International O.R. Insights: Cancer care challenges

By Martin L. Puterman, Pablo Santibanez, Scott Tyldesley and John French

Cancer care delivery systems worldwide face unprecedented pressure as a result of increased demand for services in an environment of limited resources and investment. Contributing factors include an aging population, the increased incidence and prevalence of cancer, human resource shortages, inefficient healthcare delivery systems, increased screening and the development of expensive new technologies and medications. This pressure has led to rapid increases in cancer care costs and longer wait times for services, heightening stress for patients, staff and medical professionals.

Operations research methods are well suited to address these complex issues. Recognizing this potential, researchers at the University of British Columbia’s (UBC) Sauder School of Business, in collaboration with oncologists and decision-makers at the British Columbia Cancer Agency (BCCA), obtained funding from the Canadian Institutes of Health Research (CIHR) in 2007 to create The CIHR Team in Operations Research for Improved Cancer Care (ORICC).

The BCCA, where the ORICC is based, provides a province-wide, population-based cancer control program to the 4.5 million residents of British Columbia and the Yukon. BCCA’s mandate includes screening, diagnosis, treatment, follow-up, evaluation and research. Through its programs, the BCCA seeks to reduce cancer incidence and mortality and to improve the quality of life of people living with cancer. BCCA is publicly funded and operates five regional cancer centers, providing assessment and diagnostic services, chemotherapy, radiation therapy and supportive care; a sixth will open in late 2012. In 2010, the BCCA experienced 153,000 patient ambulatory care unit visits and provided 42,000 chemotherapy and 164,000 radiation therapy treatment appointments. These volumes are projected to increase.

From its base at the BCCA, ORICC team members interact regularly with agency decision-makers, management and clinicians. These interactions provide the team with an in-depth knowledge of BCCA operations and its comprehensive province-wide data and an awareness of system challenges and improvement opportunities. The team consists of four full-time staff, three of whom have master’s degrees in operations research; the fourth is a system analyst/applications programmer. The team focuses on applied, theoretical and methodological research in O.R., as well a wide range of significant systems improvement initiatives. A unique feature of the team is the involvement of Ph.D. students, post-doctoral and clinical fellows in projects and research.

The team’s research and projects address operational, clinical and strategic issues, or a combination thereof. A description of some key projects follows. Additional details are available at the Web site www.ORinCancerCare.org/cihrteam and in references [1-6].

Chemotherapy Appointment Scheduling

The BCCA Vancouver clinic provides more than 14,000 chemotherapy appointments to over 2,000 patients each year. Each course of chemotherapy requires multiple chemotherapy appointments; some patients receive multiple courses per year. Daily volumes vary between 60 and 75 patients, and the number of patients is increasing. The unit contains nine treatment rooms and 33 treatment chairs. A nurse supervises each room and administers chemotherapy to approximately seven patients each day. Each patient’s chemotherapy treatment follows one or more of 200 different protocols. These protocols provide specific guidelines including treatment drugs, drug delivery instructions, frequency of appointments, estimated nursing time and appointment date flexibility (plus or minus two days at most). Typically, each protocol involves a series of chemotherapy appointments that are given on a weekly or monthly basis. Some protocols require treatment on consecutive days.

About 25 percent of cancer patients will receive some form of chemotherapy during their course of treatment. Because of these large patient volumes and high variability in treatment protocols, chemotherapy appointment scheduling has become a complex undertaking with system-wide consequences. Poor scheduling practices directly impact patient experience and operational performance through significant clerical re-work, inefficient resource utilization, imbalanced nursing and pharmacy workloads, and heightened stress for patients and staff.

The specific issues that motivated this research were the frequent rescheduling and late confirmation times for patient appointments. Cancer and chemotherapy can be physically and emotionally challenging for patients and their families. Uncertainty about appointment times can be stressful for patients and their families because:

• patients frequently need to coordinate transportation requiring advance notice;
• patients need sufficient notice to take pre-treatment drugs if required; and
• for patients living far away from a clinic, uncertainty can complicate travel plans and impose additional financial burdens.

To address these issues, the ORICC team employed a combination of O.R. and OM methods to identify the root causes of the problem, evaluate possible solutions, redesign the process and implement Chemo SmartBook, an innovative optimization-based scheduling system.

