- Email: [email protected]
Reducing Patient Waiting Times for Radiation Therapy and Improving the Treatment Planning Process: a Discrete-event Simulation Model (Radiation Treatment Planning)
Clinical Oncology xxx (2017) 1e7 Contents lists available at ScienceDirect
Clinical Oncology journal homepage: www.clinicaloncologyonline.net
Original Article
Reducing Patient Waiting Times for Radiation Therapy and Improving the Treatment Planning Process: a Discrete-event Simulation Model (Radiation Treatment Planning) V. Babashov *, I. Aivas y, M.A. Begen zx, J.Q. Cao y, G. Rodrigues yx, D. D’Souza y, M. Lock y, G.S. Zaric zx * Telfer
School of Management, University of Ottawa, Ottawa, Ontario, Canada Department of Radiation Oncology, London Regional Cancer Program, London Health Sciences Centre, London, Ontario, Canada z Ivey Business School, Western University, London, Ontario, Canada x Department of Epidemiology and Biostatistics, Schulich School of Medicine and Dentistry, Western University, London, Ontario, Canada y
Received 31 May 2016; received in revised form 5 January 2017; accepted 10 January 2017
Abstract Aims: We analysed the radiotherapy planning process at the London Regional Cancer Program to determine the bottlenecks and to quantify the effect of specific resource levels with the goal of reducing waiting times. Materials and methods: We developed a discrete-event simulation model of a patient’s journey from the point of referral to a radiation oncologist to the start of radiotherapy, considering the sequential steps and resources of the treatment planning process. We measured the effect of several resource changes on the ready-to-treat to treatment (RTTT) waiting time and on the percentage treated within a 14 calendar day target. Results: Increasing the number of dosimetrists by one reduced the mean RTTT by 6.55%, leading to 84.92% of patients being treated within the 14 calendar day target. Adding one more oncologist decreased the mean RTTT from 10.83 to 10.55 days, whereas a 15% increase in arriving patients increased the waiting time by 22.53%. The model was relatively robust to the changes in quantity of other resources. Conclusions: Our model identified sensitive and non-sensitive system parameters. A similar approach could be applied by other cancer programmes, using their respective data and individualised adjustments, which may be beneficial in making the most effective use of limited resources. Ó 2017 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
Key words: Discrete-event simulation; modelling; radiotherapy; waiting time
Introduction A recent report showed that many Canadians are waiting for some form of medical treatment [1]. In particular, Canadian cancer patients have experienced long waiting times in radiotherapy for many years [2e4]. Lengthy waiting times for radiation treatment may have a negative clinical impact. For example, delayed radiation treatment may Author for correspondence: M.A. Begen, Ivey Business School, Western University, 1255 Western Road, London, Ontario, N6G 0N1, Canada. Tel: þ1-519-661-4146. E-mail address: [email protected] (M.A. Begen).
increase the risk of local recurrence [5e7] and poor survival [8,9]. Several population-based studies have been conducted to investigate waiting times for radiotherapy and to determine the predictors of long waits. A higher incidence of cancer, together with an increase in the demand for radiotherapy, insufficient resources and certain patient characteristics represent some of the predictors for longer waiting times [10e12]. Conversely, radiotherapy waiting times decrease with an increase in the number of radiation therapists, medical physicists, radiation oncologists and radiation planning and therapy equipment [13].
http://dx.doi.org/10.1016/j.clon.2017.01.039 0936-6555/Ó 2017 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
Please cite this article in press as: Babashov V, et al., Reducing Patient Waiting Times for Radiation Therapy and Improving the Treatment Planning Process: a Discrete-event Simulation Model (Radiation Treatment Planning), Clinical Oncology (2017), http://dx.doi.org/10.1016/ j.clon.2017.01.039
2
V. Babashov et al. / Clinical Oncology xxx (2017) 1e7
Discrete-event simulation (DES) is a valuable tool for investigating system capacity and throughput. The use of DES models with healthcare application includes hospitals, outpatient clinics, emergency departments and pharmacies [14,15]. DES can help decision-makers to carry out a ‘whatif?’ analysis to determine good policies for scheduling patients, optimising resources, reducing waiting times of patients in clinics and improving workflows [16,17]. DES models have also been used to investigate patient scheduling challenges [18], waiting time bottlenecks, overall system throughput and system configuration in emergency rooms [19], optimal intensive care unit size [20], as well as staffing levels and bed requirements [21] in various healthcare settings. Simulation modelling has been applied in the field of radiation therapy to explore target waiting times through varying capacities [22,23] and to analyse the number of linear accelerators to achieve shorter waiting times [24]. Kapamara et al. [25] and Proctor et al. [26] used DES modelling to understand the treatment process, complexities, patient flow and bottlenecks at the radiotherapy unit. More recently, Werker et al. [27] modelled a portion of the planning process of the radiation therapy at the British Columbia Cancer Agency, with the aid of a DES model. Our work extends this study [27] by modelling the entire radiation therapy planning process, from patient arrival to treatment completion. To the best of our knowledge, this is the first paper to model and analyse the entire radiotherapy planning process at a cancer treatment facility. The primary objectives of our study were to understand how to improve the radiotherapy planning process, to understand which resources were the most important in decreasing waiting times, to increase throughput of patients treated within the 14 calendar day target (set by the
Province of Ontario and the Canadian Association of Radiation Oncology) and to provide an optimal strategy for deploying existing and new resources.
Materials and Methods We modelled a patient’s complete journey from the point of referral to radiation oncology at the London Regional Cancer Program (LRCP) to the point of starting radiation therapy, accounting for every sequential step of the treatment planning process (Fig. 1). We used the Simul8 software package (SIMUL8 Corporation), which allows for visualisation and modelling of patient flow, queues and resource utilisation. This software can project flow times and identify system bottlenecks. Our study was approved by the London Health Sciences Research Ethics Board. Data Sources Patient-level process data from 2009, available through ‘in-house’ decision support and tracking software at LRCP, served as the primary data source for our study. After a patient is referred to the LRCP and receives a consultation, a file is opened for the patient and the status of the file is tracked through this software, showing each step and each healthcare professional who works on the case. We populated the model using 3888 radiation consultation records, which corresponded to all referred patients over a 1 year period (2009e2010). We consulted with LRCP staff members, including therapists, dosimetrists, physicists and administrative staff, to ensure the face validity of the model data extracted from the decision support and tracking software database.
Pre CT Sim Delay
Reject Treatment
No
PaƟent Arrival
Contour
RadiaƟon Oncologist Consult
Dosimetry
Accept Treat?
RadiaƟon Oncologist Approval
Therapist Contour
CT SimulaƟon Yes
Physics QA
Dosimetry QA
CT SimulaƟon QA
CalculaƟon StaƟon
Treatment
Fig 1. Treatment process flow diagram of key tasks.
