Master’s Thesis
The impact of lean in health care:
How different types of variability
influence admission times
By Folkert van Zanten
S2013940
June 2015
University of Groningen, Faculty of Economics and Business MSc. Technology & Operations Management
Supervisors:
Dr. M.J. Land
Prof. dr. ir. C.T.B. Ahaus
Abstract
The purpose of this research is to determine the influences of different types of variability on the admission time in a hospital department. A case study research is done in a rheumatology department via both quantitative and qualitative research methods. Several stakeholders are interviewed and an analysis is done on the secondary historical data. The outcomes show that different types of variability have their effect on the admission times. Examples of influencing kinds of variability are the differences in capabilities and skills of the practitioners and external sources of variability. Furthermore, a significant amount of variability appears to result from managerial decisions. The main source are the changes in ratio between new and returning patients, which can, if out of balance, cause congestion of the system, which means that in the end no new patients can be accepted anymore. This phenomenon is related to the so-‐called
Table of Contents
Abstract ... 2
1. Introduction ... 4
2. Theoretical background ... 6
2.1. Lean ... 6
2.2. Lean in health care ... 7
2.3. Variability ... 7
2.4. Buffers ... 9
2.5. Admission time ... 9
2.6. Research framework ... 10
3. Methodology ... 11
3.1. Research setting ... 11
3.2. Case selection ... 11
3.3. Data collection ... 12
3.4. Data analysis ... 13
4. Analysis and results ... 16
4.1. Admission time ... 16
4.2. Capacity ... 19
4.3. Natural variability ... 20
4.4. Artificial variability ... 22
4.5. Buffers ... 25
5. Conclusion ... 26
6. References ... 28
Appendix ... 31
1. Introduction
Health care providers are increasingly put under pressure by, among others, governments, insurance companies and patients to reduce costs while maintaining or even improving the service level and safety of the offered services (Langabeer, DelliFraine, Heineke & Abbass, 2009). That is why hospital operations managers are looking for other ways of executing the health care process. A common approach is the implementation of lean practices. It aims at finding new and more efficient manners in providing care (Poksinska, 2010). Lean Manufacturing was introduced by the Toyota Production system in the automotive industry and is focused on the elimination of activities that do not contribute to the value of the customer (Joosten, Bongers & Janssen, 2009). Later, all kinds of other organizations implemented lean practices, producing both services and goods, and it became increasingly popular in health care services from mid 1990’s on (Hopp & Spearman, 2004; Joosten et al, 2009).
The evaluation of lean practices in health care has been of quite an interest to researchers. Nevertheless, it is currently hard to evaluate the influence of managerial health care decisions (Kollberg, Dahlgaard & Brehmer, 2007; Mozzacato et al., 2010). Litvak & Long (2010) agreed upon this point and they also pointed out that performance of health care operations is affected by variability, which they split up into natural variability (caused by e.g. how patients react to a treatment) and artificial variability (caused by humans). Natural variability is unavoidable and should be managed optimally. Artificial variability, however, leads to increased costs and should therefore be minimized (Litvak & Long, 2010).
Mazzocato et al. (2010) state that the application of lean thinking has positive results on facets, such as improved quality, access, efficiency, etc. Simultaneously, the researchers acknowledge that the results of lean practices might be overly positive, due to the lack of publications that address failed lean implementations. Additionally, Poksinska (2010) recognizes that lean practices have been widely researched, but that most of the papers have a speculative character, since she found only evaluations of successful lean implementations. Additionally, the influence that variability has on hospital operations is underexposed. That is why there is a need for objective evaluation of lean implementation practices.
According to Maister (2005), waiting times before the processing starts are perceived to be longer than in-‐process waiting times. In addition, shorter perceived waiting times lead to higher patient satisfaction (Thompson et al., 1996; Michael et al., 2013). Therefore, it is particularly relevant to research admission times: the time a patient has to wait between a General Practitioner’s referral and the actual appointment with the physician. This research paper attempts to understand how variability influences the performance in a health care setting. This leads to the following main research question: How do different types of variability influence the length of admission times at a hospital department?
