DELIVERY RELIABILITY IN HOSPITALS: A
CASE STUDY OF AN UROLOGY DEPARTMENT
by
Michiel de Ree
University of Groningen
Faculty of Economics and Business
ABSTRACT
There is growing interest for developing business approaches for hospital practices and research has show that it is possible to reduce the throughput times of patients. Managing the
flow of patients can to some extent draw on concepts and approaches widely used in manufacturing. This thesis consists of a case study of a Dutch hospital. Concepts like delivery reliability, capacity adjustments and priority setting as well as tools like throughput- and order
progress diagrams will be used to investigate patient throughput times and the degree to which formulated time standards are met. The findings suggest that these tools and concepts
are applicable, yet they have to be translated to fit the healthcare context. .
Key words: delivery reliability, healthcare management, throughput times, time standards, health operations management.
Research Theme: patient throughput times.
Seminar: Master’s Thesis Business Administration: Operations & Supply Chains. Seminar supervisor: dr. M.J. Land.
Word count: 8267. Table of contents: Chapter Page 1. Introduction 3 2. Theoretical Framework 6 2.1 Delivery reliability 6 2.2 Priority dispatching 8 2.3Capacity adjustments 8 3. Methodology 9
3.1 Research strategies and data gathering methods 9
3.2 Archiving data 9
4. Results and discussion 11
4.1Step 1: from Biopsy to Results Review 1 11
4.2 Step 2: from Results Review 1 to Scans 14
4.3 Step 3: from Scans to Results Review 2 17
4.4 Step 4: from Results Review 2 to Treatment 22
4.5 Summary of results 26 5. Conclusion 27 References 28 Literature 28 Websites 28 Appendices 30
Appendix 1: Flowchart of the carcinoma of the prostate care path 30
1. INTRODUCTION
Nowadays, hospitals in the Netherlands are not only competing in terms of the quality of the treatments they offer but they also compete on the waiting times that they propose to patients. From the patients’ point of view, this implies the period of time they have to wait before they receive their diagnosis and/or treatment from the hospital. Hospital activity has increased steadily in recent decades as new operations have been developed and improvements have allowed more people to be treated. At the same time, however, waiting lists have developed, as hospitals have been unable to deal promptly with all those capable of benefiting from treatment (Walshe and Smith, 2011 p.225). Moreover, in 2011, the former minister of public health, Ab Klink, predicted that waiting lists in health care would return, creating an even more burdensome problem because of an imminent labour shortage of medical personnel (Reerink and Stokmans, 2011). However, waiting is unpopular with patients and can be fatal. For example, a study on heart conditions found that for every month on the waiting list, a patient’s chances of dying increased by 5 percent (Sobolev, Fradet, Hayden et al., 2008). Therefore, regulations are used as vehicles to set performance, ‘process’ or ‘output’ targets for care providers. Though the reasons for the existence of waiting time of operations might differ from hospital to hospital, in general waiting times are seen as being caused by a sub-optimal circulation of patients. Waiting times are therefore seen as a (process) performance indicator of the hospital (Stoop, Vrangbæk and Berg, 2005). Another performance indicator is the lead- or throughput time of patients. The difference between the current throughput time and the ‘minimum’ or regulatory throughput time shows the gain to be made by for instance better patient planning (Vissers and Beech, 2005, p.112). This implies that more patients can be treated within the same period of time, hereby reducing waiting times for patients.
Reducing lead times is thus a relevant societal issue. Furthermore, recent research projects have shown that it is possible to reduce the throughput times of specific patient groups (Ekelund, Kurland, Eklund et al., 2011; Goldberg and Robbins, 2011). In order to achieve this, an adequate method to diagnose the causes of excessive throughput times is required. A method that is widely used for this diagnosis in make-to-order environments is the throughput diagram created by Wiendahl (1988). Another method is the order progress diagram
developed by Soepenberg, Land and Gaalman (2008). Managing the flow of patients through a process is important and can to some extent draw on approaches widely used in
manufacturing (Brideau, 2004). Many theories and techniques developed by business and management scholars may not be directly applicable to health operations management,
however. Nevertheless, the underlying principles may still hold and need to be translated to fit the healthcare context. This thesis should fill the gap in scientific literature on the
applicability of business and management techniques in healthcare management.