Please cite this article in press as: Babashov V, et al., Reducing Patient Waiting Times for Radiation Therapy and Improving the Treatment Planning Process: a Discrete-event Simulation Model (Radiation Treatment Planning), Clinical Oncology (2017), http://dx.doi.org/10.1016/ j.clon.2017.01.039
V. Babashov et al. / Clinical Oncology xxx (2017) 1e7
To obtain data that was not available through the tracking database, we interviewed LRCP staff to estimate the duration of several tasks, including radiation oncologist consultation, computed tomography simulation scan, therapist contouring, oncologist contouring and oncologist approval processes. Other data estimates were obtained through model calibration. Baseline parameter estimates are shown in Table 1. Resources and Shifts We modelled the following resources involved in radiation therapy planning at LRCP: radiation oncologists, computed tomography simulation machines, dosimetrists, physicists, therapists and linear accelerators. We modelled 15 radiation oncologists who specialised in the following fields: genitourinary, central nervous system, lung, breast, skin, head and neck, gynaecology, thyroid, sarcoma, lymphoma and gastrointestinal. Each radiation oncologist worked on different shift types during a given week (e.g. clinic or consult shift, contouring shift and academic/ administration shift). Shifts were split into blocks of approximately 4 h. Based on current practice at LRCP, we found that oncologists devote some of their academic/administration time to contouring. To incorporate this phenomenon into our model, we adjusted the shift duration devoted to contouring and academic/administration work for respective oncologists. We modelled the availability of oncologists over the year, accounting for days off due to vacations, Table 1 Baseline model parameters Variable Resources (number of resources) Radiation oncologist Computed tomography simulation machine Dosimetrist Physicists Therapists Linear accelerator Incoming cancer patients’ profiles (n¼3888), % Breast Gastrointestinal Genitourinary Gynaecological Head and neck Lymphoma Sarcoma Skin Thorax Central nervous system Unknown primary Treatment priority, % 1 High 2 Medium 3 Low Proceed to treatment, % Accept Reject
Value 15 2 9 2 8 10 14 11 18 6 6 3 1 18 12 2 9 4 12 84 62 38
3
conferences and other factors, as determined through an examination of the actual LRCP staff schedule. Nine dosimetrists, two physicists, two computed tomography simulation machines and eight computed tomography simulation therapists worked 5 days a week on an 8 h shift. Additional physicists and therapists were employed at the centre but were not assigned to patient care and hence they were not included in the model. Only those physicists and therapists who were actively assigned to patient care and tasks in the planning process were explicitly modelled and included in the simulation. There were 10 linear accelerators operating at the LRCP. Based on current practice and historical information, we assumed 10% downtime, including regular maintenance. The duration of patient treatment sessions varied between 10 minutes and 1 hour, and the number of treatment sessions varied between one and 60. For the modelling purposes, we represented the capacity of the linear accelerator in terms of 5 minutes treatment slots. Treatment Process Flow and Patient Attributes The radiation treatment planning process at the LRCP included consultation with a radiation oncologist, computed tomography simulation scan, therapist contour, radiation oncologist contouring, computed tomography simulation quality assurance, dosimetry planning, radiation oncologist approval and physics quality assurance. We modelled the arrival of patients using the referral date and time of each referred patient. The process flow is depicted in Fig. 1. Upon referral to the LRCP for a consultation, patients were assigned to an available radiation oncologist who could treat the patient’s specific cancer type. From this point on, the patient’s work file remained attached to this physician, and the patient’s waiting times were influenced by the assigned doctor’s availability. After seeing the radiation oncologist, the patient could then decide whether to proceed to radiation therapy, possibly after a planned delay, or to leave the system. We modelled several patient attributes including cancer type, number of treatment sessions required and priority level. Patients with level I priority were deemed an emergency, and the goal was to treat within 1 day. Patients with level II priority were deemed urgent, and the goal was to treat within 7 days. The remaining patients were level III and had a goal of treatment within 14 days. Performance Metric and Waiting Time Target for Ontario We used ready-to-treat to treatment (RTTT) as our main outcome measure. In Canada, this is a commonly used performance measure for assessing waiting time performance. Ready-to-treat is the time point when the patient is deemed ready to start treatment and all planned delays, such as the completion of other treatments or patient unavailability, have been resolved. All cancer centres in Ontario are required to report their patients’ mean RTTT and the proportion seen within 14 calendar days [28].
Please cite this article in press as: Babashov V, et al., Reducing Patient Waiting Times for Radiation Therapy and Improving the Treatment Planning Process: a Discrete-event Simulation Model (Radiation Treatment Planning), Clinical Oncology (2017), http://dx.doi.org/10.1016/ j.clon.2017.01.039
4
V. Babashov et al. / Clinical Oncology xxx (2017) 1e7
In addition to measures based on RTTT, we defined ‘utilisation’ as the proportion of time that any given resource was occupied. We measured utilisation for the following resources: radiation oncologists, dosimetrists, therapists, physicists, computed tomography simulation and linear accelerators. Validation The model was validated by comparing the mean RTTT, the proportion of patients treated within 14 calendar days and the 90th percentile of the waiting time distribution for treated patients for the simulation model versus corresponding values from the actual system. As we modelled a system wherein clinics and treatment facilities were saturated and resources were busy (i.e., as opposed to a newly opened facility), a ‘warm-up’ period [29] was necessary as part of the simulation model to allow the model to function at a capacity representative of an actively running centre. The warm-up analysis showed that the system reached a steady state after 208 days (5000 hours) of simulation time. Using 2 years of patient arrival data, we ran the model for a period of 30 replications, each with a warm-up period of 208 days and a run length of 500 days. Validation consisted of a direct comparison of the end point parameters produced by the simulation and the actual end point data.
Results Base Case Results and Validation We compared the model outputs to the actual system waiting time performance at LRCP (Table 2). The simulation model yielded a mean RTTT of 10.83 days (95% confidence interval 10.61, 11.05), which was virtually identical to the observed RTTT of 10.82 days (95% confidence interval 10.25, 11.37) at LRCP. The observed proportion of patients meeting the 14 day threshold was 83.59%, close to the simulation model’s result of 82.25%. The 90th percentile of the observed waiting time distribution was also close to that of the simulation model (21 and 19.87, respectively). The close correspondence between the simulated and observed outcomes gave us confidence that we had modelled the system correctly and that scenario analyses would provide reasonable approximations of expected system behaviour. Resource Utilisation Resources with utilisation over 30% are summarised in Fig. 2. Among all resources, the highest utilised was the pool of nine dosimetrists, who were utilised on average for 88% of the available time. The linear accelerators were busy 80% of the available time. The aggregate utilisation of two physicists
Table 2 Comparison of simulation versus actual system performance Actual system waiting time performance Mean (days) 90th percentile (days) % 14 days
Simulation model outputs 10.82 21 83.59%
Mean (days) 90th percentile (days) % 14 days
10.83 19.87 82.25%
Fig 2. Radiation Oncologists’ (ROs’) Specialty in the DES Model: Onc 15 (TH, SK, BR), Onc 13 (BR, SK, LYMP), Onc 6 (GYN, SARC & H&N) and Onc 11 (BR, GI, GU). Please cite this article in press as: Babashov V, et al., Reducing Patient Waiting Times for Radiation Therapy and Improving the Treatment Planning Process: a Discrete-event Simulation Model (Radiation Treatment Planning), Clinical Oncology (2017), http://dx.doi.org/10.1016/ j.clon.2017.01.039
V. Babashov et al. / Clinical Oncology xxx (2017) 1e7
was 52% of the available time. The most highly utilised radiation oncologist was number 15, with 51%, who had the skill set of treating breast, skin and thyroid cancer. Although we have attempted to fully capture all activities of human resources, there are probably some activities that have been missed (e.g. training students, overtime to get work done, use of academic hours for patients) as these may not be formally recorded. Thus, the utilisation values obtained from our models could be underestimating the actual utilisation levels. Sensitivity and Scenario Analysis Adding one dosimetrist to the resource pool had the greatest e albeit modest e impact on improving the number of patients treated within the 14 calendar day target, increasing the proportion of patients meeting this target from 82.25 to 84.92%. Removing one physicist from the resource pool had the greatest impact on reducing the proportion of patients treated within the set target, thus reducing the percentage of patients treated within 14 calendar days to 18.62%. Other individual scenarios, such as removing one dosimetrist, decreasing linear accelerator capacity by 15% and increasing patient arrivals by 15%, delivered a relatively significant impact on the proportion of patients treated within the target timeframe (see Table 3). Variation in the quantity of the remaining resources yielded non-significant effects on the proportion of patients meeting the target timeframe. Similarly, adding one dosimetrist had the largest impact on the mean RTTT, decreasing the average waiting time to 10.12 days. Removing one physicist had a dramatic effect on the system, increasing the mean RTTT to 31.35 days. In addition, other individual scenarios (e.g. one less dosimetrist, a 15% decrease in linear accelerator capacity and a 15% increase in patient arrivals) significantly worsened the mean RTTT, by 31.57, 11.44 and 22.53%, respectively. Changes in the remaining resources resulted in minor impacts on average waiting time.
