Outpatient appointment scheduling: policies for a chronic patient population

(1)

Outpatient appointment scheduling: policies for a

chronic patient population

Author:

Siard Runia

Student no.:

2813203

Email:

s.runia@student.rug.nl

Master program:

Technology and Operations Management

Institution:

University of Groningen

Supervisor:

Dr. ir. D.J. van der Zee

Second assessor:

Dr. H. Broekhuis

Company supervisors:

ir. T.J.J. Hoogstins

Mrs. M. Nuus

(2)

2

Abstract

Purpose: The objective of this research is to develop slot policies for the appointment scheduling of multiple chronic patient groups, that balance their timely access to care services, and staff resource efficiency. This research addresses a class of outpatient appointment systems that are characterized by a chronic heterogeneous patient population based on admission times. It aims to provide insights to efficiently match the patient population with the available capacity by means of planning windows and capacity allocation.

Method: For this research, a design science approach is used to create a viable artifact in the form of scheduling policies. Beforehand, a literature review is conducted regarding outpatient appointment systems and its elements. After which the design-science approach is executed to determine the system performance and develop redesigns to improve its performance. The proposed redesigns are tested by means of a simulation study.

Findings: Simulation results show that assuming variable availability of physicians decreases the performance of simulation outputs. The results also show that constant slot reservation for patients throughout the week is not effective when the arrival rate is variable. Symmetrical planning windows do not improve the performance compared to the default scenario. Asymmetrical windows improved the performance of the short-term and new patients, whereas acute patients don’t seem to benefit.

Conclusion: A general conclusion is that asymmetrical and symmetrical planning windows don’t improve the performance quite as much as expected. It seems that the current physician schedule and patient slot reservation is inconsistent with patient's demand/arrival rate. This is detrimental for timely access and resource efficiency.

Recommendations: Gain more accurate insight in the arrival rate of patients and structure the physician schedule, and patient slot reservation accordingly.

(3)

3

Preface

This thesis completes my student career and embarks the start of my professional career. The goal of this research is to develop an appointment system that will aid outpatient clinics in providing quick, or timely health services, and quality care, for their patients.

The process of writing this research paper was a great and amazing educational experience. I learned how to apply the design-science approach, which creates a bridge between theory and practice, and develops an artifact. Developing the simulation model improved my programming skills greatly, and taught me how to translate a real-life situation into a simulation model. The abstract thinking, which I enhanced during this process, will be a great tool for my future career.

I really enjoyed the process of my Master Thesis and it was a great experience. This was not possible without the help of certain people. First and foremost, I would like to thank Durk-Jouke van der Zee. His feedback was to the point, accommodating, and essential for the completion of his thesis. This was an important aspect during my thesis project and motivated me every time. The same goes for Ms. Broekhuis, who provided useful feedback on my earlier draft. I also thank my fellow students of the healthcare thesis theme group, who were always available for feedback or a discussion.

From the UMCG I thank Mr. Hoogstins and Ms. Nuus for the time they made available for our meetings. Their experience and understanding of the subject provided essential feedback and remarks for my thesis.

At last I would like to thank my friends and family for their support during this period. I would like to thank my brother Gerryt-Douwe Runia in particular, who always made time available to review my work and provide me with insightful comments.

Groningen, 26 June 2017

(4)

4

1. Introduction

Outpatient clinics have become more central in healthcare systems due to an emphasis on preventive medical practices, shorter hospital stays, and service provision on an outpatient basis (Cayirli & Veral, 2003). Well-designed outpatient appointment systems (OASs) have the potential to increase the utilization of expensive personnel and equipment-based medical resources (Cayirli & Veral, 2003). Appointment scheduling systems lie at the intersection of efficiency and timely access to health services (Gupta & Denton, 2008; Ahmadi-Javid, Jalali & Klassen, 2017). Not receiving timely access can pose a serious safety concern (Murray and Berwick, 2003).

This research is motivated by a scheduling problem faced by the outpatient inflammatory bowel disease (IBD) clinic at the University Medical Center of Groningen (UMCG). The clinics patient population is heterogeneous and consists out of three groups, i.e., new patients, urgent patients and control patients, each setting specific demands to their admission times. Admission time is the time between the patient’s call for an appointment and the scheduled appointment time. The patients alternate between the groups urgent- and control patients depending on flare-ups. When a patient experiences a flare-up, it is important to receive timely access. The planner struggles to efficiently match the patient population with the available capacity. The capacity of the clinic and the physicians are variable due to illness, holidays and congresses. This study will investigate how an OAS can be designed to efficiently plan a chronic heterogeneous patient population, with regards to admission times, where patients alternate between groups, with variable physician capacity.

(7)

7 patient groups, to schedule the patient population. Results showed that reserving time slots for urgent patients shortened admission times, i.e. improved performance.

Hoogstins (2008), Van Buizen (2014), and Gijsel (2016) further explored the capacity allocation through slot policies with respect to heterogeneous patient populations based on admission times. They provided evidence that slot policies can improve admission times. However, Van Buizen (2014) and Gijsel (2016) assumed constant physician capacity and they suggested to further explore slot policy configurations by allowing for e.g. asymmetrical planning windows for control patients. The aim of this research is to explore different slot policy configurations with regards to admission times and staff efficiency, thereby seeking to further exploit the notion of planning windows, whilst assuming variable physician capacity (i.e. illness, holidays or congresses). The objective of the research is formulated as follows:

The objective of this research is to develop slot policies for the appointment scheduling of multiple chronic patient groups, that balance their timely access to care services, and staff resource efficiency.

For this research, a design science approach is used to create a viable artifact in the form of slot policies (Von Alan et al., 2004) based on the guidelines set by Van Strien (1997). This approach guides the steps followed during the research: problem definition, system description, system analysis, design, testing and implementation. The thesis will be structured accordingly. The IBD clinic will serve as a case example. Computer simulation will be employed for testing alternative slot policies.

This research will contribute to the existing literature of outpatient appointment scheduling with regards to admission times, chronic heterogeneous patient populations and variable resource capacity. The practical contribution will be slot policies for outpatient clinics that are characterized by a chronic heterogeneous patient population with respect to admission times and variable resource capacity.