The structure of the paper is as follows: In the upcoming section the theoretical background will be discussed. The third section concerns among others the setting of the research, the methodology to obtain data to answer the research questions, and the justification of the selected case. The fourth section contains the results and discussion of the executed analyses. The fifth and final section consists of a conclusion based on the outcomes of the fourth section.
2. Theoretical background
2.1. Lean
The concept of lean was introduced by Taiichi Ohno, a senior executive of the car manufacturer Toyota in times of resource scarcity in Japan (Burgess & Radnor, 2013). Because of this scarcity, the focus was on the minimization of waste of valuable resources (Hopp & Spearman, 2004). Lean thinking is about recognizing a value stream for the customer throughout the production process of a good or service. It considers the process as a sum of activities and each activity should add to the customer value. Value can be explained as the capability to deliver exactly the good/service that a customer expects in the shortest possible time between the order and the delivery of the good/service at an appropriate price (Womack & Jones, 1996). Non-‐value added activities are considered to be waste, since they are mostly adding delays and require extra resources. They should therefore be eliminated from the process (Burgess & Radnor, 2013).
During the process, the activities should be executed in the right sequence and without interruptions, and they should be done with increasing effectivity (Womack & Jones, 1996). In order to achieve this, lean management initiatives have to focus on standardization and stability to be able to offer the best quality services or goods possible (Langabeer et al., 2009).
All in all, the concept of lean is based on continuous value-‐ and flow improvements and waste reduction (Burgess & Radnor, 2013). Lean in practice pays a lot of attention to the concept of waste (muda). There are, however, two more (less well known) concepts that belong to lean that cause muda. The first one is muri, or ‘excessive strain', which focuses on a good working environment. If in an organization muri is low, the circumstances are safe and do not ask unreasonable achievements from workers. The second concept is mura, and stands for ‘unevenness’ or variability. The higher the variability, the less smooth the production flow will be. Reducing muri and mura is required if one wants to reduce muda (Radnor, 2011).
health care services. This paper therefore mainly focuses on the mura concept of lean in exploring the influence of variability on admission times in health care. The next section discusses the implementation of lean in health care.
2.2. Lean in health care
Lean originates in the automotive industry in Japan in the 50’s and it found its ways to the health care industry in the 90’s. Ever since, it has been the subject of quite some papers (Poksinska, 2010). The health care industry used to have the reputation of working inefficiently and making errors regularly (Langbeer et al., 2009). Additionally, demand for care is increasing due to aging and budgets are cut, so health care providers have to serve more patients for less money (Poksinska, 2010). The reduction of (direct) waste seems to be a prerequisite for improving the efficiency of a health care organization.
It appears that health managers have picked lean as a solution to contribute to both the cost issue as well as the quality issue. Lean evaluating research papers have focused mainly on process improvement and continuous flow (Poksinska, 2010). Waiting time, however, is not often addressed in the articles that evaluate lean (Mazzocato et al., 2010), which is strange because the aforementioned negative relation between waiting times and patient satisfaction (Thompson et al., 1996; Michael et al., 2013). Because of this, no rigorous research is done on the influences of lean on the waiting time of patients, nor the admission time length (Mazzocato et al., 2010; Poksinska, 2010). Mazzocato et al. (2010) found no lean evaluation paper that focused on waiting or admission times, that had a clear and transparent research methodology in the current literature. Accordingly, no clear methodology exists that measures the influence of variability on admission times.
2.3. Variability
As discussed in section 2.1, one aspect of lean focuses on the reduction of mura or variability. Variability can be defined as the extent to which the same process is different when repeated (Joosten et al., 2009). Variability in health care operations can be divided into two categories: natural variability and artificial variability (Joosten et al, 2009; Litvak & Long, 2010).
2.3.1. Natural variability
natural variability. The first one is clinical variability, which is caused by the relative illness of the patient, the amount of possible treatments, and the way the patient responds to the treatment. The second concerns flow variability, which is caused by unevenness of the arrival rate of patients over time, i.e. the demand for care is hardly ever stable. The final reason for natural variability is professional variability and is caused by the fact that not every health care practitioner is able to provide the best possible treatment at all times.