5
Discussion We developed a DES model of the entire radiotherapy planning process at LRCP and conducted several ‘what-if?’ scenarios. Scenario analysis showed that adding one type and one more unit of a resource does not significantly improve the average RTTT or the percentage of patients treated within a 14 calendar day target. The largest improvement is achieved with the addition of a dosimetrist; 6.5% and 3.2% improvements in RTTT and the percentage of patients treated within a 14 calendar day target, respectively. On the other hand, decreasing certain resources and increasing demand both have a strong negative impact on system performance, suggesting that the LRCP does not have surplus capacity in dosimetrists, physicists or linear accelerators. These results suggest that the system is currently running well and the level of each resource is in balance with levels of other resources. The results also show that the system is running close to its capacity and thus needs to be monitored closely for changes in performance levels. Finally, to improve the system’s performance significantly one needs to consider the addition of multiple types and levels of resources. At the BC Cancer Agency, Werker et al. [27] modelled the radiotherapy treatment planning process and found that it was the variability and length of radiation oncologistrelated tasks that had the largest effect on treatment planning time. They also found that having one fewer dosimetrist-equivalent resource negatively affected their waiting times, but an additional dosimetrist had no significant effect on shortening waiting times. These findings are consistent with our observations, further validating our model and supporting our conclusions. Our model has several limitations. First, the model cannot capture human behaviour and response to incentives. For example, staff may stay at work after hours in order to meet deadlines and complete tasks on particularly concerning or complex cases. As the deadline (i.e. 14
Table 3 Results of simulation scenarios Scenarios
Mean RTTT in days (95% confidence interval) (baseline 10.83 days)
Percentage treated within 14 calendar day target (95% confidence interval) (baseline 82.25%)
One more dosimetrist One less dosimetrist One more physicist One less physicist One more therapist One less therapist One more computed tomography simulation machine One less computed tomography simulation machine One more oncologist* 15 % increase in linear accelerator 15% decrease in linear accelerator 15% increase in arrivals
10.12 14.25 10.73 31.35 10.82 10.90 10.82 10.92 10.55 10.82 12.07 13.27
84.92 58.88 82.59 18.62 82.37 81.99 82.26 82.01 81.98 82.27 74.43 67.47
(9.89, 10.35) (13.78, 14.73) (10.51, 10.95) (29.26, 33.44) (10.60, 11.04) (10.69, 11.11) (10.60, 11.04) (10.69, 11.15) (10.28, 10.82) (10.61, 11.04) (11.60, 12.53) (12.80, 13.74)
(84.38, (56.07, (81.95, (17.34, (81.70, (81.33, (81.60, (81.26, (81.20, (81.56, (72.07, (64.71,
85.46) 61.99) 83.22) 19.90) 83.03) 82.66) 82.92) 82.76) 82.75) 82.97) 76.78) 70.23)
RTTT, ready-to-treat to treatment. * With the specialisation of skin and genitourinary cancer. Please cite this article in press as: Babashov V, et al., Reducing Patient Waiting Times for Radiation Therapy and Improving the Treatment Planning Process: a Discrete-event Simulation Model (Radiation Treatment Planning), Clinical Oncology (2017), http://dx.doi.org/10.1016/ j.clon.2017.01.039
6
V. Babashov et al. / Clinical Oncology xxx (2017) 1e7
calendar days) approaches, patients may be moved forward in the queue, thereby starting the radiation treatment before the recommended deadline. Furthermore, as LRCP is an academic centre, the presence of resident trainees could either increase or decrease the time spent on a task. Centres with different deadlines or complexity of cases may require different times for each discrete event on the patient care pathway. Second, we determined and report two performance criteria (average RTTT waiting time and percentage treated within the 14 calendar day target) at the aggregate level for all patients and not by cancer type or by priority level. Third, the decision support and tracking software at LRCP is based on process flow, reflecting the amount of time a patient’s treatment plan spends at a certain work centre, rather than the actual amount of time spent by human resources to execute the task. This is a crucial issue regarding dosimetrists’ tasks specifically; in the event of an obstacle or uncertainty during the planning process, the task at hand is put on hold until consultation with the radiation oncologist or physicist can take place. In the meantime, the dosimetrist starts work on another patient’s treatment plan, but the tracking software continues to record that the previous task remains ongoing. This is an important limitation, not only of the model, but also of the data collection system identified in our study. About 38% of patients referred to LRCP for consultation in 2009e2010 either chose not to receive the radiotherapy at LRCP or were deemed ineligible for therapy. Based on anecdotal evidence, this rate seems to be higher than that in other centres. This situation may reflect a high level of second-opinion consultations or patients deciding to undertake treatment in other centres, or a high number of inappropriate referrals. This observation was discussed with the radiotherapy unit administration and noted as an area for future research. Centres with a different patient intake and referral populations may not be able use this model directly; adjustments to the model would have to be made before the simulation could be applied. In future work, we will explore modelling, reporting and validation of two performance criteria (average RTTT waiting time and percentage treated within the 14 calendar day target) by cancer types and priority levels of patients. We will also investigate simultaneous adjustments of resources to observe the extent to which the effect is synergistic versus additive and consider the possibility of having some steps carried out in parallel instead of strictly in sequence. Future work will also consider cost considerations. As is the case with every department, allocation of funds to various resources or work centres requires intelligent planning for future benefit and gain to the department. Investing in a new piece of equipment or specialised human resources is very costly. Our model can be adjusted to investigate whether a radiation therapy department, based on its caseload and history of referrals, would benefit financially from investing in a new technology or treatment platform. This kind of analysis would also allow us to investigate the efficiency of the system. Our model, which was developed and validated to analyse the entire radiotherapy planning process at a cancer
treatment facility, can be used to assess the impact of equipment levels, staff assignments and absences, changes in patient demand and changes in the number of staff. The model can also be used in scenario analyses to determine necessary staffing and resource levels for a given (desired) service level and RTTT time, as well as to identify bottlenecks, suggest improvement and estimate the performance metrics (e.g. RTTT, throughput) of a new proposed system. This model allows for an impact assessment of potential solutions before real-life implementation. A similar approach, with some learning and set-up costs, can be used in most radiation treatment facility/departments, and this could be especially important for cancer centres in developing countries that are facing increased demand and limited resources. These results, combined with costing data, can help decision-makers in policy-making to determine costeffective ways to improve the system or determine a configuration of the system (resource and staffing levels) to achieve a desired level of service within a limited budget. Furthermore, real-life experimentation is costly. There is a one-time chance to collect data and observe results and making adjustments in the parameters affecting patient care is risky. It is also difficult to appreciate the impact of such decisions on real-life outcomes.