(8)

8

2. Problem statement and research set-up

This chapter gives an overview of the overall research. Chapter 2.1 will introduce the research background. Chapter 2.2 will clarify the research objective and chapter 2.3 the conceptual model. Chapter 2.4 will discuss the research design and section 2.5 concludes with data sources.

2.1 Research background

In the UMCG a new care path was applied which lead to an increase in the patient population of the IBD clinic. The increase of the patient population created a challenge for the scheduler as he has to plan a larger patient population with the same capacity whilst ideally fulfilling timely service for the patients. Currently the clinic is unable to provide timely health services for acute and new patients. The UMCG clinic will be used as a case example to address a class of systems. It will provide a clinic design and provides required data to perform a simulation study and test alternative slot policies.

Initial research of Hoogstins (2008) developed insights on how the serve this increasing patient population. Van Buizen (2014) continued, based on the findings of Hoogstins (2008), with researching the possibilities of slot policies and their effect on performance. A general conclusion of Van Buizen (2014) his research is that slot policies can be used to efficiently schedule the patient population with respect to admission times and the available capacity.

(9)

9 wherein the appointment has to be scheduled). It will look at different slot policies and their effect on the performance of the OAS based on a simulation study.

2.2 Research objective

The objective of this research is to develop slot policies for the appointment scheduling of multiple chronic patient groups, that balance their timely access to care services, and staff resource efficiency.

Timely access of patients’ service is related to norms set on admission times for each of the patient groups. Admission times refer to the times between the days patients call for an appointment and the actual appointment dates (Cayirli, Veral & Rosen 2008). There exist different criteria per patient group with respect to the admission times. Control patients require regularity in their admission time in term of pre-specified treatment intervals (e.g. once every half year) and are ideally scheduled close to the ideal appointment date. New and urgent patients require almost immediate health service and require short admission times (i.e. in a matter of days).

The staff resource efficiency focuses on the capacity of physicians. It will emphasize the utilization of expensive personnel, i.e. the total time scheduled for consultation versus the actual time in consultation. It is important to note that the capacity of physicians is not constant and fluctuates over time due to illnesses, congresses and holidays.

2.3 Conceptual model

(10)

10

Figure 1: Conceptual model.

2.4 Research design

The objective of this research is to develop an OAS. Therefore, a design science approach will be used to create such an artifact (Von Alan et al., 2004) where other research methodologies do not allow for this. This approach describes a set of steps to follow during the research: (1) problem definition, (2) system description, (3) system analysis, (4) design, (5) evaluation and (6) implementation (Van Strien, 1997). Table 1 presents all the steps as described by the regulative cycle of Van Strien (1997) and are elaborated after presenting the table.

The case example, to address a class of systems, is the IBD clinic at the UMCG. It supplies data on the system set-up and operations. Their clinic setup is used as a test environment for the proposed redesigns in chapter 6. The IBD clinic at the UMCG characterizes itself with a heterogeneous patient population that require chronic health services. The patients alternate between the different patient groups based on the state of their disease. The population is served by physicians that have variable availability. A complete system description will be given in chapter 4.

(11)

11 windows. This step is not explicitly mentioned in the regulative cycle but is incorporated and will be discussed in chapter 3.

Step in design cycle Chapter Research question or purpose of respective chapter

3

Theoretical background that discusses outpatient clinics, heterogeneous patient populations, capacity allocation and planning windows.

I System description 4

What are the main characteristics of the system and how is appointment scheduling organized?

II System analysis 5 What is the current performance of the system and _{what influences the performance?}

III Design 6 How can the current system be redesigned to improve _{the performance?}

IV Evaluation 7

How do the alternative slot policies for the system perform with regards to timely service and resource effectiveness?

8 Conclusions, recommendations, limitations and future _{research suggestions.}

Table 1: Research questions per step in the design cycle.

I. System description: this step will give a description of the IBD clinic and its characteristics to specify the problem and the current performance of the system. The system will be characterized based on contextual factors regarding OASs. This is done by observations of the actual system and by interviews with actors in the system. Previous studies that used the IBD clinic as a research vehicle will also attribute to the system description by providing the initial description of the system which can be enhanced or changed.

II. System analysis: this step will explore to which extent the current setup of the appointment system succeeds in satisfying its objectives. This step will be executed by investigating company and performance data, literature and unstructured interviews with actors in the IBD clinic.

(12)

12 Van Buizen (2014) and Gijsel (2016). They provided recommendations that will improve the performance of the OAS and suggestions for future research that should be explored.

IV. Evaluation: the proposed slot policies will be evaluated with a simulation. A simulation is used because probability distributions are used to mimic the behavior of key elements in the system to carry out numerical experiments where the system’s behavior is simulated over a certain period (Zonderland, 2014). For the simulation, it is essential to collect data that can be used for the conceptual modelling of the simulation. The conceptual model attributes to the data requirements, the validity of the model, and the confidence that is placed in the simulation results (Robinson, 2004). This data will be collected in step 1 and 2 of the design cycle. Finally, the simulation will compare the default situation with alternative configurations based on relevant performance indicators.

2.5 Data sources

(13)

13

3. Theoretical background

This chapter will provide a theoretical background on outpatient appointment systems and aspects mentioned in the conceptual model. Section 3.1 will introduce a framework to typify outpatient appointment scheduling systems and classify the system in research. Section 3.2 will elaborate on heterogeneous patient populations. Section 3.3 will zoom in on appointment scheduling regarding capacity allocation and planning windows. Section 3.4 concludes with an overview of the findings.

3.1 Outpatient appointment scheduling systems

Outpatient clinics have become more central in healthcare systems due to an emphasis on preventive medical practices, shorter hospital stays, and service provision on an outpatient basis (Cayirli & Veral, 2003). The role of outpatient appointment systems is to smooth out the variability in demand to match the existing capacity (Cayirli & Gunes, 2014) and increase the utilization of expensive personnel and resources (Cayirli & Veral, 2003).