The study began with a review of chemotherapy scheduling operations through direct observation, interviews with staff and patients and extensive data analysis. The data comprised more than 19,000 appointment records for 2008 and 2009. Analysis revealed that over 40 percent of chemotherapy patients received their final appointment notification less than one week in advance. Linking the appointment data with resource utilization and case complexity enabled the team to:

• characterize demand by daily volume, nursing time requirements and treatment protocol;
• quantify daily waitlists and their evolution;
• estimate resource utilization; and
• produce detailed appointment booking process maps.

This process review identified the following significant issues:

• an inflexible capacity reservation system;
• inefficient use of the flexibility in treatment date;
• uncontrolled wait list size; and
• variable day-to-day and within day nursing and pharmacy workloads.

A discrete-event computer simulation of the booking process evaluated the impact of alternative scheduling practices on patient notification, chemotherapy clinic operations and pharmacy workloads. Performance metrics included notification time, waitlist size and chemotherapy clinic utilization. Recognizing that over-constrained capacity and inflexibility in appointment date were major issues, scenarios considered different capacity reservation rules, flexibility in treatment date and waitlist limits. As a result of the simulation analysis, the process redesign focused on:

• removal of most booking constraints through consolidation of appointment types;
• increased flexibility in booking appointments by formalizing and accounting for clinically approved booking tolerance when faced with limited capacity for the requested date;
• waitlist control through limiting the maximum daily wait list size; and
• implementation of a two-stage process, in which the appointment date is assigned at the time the appointment is made, and the appointment time is determined one-week in advance of the treatment date.

The redesigned process could not have been implemented without the development of Chemo SmartBook, an optimization-based scheduling tool that assigns appointment times and balances workload and patient preferences. The tool considers workload balance for nurses and pharmacy simultaneously with patients’ treatment requirements and time preferences.

Figure 1: Chemo SmartBook is a Web-based scheduling tool that uses optimization to assign the nurse and appointment time to each patient, balancing workload for nurses and pharmacy.

Chemo SmartBook is a Web-based application (Figure 1) linked to the appointment information system. The chemotherapy appointment-scheduling model is a multi-objective, integer goal-program. The main decision, which is represented through a binary variable, is whether or not an appointment is scheduled to start in a given time slot for a specific nurse. The main resource constraint is nursing time used by the appointments scheduled under each nursing shift. A series of goals or soft constraints are imposed to measure deviations from capacity targets (e.g., drug preparation capacity for pharmacy) or scheduling preferences (e.g., start appointment after 11 a.m.). All deviations are penalized according to an importance factor and aggregated into a single objective to be minimized.

Implementation required process change, review of protocol guidelines, information system modifications, software development, staff training and buy-in, and communication of new practices to patients and staff. Management and staff participation and involvement in all aspects of the project were crucial to its adoption and ultimate success. The new system went live in June 2010 (Figure 2) just 15 months after the study began.

Figure 2: Chemo SmartBook is used on a daily basis by a scheduling clerk at the chemotherapy clinic to determine the best arrangement of appointments for more than 60 patients.

Project impact was measured from the perspectives of patients, staff and the organization, and included analysis of appointment data and structured management interviews and patient and staff satisfaction surveys. Results demonstrate that the changes implemented to scheduling practices significantly improved the process across a range of metrics (Figure 3):

• late patient appointment notification was reduced by 58 percent;
• waitlist size decreased by 83 percent;
• pharmacy drug preparations volumes were more evenly spread throughout the day;
• patient surveys confirmed increased satisfaction with confirmation times; and
• staff feedback reported improved nursing workload distribution and reduced stress levels.

Figure 3: Results from the project exceeded expectations. One of the evaluation metrics, waitlist size one week before the treatment date decreased by 83 percent from a daily median of 24 to 4. Comparable periods of time were used for evaluation to avoid seasonality effects.

In 2011, the project was recognized with the “Excellence in BC Health Care Award” for innovation, presented by the Health Employers Association of British Columbia. Further, the team was awarded a CIHR Knowledge Translation Grant to continue the implementation and development of Chemo SmartBook. Reviewers from the CIHR panel praised the project; one described it as “by far the most exciting application I have read in a while,” with “great potential for significant improvement in the health care system for chemotherapy patients and could possibly revolutionize the appointment process of other services in the future.” Presently, the redesigned process and new scheduling tool are being implemented in other BCCA chemotherapy clinics, and new scheduling features are being added to Chemo SmartBook. Broader applicability is also under study.