Acknowledgements The authors would like to thank all the staff at LRCP who provided feedback and contributed to this study. In particular, the authors thank Reza Mahjoub, Derrick Fournier and Tomasz Bielecki for providing considerable assistance throughout this project. Vusal Babashov was funded by the Natural Sciences and Engineering Research Council of Canada. Gregory Zaric was supported through the Canada Research Chairs Program.
References [1] Barua B, Rovere M, Skinner BJ. Waiting your turn: wait times for health care in Canada 2011 report. Studies in Health, Fraser Institute; 2011. [2] Kavanagh AG, Lee JC, Donnelly B. Time to treatment of prostate cancer through the Calgary Prostate Institute rapid access clinic. Can J Urol 2008;15(2):3975e3979. [3] Belyea J, Rigby M, Jaggi R, Hart RD, Trites J, Mark Taylor S. Wait times for head and neck cancer patients in the Maritime provinces. J Otolaryngol Head Neck Surg 2011;40(4):318e322. [4] Mackillop WJ, Zhou Y, Quirt CF. A comparison of delays in the treatment of cancer with radiation in Canada and the United States. Int J Radiat Oncol Biol Phys 1995;32(2):531e539. [5] Hebert-Croteau N, Freeman CR, Latreille J, Rivard M, Brisson J. A population-based study of the impact of delaying radiotherapy after conservative surgery for breast cancer. Breast Cancer Res Treat 2004;88(2):187e196. [6] Huang J, Barbera L, Brouwers M, Browman G, Mackillop WJ. Does delay in starting treatment affect the outcomes of radiotherapy? A systematic review. J Clin Oncol 2003;21(3):555e563. [7] Chen Z, King W, Pearcey R, Kerba M, Mackillop WJ. The relationship between waiting time for radiotherapy and clinical outcomes: a systematic review of the literature. Radiother Oncol 2008;87(1):3e16.
Please cite this article in press as: Babashov V, et al., Reducing Patient Waiting Times for Radiation Therapy and Improving the Treatment Planning Process: a Discrete-event Simulation Model (Radiation Treatment Planning), Clinical Oncology (2017), http://dx.doi.org/10.1016/ j.clon.2017.01.039
V. Babashov et al. / Clinical Oncology xxx (2017) 1e7 [8] Do V, Gebski V, Barton MB. The effect of waiting for radiotherapy for grade III/IV gliomas. Radiother Oncol 2000;57(2): 131e136. [9] Dahrouge S, Samant R, Mirzaei A, Price J. Radical radiotherapy for cervix cancer: the effect of waiting time on outcome. Int J Radiat Oncol Biol Phys 2005;61(4):1071e1077. [10] Fortin B, Goldberg MS, Mayo NE, Valois MF, Scott SC, Hanley J. Waiting time for radiation therapy in breast cancer patients in Quebec from 1992 to 1998: a study of surgically treated breast cancer patients in Quebec documents and helps to explain increased waiting times for radiation therapy. Healthcare Policy 2006;1(2):152e167. [11] Benk V, Przybysz R, McGowan T, Paszat L. Waiting times for radiation therapy in Ontario. Can J Surg 2006;49(1):16e21. [12] Benk V, Ho V, Fortin PR, Zhang G, Levinton C, Freeman CR. Predictors of delay in starting radiation treatment for patients with early stage breast cancer. Int J Radiat Oncol Biol Phys 1998;41(1):109e115. [13] Cooke AL, Appell R, Suderman K, Fradette K, Latosinsky S. Radiation treatment waiting times for breast cancer patients in Manitoba, 2001 and 2005. Curr Oncol 2009;16(5):58e64. € nal MM, Pidd M. Discrete event simulation for perfor[14] Gu mance modelling in health care: a review of the literature. J Simulation 2010;4(1):42e51. [15] Jacobson SH, Hall SN, Swisher JR. Discrete-event simulation of health care systems, Patient flow: reducing delay in healthcare delivery. New York: Springer; 2006. p. 211e252. [16] Santibanez P, Chow VS, French J, Puterman ML, Tyldesley S. Reducing patient wait times and improving resource utilization at British Columbia Cancer Agency’s ambulatory care unit through simulation. Health Care Manag Sci 2009;12(4):392e407. [17] VanBerkel PT, Blake JT. A comprehensive simulation for wait time reduction and capacity planning applied in general surgery. Health Care Manag Sci 2007;10(4):373e385.
7
[18] Everett JE. A decision support simulation model for the management of an elective surgery waiting system. Health Care Manag Sci 2002;5(2):89e95. [19] Duguay C, Chetouane F. Modeling and improving emergency department systems using discrete event simulation. Simulation 2007;83(4):311e320. [20] Kim S-C, Horowitz I, Young KK, Buckley TA. Analysis of capacity management of the intensive care unit in a hospital. Eur J Operational Res 1999;115(1):36e46. [21] Akkerman R, Knip M. Reallocation of beds to reduce waiting time for cardiac surgery. Health Care Manag Sci 2004;7(2): 119e126. [22] Munro A, Potter S. Waiting times for radiotherapy treatment: not all that mysterious and certainly preventable. Clin Oncol 1994;6(5):314e318. [23] Thomas S, Williams M, Burnet N, Baker C. How much surplus capacity is required to maintain low waiting times? Clin Oncol 2001;13(1):24e28. [24] Thomas S. Capacity and demand models for radiotherapy treatment machines. Clin Oncol 2003;15(6):353e358. [25] Kapamara T, Sheibani K, Petrovic D, Haas O, Reeves C. A simulation of a radiotherapy treatment system: a case study of a local cancer centre. ORP3 meeting; 2007. [26] Proctor S, Lehaney B, Reeves C, Khan Z. Modelling patient flow in a radiotherapy department. OR Insight 2007;20(3):6e14. [27] Werker G, Saure A, French J, Shechter S. The use of discrete-event simulation modelling to improve radiation therapy planning processes. Radiother Oncol 2009;92(1): 76e82. [28] Cancer Care Ontario Radiation Treatment Wait Time Targets. Available at: https://www.cancercare.on.ca/ocs/wait-times/ radiationwt/wttargets/ (accessed 28 May 2015). [29] Kelton WD, Sturrock DT. Simulation with Arena, 4th edn. New York: McGraw Hill; 2007.