(14)

14

Decision type Decision level Name

Design decisions Strategic Access policy

Number of servers/resources Policy on acceptance of walk-ins Type of scheduling

Planning decisions Tactical Allocation of capacity

Appointment interval (slot policies) Appointment scheduling window Block size

Number of appointments in consultation session Panel size

Priority of patient groups

Operational Allocation of patients to servers/resources

Appointment day Appointment time

Patient acceptance/rejection Patient selection from waiting list Patient sequence

Environmental factors Patient preference

Patient heterogeneity Random service time

Type of appointment required by patients

Table 2: OAS characteristics framework of Ahmadi-Javid, Jalali & Klassen (2017).

The framework is structured according to three decision types and their respective decision level(s). The design decisions, which are on a strategic level, are usually regarded as the inputs and determine the main structure of an OASs. The second decision type considers the planning decisions consisting out of the tactical- and operational decision levels. Tactical decisions determine how patients as a whole are scheduled or how groups of patients are processed. Their goal is to maximize resource utilization and to ensure accessibility of health services. The operational decision level reflects the individual patient level and their respective scheduling. The last decision level considers the environmental decisions. These decisions relate to internal and external environmental factors.

(15)

15 known, such as consultation time and treatment type based upon which classifications can be made (Ahmadi-Javid et al., 2017). Capacity allocation deals with the problem of how the available capacity should be divided among different patient groups (Ahmadi-Javid et al., 2017). Section 3.2 will further elaborate on literature regarding a heterogeneous patient population whereas section 3.3 will further explore literature about capacity allocation (i.e. slot policies). The remainder of the classifications from the framework in table 2 will not be discussed in the theoretical background. They are further elaborated and explained in the system description in chapter 4. For an extensive explanation of the framework and all the characteristics I refer to the article of Ahmadi-Javid et al. (2017).

3.2 Heterogeneous patient population

According to Cayirli & Veral (2003), the majority of studies regarding OASs assume a homogeneous patient population instead of a heterogeneous patient population. Gupta and Denton (2008) also suggested to further research the development of multi patient group scheduling and capacity reservation models that account for the variability among classes in patients’ needs and resource requirements. Acknowledging heterogeneous patient populations and therefore multiple patient groups raises the issue whether this distinction can improve appointment systems (AS). The requirement is that the patient population can be distinctly classified into groups based on consultation time characteristics, such as new/return procedures (Cayirli, Veral, & Rosen, 2008; Ahmadi-Javid et al., 2017). Also, patient characteristics like priority level, consultation time, disease type, treatment type and no-show probability are often known. Zacharias & Pinedo (2014) concluded that acknowledging patients’ heterogeneity has a signiﬁcant impact on the optimal schedule and should be taken under consideration.

(16)

16 improved, often by the use of simple heuristic policies (Green, Savin, & Wang, 2006). Unfortunately, it seems that there isn’t a single overall capacity management policy that works well under all circumstances. Heuristics heavily depend on the characteristics of the healthcare facility and policy decisions of the facility.

This research addresses outpatient clinics with a chronic heterogeneous patient population. According to Zonderland (2014) you can distinguish three different patient types based on urgency: elective patients, urgent patients and patients that are neither elective nor urgent. This research distinguishes patient classifications with regards to the urgency in admission times. Indirect waiting times, also admission times, refer to the times between the days patients call for an appointment and the actual appointment dates (Liu, Ziya & Kulkarni, 2010). Figure 2 depicts a basic outpatient clinic and its time-based performance indicators among which the admission time. The admission time, as depicted in figure 2, is the time between the patient request and the planned starting time. This definition of admission time will be used throughout the research.

Figure 2: Basic outpatient clinic and time-based performance indicators (Zonderland, 2014).

(17)

17 frequently recommended admission times as an important performance indicator and a direction for future research (Gupta & Denton, 2008; Ahmadi-Javid et al., 2017). Literature about admission times is, to the author’s knowledge, still very scarce and a topic that is not addressed enough in the OAS literature. Addressing admission times as a performance measure is important because timely access, and thus short admission times, is important for realizing good medical outcomes (Murray and Berwick, 2003; Gupta & Denton, 2008). Even more so when the system serves a chronic patient population which requires regularity in the admission time. Acknowledging different patient groups conveys the issue on how the available capacity should be divided among these groups to generate an efficient appointment system. This will be addressed in the subsequent section.

3.3 Appointment scheduling

Appointment scheduling deals with the problem of how the available capacity should be divided among different patient groups (Ahmadi-Javid et al., 2017). Outpatient clinics constantly make capacity management decisions regarding human resources, which is difficult because the variable patient demand must be balanced with variable resource availability (Cayirli & Veral, 2003; Cayirli, Veral & Rosen 2008; White, Froehle & Klassen, 2011). Even a short vacation or illness leads to immediate and significant deterioration of the resource efficiency (Vermeulen et al., 2009). Variable capacity is a research area that is not often considered and is suggested for future research by Vermeulen et al. (2009), White et al. (2011), Van Buizen (2014) and, Gijsel (2016). Variable capacity generates more realistic results because variation in human resources is almost inevitable.

3.3.1 Capacity allocation

(18)

18 scheduling problem to that of ﬁnding a suitable match among the available time slots of providers in the clinic, provider prescribed restrictions on how available slots may be ﬁlled and patients’ preferences for day/time of week as well as for a particular service provider (Gupta & Denton, 2008; Klassen & Rohleder, 1996). The scheduling of patients to appointments (slots) consists out of capacity allocation, appointment scheduling and short-term decisions on the day of service (Schuetz & Kolisch, 2012). This research focuses on the capacity allocation and appointment scheduling where short-term considerations like patient unpunctuality, physician lateness, interruption and patient no-show and cancellation are not questioned.

Klassen & Rodleder (1996) employed the idea of using capacity allocation for urgent clients because of their unpredictable nature. Their results show that there is no general solution and policies depend on the goals of the clinic. For example, if a clinic focuses on client waiting time then earlier slots should be left open and if they want to reduce server idle-time later slots are desirable. It did show that leaving slots open for urgent patients in the middle of the session is best in many cases because it balances client waiting time and server idle-time. A more recent study of Klassen & Rodleder (2004) showed that spreading slots for urgent patients evenly over the day performed best which is also concluded by Cayirli & Gunes (2014).

