A decision support system for home dialysis visit scheduling and nurse routing

https://doi.org/10.1016/j.dss.2019.113224Get rights and content

Highlights

  • A home dialysis visit scheduling and nurse routing problem is considered.

  • The development of a user-friendly decision support system is described.

  • The computer-based tool employs a multi-criteria mixed-integer programming model.

  • The system supports the creation of nurse schedules and daily driving itineraries.

  • Results show the total distance travelled can be potentially reduced by more than 25.

Abstract

Over the last years home dialysis has become the preferred treatment option for some patients with kidney failure. However, determining efficient and effective daily home-dialysis service plans imposes multiple challenges to hospital administrators, as it is a complex task with a number of interrelated decisions. These decisions are about the number of nurses required for daily visits and their travel itineraries, and involve multiple (often conflicting) objectives.

Working together with physicians and administrators from The Ottawa Hospital (TOH) in Canada, we have developed a system intended to support administrators of nephrology departments in creating daily visit schedules and routes for nurses assisting with dialysis treatment in patients' homes. The decision support system, called Home Dialysis Scheduler System (HDSS), employs a mixed-integer linear programming model to create daily nurse itineraries that minimize the cost of providing home dialysis for a pre-specified group of patients. In developing this model, we also considered nurses' workload balance, overtime work, need for mealtime breaks (lunch or dinner, depending on shift times), restrictions and preferences associated with the time of the visits, and different types of services provided to patients. The model was validated using data provided by the Division of Nephrology at TOH. The interface of the HDSS was developed following the principles of user-centred design and validated with a group of end-users. In the validation stage, daily visit schedules and nurse routes generated by the HDSS were compared to nurse itineraries created manually by hospital administrators. The use of the HDSS resulted in improved workload distribution among nurses, simpler routes, and reduced total distance travelled — which translates into lower costs for the home dialysis program. In this paper, we provide details about the mixed-integer linear programming model, describe the HDSS, and discuss its implementation results and managerial implications.

Introduction

Dialysis is a life-sustaining treatment for patients with end-stage kidney disease. There are two basic types of dialysis: hemodialysis and peritoneal dialysis. Hemodialysis involves purifying blood directly through an extracorporeal dialysis machine, whereas peritoneal dialysis involves filling up the peritoneal cavity with sterile fluid and allowing the peritoneal membrane to act as a natural filter. A patient can complete peritoneal dialysis at home while hemodialysis is usually done by trained healthcare professionals in a dialysis centre or hospital. Although clinical outcomes of both types of dialysis are similar, peritoneal dialysis offers the possibility of a more flexible treatment schedule, greater convenience for patients, and represents a more cost-effective option [14, 18, 34] . However, despite its benefits, the broader use of in-home dialysis is limited by the inability of some older and frail patients to perform the treatment independently [3, 15, 36] . For this reason, and with the ultimate goal of expanding the delivery of in-home peritoneal dialysis (hereafter simply called “home dialysis”), many healthcare authorities around the world provide hospitals with additional funding to support regional home-dialysis programs. Patients in these programs receive nursing visits up to twice a day to assist them with the use of the dialysis equipment [7, 21, 22] . This additional financial support has resulted in an increase in the utilization of home dialysis services, for example, in Ontario, Canada. However, recent changes in funding has shifted the responsibility for providing home dialysis visits from regional healthcare authorities to hospitals, thereby increasing the need for decision support tools that help make an efficient use of the resources available at these programs.

Many home dialysis patients receive daily visits from nurses. A nursing visit is required to set up a peritoneal cycler (dialysis machine) and, in some cases, to connect and disconnect the patient from the cycler. Some of these visits are time sensitive as patients need to be connected to the cycler for a specific period of time. Home dialysis patients quickly become familiar with the nurses who visit them at home and, consequently, continuity of care by the same care provider plays an important role in patient satisfaction. If, on top of these considerations, we take into account other aspects of home dialysis delivery such as geographical location, shift times of the nurses, mandatory mealtime breaks (lunch or dinner, depending on the specific shift times), overtime work, and workload balance, then determining an efficient daily home dialysis visit schedule becomes a challenging task that is very difficult and time consuming if done manually.

This paper describes a decision support system, called Home Dialysis Scheduler System (HDSS), designed to help hospital administrators with the scheduling of home dialysis visits and the routing of nurses. This system uses a mixed-integer linear programming (MILP) model to optimize nurse itineraries with an objective function that combines multiple criteria: total distance travelled by all nurses, travel and overtime costs, number of nurses required to visit all patients, and workload balance across nurses. The objective function is minimized and each individual criterion is weighted to reflect its relative importance as requested by hospital administrators. The optimization model also considers patient-nurse compatibility restrictions (e.g., required skill level and case load complexity), nurse availability (i.e., shift times, mealtime breaks, and overtime work), patient preferences for specific visit times, and visit durations. Hospital administrators interact with the HDSS using a spreadsheet-like user interface. To the best of our knowledge, the HDSS is the first-of-its-kind decision support system available to hospital administrators in charge of home dialysis programs.