2.3.2. Artificial variability
Artificial variability, however, can be related to controllable factors in the management and the design of health care operations (Joosten et al, 2009; Litvak & Long, 2010). Variability can lead to an increased WIP and lead-‐time, and a decrease of throughput. Hopp & Spearman (2008) even state that increased variability always leads to reduced performance of a production system.
One phenomenon that can be caused by artificial variability is the service bullwhip effect. It occurs when the variation of the demand pattern that is coming out of the process is larger than the variation in demand that came into the process. The service bullwhip effect can be caused by mismanagement; for instance inadequate planning or information. These inadequacies can amplify further down in the service supply chain, causing an organization to be is less able fulfilling new customer demands (Akkermans & Voss, 2013).
One important variable that can be used to mitigate the chance of the occurrence of a service bullwhip effect is the capacity. Depending on whether the variability is planned or unplanned, and the question if the capacity can easily be increased, there are different strategies that enable prevention of the service bullwhip effect, as can be seen in figure 1 (Akkermans & Voss, 2013).
Figure 1: Capacity strategies to prevent and mitigate the Service Bullwhip Effect (Akkermans & Voss, 2013)
2.4. Buffers
Hopp & Spearman (2008) elucidate that variability in a production process will always be dealt with using buffers. They mention three kinds of buffers that can be applied in a manufacturing setting. These are inventory buffers, capacity buffers and time buffers. Buffers, in general, aim at coping with variations in both production and demand. Capacity buffers do so using an overestimated production capacity, so that extra resources can be deployed in case of e.g. sudden high demand. One example of a reflection of inventory buffers is the creation of a safety stock of raw materials or finished goods. Time buffers are expressed as the time between the appearance and the satisfaction of the demand (Hopp & Spearman, 2004; Hopp & Spearman, 2008). In a service creation process, and thus in health care operations, inventory buffers are not used because of the fact that services cannot be stored.
Hopp et al. (2007) introduce a third variability buffer for services: a quality buffer. A service provider can, for example, adjust the quality of his/her output to manage his/her workload. The reason for this is that in services one can determine the amount of time he/she spends on the process, since outcome criteria are difficult to standardize (Hopp et al., 2007).
A quality buffer is used in a health care setting if the inflow of patient is exceeding capacity, resulting in a shorter time to serve the patient. This does, however, not necessarily mean that the consult has an inferior quality. Capacity buffers would result in an increased capacity of resources to cope with peaks in demand. This would mean that a larger number of physicians, nurses, equipment, etcetera should be present at the department. However, this might not in every case be feasible, since increased capacity will obviously lead to higher costs. Time buffers in a hospital setting lead to waiting time for the patient. This might not be desirable for patients that require urgent care. The next section elaborates on the specification of the buffer that is researched in this paper.
2.5. Admission time
satisfaction is an important factor when evaluating the quality of the services in a hospital (Majid et al., 2013; Farley et al., 2014). That is why this research paper focuses on determining the exact influence of variability on the admission time for outpatients, i.e. the time that a patient has to wait between the referral of the General Practitioner and the appointment with the required physician (e.g. meeting a rheumatologist or internist).
2.6. Research framework
Different kinds of variability can only be dealt with using buffers (Hopp & Spearman, 2004). The different kinds of possible buffers have different effects on admission times. The time buffer leads to a longer admission time, since the patient has to wait for the appointment, so for available capacity. Quality buffers can lead to shorter admission times, since when there is a high inflow of patients, the time per patient decreases, and therefore productivity is enhanced (Hopp et al, 2007). The capacity buffer can lead to a shorter admission time as well, since an increase in capacity will enable serving more patients in the same period of time (Hopp et al., 2007).
Figure 2: Conceptual model
Buffers
Natural variability § Professional variability § Flow variability § Clinical variability3. Methodology
This study attempts to find how different sources of variability influence the length of admission times in a hospital setting. At this point of time no such information exists in the literature. The research in this paper will therefore be exploratory and theory building. A case study has been widely recognized as being adequate when attempting to answer how and why questions (Yin, 1994). Additionally, a case study has a high validity in practice, since the case is researched in its natural setting (Karlsson, 2009). Multiple sources of data are used, enabling data triangulation (Karlsson, 2009). The upcoming sections discuss the combination of methods with which the research question is answered.