Hoogstins (2008) also suggested that capacity allocation with slot policies would possibly improve the performance with regards to admission times at a clinic that serves multiple patient groups with respect to admission times. However, in his results, slot policies showed no immediate improvements and due to time restrictions, he could not further explore this idea. It did show that when the occupancy rate is high and slots are reserved for urgent patients, the urgent patients will benefit at the expense of control patients. He suggested that future research should reveal and determine an optimal slot policy.

(19)

19 near the end or near the beginning of the week, or evenly distributed, to improve admission times. Also, combination slots (i.e. the slot may be occupied by either new or acute patients) are preferred over single patient group slots. For future research, he suggests investigating variable resources since they assumed that resources (physicians) are constant while this is not actually the case due to illness, holidays and congresses. Vermeulen et al., 2009 also state that even a short vacation or illness leads to immediate and significant deterioration of the resource efficiency. In this research, variable capacity will be considered alongside the capacity allocation with slot policies.

3.3.2 Planning window

Acute and new patients are characterized by the fact that they require health services as quickly as possible because they experience a flare-up or are new to the symptoms they feel. However, control patients are characterized by the return procedure for their consultation and need regularity in their care over long intervals, where it is medically permissible to schedule the patient some weeks before- or after the ideal appointment date (figure 3). This specific characteristic enables sequencing rules for a heuristic as discussed in section 3.2.

Figure 3: Symmetrical- and asymmetrical planning windows.

(20)

20 rule would consider the workload per week and schedule patients in the least utilized week. According to Van Buizen (2014), the planning window reduces the variability in the number of appointments each week, but also postpones appointments for control patients. Using a planning window is almost non-existent in the healthcare literature except for the research of Van Buizen (2014). Other fields of research are also consulted, but nothing similar is found when looking into planning windows or the balancing of workload. Van Buizen (2014) suggested further researching planning windows for an OAS and in specific considering asymmetrical planning windows.

3.4 Summary

The role of AS is to smooth out the variability in demand to match the existing capacity. Considering variable capacity (i.e. illness, congress and vacations) is not yet addressed in literature while this corresponds with the specific role of an AS. Also, waiting time was the most dominant performance indicator while admission times are (almost) not addressed. Short admission times are required for timely access to health services and timely access is important for good medical outcomes. Heterogeneous patients are also not frequently considered in literature while this is often the case and they affect performance of an outpatient system. This research focuses on clinics that acknowledge the importance of timely access and multiple patient groups.

(21)

21

4. System description

To address the class of systems the case of the IBD clinic at the UMCG is used. This chapter will describe the case by reviewing the AS based on the framework of Ahmadi-Javid, Jalali & Klassen (2017), which is briefly discussed and displayed in section 3.1. This chapter looks to answer the following research question:

What are the main characteristics of the system and how is appointment scheduling organized?

Section 4.1 will elaborate on the service of the system. Section 4.2 will elaborate on the patient population. Section 4.3 will discuss the system on a strategic level, section 4.4 on a tactical level and section 4.5 on an operational level. The last section will briefly explain the system workings.

4.1 System overview

In figure 4, a system overview of the IBD clinic is depicted. Inflammatory bowel disease (IBD) is a term for a group of chronic diseases concerning inflammations in the stomach and bowel. Diseases are characterized by episodes of flare-ups alternating with periods of remission (Lin et al., 2014). Main activities of the clinic are the provision of health services and their scheduling. Care services concern consultations with gastro-entomologists where test results and medications are discussed.

(22)

22

4.2 Patient population

4.2.1 Patient heterogeneity

The IBD clinic serves a patient population consisting out of 4 patient groups: new, acute, short-term control and long-term control. These groups are based on the patients’ state of the disease and the required admission time. The state of the patients’ disease, which can change over time, determines the group classification and a patient can therefore alternate between the different groups. Table 3 gives an overview of the patient groups with their associated admission norm and the required consultation length.

Patient group Consult length Admission time

Acute patients 20 minutes < 7 days

New patients 40 minutes < 14 days

Short-term control patients 20 minutes 6 – 12 weeks

Long-term control patients 20 minutes 12 – 52 weeks

Table 3: Patient groups of the IBC clinic.

The most common patient groups in the system are the control patients. The short-term control patients are most often former acute patients and require higher frequencies in their consults and thus lower admission times. The range of the required admission times lies between 6 – 12 weeks. The long-term control patients’ disease is in a state of remission where the consult is a form of control where complaints can be addressed and physicians can re-assess the state of the disease. The associated admission times lie within 3-12 months and are less strict because of the long interval. Currently there is no specified planning to schedule control patients. Both short- and long-term control patients require a consultation session of 20 minutes.

(23)

23 most often scheduled with interns so that the physicians can be coupled with their own patient population. For new patients, scheduling rules are released when they exceed the norm of 28 days, which is regulated by The Dutch government. This is not set at 14 days because the urgency is not necessarily that high.

The last group in the system are is the acute patient group that requires quick consultation. These are control patients that experience a flare-up of their disease or are general acute patients (e.g. from a car accident). General acute patients are ideally seen within a day whereas control patients who experience a flare-up have to be seen in less than 7 days. Acute patients are ideally treated by their own physician. This restriction is relaxed when the admission time norm will be exceeded. Both acute cases require a consultation session of 20 minutes.

As a way to shrink the population that requires physical healthcare service at the clinic, they introduced a procedure which does not require a visit to the clinic. Patients, whose disease is in a state or remission, take blood samples at a location near their home. The results are sent to the physicians at the hospital and analyzed. Depending on the results the patients’ need to visit the hospital and if the results are ‘good’ the patient can stay at home and wait for the next sample time. This group is not accounted for in the simulation because they don’t require appointment slots.

4.2.2 Patient preference

(24)

24

4.3 Appointment scheduling on a strategic level

4.3.1 The access policy

The access policy, which is the type of appointments, is hybrid. This means that the clinic uses both same-day scheduled patients, which do not happen frequently, and pre-scheduled patients. The same-same-day scheduled patients are most often urgent or new patients where pre-scheduled patients are control patients. Emergency patients are also required to make an appointment at the clinic which corresponds with the walk-in policy in chapter 4.3.3.