The remainder of this paper is organized as follows: Section 2 describes home dialysis delivery as implemented by the Division of Nephrology at The Ottawa Hospital (TOH) in Ontario, Canada. Section 3 reviews related relevant work and Section 4 introduces the MILP model that is at the core of the HDSS. Section 5 describes the development of the HDSS and Section 6 discusses the use of the system for visit scheduling and nurse routing in Ottawa, Ontario, Canada. Section 6 also reviews some of the managerial implications associated with the use of the system. Finally, in Section 7, the paper concludes with final remarks.

Section snippets

Problem definition

The home dialysis program at TOH is one of the largest dialysis programs in North America. It currently serves more than 220 patients and involves close to 18 full-time nurses. Before the development of the HDSS, visit scheduling and nurse routing decisions were made manually by the home-dialysis clinical care facilitator.

Every day, the care facilitator first needs to prepare a daily schedule for nurses either to provide in-home patient visits or to deliver teaching and clinical support to

Related work

In general, home health care (HHC) involves visits to patients' homes by different care providers (social workers, physiotherapists, nurses, etc.) and often represents an interesting scheduling and routing problem (SRP). Recent research on SRP in HHC is summarized in Refs. [11], [5], and [6]. These review papers present problem classifications based on solution methodology, planning time horizon, objectives, and constraints.

The SRP is classified as either single-period or multi-period depending

Mathematical model

We consider a single-period model that allows us to determine a set of daily nurse itineraries. A nurse itinerary defines when a specific nurse should leave home, visit each patient assigned to him/her, have a mealtime break (if any), and return home. The sets of patients and nurses are denoted by P = {1,2,…,p} and N = {1,2,…,n}, respectively, where p is the total number of patients to be visited on a given day and n is the total number of nurses available that day. The travel distance and the

HDSS

The HDSS was developed using rapid prototyping combined with user-centred interface design. At each stage, the HDSS prototype together with the user interface option were presented to the hospital administrator (immediate end-user), management, and clinical leadership of the home dialysis program. Each modelling assumption and associated course of action was validated and each user interface option was presented and discussed with the help of mock-ups. Once consensus among the group members was

Comparative evaluation

At the end of the system development cycle, we conducted a comparison of the schedules generated by the HDSS with those developed manually by the hospital administrator. All computations reported here were executed on a stationary computer with an Intel Core i3-2120 3.30 GHz processor with 4 GB RAM running the operating system Windows 7 Enterprise (64-bit). The execution times never exceeded 3 min.

Conclusion

The use of home dialysis has increased over the last years because of the flexibility it provides to patients and the fact that is a cost-effective option for healthcare systems. Determining good visit schedules and nurse routes to meet patient requirements is one of the key tasks performed by home-dialysis care facilitators. This is because these decisions drive the program's operating cost, impact the quality of the service provided to patients, and influence nurses' job satisfaction.

In this

Acknowledgments

The authors thank all the staff of the Home Dialysis Program at the Division of Nephrology at TOH for their insights about home dialysis and help with the development of the HDSS. This research was partially supported by the Natural Sciences and Engineering Research Council of Canada (NSERC) and the Telfer Health Transformation Exchange at the Telfer School of Management.

Ahmet Kandakoglu is a Research Associate at the Telfer School of Management at the University of Ottawa. His research interests span the areas of combinatorial optimization, decision-making under uncertainty, decision support systems, and applied artificial intelligence. He has over fifteen years of professional experience in positions of increasing responsibility encompassing the areas of advanced analytics, information technologies, and project management. He has assisted in the development

References (1)

    Cited by (36)

    • A multi-objective home healthcare delivery model and its solution using a branch-and-price algorithm and a two-stage meta-heuristic algorithm

      2023, Transportation Research Part C: Emerging Technologies
      Citation Excerpt :

      The above discussion makes it clear that various formulations of the problem have been addressed in the literature, taking into account numerous considerations such as nurse qualifications or consistency of visits for each patient (Grenouilleau et al., 2020). In addition, different sizes of the problem have been solved with several solution techniques, including branch-and-price (Yuan et al., 2015), various heuristics (Martin et al., 2020; Fikar et al., 2016; Hiermann et al., 2015) and even Decision Support Systems (DSS) (Begur et al., 1997; Kandakoglu et al., 2019). However, most of the existing papers focus on solving the assignment problem (Lanzarone and Matta, 2014; Hertz and Lahrichi, 2009; Carello and Lanzarone, 2014) and then performing routing (Hiermann et al., 2015) or simply solving the routing problem (Mankowska et al., 2014; Kergosien et al., 2009).