3.1. Research setting
The organization where the research is done is a hospital in the northern part of the Netherlands with 643 beds (2013). The hospital aims at becoming a top-‐notch care provider, and one of its spearheads is the optimization of processes using Lean Six Sigma (LSS). The hospital trains it personnel to become yellow, orange or green belt to achieve continuous improvement on the processes by reducing waste, waiting times and improve flow. One aspect that the hospital focuses on is using lean to reduce admission times at several outpatient departments. A schematic overview of the process that a new patient goes through can be found in Appendix 1.
3.2. Case selection
The unit of analysis is on the departmental level of a hospital. There have been several lean implementation projects on different departments in the hospital, among others to reduce admission times. This paper focuses on one of those projects:
The case on which this paper focuses is the rheumatology department. The rheumatology department is an outpatient department characterized by a relatively clear sequence of activities that is always executed; if a patient has problems with his/her joints, the patient visits a GP, who refers him/her to the rheumatologist. Since rheumatism is in many cases a chronic disease, the rheumatologist will not be able to solve the problem in a clinical department. Instead, the rheumatologist executes an anamnesis and a physical examination. If needed, the physician requests for further investigation, such as an ultrasound examination, X-‐ray or blood test.
straightforward process sequence and the absence of a clinical rheumatology department, the influence of different types of variability on the admission times should therefore be rather well visible in this case. In the rheumatology department the admission times where rather long (up to 16 weeks) and therefore the board of the hospital initiated a lean improvement project. The green belt project leader investigated the situation using quantitative and qualitative data and made adjustments in the allocation of human resources. This research investigates the admission times as well, however, with a more theory-‐based viewing angle. Table 1 represents a few relevant statistics about the department.
Case Rheumatology
Number of specialists 2
Number of other practitioners 2 Physician assistants or interns? Yes Number of major diseases treated ±10 Number of patient served in 2013 9.277
In-‐ or outpatient? Outpatient
Issue Long admission times
Project leader background Green belt Table 1: Case facts
3.3. Data collection
When the hospital was looking for ways to reduce the admission times at a department using LSS, it collected numerical data about Key Performance Indicators (KPIs), such as the admission time, utilization and capacity for a period of 1 year (2013). This already available secondary data enables a quantitative research on the admission times and causes of variability. A second dataset is available with the agendas of the practitioners in the same period of time.
Because not every type of variability can be deduced from quantitative data, data are also collected by observations and interviews with stakeholders, such as physicians, nurses, LSS project leaders, etcetera. These interviews also discuss which sources of variability there are and how they are coped with. The interviews are semi-‐structured using an interview protocol, which enhances the validity and reliability of the data (Karlsson, 2009). Notes are made during the interviews, and the conversation is recorded as well to enable accurate retention.
same phenomenon will increase the reliability of the results of the analysis (Karlsson, 2009). This research aims at triangulation by using both quantitative and qualitative information sources.
3.4. Data analysis
The current and historical data about the appointments at the rheumatology department are numerical, and are therefore measured statistically. The relevant factors (e.g. admission time, number of patients served, etcetera) are structured in MS Excel and analyzed. The following paragraphs describe the measurement methods that are used for the analysis of the different constructs.
3.4.1. Admission times
The admission times of the department are operationalized as the time between the GP referral and the first visit at the physician. This paper uses multiple ways to calculate the admission time based on the historical data from the hospital department from the year 2013. The first measurement is the median in admission time per week for the served patients in that week. The median is a reliable measure for determining a dataset, since it is less sensitive for outliers compared to the average admission time. In addition, the mean is calculated from the admission times per week.
Besides the median and mean, the 90th percentile will be calculated per week as well, to see the admission times of 90% of the served new patients of that certain week.
The final used method for analyzing the admission times is throughput diagramming as used by Soepenberg, Land & Gaalman (2008). This provides a framework to analyze the cumulative inflow and outflow of the process, and displays the admission time length and the amount of people that are waiting for an appointment with the specialist.