4.3.2 Number of resources

The number of resources in the system is 4 physicians, and most ideally every physician has his own intern that he accompanies. The 4 physicians serve their own patient population most of the time whereas interns are used for new patients. After the internship, the patients are ideally distributed over the internship supervisor. The clinic manager currently works on a program for ‘specialized nurses’ that will be able to perform consultation sessions as a replacement for the interns and physicians. This initiative is launched due to the limited capacity of the physicians, increasing population, and high utilization rate.

4.3.3 Walk-in policy

(25)

25

4.3.4 Type of scheduling

The last strategic consideration is the type of scheduling. The IBD clinic has an online scheduling approach where patients are scheduled immediately upon their request. Patients are scheduled based on the first-come first-serve method. Patients are most often scheduled in the first available slot closest to the appointment date indicated by the physician. If the scheduler isn’t busy with other activities, he will make the effort to plan the patient within the medical boundaries on a day with the lowest utilization.

4.4 Appointment scheduling on a tactical level

4.4.1 Allocation of capacity to patient groups

The first tactical consideration is the allocation of capacity to patient groups. The clinic serves four patient groups, namely: new, acute, short-term and long-term control. The available consultation slots have no slots specified for control and acute patients whereas there are slots specified for new patients but these are often used for control patients. The slots specified for new patients are most often served by interns and physicians serve the majority of the control patients.

4.4.2 Appointment interval (slot policies)

(26)

26

Day Number of slots Served by

Monday 9 Physician 1 Tuesday 4 Physician 1 10 Intern 1 10 Intern 2 Wednesday 4 Physician 2 10 Intern 1 Thursday 4 Physician 3 10 Intern 2 8 Physician 4 Friday 8 Physician 2

Table 4: Number of consultation slots per physician.

4.4.3 Scheduling window

The appointment scheduling window, not to be confused with the planning window, is how far into the future an appointment can be scheduled. This depends on the state of the patients’ disease. The physician determines, after a consultation session, based on the consult when the patient requires a new session. This can range from 1 week to 52 weeks depending on the patient. This also determines the patient’s patient group which is discussed in section 4.2.1.

4.4.4 Block size

(27)

27

4.4.5 Number of appointments in consultation session

The number of consultation sessions per day, as can be seen in table 2, depends on the day of the week. This schedule is fixed but it does happen that, due to for example illness, vacations or congresses, consult sessions expire and are moved to other available slots.

4.4.6 The panel size

The panel size, the population size that the facility is committed to provide, of the IBD clinic grows every year. Table 5 shows the increase of unique patients served per year except in 2016. This trend in unique patients served per year is also seen at the graph of number of appointments per year (section 5.2.2). In section 5.3 it is discussed why in 2016 there is a decrease in the data. Even though the panel size increases over time, the capacity roughly remains the same depending on the number of interns. An increase in unique patients served and number of appointments per year while roughly remaining at the same level of capacity results in a high utilization rate. This can complicate the scheduling of all the patients. This will be further discussed in chapter 5.

Year 2013 2014 2015 2016

Number of unique patients served 827 905 1005 789

Table 5: Unique patients served per year.

4.4.7 Priority of patient groups

(28)

28

4.5 Appointment scheduling on an operational level

4.5.1 Allocation of patient to resources

The IBD clinic serves a chronic patient population that requires repetitive care. For this reason, they try to schedule patients with the same physician each consult, also known as patient-physician coupling. This allocation of patients to resources is considered an important performance measurement of the clinic. Previous research by Van Buizen (2014) showed promising results with respect to the admission times of patient groups by releasing the patient-physician coupling. Relaxing the patient-physician coupling simplifies the appointment scheduling but reduces its quality.

4.5.2 Appointment day

The appointment day of a patient is determined by the physician at the latest consult. After the consultation session, the patient will be scheduled immediately based on this appointment date. The scheduler initially plans the patient on the first available slot. If time permits it, he will schedule the patient on a consultation day which is underutilized compared to other days to balance capacity. So, the scheduler has the possibility to use a planning window around the appointment day which can balance the capacity.

4.5.3 Appointment time

The appointment time decision problem deals with finding a specific time when a patient is scheduled to start receiving care, so that a performance criterion is optimized (Ahmadi-Javid, Jalali & Klassen, 2017). The IBD clinic uses predetermined slots to schedule patients whereby specific appointment times are not necessary and not considered in this research.

4.5.4 Patient acceptance/rejection

(29)

29 the pre-determined slots or the patient will be double-booked with another patient. This means that the physician serves 2 patients in a consultation slot which prevents patient rejections.

4.5.5 Patient selection from waiting list

At this moment, the clinic does not use a patient waiting list. However, as mentioned earlier in chapter 4.4.6, the panel size of the clinic is increasing each year. Specifics on these numbers can be found in chapter 5, the system analysis. It shows that the clinic has an increasing patient population whereas the capacity remains almost the same, depending on the number of interns. On that note, it is plausible that it is only a matter of time that a waiting list is required if no alternative solutions are proposed.

4.5.6 Patient sequence

Patient sequencing, the order in which patients should be scheduled, does not apply because it does not occur that a list of patients’ needs to be scheduled. Patients are scheduled first-come first-appointment and the process of scheduling is immediately executed after the patient’s consult session.

4.6 System scheduling workings

(30)

30

5. System analysis

The previous chapter discussed the as-is situation of the outpatient appointment system of the IBD clinic in the UMCG. Chapter 5, system analysis, will further explore the system with regards to the performance and objectives of the clinic. Section 5.1 will explain the approach for measuring the performance of the system. Consequently, the actual performance will be presented in section 5.2. Section 5.3 will identify possible root causes for the performance of the system. The main question in this chapter is:

What is the current performance of the system and what influences the performance?

5.1 Approach

To analyze the performance of the IBD clinic, it is important to keep the objective of the research in mind. The objective is to assure timely access to health services for the patient population with the efficient use of staff resources. This translates into 2 objectives with respect to relevant performance indicators. To analyze the system of the IBD, clinic data from the year 2013-2016 is used. This data is used to calculate the performance of the system. The data is presented per year and otherwise the average or the sum of those 4 years is used.

