmeci EL, Kunduzcu D (2015) Preventing and diagnosing colorectal cancer with a limited colonoscopy resource. Prod Oper Manag 24(1):1–20
Haase K (2009) Discrete location planning. Technical Report WP-09-07, Institute of Transport and Logistics Studies, University of Sydney
Haase K, Müller S (2013) Management of school locations allowing for free school choice. Omega 41(5):847–855
Haase K, Müller S (2014) A comparison of linear reformulations for multinomial logit choice probabilities in facility location models. Eur J Oper Res 232(3):689–691
Haase K, Müller S (2015) Insights into clients’ choice in preventive health care facility location planning. OR Spectr 37(1):273–291
Harewood GC (2005) Relationship of colonoscopy completion rates and endoscopist features. Dig Dis Sci 50(1):47–51
Harper PR, Shahani AK, Gallagher JE, Bowie C (2005) Planning health services with explicit geographical considerations: a stochastic location-allocation approach. Omega 33(2):141–152
Hol L, de Bekker-Grob EW, van Dam L, Donkers B, Kuipers EJ, Habbema JDF, Steyerberg EW, van Leerdam ME, Essink-Bot M-L (2010) Preferences for colorectal cancer screening strategies: a discrete choice experiment. Br J Cancer 102(6):972–980
Killaspy H, Banerjee S, King M, Lloyd M (2000) Prospective controlled study of psychiatric out-patient non-attendance. Br J Psychiatry 176(2):160–165
Koppelman FS, Bhat C (2006) A self instructing course in mode choice modeling: multinomial and nested logit models. US Department of Transportation, Federal Transit Administration
Lacy NL, Paulman A, Reuter MD, Lovejoy B (2004) Why we don’t come: patient perceptions on no-shows. Ann Fam Med 2(6):541–545
Lancsar E, Louviere J (2008) Conducting discrete choice experiments to inform healthcare decision making. PharmacoEconomics 26(8):661–677
Liu N, Finkelstein SR, Kruk ME, Rosenthal D (2018) When waiting to see a doctor is less irritating: understanding patient preferences and choice behavior in appointment scheduling. Manag Sci 64(5):1975–1996
Ljubić I, Moreno E (2018) Outer approximation and submodular cuts for maximum capture facility location problems with random utilities. Eur J Oper Res 266(1):46–56
Madadi M, Zhang S, Henderson LM (2015) Evaluation of breast cancer mammography screening policies considering adherence behavior. Eur J Oper Res 247(2):630–640
Marianov V, Ríos M, Icaza MJ (2008) Facility location for market capture when users rank facilities by shorter travel and waiting times. Eur J Oper Res 191(1):32–44
McFadden D (2001) Economic choices. Am Econ Rev 91(3):351–378
Moscelli G, Siciliani L, Gutacker N, Gravelle H (2016) Location, quality and choice of hospital: evidence from England 2002–2013. Reg Sci Urban Econ 60:112–124
Müller S, Haase K (2014) Customer segmentation in retail facility location planning. Bus Res 7(2):235–261
Müller S, Haase K (2016) On the product portfolio planning problem with customer-engineering interaction. Oper Res Lett 44(3):390–393
Osadchiy N, Diwas K (2017) Are patients patient? The role of time to appointment in patient flow. Prod Oper Manag 26(3):469–490
QGIS Development Team (2015). QGIS Geographic Information System. Open Source Geospatial Foundation Project
R Core Team (2017) R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria
Rahmaniani R, Crainic TG, Gendreau M, Rei W (2017) The benders decomposition algorithm: a literature review. Eur J Oper Res 259(3):801–817
Rusmevichientong P, Shmoys D, Tong C, Topaloglu H (2014) Assortment optimization under the multinomial logit model with random choice parameters. Prod Oper Manag 23(11):2023–2039
Ryan M, Gerard K, Amaya-Amaya M (2008a) Discrete choice experiments in a nutshell. In: Ryan M, Gerard K, Amaya-Amaya M (eds) Using discrete choice experiments to value health and health care, volume 11 of the economics of non-market goods and resources. Springer, Dordrecht, pp 13–46
Ryan M, Gerard K, Watson V, Street DJ, Burgess L (2008b) Practical issues in conducting a discrete choice experiment. In: Ryan M, Gerard K, Amaya-Amaya M (eds) Using discrete choice experiments to value health and health care, volume 11 of the economics of non-market goods and resources. Springer, Dordrecht, pp 73–97