3.4.2. Capacity
3.4.3. Natural variability
As mentioned before, a distinction is made between natural and artificial variability. This section discusses the measurement of the three types of natural variability.
3.4.3.1. Professional variability
The level of professional variability operationalized as the extent to which the physicians are able to treat a patient uniformly and is determined qualitatively. The reason for this is that it can hardly be deducted from quantitative data. The level of professional variability is asked for using interviews. It might, however, be hard to find an objective answer since the subject can be sensitive to the interviewees.
3.4.3.2. Flow variability
The amount of flow variability is the extent to which the demand for care varies. It is determined quantitatively by measuring the amount of new patients that wish to have an appointment with a physician per week. Furthermore, an overview is made to see the spread of new patients coming in the system per week.
3.4.3.3. Clinical variability
Clinical variability is the extent to which the time to treat a patient on the first visit fluctuates. It is measured both quantitatively and qualitatively, so the amount of time spent per patient is deducted from the data, and the practitioners are asked how the time varies in practice.
3.4.4. Artificial variability
Artificial variability is variability caused by manners of working, behavior or the way of organizing a process. Examples of artificial variability include, among others, the planned absence of a physician because of e.g. a holiday, scheduling methods, and other ways of organizing the process. The level of artificial variability is measured both quantitatively and qualitatively. The quantitative measurements include an analysis of a dataset that contains the absenteeism of the practitioners for one year. The measurement is similar to the measurement method of the capacity, so the distinction between half days off and full days off is made here as well.
3.4.4.1. Service Bullwhip Effect
The service bullwhip effect can be caused by managerial decisions that appear to be rational at short notice, but work counterproductively in the long-‐term. One important factor that can be influenced by the management decisions is the weekly ratio of new patients and returning patients. The service bullwhip effect is measured both qualitatively and quantitatively. A dataset is analyzed to find these ratios. Also the absolute numbers of new patients and returning patients are analyzed. During the interviews, the perception of the ratio between new patients and returning patients is discussed with the stakeholders at the department as well. The goal is to find whether the influence of this ratio might cause constipation in the system when numbers of certain patient types suddenly increase or decrease.
3.4.5. Buffers
As mentioned earlier, variability is always dealt with using buffers. The following paragraphs describe the operationalization of the three types of buffers.
3.4.5.1. Time buffers
The time buffer is the length of time a new patient has to wait between the referral of the GP and the actual first appointment with the physician. It is measured quantitatively using a dataset with the information per patient for one year.
3.4.5.2. Capacity buffers
Capacity buffers are used when the capacity normally exceeds the demand, to cope with high peaks in demand. Capacity buffers are harder to measure quantitatively than time buffers. The reason is that most of the time the appointments that are made out of the planned consulting time are not distinguishable from the dataset. That is why this construct is measured qualitatively, by asking multiple stakeholders.
3.4.5.3. Quality buffers
4. Analysis and results
This section describes the outcomes of the quantitative and qualitative analyses as discussed in the previous chapter. The sequence of subjects of chapter 3.4 is used in this section. Each paragraph is structured similarly: first the results are displayed; second, a brief explanation is given of how to view the results. Third, possible conspicuities are described and finally, the striking phenomena are attempted to be clarified.
4.1. Admission time
The admission times are calculated in several different ways: 4.1.1. Median and mean
The median and the mean of the admission time over time can be seen in figure 3. The week numbers are displayed on the x-‐axis and the admission time in weeks is displayed on the y-‐axis. A large difference is visible between the first and second part of the year. In week 43, there is a peak visible, because suddenly the admission time rises to 10 weeks and decreases rather quickly afterwards. It can be explained by the data: there were only 6 appointments in that week; 3 of which had an admission time of around 100 days and the remaining 3 had an admission time of less than 10 days.
Figure 3: Median and mean admission time per week
The mean weekly admission time decreases from week 31 on as well. The difference between the mean and median can be explained by the fact that in the first part of the year, the major part of the appointments had a relatively long admission time. For example in week 25, 55% of the appointments had an admission time of 14 weeks or longer. The largest part of the appointments in the second part of the year had a
relatively short admission time. For example in week 35, 69% of the appointments had an admission time of less than 2 weeks.