(31)

31 4.5.1), the number of unique physicians a patient has seen will be calculated. A sample of the population will be generated to check if patients are mostly seen by 1 physician or if they have different physicians for their appointments.

For the efficient use of staff resources, it is important to have available resources that can perform the required services. The physicians experience variable availability due to illness, holidays, congresses, and other activities. To calculate the actual availability, the canceled availability is subtracted from the initial availability. The initial availability equals all the available slots without cancellation of any physician. These numbers are expressed in slots of 20 minutes each. This will provide a percentage of the time a physician is actually available. From the actual availability, the utilization is calculated. Also, the average number of unique appointments in a year is calculated. Table 6 gives a summary of all the previous mentioned performance measures and how they’re quantified.

Performance measure Quantification

Admission time in days Average in days and % within norm

Patient physician coupling Unique count of physicians seen

Availability of the physicians Initial capacity versus actual capacity in slots of 20

minutes

Utilization of the physicians Used physician capacity

System activity Number of appointments per year

Table 6: Performance measures.

(32)

32

5.2 Performance

5.2.1 Admission times

Table 7 shows the average admission time per patient group over the course of 2013-2016. Section 4.2.1 discussed norms per patient group and acute patients are ideally treated immediately. Over the course of 4 years, the average admission time for acute patients is 18 days where the norm is <7 days. Acute patients, where quick care is of utmost importance, have to wait more than twice the norm. During this period only 22% of the 582 acute patients are treated within the norm.

Average patient admission time in days (norm)

Patient type 2013 2014 2015 2016 Avg. Norm % in norm

Acute 18 (22%) 17 (22%) 18 (23%) 18 (20%) 18 <7 days 22%

New 36 (24%) 34 (26%) 42 (36%) 20 (55%) 32 <14 days 36%

Control short-term 56 56 56 57 56 N/A -

Control long-term 198 204 208 162 195 N/A -

Table 7: Average admission time per patient group.

New patients also benefit from quick access to health services but the urgency is less important compared to acute patients. The average admission time for new patients is 32 days where the norm is <14 days. A total of 899 patients is treated over the course of 2013-2016 of which 326 (36%) within the set norm. The admission time exceeds the norm significantly and the percentage of patients treated within the norm is also low. This is not in line with the goal of providing timely treatment for new patients. However, it is observed that the percentage of new patients seen within the norm increased over the course of 2015-2016.

(33)

33 advance opposed to acute and new patients who require immediate services after they call for an appointment. This hints at the idea that control patients obtain quite some capacity in the future because of the fact that they’re scheduled weeks in advance. Table 7 does show a decrease in admission time for long-term control patients in 2016.

Section 4.2.2 and section 4.5.1 discussed that both patients and physicians ideally are coupled. The manager of the department confirmed that this is a goal of the clinic and desired for both patients and physicians. Table 8 shows the part of the population and the number of unique physicians they’ve seen. The total population consists roughly out of 1900 patients, where 78% of the population has not seen more than 3 different physicians. More than 50% has not seen more than 2 physicians and only 27% of the population has only seen 1 physician during the period 2013-2016. Based on a sample of 200 patients it showed that, when a patient has multiple appointments per year, the patient is generally treated by the same physician.

# of physicians 1 2 3 4 5 >6

# of patients 523 553 419 249 118 45

Table 8: Treated by # different physicians.

5.2.2 Resource effectiveness

(34)

34

Year Initial availability Canceled availability Availability

2013 4104 737 0.82

2014 4511 939 0.79

2015 4979 1026 0.79

2016 6028 2563 0.57

Table 9: Availability of the physician’s 2013-2016.

Appendix I shows the distribution of the lost capacity over the course of 2013-2015 for each weekday. It shows that for Monday-Wednesday and Friday the lost capacity is around 50-100 slots per year, whereas for Thursday this ranges between the 500-600 slots. In 2016 the number of lost slots per weekday is higher. This is due to the fact that in 2016 there is more canceled availability. In 2016, the number of lost slots per weekday shows an identical distribution across the weekdays compared to the previous years.

The actual available capacity (initial availability – canceled availability) is used to schedule the patient population. In table 10 the utilization rates for 4 months are calculated based on the available hours and the hours that are actually scheduled with patients. The utilization rate averages at a ratio of .94. A utilization rate of .94, which is considered high, worsens the ability to cope with the demand variability for new and acute patients.

Month Available capacity (hours) Scheduled capacity (hours) Utilization

January 86,9 81,4 0,94

February 62,2 55,6 0,90

March 100,2 95,7 0,96

April 66,7 63,3 0,95

Table 10: Utilization rate.

(35)

35 the demand of the patient population approximately remains the same. Because there was less capacity that year a lot of patients had to be scheduled even later than the proposed ideal appointment date. In appendix II the moving average of the number of appointments per week is depicted in a graph. This graph shows that the workload is highly variable per week.

Year Number of appointments Difference available slots and # appointments

2013 2743 624

2014 2813 759

2015 3009 944

2016 2664 801

Table 11: Appointments per year 2013-2016.

When comparing the number of appointments per year and the actual available appointment slots from table 9, there’s a gap of 782 slots on average. This is a large portion of the total available slots and doesn’t correspond with the utilization calculated in table 10.

5.3 Root causes for the performance

The previous sections showed that the IBD clinic does not perform optimal with respect to the performance indicators. This section will explain possible causes for the lacking performance of the system.

5.3.1 Admission time

(36)

36 scheduled more than 6 weeks in advance and therefore seize all the capacity before new and acute patients call for an appointment.

To cope with this problem, especially for acute patients, the clinic resorts to double-booking acute patients with control patients or even schedule them outside regular clinic hours. This is not ideal because the available consultation slot has to be divided in 2. This means that the physicians have a reduced amount of time for their consult per patient. Scheduling outside the regular clinic hours is also not desirable because physicians also have other obligations in the hospital they have to attend. This also affects the performance with respect to resource effectiveness and the workload of the physicians per week.

In 2015 and 2016 an increase in percentage of patients seen within norm for new patients is observed. This is due to the fact that the schedulers consciously plan new patients with interns which leads to admission time reduction. Also, the gatekeeping for new patients is increased. This decreases the inflow of potential new patients who aren’t necessarily IBD patients, and thus don’t acquire available capacity.