Ryan M, Skåtun D, Major K (2008c) Using discrete choice experiments to go beyond clinical outcomes when evaluating clinical practice. In: Ryan M, Gerard K, Amaya-Amaya M (eds) Using discrete choice experiments to value health and health care, volume 11 of the economics of non-market goods and resources. Springer, Dordrecht, pp 101–116
Schrag D, Cramer LD, Bach PB, Cohen AM, Warren JL, Begg CB (2000) Influence of hospital procedure volume on outcomes following surgery for colon cancer. JAMA 284(23):3028–3035
Street DJ, Burgess L (2007) The construction of optimal stated choice experiments: theory and methods. Wiley, Hoboken
Street DJ, Burgess L, Viney R, Louviere J (2008) Designing discrete choice experiments for health care. In: Ryan M, Gerard K, Amaya-Amaya M (eds) Using discrete choice experiments to value health and health care, volume 11 of the economics of non-market goods and resources. Springer, Dordrecht, pp 47–72
Tiwari V, Heese HS (2009) Specialization and competition in healthcare delivery networks. Health Care Manag Sci 12(3):306–324
Train KE (2009) Discrete choice methods with simulation, 2nd edn. Cambridge University Press, Cambridge
Truong V-A (2015) Optimal advance scheduling. Manag Sci 61(7):1584–1597
van Dam L, Hol L, de Bekker-Grob EW, Steyerberg EW, Kuipers EJ, Habbema JDF, Essink-Bot M-L, van Leerdam ME (2010) What determines individuals’ preferences for colorectal cancer screening programmes? A discrete choice experiment. Eur J Cancer 46(1):150–159
van der Pol M, Currie G, Kromm S, Ryan M (2014) Specification of the utility function in discrete choice experiments. Value Health 17(2):297–301
Verter V, Lapierre SD (2002) Location of preventive health care facilities. Ann Oper Res 110(1–4):123–132
Vidyarthi N, Kuzgunkaya O (2015) The impact of directed choice on the design of preventive healthcare facility network under congestion. Health Care Manag Sci 18(4):459–474
Viiala CH, Tang KW, Lawrance IC, Murray K, Olynyk JK (2007) Waiting times for colonoscopy and colorectal cancer diagnosis. Med J Aust 186(6):282–285
Viney R, Lancsar E, Louviere J (2002) Discrete choice experiments to measure consumer preferences for health and healthcare. Expert Rev Pharmacoecon Outcomes Res 2(4):319–326
Wang R, Wang Z (2017) Consumer choice models with endogenous network effects. Manag Sci 63(11):3944–3960
Wang Q, Batta R, Rump CM (2002) Algorithms for a facility location problem with stochastic customer demand and immobile servers. Ann Oper Res 111(1–4):17–34
Wang D, Morrice DJ, Muthuraman K, Bard JF, Leykum LK, Noorily SH (2018) Coordinated scheduling for a multi-server network in outpatient pre-operative care. Prod Oper Manag 27(3):458–479
Zhang Y, Berman O, Verter V (2009) Incorporating congestion in preventive healthcare facility network design. Eur J Oper Res 198(3):922–935
Zhang Y, Berman O, Marcotte P, Verter V (2010) A bilevel model for preventive healthcare facility network design with congestion. IIE Trans 42(12):865–880
Zhang Y, Berman O, Verter V (2012) The impact of client choice on preventive healthcare facility network design. OR Spectr 34(2):349–370
Acknowledgements
The authors are very grateful for the very helpful suggestions of two anonymous reviewers to improve the paper’s quality.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
Appendix 1: Tighter bounds on variables
We can improve the upper bound for \(X_{i,j,m}\) in (25): If it is certain that r facilities will exist, we do not need to do the calculation as if (j, m) was the only facility, but we can assume that \(r-1\) other facilities will be established as well. We replace \(p_{i,j,m}\) and get
where \(\mathscr {J}_i^{\text {min}}\) is the set of the \(r-1\) least attractive other facility locations for demand node i and \(\check{m}\) represents the globally least attractive mode. This change reduces solution times remarkably—a finding that is also discussed in Freire et al. (2016).