4.1.2. 90th percentile
Figure 4 represents the admission time of 90% of all the patients that are served. The horizontal axis represents the week numbers, and the vertical axis represents the admission time of the 90th percentile. In the 90th percentile, the decrease in admission time from week 31 is not clearly visible compared to the mean and median.
Figure 4: 90th percentile of the admission time per week
This means that even though the mean and median of the admission time are decreasing, there is still a relatively large group (40% compared to the median) that is served within a much longer period of time compared to the median from week 31 until week 46. From week 47 and on, the differences between the median and the 90th percentile are much smaller, so the admission times of the majority of the patients is decreased in the final weeks of the year.
4.1.3. Throughput diagram
The final measurement of the admission time is the throughput diagram as can be seen in figure 5. On the horizontal axis, the week numbers are displayed. On the vertical axis, the patient number is displayed. The upper line represents the cumulative inflow of patients and the lower line represents the cumulative outflow of patients.
Figure 5: Throughput diagram
The horizontal difference between the two lines can be seen as the mean admission time. The 400th patient, for example, has an admission time of 12 weeks, whereas the 800th patient has a mean admission time of 2 weeks. One assumption that has to be made when developing a throughput diagram is that it is based on the First Come First Serve (FCFS) principle. In practice, this scheduling method is not used at the rheumatology department. However, using a throughput diagram still provides a good impression of the average admission times if the FCFS scheduling method was used. Figure 6 represents the average admission time based on the throughput diagram. When compared to the mean admission time per week (figure 3), the decrease in the second part of the year is much less steep. The reason is the assumption of the FCFS scheduling method as well. In practice, a patient might come at the department with an admission time of 12 weeks, and on the same day another patient might be served with an admission time of 2 weeks. However, it still provides an insight of the mean admission time if FCFS would be used in the system.
Figure 6: Mean admission time based on the throughput diagram
0 200 400 600 800 1000 1200 1 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 51 Pa )en t i nflo w & o u6l ow Weeknumber
Inflow Ou?low 00 02 04 06 08 10 12 14 1 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 51
For both figure 5 and figure 6 holds that the inflow line stops at the end of week 47, since there is only data about the year 2013. The inclusion of the data of the rest of the year would make the diagram less reliable. The reason is that appointments made in week 48 until week 53 could occur in the year 2014, so they would not be taken into account in this diagram.
4.2. Capacity
The capacity at the rheumatology department is discussed in this paragraph. Figure 7 represents the capacity per week in man-‐days. The horizontal axis shows the week numbers, whereas the vertical axis displays the weekly amount of man-‐days that is worked at the department on one side, and the admission time length in weeks on the other side. The capacity can be increased using a physician assistant and an intern. The productivity of a physician-‐assistant is the same as the physician, but the intern needs 3 times as much time to see a new patient. Figure 7 therefore displays the weekly amount of man-‐days by the intern as one third of the actual hours in order to make a fair calculation of the capacity. Table 2 displays the weekly presence of the different functions at the department.
Figure 7: Weekly capacity in man-‐days and median of the admission time
Table 2: Present per function throughout the year
The bar chart shows that there is a relatively large amount of weeks in which the capacity does not reach 10 man-‐days. From week 27 until 29, for instance, one of the physicians is on vacation.
In the second part of the year the capacity is expanded. From week 31 there is an intern active at the department until the end of the year, and from week 33 until week 39 there is a physician-‐assistant as well. Mainly the presence of the physician-‐assistant results in a much larger capacity. Once the physician-‐assistant has left the department, the capacity starts to vary again, as in the first part of the year. It especially comes forward in week 43 and 44, where the physicians are absent for one week each.
A clear negative correlation between the capacity and the median of the admission time is visible. When the capacity is increased, the admission time starts to drop immediately.
4.3. Natural variability
The following section discusses the analysis on the three types of natural variability. 4.3.1. Professional variability
From the interviews it appeared that professional variability can be interpreted twofold. The first perception is the difference between physicians in the ability to treat a patient. According to the interviewed stakeholders, this phenomenon does not appear to occur at the researched rheumatology department.