This research project was initiated by collaboration between the University of Groningen and a hospital in the Netherlands. It focuses in particular on the carcinoma of the prostate care pathway at the urology department. During the first conversation at the hospital, the specialised nurse (SN) stated that the only performance indicator currently available is the result of patient satisfaction enquiries. This indicates whether patients are satisfied with the care and treatments they have received but it does not provide information about whether the time standards are met. Therefore, the urology department aims to gain insight in the
realisation of their formulated time standards regarding the carcinoma of the prostate care pathway. A care pathway determines locally agreed, multidisciplinary practice based on guidelines and evidence where available, for a specific patient or client group. It forms all or part of the clinical record, documents, the care given and facilitates the evaluation of
standards are not met, the causes of these delays should be identified. At the same time, the Faculty of Economics and Business of the University of Groningen aims to investigate the applicability of business diagnosis methods in healthcare environments.
Therefore the objective of this research is to determine the degree to which formulated time standards are realised in the carcinoma of the prostate care pathway, and to determine the causes for delays. By using business diagnosis methods in this analysis, the applicability of these methods in a healthcare context will be explored simultaneously. Processes play an important role in the development of quality management systems and are often the starting point for the improvement of performance (Vissers and Beech, 2005 p.101). This implies that the time dimension of delivery reliability, the degree to which the formulated time standards are met, of the process must be addressed. This results in the central research question for this study:
‘How can the delivery reliability of the carcinoma of the prostate care pathway at the investigated hospital be improved?’
In order to assess the delivery reliability of the care path, the realized throughput times need to be identified first. These throughput times can than be compared to the standard times to determine the degree to which these standard times are met. After the delivery reliability is assessed, the factors that influence the delivery reliability need to be identified. When these factors are identified, suggestions for improvement can be provided. The delivery reliability is ought to be improved when the major factors that cause delays in patient flows are identified and improved. Based on the maturation and increase of the population, the amount of men by whom a carcinoma of the prostate is found is expected to increase with 64% in the period from 2000 to 2020 (Van Oers, 2002). Improving the delivery reliability of this care pathway could thus prove to be crucial in the coming years. For make-to-order companies, production planning and control (PPC) decisions can best be divided into input and output control
decisions. Input control decisions are acceptance/delivery date promising, release and priority dispatching (Soepenberg, 2010). Once orders are accepted and released, they remain in the process. Priority dispatching rules will then direct them along their downstream activities (Henrich, Land and Gaalman, 2004). These input decisions can be accompanied by output decisions in terms of capacity adjustments. Capacity refers to the ability of a resource to generate production, measured in the amount of patients per unit of time (Vissers and Beech, 2005 p.52). These input and output decisions influence the realised throughput time and therefore also the degree to which the formulated time standards are met, the delivery
Priority Setting
Capacity Adjustments
Delivery Reliability
+
Figure 1: Conceptual model
In order to answer the central research question about delivery reliability, a set of sub-questions has to be derived from the variables in the conceptual model (Figure 1). Once an answer to these sub-questions is found, the central research question can be answered:
How is priority assigned to patients in the carcinoma of the prostate care pathway and how does this influence delivery reliability?
What capacity adjustments are available and how do they influence delivery reliability?
2. THEORETICAL FRAMEWORK
Recall from the introduction that throughput diagrams and orders progress diagrams are useful tools for assessing the delivery reliability of a process. These tools will be explored in the next section. Hereafter, the concepts of priority setting and capacity adjustments will be explored as well as adjusted to fit the context of a hospital.
2.1 Delivery reliability
According to Soepenberg et al. (2008), achieving high delivery reliability is a combination of controlling the average lateness as well as the variance of lateness. The average lateness is the difference between the average realized throughput time and the average promised delivery time. The average lateness will increase when a larger number of order have to produced in a certain period assuming both capacity and promised delivery times remain the same
(Soepenberg et al., 2008). The throughput diagram is regarded as a useful tool to facilitate diagnosis of performance in terms of the average lateness (Soepenberg et al., 2008). The general layout of a throughput diagram is provided by Wiendahl (1988) and is displayed in Figure 2. The top line represents the cumulative input. Once an order, or in this case a patient, completes the first step in a process, the curve increases by one unit. The same applies to the bottom line in Figure 2, which represents the cumulative output. Once an order completes the second step, the cumulative output increases by one unit. The difference between the
cumulative output and the cumulative input can thus be regarded as the throughput time between the two steps. However, only for First Come First Served (FCFS) discipline the horizontal distances between both curves indicates the throughput times of individual orders (Soepenberg, 2010). The amount of WIP can be regarded as the vertical distance between the two curves (Soepenberg et al., 2008). This represents the amount of orders waiting for the second step.