Long-term control patients also seem to have a decreased admission time in 2016 of approximately 40 days. The underlying reasons for this occurrence are the new introduced policies. These policies concern the necessity for long-term control patients needing a consult and the remote monitoring which forestall the need for a hospital visit.

5.3.2 Resource availability

(37)

37 had to be rescheduled which lead to an increase in admission times and fewer appointments in 2016. In section 5.2 it is explained that the canceled availability peaks at Thursday and for the remainder of days it remains the same. The reason for the peak being at Thursdays is because the appointment interval each week is fixed where most capacity is made available on Thursdays. Thus, making it more likely that availability is lost. The weekly allocation of capacity does change sometimes, but Thursday always has the most available capacity.

The available capacity has to be divided over the patient population. Table 10 showed that the utilization of the availability is generally 95%. A high utilization rate worsens the ability to cope with variability in for example the arrivals of new and acute patients because there is not much “wiggle room” in the schedule. This makes it harder to schedule new and acute patients because there is no capacity left to schedule patients. The capacity is most of the time taken by control patients who are scheduled more than 6 weeks in advance.

(38)

38

5.4 Summary

Chapter 5 reviewed the performance of the IBD clinic as adequately as possible with the available data regarding relevant performance indicators based on the research objective. The main cause for the lacking performance is the allocation of capacity for acute- and new patients who represent a large portion of the total population. Also, the control patients, who require consults on a large interval where some slack is medically allowed, enable the use of planning windows and specific scheduling policies to balance the workload.

Observation Cause

Long admission times for acute- and new patients No reserved slots for acute- and new patients. Double-booking acute- and new patients and

scheduling outside clinic hours

No reserved slots for acute- and new patients.

Variable workload per week Strictly adhering to the ideal appointment date set

by the physician, not using planning windows. A lot of lost availability in consult sessions Clinic depends on the availability of interns and

physicians.

Patient-physician coupling is met most of the time Number of consults per patient are done by the same physician most of the time.

Large gap in appointments executed and net. available appointments.

(39)

39

6. Design

In chapter 5 the performance of the IBD clinic is measured according to relevant performance indicators. The results show that the clinic doesn’t perform optimally where causes lie in the appointment scheduling of the clinic. This chapter will develop slot policies for the appointment scheduling that possibly improve the performance of the system. The following research questions will be covered in this chapter:

How can the current system be redesigned to improve its performance?

As previously mentioned, chapter 5 discovered possible causes for the current unsatisfactory performance. Based on these causes, redesign proposals are discussed. The clinic divided available provider time into equal length time slots such that patients’ needs can be accommodated in a standard appointment slot. However, the clinic doesn’t distinguish specific slot allocation for specific patient groups. Therefore, possible solutions will be discussed with regards of capacity allocation design decisions. This considers the amount of slots, type of slots and position of slots. These will be discussed in section 6.1, 6.2, and 6.3, respectively. The system analysis also revealed inconsistencies in the workload per week due to physician unavailability and the lack of scheduling policies. So, section 6.4 will cover aspects regarding the scheduling policies and section 6.5 gives an overview of the proposed redesigns.

6.1 Amount of slots

(40)

40 admission times and percentage of patients seen within the norm. Therefore, this research will consider slightly more and less than the long-term average number of slots per patient group.

The system analysis showed that on average, over the course of 4 years, physicians are available 80% of the time. This decreased the available slots for the IBD clinic. Previous research often assumed constant availability. This assumption doesn’t correspond with reality and will influence results significantly. So, to test the alternative designs, both variable and constant availability will be simulated. The constant availability will use the schedule in section 4.4.2 for every week throughout the simulation. To recreate the variable availability, the simulation will randomly set days to unavailable until 25% of the days is unavailable (see section 5.5.2).

6.2 Type of slots

According to Van Buizen (2014) slots can be dedicated or reserved. Dedicated slots are for a specific patient group who can only be planned in these slots. Reserved slots are available for a certain patient group, but these patients may also be planned in other type of slots if necessity breaks law. A combination of dedicated- and reserved slots also exits, i.e. reserved slots for multiple patient groups. These patient groups are ideally scheduled in specified slots, but can also be scheduled in other slots if necessary to provide care services within the norms. Van Buizen (2014) his results show that combination slots use capacity more efficient compared to dedicated and reserved slots. The admission time for acute patients is reduced significantly whereas the admission times for new patients only slightly increases. As for the percentage of patients that is seen within the norm, it increases for both patient groups new and acute. In this research only combination slots will be considered since they proved the best results. Combination slots are reserved for new and acute patients. The remainder of slots are for control patients.

6.3 Position of slots

(41)

41 placement throughout a week, Klassen & Rohleder (2004) suggest that uneven slot placement might perform better. It is mentioned that an ideal positioning of the slots is dependent on the goal of the clinic as discussed in section 3.3. Van Buizen (2014) incorporated this idea in his research at the IBD clinic in the UMCG. Van Buizen (2014) his results show that either positioning slots evenly, concentrated near the beginning or near the end of the week, or a combination of the two all performed identically and improved the admission time significantly. This research distributes the slots evenly throughout the week and does not question different slot positions throughout the week.

6.4 Scheduling policies

The previous sections concerned the allocation of the available capacity. Scheduling policies concern the actual scheduling of the patients and which slot they acquire. The current and default situation schedules all patients first-come first serve (FCFS). Acute and new patients make a call for an appointment and are scheduled FCFS. This makes sense because they require quick access to healthcare services. Control patients receive an indicated appointment date from their physician and are than planned as close as possible to this date. For either short- and long-term control patients it is less urgent to receive care as close as possible to the indicated date. It is acceptable, according to the physician and manager, to schedule either 4 weeks before or ahead of this date for long-term control patients. For short-term control patients, a window of 2 weeks is allowed. This provides a planning window which can be used to evenly spread workload opposed to the current situation where the workload fluctuates a lot per week because planners strictly adhere to the ideal appointment date.

(42)

42 seen within the set norm. In his research, he assumed symmetrical planning windows and determined utilizations per week instead of per day. Therefore, this research will explore the effects of symmetrical and asymmetrical planning windows. Also, a scheduling rule that will schedule based on least utilization per day instead of per week will be used.