Radiation Therapy Workforce Planning and Scheduling

The radiation therapy (RT) delivery process is highly labor-intensive and requires staff with a wide range of skills. Radiation therapists develop treatment plans, deliver treatment and schedule appointments, and perform other technical and operational functions. With a wide range of responsibilities and ever-changing technology, radiation therapists often require more than five years to gain the skills necessary to perform high-level tasks. Therapists acquire needed skills through experience; a specific goal of this project is to ensure a sufficient number of trained staff to deliver treatment both today and into the future.

Workforce planning and scheduling for radiation therapy poses both strategic and operational challenges. At the strategic level, managers require a method to ensure that radiation therapists possess the appropriate skills available to meet care needs. At the operational level, managers need to schedule radiation therapists to specific areas on a day-by-day basis.

Prior to the team’s research, both of these activities were time consuming tasks carried out by managers using rudimentary tools. The result was considerable time and effort taken away from other managerial activities and sub-optimal and inflexible plans and schedules. At the BCCA Vancouver Centre, more than 80 full-time and part-time radiation therapy staff must be scheduled to over 30 specific tasks each day. Similar challenges were faced at other BCCA cancer centers throughout British Columbia.

Two projects were undertaken in this area. A strategic planning tool was developed and used to support multi-year workforce planning at BCCA Vancouver Centre. It assigns staff to broad task areas on a quarterly basis, taking into account demand, job precedence requirements, acquired skills and staff preferences. Another challenge in this area relates to creating the daily schedule of radiation therapist, currently a complex and time consuming process. All task areas require a minimum number of staff while meeting certain constraints including minimum experience to work in a specific task area, staff availability due to scheduled vacations, maternity and other leaves, and staff preferences to work in a specific task area. To address these issues, the team has developed and validated metrics to assess schedule quality and is developing mathematical programming-based interactive tools that provide high quality schedules efficiently (Figure 4).

Figure 4: The staff scheduling tool uses an integer programming model to create the daily assignment of specific tasks within the radiation therapy department to each radiation therapist based on care delivery, staff and organizational considerations.

Optimal Radiation Therapy (RT) Start Times

Intermediate- and high-risk prostate cancer patients are often treated with combined hormone therapy and RT, with several months of hormone treatment given prior to the start of RT. However, optimal duration of hormone treatment prior to radiation therapy is not known. Finding the best time to initiate RT based on the patient’s maximum response to hormone treatment is a key question faced by clinicians. This research addressed that question by developing a dynamic statistical model of a patient’s prostate specific antigen (PSA) levels, which is used to predict the time of maximal tumor regression. The model and a prototype decision-support tool updated predictions of this time as each new PSA reading became available and provided the clinician with a probability distribution of this target treatment time. The research showed that two clinically implementable policies based on the model outperform the approach that is currently used. This research was awarded the 2010 INFORMS Health Section Pierskalla Prize.

Modeling Radiation Therapy Operations

Radiation therapy is widely used to treat cancer; roughly 52 percent of cancer patients will require RT at least once during their cancer treatment. A course of external beam RT may range from one day to daily treatments over several weeks. The RT process is complex, requiring several steps prior to the start of treatment, including consultations, planning, treatment simulation and patient preparation and education. With 30 treatment units managed by the BCCA, maintaining efficiency in daily operations is a challenging and complex task, treatment machine capability varies, patients require different courses of treatment, and machine break down and required maintenance may decrease effective capacity. Service delays arise not only from an imbalance between capacity and demand, but also as a result of inefficient patient scheduling practices. The focus of this project is to optimize the daily treatment process. This involves determining operational hours for each treatment unit, deciding on the mix of patients to be treated in each unit and the duration of individual appointments. An extensive database has been established from machine records, and a simulation model has been developed to address the complexity of the operations, the high variability in the process, and to evaluate the performance of the system under different operating scenarios.