Figure 2: The basics of a throughput diagram
orders are on the horizontal axis, some are above and some are below, which reflects the variance of lateness.
Figure 3: Order scatter diagram
An example of an OPD is displayed below in Figure 4. The horizontal axis shows the realised dates for the different stages of the process in terms of working days. The vertical axis shows the estimated lateness in different stages of the process. The estimated lateness after each stage is defined by Soepenberg et al. (2008) as the difference between the realised completion date of that stage and the virtual due date of that stage, assuming average throughput times for later stages until the final due date. Estimated positive or negative lateness after completing a certain stage means that an order will be delivered late or early, respectively. If the line segment is heading up, realised throughput times are longer than average throughput times in that stage and vice versa (Soepenberg et al., 2008). These upward or downward sloping line segments could be caused by the applied dispatching rules or output control decisions in the considered stage.
2.2 Priority setting
Priority dispatching is a relatively weak input control decision according to Kingsman and Hendry (2002). Once an accurate release decision has been made, priority dispatching has a limited influence on both the average lateness and the variance of lateness. However, some dispatching rules exist that still improve the delivery reliability (Soepenberg, 2010). Some operations serve customers in exactly the sequence they arrive. This is called first in first out sequencing (FIFO) or sometimes first come, first served (FCFS) (Slack, Chambers and Johnston, 2007 p.300). Yet, in a healthcare context it is more suitable to use the term priority setting. Admission planning refers to the operational planning of patients who need to be admitted as patients to a hospital (Kusters and Groot, 1996). Patients to be admitted to a hospital can be classified as elective, urgent or emergency based on their priority (Vissers and Beech, 2005 p.265). Elective patients do not have to be treated immediately, can be put on a waiting list, or can be given an appointment for admission. The waiting lists for elective patients are used as buffers for variations in the level of demand. The patients are then scheduled by picking them from the waiting list in some priority order. This philosophy of maximum resource use is increasingly viewed as unacceptable because priority is given to optimization of resource use without considering the consequences for the service level. Urgent patients need to be admitted on short notice, which is usually as soon as possible. Finally, emergency patients need to be admitted immediately. Emergency patients are thus given the highest priority, followed by urgent patients and finally the elective patients.
2.3 Capacity adjustments
Because of the stochastic nature of emergency patients flow, the workload will show huge variations over time. These variations can be handled only by deferring patients in the case of a fully occupied hospital or by creating over-capacity (Vissers and Beech, 2005 p.278). Soepenberg et al. (2008) state that output control decisions can dedicate capacity to those capacity groups where orders are congesting. They usually focus on controlling the average lateness but, by chasing specific requirement peaks across time, the variance of lateness may be reduced as well. However, most hospitals do not have a formalised and accepted method of allocating resources and revise these allocations when the annual contracts with purchasers or commissioners have changed (Vissers and Beech, 2005 p.116). Furthermore, the current practice of allocating resources to specialties within a hospital results in capacity losses. There are three reasons for this. First, the amount of a resource available to a specialty may not be in balance with the demand for that resource. Second, the timing of the allocations in terms of the periods that a resource is made available may lead to peaks and troughs in workloads. The third form of capacity loss may arise when the capacities of different resources that are
3. METHODOLOGY
The theoretical framework in the previous chapter described the relationships between the different concepts and their influence on delivery reliability. This chapter will explain which methodology was used to find an answer to the research questions. First the research strategy and the data gathering methods are discussed in paragraph 3.1 Hereafter an description of the data analysis and data processing will be presented in paragraph 3.2.
3.1 Research strategies and data gathering methods
This research will consist of a case study at the urology department of the hospital, a general hospital that offers a broad range of specialized care. It currently employs over 2500
employees and provides specialized treatments in 31 areas. One of these areas is the urology department. Usual practices include treating urinary infections, fertility issues and treating carcinomas in the kidney, prostate, bladder and testicles. Note that this study will only focus on the carcinoma of the prostate care path. A flowchart of the carcinoma of this care path can be found in Appendix 1. The used data gathering methods will be explained in Table 1. The table also explains what data gathering methods are used to analyse the different concepts depicted in the conceptual model.