The second perception is the difference in ability of a practitioner in general, so this includes the nurses as well. Professional variability in this sense does seem to occur at the department, because the physicians are educated and capable to see a wider range of new patients compared to the nurses. The nurses serve only a predetermined, smaller group of patients. The percentage of new patients that is seen by the two nurses is 23,6%, while the remainder of new patients (76,5%) is seen by the two physicians. One of the interviewees mentioned that currently the utilization rate of the two physicians is very high. Simultaneously, the utilization rate of the nurses is relatively low.
Quartile Weekly inflow new patients 0% 0 25% 11 50% 16 75% 27 100% 43
foresee and prevent possible problems with e.g. interactions with different kinds of medication that a patient has to take for other diseases.
The unit head of the rheumatology department, on the other hand, argues that in many cases the nurse can still be used for seeing a larger group of patients compared to the current situation. Additionally, he argues that dividing the patients more evenly would release the pressure on the physicians a little.
4.3.2. Flow variability
In the inflow of new patients at the rheumatology department, a distinction is made between different patient groups. The incoming consult requests are judged by a rheumatologist: the triage. There are three different categories, depending on the urgency of the disease: regular new patients, semi-‐urgent patients and urgent patients. The admission time standard of urgent patients and semi-‐urgent patients is set to two and four weeks respectively. From all the new patients is around 7% urgent and around 6% semi-‐urgent. There is currently no standard compliance concerning the admission times for regular patients. This research focuses on the admission time of regular patients, since those fluctuations are relatively high.
Figure 8 displays the weekly inflow of new patients. The horizontal axis represents the week numbers and the horizontal axis shows the amount of new patients that are coming in the system per week. Table 3 contains information about the spread of the weekly inflow of patients using quartiles.
Figure 8: Inflow of new patients per week
As can be seen in figure 8, the inflow of regular new patients changes weekly and the numbers of patients entering the system are also widespread, as can be seen in table 3. In the second part of the year the weekly new patient inflow clearly increases. According to the interviewees, the sudden increase of new patient inflow is due to the increased
0 10 20 30 40 50 1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 Num be r of ne w pa) ents Weeknumber Inflow/ week
capacity as discussed in paragraph 4.2. When the capacity increased, the referrers immediately start to send patients to the department in question. The interviewees mention therefore that there is always enough demand for an appointment with a rheumatologist to fill the schedules at all times.
One other, less obvious factor influencing the flow variability was mentioned during the interviews. In the town in which the researched hospital is located, there is another large hospital as well. There is said to be a ‘market of admission times’ in the town: if hospital “A” has relatively long admission times (e.g. 16 weeks) and hospital “B” has shorter admission times (e.g. 6 weeks), referrers tend to send the patients to the hospital that has the shortest admission times. This causes that suddenly a lot of patients come in the system of hospital “B”. As a result, the capacity gets filled and as a result, the admission times of hospital “B” increase. Due to the smaller demand for appointments in hospital “A”, the admission times in hospital “A” decrease. Several different stakeholders mentioned this ‘suction effect’ during the interviews.
4.3.3. Clinical variability
Clinical variability, the extent to which the time to treat a patient varies, does not appear to occur at the rheumatology department. As said, there are standard time slots in which a patient is seen. For physicians and physician-‐assistants it is 30 minutes, for an intern it is 90 minutes. Besides that, all of the interviewees mentioned that the scheduled times to serve a new patient are adequate in practice. So, again, clinical variability does not seem to play a role at this department.
4.4. Artificial variability
The following section discusses the occurrence of variability caused by the way of working or organizing a process.
4.4.1. Agenda
The first factor that influences the artificial variability is caused by the absence of the rheumatologists, so by blockages in the agenda. In 45% of all the weeks in the year, the rheumatologists both are present the entire week, so 5 days per person. In the rest of the weeks (55%), at least one rheumatologist is absent for at least half a day.
When a physician-‐assistant is active at the department as well, the effects of absenteeism appear to become less radical. If one physician is absent, now the influence of absenteeism on the capacity is lower. The capacity of the department is now lowered by one third instead of 50% compared to the situation with two physicians. Having a small capacity thus makes the department relatively vulnerable to absenteeism.