Please cite this article in press as: Babashov V, et al., Reducing Patient Waiting Times for Radiation Therapy and Improving the Treatment Planning Process: a Discrete-event Simulation Model (Radiation Treatment Planning), Clinical Oncology (2017), http://dx.doi.org/10.1016/ j.clon.2017.01.039
Clinical Oncology journal homepage: www.clinicaloncologyonline.net
Original Article
Reducing Patient Waiting Times for Radiation Therapy and Improving the Treatment Planning Process: a Discrete-event Simulation Model (Radiation Treatment Planning) V. Babashov *, I. Aivas y, M.A. Begen zx, J.Q. Cao y, G. Rodrigues yx, D. D’Souza y, M. Lock y, G.S. Zaric zx * Telfer
School of Management, University of Ottawa, Ottawa, Ontario, Canada Department of Radiation Oncology, London Regional Cancer Program, London Health Sciences Centre, London, Ontario, Canada z Ivey Business School, Western University, London, Ontario, Canada x Department of Epidemiology and Biostatistics, Schulich School of Medicine and Dentistry, Western University, London, Ontario, Canada y
Received 31 May 2016; received in revised form 5 January 2017; accepted 10 January 2017
Abstract Aims: We analysed the radiotherapy planning process at the London Regional Cancer Program to determine the bottlenecks and to quantify the effect of specific resource levels with the goal of reducing waiting times. Materials and methods: We developed a discrete-event simulation model of a patient’s journey from the point of referral to a radiation oncologist to the start of radiotherapy, considering the sequential steps and resources of the treatment planning process. We measured the effect of several resource changes on the ready-to-treat to treatment (RTTT) waiting time and on the percentage treated within a 14 calendar day target. Results: Increasing the number of dosimetrists by one reduced the mean RTTT by 6.55%, leading to 84.92% of patients being treated within the 14 calendar day target. Adding one more oncologist decreased the mean RTTT from 10.83 to 10.55 days, whereas a 15% increase in arriving patients increased the waiting time by 22.53%. The model was relatively robust to the changes in quantity of other resources. Conclusions: Our model identified sensitive and non-sensitive system parameters. A similar approach could be applied by other cancer programmes, using their respective data and individualised adjustments, which may be beneficial in making the most effective use of limited resources. Ó 2017 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
Key words: Discrete-event simulation; modelling; radiotherapy; waiting time
Introduction A recent report showed that many Canadians are waiting for some form of medical treatment [1]. In particular, Canadian cancer patients have experienced long waiting times in radiotherapy for many years [2e4]. Lengthy waiting times for radiation treatment may have a negative clinical impact. For example, delayed radiation treatment may Author for correspondence: M.A. Begen, Ivey Business School, Western University, 1255 Western Road, London, Ontario, N6G 0N1, Canada. Tel: þ1-519-661-4146. E-mail address: [email protected] (M.A. Begen).
increase the risk of local recurrence [5e7] and poor survival [8,9]. Several population-based studies have been conducted to investigate waiting times for radiotherapy and to determine the predictors of long waits. A higher incidence of cancer, together with an increase in the demand for radiotherapy, insufficient resources and certain patient characteristics represent some of the predictors for longer waiting times [10e12]. Conversely, radiotherapy waiting times decrease with an increase in the number of radiation therapists, medical physicists, radiation oncologists and radiation planning and therapy equipment [13].
http://dx.doi.org/10.1016/j.clon.2017.01.039 0936-6555/Ó 2017 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
Please cite this article in press as: Babashov V, et al., Reducing Patient Waiting Times for Radiation Therapy and Improving the Treatment Planning Process: a Discrete-event Simulation Model (Radiation Treatment Planning), Clinical Oncology (2017), http://dx.doi.org/10.1016/ j.clon.2017.01.039
2
V. Babashov et al. / Clinical Oncology xxx (2017) 1e7
Discrete-event simulation (DES) is a valuable tool for investigating system capacity and throughput. The use of DES models with healthcare application includes hospitals, outpatient clinics, emergency departments and pharmacies [14,15]. DES can help decision-makers to carry out a ‘whatif?’ analysis to determine good policies for scheduling patients, optimising resources, reducing waiting times of patients in clinics and improving workflows [16,17]. DES models have also been used to investigate patient scheduling challenges [18], waiting time bottlenecks, overall system throughput and system configuration in emergency rooms [19], optimal intensive care unit size [20], as well as staffing levels and bed requirements [21] in various healthcare settings. Simulation modelling has been applied in the field of radiation therapy to explore target waiting times through varying capacities [22,23] and to analyse the number of linear accelerators to achieve shorter waiting times [24]. Kapamara et al. [25] and Proctor et al. [26] used DES modelling to understand the treatment process, complexities, patient flow and bottlenecks at the radiotherapy unit. More recently, Werker et al. [27] modelled a portion of the planning process of the radiation therapy at the British Columbia Cancer Agency, with the aid of a DES model. Our work extends this study [27] by modelling the entire radiation therapy planning process, from patient arrival to treatment completion. To the best of our knowledge, this is the first paper to model and analyse the entire radiotherapy planning process at a cancer treatment facility. The primary objectives of our study were to understand how to improve the radiotherapy planning process, to understand which resources were the most important in decreasing waiting times, to increase throughput of patients treated within the 14 calendar day target (set by the
Province of Ontario and the Canadian Association of Radiation Oncology) and to provide an optimal strategy for deploying existing and new resources.
Materials and Methods We modelled a patient’s complete journey from the point of referral to radiation oncology at the London Regional Cancer Program (LRCP) to the point of starting radiation therapy, accounting for every sequential step of the treatment planning process (Fig. 1). We used the Simul8 software package (SIMUL8 Corporation), which allows for visualisation and modelling of patient flow, queues and resource utilisation. This software can project flow times and identify system bottlenecks. Our study was approved by the London Health Sciences Research Ethics Board. Data Sources Patient-level process data from 2009, available through ‘in-house’ decision support and tracking software at LRCP, served as the primary data source for our study. After a patient is referred to the LRCP and receives a consultation, a file is opened for the patient and the status of the file is tracked through this software, showing each step and each healthcare professional who works on the case. We populated the model using 3888 radiation consultation records, which corresponded to all referred patients over a 1 year period (2009e2010). We consulted with LRCP staff members, including therapists, dosimetrists, physicists and administrative staff, to ensure the face validity of the model data extracted from the decision support and tracking software database.
Pre CT Sim Delay
Reject Treatment
No
PaƟent Arrival
Contour
RadiaƟon Oncologist Consult
Dosimetry
Accept Treat?
RadiaƟon Oncologist Approval
Therapist Contour
CT SimulaƟon Yes
Physics QA
Dosimetry QA
CT SimulaƟon QA
CalculaƟon StaƟon
Treatment
Fig 1. Treatment process flow diagram of key tasks.