The sequencing rules within the planning window are as follows. First the available slots for control patients in the planning window will be filled first-come first-serve. When all these slots for control patients are occupied, the least utilized day in the window will be prioritized to spread the workload. Then slot type is neglected and control patients can also be scheduled on slots for acute and new patients. The alternative scheduling rule schedules control patients on the least utilized day, in their available planning window, to distribute workload.

(43)

43

6.5 Summary

Table 12 provides an overview of all the redesign decisions discussed in the previous sections. It summarizes the decisions based on capacity allocation aspects, scheduling policies and the norms of the clinic. The key redesigns are the planning windows and the accompanied scheduling rule for these windows.

Elements of appointment system New Acute Control

short-term Control long-term Sl o t p o lic ie s Cap a c ity a llo c a ti o n

Type of slots Combination slots No reserved slots Number of slots 1. Less than long-term average

2. More than long-term average

Remainder of the slots

Position of slots 1. Evenly distributed. Remainder of the slots Available capacity 1. Assuming constant capacity.

2. Assuming variable capacity.

Sc h e d u lin g p o lic ie s

Scheduling rule FCFS, slot closest to indicated date. 1. FCFS 2. Balanced

3. Least utilized day. Planning window Release coupling restriction if

admission time will exceed the respective norm.

Symmetrical and asymmetrical configurations. 1. +/- 1 week 2. +/- 2 weeks 1. +/- 1 week 2. +/- 2 weeks 3. +/- 3 weeks 4. +/- 4 weeks Norms of the clinic Patient

physician-coupling

Patient-physician coupling is considered for all patient groups, where new patients are only scheduled with interns.

(44)

44

7. Evaluation

In chapter 6 alternative slot policies are determined. These designs are tested by means of a simulation. A simulation is used because the redesigns are complex and practical testing is not possible. A simulation study to investigate appointment systems is also dominant throughout literature (Cayirli & Veral, 2003; Ahmadi-Javid, Jalali & Klassen, 2017). The aim of the simulation is to gain insights in potential OAS configurations for clinics that are characterized by a chronic heterogeneous patient population. The research question leading this chapter is:

How do the alternative slot policies for the system perform with regards to timely service and resource effectiveness?

Section 7.1 will address the experimental design and section 7.2 the simulation modelling. Lastly, section 7.3 will provide an analysis/a discussion of the simulation results.

7.1 Experimental design – OAS configurations

The set-up of the simulated clinic is based on the following fixed factors. The clinic provides 1 service which is served by 4 physicians and 4 interns. The service time for the appointments is fixed. The arrival times are modelled with an empirical distribution that provides single patients as input. Patient type is determined with a fixed distribution.

The experimental factors for the simulation study are determined in chapter 6 and are summarized in section 6.5, table 12. The only experimental factor that’s not mentioned, is the input of the simulation; the number of patient requests per year. This factor is used to increase the utilization of the system and test redesigns against more extreme conditions.

(45)

45 slot type or patient physician coupling) are relaxed to keep admission times within the set norm. For an extensive overview of the simulation model see Appendix II, the conceptual model.

7.2 Simulation modelling

Tecnomatix Plant Simulation 12, version 12.0.9 is used to carry out experiments. The warm-up length of the simulation is determined with the Welch method (Robinson, 2004). The warm-up length is set to .5 year and the run length is set to 6 years. The required number of replications for valid results is set to 10 repetitions per experiment. As determined by the confidence interval method (Robinson, 2004). The calculations are listed in Appendix III.

7.3 Analysis of simulation results

In this section, the outcomes of the simulation study will be analyzed. Only the best performing experiments are listed. For an overview of all experiments, see appendix V and for all corresponding results, see appendix VI. Short-term patients will be referred to with “ST” and long-term patients with “LT” in all of these sections’ tables. Table 13 presents the default scenario which is used for benchmarking. The default scenario resembles the actual set-up of the clinic where they reserved no slots for patient groups and do not use planning windows for the scheduling of control patients.

System Acute patients New patients Control patients

Input (avg. app. per year)

Mean utilization of physicians Avg. admission time Norm Coupling restriction lifted Avg. admission time Norm Coupling restriction lifted Avg. days right of ideal app. day (ST) Avg. days right of ideal app. day (LT) 3699 0.97 5.01 100% 69% 13.33 63% 21% 8.14 5.41

Table 13: Default scenario for benchmarking.

(46)

46

7.3.1 Constant versus variable availability

The next experiment will show the effect between constant and variable availability of physicians. Section 6.1 discussed that assuming constant availability is not in line with real life situations. Assuming variable availability should result in a more accurate simulation and performance. This experiment will assume constant availability opposed to the default scenario where variable capacity is considered. This will randomly select days throughout the simulation and set all slots on that day to unavailable. Approximately 20-25% of the available days will be set to unavailable, this is based on the analysis in section 5.2.2.

System Acute patients New patients Control patients

Experiment Input (avg. app. per year) Mean utilization of physicians Avg. admission time Norm Coupling restriction lifted Avg. admission time Norm Coupling restriction lifted Avg. days right of ideal app. day (ST) Avg. days right of ideal app. day (LT) Default 3699 0.98 6.84 49% 64% 14.34 58% 22% 8.14 5.41 Constant availability 3699 0.78 2.62 100% 14% 6.84 97% 0% 3.78 3.08

Table 14: Constant versus variable availability.

The performance of the simulation improves when availability is constant (table 14). The admission time for acute patients decreased with more than 50% (6.84 vs 2.62) and the percentage of patients seen within the norm almost doubled (49% vs 92%). The admission time for new patients is halved from 14.34 days to 6.84 days and the percentage of patients seen within the norm is almost 100%. Control patients also benefit greatly. Both patient groups experience a huge decrease in the average of days distant from their ideal appointment date, which is less than 4 days for both groups.

These results are in line with expectations where less availability leads to a worse performance and full availability on the other hand improves the performance. However, constant availability leads to lower utilization (.98 vs .78) which means that resources are not used optimally.

7.3.2 Capacity allocation

Outpatient appointment scheduling: policies for a chronic patient population