4.4.2. Ratio new patients/returning patients
At the rheumatology there are two main outpatient groups that are served: new patients, that visit the department for the first time and returning patients, that visited the rheumatology department at least once in the past. Currently, 76,5% of the patients that see a rheumatologist has been at the department before. The remainder (23,5%) consists of new patients. At this moment, if a patient enters the department for the first time, he/she has on average four follow-‐up appointments in the next 365 days.
The ratio of new and returning patients is a source of artificial variability as well. As mentioned before, from week 31 and on the capacity is larger than in the first part of the year. Figure 9 displays the weekly ratio between new patients (NP) and returning patients (CP) as percentages of the total number of appointments. On the horizontal axis the week numbers are shown, and on the vertical axis the percentages can be seen of the share of new and returning patients.
Figure 9: Ratio new patients and returning patients served per week
As can be seen from the diagram, in the first 30 weeks, the share of new patients is around 20% and is increased afterwards. As discussed in paragraph 4.2, the capacity in week 43 and 44 is relatively low due to absenteeism and the share of new patients suddenly increases. It therefore appears that there is a positive relationship between the relative number of new patients seen and the capacity. In this way the management influences the ratio between new and returning patients, and therefore it can be seen as
a source of artificial variability. The following paragraph elaborates on the effects that a changing ratio can have on the long term.
4.4.3. Service bullwhip effect
Figure 10 represents the amount of served new patients that are flowing out of the system each week. It looks similar to figure 8, but in that chart the inflow of new patients that are waiting for an appointment is shown. The horizontal axis represents the week numbers, and the vertical axis the number of new patients that the physicians have seen during that week.
Figure 10: Served new patients per week
An increase of inflow of new patients is visible from week 31. The interviewees mention that the increase in capacity is initiated by the management in order to reduce the admission times. As discussed in paragraph 4.4.2, increased capacity leads to a relatively higher number (absolute and relatively) of new patients entering the department. The historical returning rate suggests that if 10 more new patients are seen in a period of time, it leads to an increase of 40 follow-‐up appointments in the year that follows. This increase in follow-‐up appointments can only be dealt with if the capacity is growing as well, if the department wants to remain seeing a certain amount of new patients each week. When the capacity temporarily decreases in the weeks 43, 44 and 45, as can be seen in figure 9, the share of returning patients reaches around 90%. It is therefore not unlikely that the system will congest in the future if the extra temporary capacity (physician-‐assistant and the intern) is not working at the department anymore. The physicians would then only see returning patients, since the limited capacity would not allow seeing any new patients.
4.5. Buffers
At the rheumatology department, several types of buffers are used. The following part discusses the three types of buffers that are mentioned earlier. Again, a distinction is made between different kinds of patients, namely between regular new patients, semi-‐ urgent patients and urgent new patients. As mentioned, the prioritization of the new patients is done by the physician, at the triage phase.
4.5.1. Time buffers
According to the interviews, the most used buffer at the rheumatology department is the time buffer. It is expressed in the admission time of the new patients. These admission times have been discussed in paragraph 4.1. So there is no question if the time buffer is used, however, it appears that the time buffer only is used for one particular patient group. The admission time is usually only applicable to regular new patients. If the physician is triaging patient’s referral letters, and the physical complaint of a patient does not have to be treated in a relatively short period of time (i.e. the physical complaint does not have to be treated urgently), the patient has to wait for the length of the admission time of that moment.
4.5.2. Capacity and quality buffers
If the physician triages the physical deems the physical complaint as urgent, he/she prioritizes the patient as urgent or semi-‐urgent. The admission times for urgent and semi-‐urgent patients are in general shorter than the admission times for regular new patients.
The interviewees did mention that no empty time slots are scheduled to see urgent and semi-‐urgent patients during the weeks. Nor did they directly mention the existence or the use of capacity and quality buffers. However, after a thorough analysis of the interview scripts, it appears that they seem to be used anyway.
Most of the time, the schedule is already filled with appointments with returning and regular new patients, so with no time left to see potential urgent patients. Urgent patients are thus scheduled in a different way compared to regular new patients. This happens in two ways.