Please cite this article in press as: Babashov V, et al., Reducing Patient Waiting Times for Radiation Therapy and Improving the Treatment Planning Process: a Discrete-event Simulation Model (Radiation Treatment Planning), Clinical Oncology (2017), http://dx.doi.org/10.1016/ j.clon.2017.01.039
V. Babashov et al. / Clinical Oncology xxx (2017) 1e7
To obtain data that was not available through the tracking database, we interviewed LRCP staff to estimate the duration of several tasks, including radiation oncologist consultation, computed tomography simulation scan, therapist contouring, oncologist contouring and oncologist approval processes. Other data estimates were obtained through model calibration. Baseline parameter estimates are shown in Table 1. Resources and Shifts We modelled the following resources involved in radiation therapy planning at LRCP: radiation oncologists, computed tomography simulation machines, dosimetrists, physicists, therapists and linear accelerators. We modelled 15 radiation oncologists who specialised in the following fields: genitourinary, central nervous system, lung, breast, skin, head and neck, gynaecology, thyroid, sarcoma, lymphoma and gastrointestinal. Each radiation oncologist worked on different shift types during a given week (e.g. clinic or consult shift, contouring shift and academic/ administration shift). Shifts were split into blocks of approximately 4 h. Based on current practice at LRCP, we found that oncologists devote some of their academic/administration time to contouring. To incorporate this phenomenon into our model, we adjusted the shift duration devoted to contouring and academic/administration work for respective oncologists. We modelled the availability of oncologists over the year, accounting for days off due to vacations, Table 1 Baseline model parameters Variable Resources (number of resources) Radiation oncologist Computed tomography simulation machine Dosimetrist Physicists Therapists Linear accelerator Incoming cancer patients’ profiles (n¼3888), % Breast Gastrointestinal Genitourinary Gynaecological Head and neck Lymphoma Sarcoma Skin Thorax Central nervous system Unknown primary Treatment priority, % 1 High 2 Medium 3 Low Proceed to treatment, % Accept Reject
Value 15 2 9 2 8 10 14 11 18 6 6 3 1 18 12 2 9 4 12 84 62 38
3
conferences and other factors, as determined through an examination of the actual LRCP staff schedule. Nine dosimetrists, two physicists, two computed tomography simulation machines and eight computed tomography simulation therapists worked 5 days a week on an 8 h shift. Additional physicists and therapists were employed at the centre but were not assigned to patient care and hence they were not included in the model. Only those physicists and therapists who were actively assigned to patient care and tasks in the planning process were explicitly modelled and included in the simulation. There were 10 linear accelerators operating at the LRCP. Based on current practice and historical information, we assumed 10% downtime, including regular maintenance. The duration of patient treatment sessions varied between 10 minutes and 1 hour, and the number of treatment sessions varied between one and 60. For the modelling purposes, we represented the capacity of the linear accelerator in terms of 5 minutes treatment slots. Treatment Process Flow and Patient Attributes The radiation treatment planning process at the LRCP included consultation with a radiation oncologist, computed tomography simulation scan, therapist contour, radiation oncologist contouring, computed tomography simulation quality assurance, dosimetry planning, radiation oncologist approval and physics quality assurance. We modelled the arrival of patients using the referral date and time of each referred patient. The process flow is depicted in Fig. 1. Upon referral to the LRCP for a consultation, patients were assigned to an available radiation oncologist who could treat the patient’s specific cancer type. From this point on, the patient’s work file remained attached to this physician, and the patient’s waiting times were influenced by the assigned doctor’s availability. After seeing the radiation oncologist, the patient could then decide whether to proceed to radiation therapy, possibly after a planned delay, or to leave the system. We modelled several patient attributes including cancer type, number of treatment sessions required and priority level. Patients with level I priority were deemed an emergency, and the goal was to treat within 1 day. Patients with level II priority were deemed urgent, and the goal was to treat within 7 days. The remaining patients were level III and had a goal of treatment within 14 days. Performance Metric and Waiting Time Target for Ontario We used ready-to-treat to treatment (RTTT) as our main outcome measure. In Canada, this is a commonly used performance measure for assessing waiting time performance. Ready-to-treat is the time point when the patient is deemed ready to start treatment and all planned delays, such as the completion of other treatments or patient unavailability, have been resolved. All cancer centres in Ontario are required to report their patients’ mean RTTT and the proportion seen within 14 calendar days [28].
Please cite this article in press as: Babashov V, et al., Reducing Patient Waiting Times for Radiation Therapy and Improving the Treatment Planning Process: a Discrete-event Simulation Model (Radiation Treatment Planning), Clinical Oncology (2017), http://dx.doi.org/10.1016/ j.clon.2017.01.039
4
V. Babashov et al. / Clinical Oncology xxx (2017) 1e7
In addition to measures based on RTTT, we defined ‘utilisation’ as the proportion of time that any given resource was occupied. We measured utilisation for the following resources: radiation oncologists, dosimetrists, therapists, physicists, computed tomography simulation and linear accelerators. Validation The model was validated by comparing the mean RTTT, the proportion of patients treated within 14 calendar days and the 90th percentile of the waiting time distribution for treated patients for the simulation model versus corresponding values from the actual system. As we modelled a system wherein clinics and treatment facilities were saturated and resources were busy (i.e., as opposed to a newly opened facility), a ‘warm-up’ period [29] was necessary as part of the simulation model to allow the model to function at a capacity representative of an actively running centre. The warm-up analysis showed that the system reached a steady state after 208 days (5000 hours) of simulation time. Using 2 years of patient arrival data, we ran the model for a period of 30 replications, each with a warm-up period of 208 days and a run length of 500 days. Validation consisted of a direct comparison of the end point parameters produced by the simulation and the actual end point data.
Results Base Case Results and Validation We compared the model outputs to the actual system waiting time performance at LRCP (Table 2). The simulation model yielded a mean RTTT of 10.83 days (95% confidence interval 10.61, 11.05), which was virtually identical to the observed RTTT of 10.82 days (95% confidence interval 10.25, 11.37) at LRCP. The observed proportion of patients meeting the 14 day threshold was 83.59%, close to the simulation model’s result of 82.25%. The 90th percentile of the observed waiting time distribution was also close to that of the simulation model (21 and 19.87, respectively). The close correspondence between the simulated and observed outcomes gave us confidence that we had modelled the system correctly and that scenario analyses would provide reasonable approximations of expected system behaviour. Resource Utilisation Resources with utilisation over 30% are summarised in Fig. 2. Among all resources, the highest utilised was the pool of nine dosimetrists, who were utilised on average for 88% of the available time. The linear accelerators were busy 80% of the available time. The aggregate utilisation of two physicists
Table 2 Comparison of simulation versus actual system performance Actual system waiting time performance Mean (days) 90th percentile (days) % 14 days
Simulation model outputs 10.82 21 83.59%
Mean (days) 90th percentile (days) % 14 days
10.83 19.87 82.25%
Fig 2. Radiation Oncologists’ (ROs’) Specialty in the DES Model: Onc 15 (TH, SK, BR), Onc 13 (BR, SK, LYMP), Onc 6 (GYN, SARC & H&N) and Onc 11 (BR, GI, GU). Please cite this article in press as: Babashov V, et al., Reducing Patient Waiting Times for Radiation Therapy and Improving the Treatment Planning Process: a Discrete-event Simulation Model (Radiation Treatment Planning), Clinical Oncology (2017), http://dx.doi.org/10.1016/ j.clon.2017.01.039
V. Babashov et al. / Clinical Oncology xxx (2017) 1e7
was 52% of the available time. The most highly utilised radiation oncologist was number 15, with 51%, who had the skill set of treating breast, skin and thyroid cancer. Although we have attempted to fully capture all activities of human resources, there are probably some activities that have been missed (e.g. training students, overtime to get work done, use of academic hours for patients) as these may not be formally recorded. Thus, the utilisation values obtained from our models could be underestimating the actual utilisation levels. Sensitivity and Scenario Analysis Adding one dosimetrist to the resource pool had the greatest e albeit modest e impact on improving the number of patients treated within the 14 calendar day target, increasing the proportion of patients meeting this target from 82.25 to 84.92%. Removing one physicist from the resource pool had the greatest impact on reducing the proportion of patients treated within the set target, thus reducing the percentage of patients treated within 14 calendar days to 18.62%. Other individual scenarios, such as removing one dosimetrist, decreasing linear accelerator capacity by 15% and increasing patient arrivals by 15%, delivered a relatively significant impact on the proportion of patients treated within the target timeframe (see Table 3). Variation in the quantity of the remaining resources yielded non-significant effects on the proportion of patients meeting the target timeframe. Similarly, adding one dosimetrist had the largest impact on the mean RTTT, decreasing the average waiting time to 10.12 days. Removing one physicist had a dramatic effect on the system, increasing the mean RTTT to 31.35 days. In addition, other individual scenarios (e.g. one less dosimetrist, a 15% decrease in linear accelerator capacity and a 15% increase in patient arrivals) significantly worsened the mean RTTT, by 31.57, 11.44 and 22.53%, respectively. Changes in the remaining resources resulted in minor impacts on average waiting time.