    • An integrated routing and scheduling problem for home healthcare delivery with limited person-to-person contact

      2022, European Journal of Operational Research
      Citation Excerpt :

      A similar conclusion can be reached by analyzing the recent works that are summarized in Table 1. While optimizing the operational cost (Ros-McDonnell, Szander, de-la-Fuente-Aragón & Vodopivec, 2019) is still a major focus for the researchers, incorporation of additional objectives using multi-objective formulation (Trautsamwieser, Gronalt & Hirsch, 2011) to reflect various hidden costs like leaving of a competent staff because of unbalanced workload (Kandakoglu et al., 2020) or cost of an unsatisfied patient by reduced service quality (Rest & Hirsch, 2016) is becoming more common in the literature. In addition to the cost consideration, some unique objectives like minimizing the outsourced patient (Nasir & Dang, 2018) and optimizing the assignment based on patient preferences (Erdem & Koç, 2019) have also been considered in recent years.

    • A cluster-based algorithm for home health care planning: A case study in Australia

      2022, Transportation Research Part E: Logistics and Transportation Review
      Citation Excerpt :

      Various HHC applications in the operational research (OR) literature have been comprehensively reviewed by Genet et al. (2011), Fikar and Hirsch (2017) and Cissé et al. (2017) in the recent years. Different deterministic mathematical models have been developed by scholars aim to minimise the total travel time by caregivers (Castillo-Salazar et al., 2016; Liu et al., 2014; Decerle et al., 2018; Eveborn et al., 2006; Liu et al., 2017; Trautsamwieser and Hirsch, 2011; Liu et al., 2013; Wang et al., 2019; Masmoudi et al., 2018), total cost of operation (Nickel et al., 2012; Rasmussen et al., 2012; Mankowska et al., 2014; Yuan et al., 2015; Braekers et al., 2016; Zhan and Wan, 2018; Kandakoglu et al., 2020; Shi et al., 2019), travel distance (Duque et al., 2015; Shi et al., 2017; Riazi et al., 2018; Dengiz et al., 2019; Moussavi et al., 2019), and clients’ waiting time (Lanzarone and Matta, 2014; Zhan and Wan, 2018; Shehadeh et al., 2019). HHCRSP is mainly required to solve the problem where clients must be serviced at a specific time due to their requirements and preferences (Trautsamwieser and Hirsch, 2011; Rasmussen et al., 2012; Nickel et al., 2012; Liu et al., 2013; Martin et al., 2020; Çakırgil et al., 2020; Pereira et al., 2020).

    View all citing articles on Scopus

    Ahmet Kandakoglu is a Research Associate at the Telfer School of Management at the University of Ottawa. His research interests span the areas of combinatorial optimization, decision-making under uncertainty, decision support systems, and applied artificial intelligence. He has over fifteen years of professional experience in positions of increasing responsibility encompassing the areas of advanced analytics, information technologies, and project management. He has assisted in the development of award-winning decision support systems and received prestigious awards from several communities.

    Antoine Sauré is an Assistant Professor at the Telfer School of Management at the University of Ottawa. His research interests include stochastic modelling, dynamic optimization, and decision-making under uncertainty. He has more than twelve years of experience developing and applying advanced analytics techniques to large-scale problems in several industries. He has worked on the development of numerous planning and scheduling systems aimed to provide timely access to quality cancer care.

    Wojtek Michalowski is a Full Professor of Health Informatics at the Telfer School of Management at the University of Ottawa. He is a founding member of the MET Research Laboratory at the University of Ottawa, and Adjunct Research Professor at the Sprott School of Business, Carleton University. Dr. Michalowski is member of several editorial boards and for a number of years he was a Chair of the Awards Committee of the International Society on Multiple Criteria Decision Making. His research interests include computer-interpretable clinical practice guidelines, decision analysis and medical decision making, clinical decision support systems, and computer modelling of interdisciplinary healthcare teams.

    Michael Aquino is the Technical Manager, Biomedical Engineering for the Nephrology Program at The Ottawa Hospital. Michael received his Bachelor of Applied Science in Chemical Engineering from the University of Toronto. He has provided administrative and research support to multiple quality improvement projects in the area of nephrology access at the Ottawa Hospital.

    Janet Graham has extensive experience in Nephrology in the areas of hemodialysis, peritoneal dialysis and renal transplantation holding a variety of positions including the Director of Clinical Education and Quality Improvement at the Ontario Renal Network. Janet is currently the Director of Acute Medical Care at The Ottawa Hospital and the Regional Director of Nephrology at the Champlain LHIN. Her involvement in research has focused on dialysis access having presented throughout North America and having published a number of articles in this area.

    Dr. Brendan McCormick is an Assistant Professor of Medicine in the Division of Nephrology at The Ottawa Hospital and the University of Ottawa. His major clinical interests include resistant hypertension, peritoneal dialysis and hemodialysis. He is medical director of the satellite dialysis units at St. Vincent’ s Hospital, Winchester Hospital and the Queensway Carleton Hospital. Dr. McCormick received his Bachelor of Science in Biochemistry from QueenŠs University and completed his MD at the University of Toronto where he also completed Core Internal Medicine training followed by a fellowship in Nephrology.

    View full text