It is also helpful to determine bounds for the no-choice probability variables \(Z_i\). They can be easily computed by assuming worst case or best case scenarios. Let \(\mathscr {J}_i^{\text {worst}}\) and \(\mathscr {J}_i^{\text {best}}\) be the sets of the r least or most attractive alternatives j for demand node i, and \(\hat{m}\) the most attractive mode. Then we get upper bounds \(Z_i^{\text {UB}}\) and lower bounds \(Z_i^{\text {LB}}\):
If we also decided about the number of established facilities r with a variation of (30) to
our model is harder to solve and (34) and (35) would not be applicable.
Appendix 2: Modeling framework to derive a lower bound
Since it may take a long time to find a first integer solution within the solving process, it is expedient to determine a lower bound for our linear model PHCFLPP (23)–(33). We present three models that altogether yield a lower bound. This approach is basically introduced in Haase and Müller (2015). It relies on the constant part of individuals’ utility functions, which mainly depends on distances.
2.1 Step I
With the first auxiliary model the highest choice probabilities for each single demand node are chosen. I.e., the most attractive facilities for them are identified. We continue our nomenclature:
Additional variables
- \(\Upsilon _{i,j,m}\):
-
= 1 if demand node i is assigned to location j in mode m; 0, otherwise
- \(F^{\mathrm {I}}\):
-
Objective function value of Model Step I (cumulated r highest choice probabilities)
subject to
The objective function (38) maximizes the cumulated chosen choice probabilities subject to (39), which means that the at most r highest choice probabilities for each demand node i are selected. (40) ensures that a facility can only be demanded in exactly one mode m.
2.2 Step II
In the second auxiliary model, the r overall most attractive facilities are established. The specific mode is not of interest here and the best one is always chosen. The decision is based on the remaining influences, mainly the distance between demand and supply nodes. An additional parameter that makes use of the solution to Step I sums up for how many demand nodes a certain facility belongs to the most attractive ones. The more demand nodes assigned to a facility the higher its attraction by this definition. We further extend our nomenclature:
Additional parameter
- \(b_{j,m}\):
-
Attractiveness value for each facility at location j in mode m with \(b_{j,m} = \sum _{i \in \mathscr {I}} {\Upsilon _{i,j,m}^{*}}\) where \({\Upsilon _{i,j,m}^{*}}\) is the optimal solution to Step I
Additional variables
- \(\tilde{Y}_{j,m}\):
-
= 1 if facility at location j is specified to offer healthcare service in mode m; 0, otherwise
- \(F^{\mathrm {II}}\):
-
Objective function value of Model Step II (attractiveness of located facilities)
subject to
The objective function (42) maximizes the cumulated overall attractiveness and thereby chooses the r most demanded facilities according to (43). The result represents a solution that depends on the individuals’ constant part of their deterministic utility function disregarding capacities.
2.3 Step III
The last auxiliary model is used to determine the final modes for the facility locations previously identified in Step II. We consider
Additional parameter
- \(\tilde{Y}_{j,m}^{*}\):
-
optimal solution to Step II
Additional variable
- LB:
-
objective function value of Model Step III (a lower bound for PHCFLPP)
subject to (24)–(28), (31)–(33) as well as
The first restriction blocks are in fact PHCFLPP without two redundant constraints of which information is already included in Step I and Step II. The information is that only one mode per facility can be present as well as that r facilities are established. (45) maximizes the expected participation by now selecting the final modes of the predestined facility locations via (46). The objective function value is a lower bound for PHCFLPP as it is a feasible integer solution to it. It is allowed to add up the pre-defined locations on the right hand side of (46), because at most one value can equal 1 due to (40). Thus, (46) is a substitute for (29) and a linking constraint in addition. Those constraints can be left out in Step III, as well as (30), because of (43). So we can perform a MIP start with PHCFLPP after solving the Steps I, II and III in a row.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Krohn, R., Müller, S. & Haase, K. Preventive healthcare facility location planning with quality-conscious clients. OR Spectrum 43, 59–87 (2021). https://doi.org/10.1007/s00291-020-00605-w
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00291-020-00605-w