5
Discussion We developed a DES model of the entire radiotherapy planning process at LRCP and conducted several ‘what-if?’ scenarios. Scenario analysis showed that adding one type and one more unit of a resource does not significantly improve the average RTTT or the percentage of patients treated within a 14 calendar day target. The largest improvement is achieved with the addition of a dosimetrist; 6.5% and 3.2% improvements in RTTT and the percentage of patients treated within a 14 calendar day target, respectively. On the other hand, decreasing certain resources and increasing demand both have a strong negative impact on system performance, suggesting that the LRCP does not have surplus capacity in dosimetrists, physicists or linear accelerators. These results suggest that the system is currently running well and the level of each resource is in balance with levels of other resources. The results also show that the system is running close to its capacity and thus needs to be monitored closely for changes in performance levels. Finally, to improve the system’s performance significantly one needs to consider the addition of multiple types and levels of resources. At the BC Cancer Agency, Werker et al. [27] modelled the radiotherapy treatment planning process and found that it was the variability and length of radiation oncologistrelated tasks that had the largest effect on treatment planning time. They also found that having one fewer dosimetrist-equivalent resource negatively affected their waiting times, but an additional dosimetrist had no significant effect on shortening waiting times. These findings are consistent with our observations, further validating our model and supporting our conclusions. Our model has several limitations. First, the model cannot capture human behaviour and response to incentives. For example, staff may stay at work after hours in order to meet deadlines and complete tasks on particularly concerning or complex cases. As the deadline (i.e. 14
Table 3 Results of simulation scenarios Scenarios
Mean RTTT in days (95% confidence interval) (baseline 10.83 days)
Percentage treated within 14 calendar day target (95% confidence interval) (baseline 82.25%)
One more dosimetrist One less dosimetrist One more physicist One less physicist One more therapist One less therapist One more computed tomography simulation machine One less computed tomography simulation machine One more oncologist* 15 % increase in linear accelerator 15% decrease in linear accelerator 15% increase in arrivals
10.12 14.25 10.73 31.35 10.82 10.90 10.82 10.92 10.55 10.82 12.07 13.27
84.92 58.88 82.59 18.62 82.37 81.99 82.26 82.01 81.98 82.27 74.43 67.47
(9.89, 10.35) (13.78, 14.73) (10.51, 10.95) (29.26, 33.44) (10.60, 11.04) (10.69, 11.11) (10.60, 11.04) (10.69, 11.15) (10.28, 10.82) (10.61, 11.04) (11.60, 12.53) (12.80, 13.74)
(84.38, (56.07, (81.95, (17.34, (81.70, (81.33, (81.60, (81.26, (81.20, (81.56, (72.07, (64.71,
85.46) 61.99) 83.22) 19.90) 83.03) 82.66) 82.92) 82.76) 82.75) 82.97) 76.78) 70.23)
RTTT, ready-to-treat to treatment. * With the specialisation of skin and genitourinary cancer. Please cite this article in press as: Babashov V, et al., Reducing Patient Waiting Times for Radiation Therapy and Improving the Treatment Planning Process: a Discrete-event Simulation Model (Radiation Treatment Planning), Clinical Oncology (2017), http://dx.doi.org/10.1016/ j.clon.2017.01.039
6
V. Babashov et al. / Clinical Oncology xxx (2017) 1e7
calendar days) approaches, patients may be moved forward in the queue, thereby starting the radiation treatment before the recommended deadline. Furthermore, as LRCP is an academic centre, the presence of resident trainees could either increase or decrease the time spent on a task. Centres with different deadlines or complexity of cases may require different times for each discrete event on the patient care pathway. Second, we determined and report two performance criteria (average RTTT waiting time and percentage treated within the 14 calendar day target) at the aggregate level for all patients and not by cancer type or by priority level. Third, the decision support and tracking software at LRCP is based on process flow, reflecting the amount of time a patient’s treatment plan spends at a certain work centre, rather than the actual amount of time spent by human resources to execute the task. This is a crucial issue regarding dosimetrists’ tasks specifically; in the event of an obstacle or uncertainty during the planning process, the task at hand is put on hold until consultation with the radiation oncologist or physicist can take place. In the meantime, the dosimetrist starts work on another patient’s treatment plan, but the tracking software continues to record that the previous task remains ongoing. This is an important limitation, not only of the model, but also of the data collection system identified in our study. About 38% of patients referred to LRCP for consultation in 2009e2010 either chose not to receive the radiotherapy at LRCP or were deemed ineligible for therapy. Based on anecdotal evidence, this rate seems to be higher than that in other centres. This situation may reflect a high level of second-opinion consultations or patients deciding to undertake treatment in other centres, or a high number of inappropriate referrals. This observation was discussed with the radiotherapy unit administration and noted as an area for future research. Centres with a different patient intake and referral populations may not be able use this model directly; adjustments to the model would have to be made before the simulation could be applied. In future work, we will explore modelling, reporting and validation of two performance criteria (average RTTT waiting time and percentage treated within the 14 calendar day target) by cancer types and priority levels of patients. We will also investigate simultaneous adjustments of resources to observe the extent to which the effect is synergistic versus additive and consider the possibility of having some steps carried out in parallel instead of strictly in sequence. Future work will also consider cost considerations. As is the case with every department, allocation of funds to various resources or work centres requires intelligent planning for future benefit and gain to the department. Investing in a new piece of equipment or specialised human resources is very costly. Our model can be adjusted to investigate whether a radiation therapy department, based on its caseload and history of referrals, would benefit financially from investing in a new technology or treatment platform. This kind of analysis would also allow us to investigate the efficiency of the system. Our model, which was developed and validated to analyse the entire radiotherapy planning process at a cancer
treatment facility, can be used to assess the impact of equipment levels, staff assignments and absences, changes in patient demand and changes in the number of staff. The model can also be used in scenario analyses to determine necessary staffing and resource levels for a given (desired) service level and RTTT time, as well as to identify bottlenecks, suggest improvement and estimate the performance metrics (e.g. RTTT, throughput) of a new proposed system. This model allows for an impact assessment of potential solutions before real-life implementation. A similar approach, with some learning and set-up costs, can be used in most radiation treatment facility/departments, and this could be especially important for cancer centres in developing countries that are facing increased demand and limited resources. These results, combined with costing data, can help decision-makers in policy-making to determine costeffective ways to improve the system or determine a configuration of the system (resource and staffing levels) to achieve a desired level of service within a limited budget. Furthermore, real-life experimentation is costly. There is a one-time chance to collect data and observe results and making adjustments in the parameters affecting patient care is risky. It is also difficult to appreciate the impact of such decisions on real-life outcomes.