Strategic Configuration of Cancer Care Services and Resources

A key component of quality cancer care is timely access to clinical, diagnostic and therapeutic resources. This requires the efficient coordination of services and utilization of resources: the ability to provide the correct mix of services and resources at the right time and in the right location. Configuration of an effective network of cancer care services across a large geographic region presents complex managerial and planning challenges and has a direct impact on the quality of patient care. Primary decisions include where to locate screening, diagnostic and treatment facilities, and how to allocate highly specialized resources and staff to each location. Secondary decisions affect practice at the center level and range from determining the mix of clinical services to operational parameters such as work hours and staffing levels.

The ORICC’s research in this area focused on developing a methodological framework to determine the strategic configuration of provincial cancer treatment services. This includes the formulation and implementation of analytical models that integrate clinical, operational, demographic and geographic considerations (Figure 5).

Figure 5: The configuration of cancer care services depends on clinical, operational, demographic and geographic considerations. An optimization model is used to determine location-allocation decisions which are then represented visually on a map in the form of catchment areas for each cancer centre under different planning assumptions.

This research uses the following methodologies to determine location-allocation configuration decisions:

• forecasting models to project future demand for services and resource requirements under different scenarios;
• combinatorial optimization models to design and evaluate multiple service configuration options; and
• decision trees and stochastic programming models to incorporate uncertainty related to clinical practice, population demand, and treatment technology.

Conclusion

Through the above projects, the ORICC team has introduced an innovative approach to addressing operational, strategic and clinical issues in a large British Columbia healthcare agency. The BCCA, as is typical of most healthcare institutions, was previously unaware of the benefits of advanced analytics. This research has demonstrated to BCCA managers, clinicians and decision-makers the potential that O.R. brings to improving patient care delivery, streamlining organizational processes, increasing patient and staff satisfaction, maximizing the benefits of human, facility and financial resources and supporting clinical decision-making. Initiatives such as ours are especially timely in the face of projected increased demands and costs for health services both in Canada and worldwide.

Martin L. Puterman (martin.puterman@sauder.ubc.ca) is advisory board professor of operations in the University of British Columbia’s Sauder School of Business. He is recipient of the INFORMS Lanchester Prize and an INFORMS Fellow.

Pablo Santibanez is an operations research scientist at the British Columbia Cancer Agency and adjunct professor at the Sauder School of Business. He holds an industrial engineering degree from the University of Chile and a master’s in operations research from UBC.

Dr. Scott Tyldesley is a radiation oncologist at the BC Cancer Agency and clinical associate professor in radiation oncology at UBC. He is a Michael Smith Foundation Scholar in health services research, which includes radiation oncology (genitourinary, breast); health policy, health services research, operations research and needs assessment.

John French is the senior director for operations, business and strategic planning for the Provincial Radiation Therapy Program in British Columbia. He has more than 25 years experience in healthcare in the United Kingdom, Toronto and British Columbia.

The authors lead the CIHR Team for Operations Research in Improved Cancer Care.

References

1. Lavieri, M., Puterman, M.L, Tyldesley, S. and Morris, W.J., “When to treat prostate cancer patients based on their PSA dynamics,” IIE Transaction in Healthcare Engineering, in press.
2. Santibáñez P., Chow V.S., French J., Puterman M.L. and Tyldesley S., 2009, “Reducing patient wait times and improving resource utilization at British Columbia Cancer Agency’s ambulatory care unit through simulation,” Health Care Management Science, Vol. 12, pp. 392–407, DOI: 10.1007/s10729-009-9103-1.
3. Santibáñez P., Aristizabal R., Chow V.S., Huang W., Kollmannsberger C., Nordin T., Runzer N., Puterman M.L. and Tyldesley S., “Improving chemotherapy scheduling and delivery through process redesign and advanced analytics,” manuscript submitted for publication.
4. Sauré A., Patrick J., Tyldesley S. and Puterman M.L., “Dynamic multi-appointment patient scheduling for radiation therapy,” manuscript submitted for publication.
5. Werker G., Sauré A., French J. and Shechter S., 2009, “The use of discrete-event simulation modelling to improve radiation therapy planning processes,” Radiotherapy and Oncology, Vol. 92, No. 1, pp. 76–82, DOI:10.1016/j.radonc.2009.03.012.
6. Werker G. and Puterman M., “Strategic planning of radiation therapists: An integer programming approach that includes duration and experience constraints,” manuscript submitted for publication.