Acknowledgements The authors would like to thank all the staff at LRCP who provided feedback and contributed to this study. In particular, the authors thank Reza Mahjoub, Derrick Fournier and Tomasz Bielecki for providing considerable assistance throughout this project. Vusal Babashov was funded by the Natural Sciences and Engineering Research Council of Canada. Gregory Zaric was supported through the Canada Research Chairs Program.
References [1] Barua B, Rovere M, Skinner BJ. Waiting your turn: wait times for health care in Canada 2011 report. Studies in Health, Fraser Institute; 2011. [2] Kavanagh AG, Lee JC, Donnelly B. Time to treatment of prostate cancer through the Calgary Prostate Institute rapid access clinic. Can J Urol 2008;15(2):3975e3979. [3] Belyea J, Rigby M, Jaggi R, Hart RD, Trites J, Mark Taylor S. Wait times for head and neck cancer patients in the Maritime provinces. J Otolaryngol Head Neck Surg 2011;40(4):318e322. [4] Mackillop WJ, Zhou Y, Quirt CF. A comparison of delays in the treatment of cancer with radiation in Canada and the United States. Int J Radiat Oncol Biol Phys 1995;32(2):531e539. [5] Hebert-Croteau N, Freeman CR, Latreille J, Rivard M, Brisson J. A population-based study of the impact of delaying radiotherapy after conservative surgery for breast cancer. Breast Cancer Res Treat 2004;88(2):187e196. [6] Huang J, Barbera L, Brouwers M, Browman G, Mackillop WJ. Does delay in starting treatment affect the outcomes of radiotherapy? A systematic review. J Clin Oncol 2003;21(3):555e563. [7] Chen Z, King W, Pearcey R, Kerba M, Mackillop WJ. The relationship between waiting time for radiotherapy and clinical outcomes: a systematic review of the literature. Radiother Oncol 2008;87(1):3e16.
Please cite this article in press as: Babashov V, et al., Reducing Patient Waiting Times for Radiation Therapy and Improving the Treatment Planning Process: a Discrete-event Simulation Model (Radiation Treatment Planning), Clinical Oncology (2017), http://dx.doi.org/10.1016/ j.clon.2017.01.039
V. Babashov et al. / Clinical Oncology xxx (2017) 1e7 [8] Do V, Gebski V, Barton MB. The effect of waiting for radiotherapy for grade III/IV gliomas. Radiother Oncol 2000;57(2): 131e136. [9] Dahrouge S, Samant R, Mirzaei A, Price J. Radical radiotherapy for cervix cancer: the effect of waiting time on outcome. Int J Radiat Oncol Biol Phys 2005;61(4):1071e1077. [10] Fortin B, Goldberg MS, Mayo NE, Valois MF, Scott SC, Hanley J. Waiting time for radiation therapy in breast cancer patients in Quebec from 1992 to 1998: a study of surgically treated breast cancer patients in Quebec documents and helps to explain increased waiting times for radiation therapy. Healthcare Policy 2006;1(2):152e167. [11] Benk V, Przybysz R, McGowan T, Paszat L. Waiting times for radiation therapy in Ontario. Can J Surg 2006;49(1):16e21. [12] Benk V, Ho V, Fortin PR, Zhang G, Levinton C, Freeman CR. Predictors of delay in starting radiation treatment for patients with early stage breast cancer. Int J Radiat Oncol Biol Phys 1998;41(1):109e115. [13] Cooke AL, Appell R, Suderman K, Fradette K, Latosinsky S. Radiation treatment waiting times for breast cancer patients in Manitoba, 2001 and 2005. Curr Oncol 2009;16(5):58e64. € nal MM, Pidd M. Discrete event simulation for perfor[14] Gu mance modelling in health care: a review of the literature. J Simulation 2010;4(1):42e51. [15] Jacobson SH, Hall SN, Swisher JR. Discrete-event simulation of health care systems, Patient flow: reducing delay in healthcare delivery. New York: Springer; 2006. p. 211e252. [16] Santibanez P, Chow VS, French J, Puterman ML, Tyldesley S. Reducing patient wait times and improving resource utilization at British Columbia Cancer Agency’s ambulatory care unit through simulation. Health Care Manag Sci 2009;12(4):392e407. [17] VanBerkel PT, Blake JT. A comprehensive simulation for wait time reduction and capacity planning applied in general surgery. Health Care Manag Sci 2007;10(4):373e385.
7
[18] Everett JE. A decision support simulation model for the management of an elective surgery waiting system. Health Care Manag Sci 2002;5(2):89e95. [19] Duguay C, Chetouane F. Modeling and improving emergency department systems using discrete event simulation. Simulation 2007;83(4):311e320. [20] Kim S-C, Horowitz I, Young KK, Buckley TA. Analysis of capacity management of the intensive care unit in a hospital. Eur J Operational Res 1999;115(1):36e46. [21] Akkerman R, Knip M. Reallocation of beds to reduce waiting time for cardiac surgery. Health Care Manag Sci 2004;7(2): 119e126. [22] Munro A, Potter S. Waiting times for radiotherapy treatment: not all that mysterious and certainly preventable. Clin Oncol 1994;6(5):314e318. [23] Thomas S, Williams M, Burnet N, Baker C. How much surplus capacity is required to maintain low waiting times? Clin Oncol 2001;13(1):24e28. [24] Thomas S. Capacity and demand models for radiotherapy treatment machines. Clin Oncol 2003;15(6):353e358. [25] Kapamara T, Sheibani K, Petrovic D, Haas O, Reeves C. A simulation of a radiotherapy treatment system: a case study of a local cancer centre. ORP3 meeting; 2007. [26] Proctor S, Lehaney B, Reeves C, Khan Z. Modelling patient flow in a radiotherapy department. OR Insight 2007;20(3):6e14. [27] Werker G, Saure A, French J, Shechter S. The use of discrete-event simulation modelling to improve radiation therapy planning processes. Radiother Oncol 2009;92(1): 76e82. [28] Cancer Care Ontario Radiation Treatment Wait Time Targets. Available at: https://www.cancercare.on.ca/ocs/wait-times/ radiationwt/wttargets/ (accessed 28 May 2015). [29] Kelton WD, Sturrock DT. Simulation with Arena, 4th edn. New York: McGraw Hill; 2007.
Please cite this article in press as: Babashov V, et al., Reducing Patient Waiting Times for Radiation Therapy and Improving the Treatment Planning Process: a Discrete-event Simulation Model (Radiation Treatment Planning), Clinical Oncology (2017), http://dx.doi.org/10.1016/ j.clon.2017.01.039