Table 1: Research strategies, data gathering methods and concepts Research strategy Data gathering method Concepts
Case study Archiving data Delivery reliability
Interviews Priority dispatching
Capacity adjustments
First, theoretical research was conducted to support this study. Databases were used to find academic articles in order to gain understanding of the topic and tools used in this paper. Academic articles were also used to design the research process and report. Consulted databases are Business Source Premier, EBSCOhost COMPLETE and Informa Healthcare. Second, interviews were used to gain insight in the situation at the urology department. The first interviews of this research were unstructured and mostly explorative. During the first meetings at the hospital a clear overview of the care path, research objectives and available data was achieved. The first conversations were attended by an organizational manager, the head of planning and a Specialised Nurse (SN). This implies that multiple perspectives were available, providing a more objective view of the situation. During later conversations, a functional application manager (FAM) was also consulted for the delivery and interpretation of the data sets. Open interviews were held later on in this study in order to gain insight in the ‘why and how’ of the concepts depicted in Table 1. Planners from both the radiology
department as well as the urology department were consulted. In the third place a data set provided by the hospital was archived and analysed. This data archiving process is explained in the next paragraph. By combining both qualitative and quantitative research, a
comprehensive picture of the situation at the hospital is achieved. 3.2 Archiving data
4. RESULTS AND DISCUSSION
This chapter will present and discuss the results of the data analysis for the carcinoma of the prostate care pathway. A flowchart of this care pathway can be found in Appendix 1. In consultation with the SN, the care pathway was divided into four steps. These steps are from Biopsy to Results Review 1, from Results Review 1 to Scans, from Scans to Results Review 2, and finally from Results Review 2 to Treatment. Every patient with a carcinoma of the prostate is considered to pass through these steps at least once. Because the amount of data about patients who consecutively went through all these steps (only once) is limited, however, these steps need to be investigated separately instead of altogether. Calendar days are used in all of the following analyses. For all steps a throughput diagram, the distribution of lateness for the actual situation and the distribution of lateness in a First Come First Served (FCFS) simulation will be provided in paragraph 4.1 to 4.4. A throughput diagram will be the starting point for the analysis of every step. It should provide insight in the arrival and ‘departure’ of patients for each step. By adding the time standard of each individual step to the dates on which a patient arrived, a curve representing the virtual due dates can also be made. This should aid the appraisal of the deviation between the input and output of patients. Both the distribution of lateness for the actual situation and the FCFS simulation should provided insight in the variance of lateness for each step. Moreover, the FCFS simulation should provide understanding of how the variance of lateness could be altered if patients were treated in exactly the same sequence they arrived. This simulation is made by first looking at the date on which one additional patient receives a biopsy. Hereafter, the date at which one additional patient received Results Review 2 is registered. The time between these two dates can than be calculated and used to determine the lateness. Because the different time standards apply to one step instead of the care pathway as whole, an Order Progress Diagram is not used in this analysis. Therefore, a scatter diagram will provided for every step individually as well. This should provide insight in the variance of lateness over the investigated period of time. Also, if relevant, comments from conversations with the planners from the radiology and the urology departments will be discussed for every step. Finally, a summary of the results will be given in paragraph 4.5.
4.1 Step 1: from Biopsy to Results Review 1
In total 403 patients received 419 biopsies. Of these patients, 46 patients were removed because they did not receive a consultation at the hospital after their biopsy. This consultation is defined as Results Review 1 and is needed to calculate the throughput for this step. After removing these patients, 357 unique patients are left. For this step, 11 patients who were delayed more then 10 days and thus had a throughput time greater than 20 days (the time standard of 10 days together with a positive lateness of 10 days) are considered to be
the amount of work in process and the horizontal space represents the throughput time if First Come First Serve is applied. Both the lines for the Biopsy and Results Review 1 in the
throughput diagram show no major deviations from each other and run rather parallel to each other. Ideally, these two lines run parallel to each other (Soepenberg et al., 2008), implying that that the WIP and throughput times for this step are constant and controllable. Also, the red and blue lines tend to run on top of each other indicating that, in most cases, Results Review 1 was performed within the time standard of 10 days.
Another assessment of Step 1 can be made by looking at the distribution of lateness. This was done by first subtracting the dates for the Biopsy from the corresponding dates for Results Review 1. This results in an overview of the realized throughput times. The mean of all these realized throughput times is 8.90 days and the median is 8 days. Considering the time
standard of 10 days, these findings suggest that most patients are finished on time. On the other hand, some patients are finished late. Therefore, an overview of the variance of lateness will provide a more complete picture. This is provided in Figure 6. Figure 6 was created by subtracting the time standard of 10 days from the actual throughput time and counting the times a given lateness occurs. For example, there are 48 cases where the patient was informed after the biopsy in 6 days (6 – 10 = -4). The distribution of lateness for the FCFS simulation is displayed in Figure 6 as well to find out what would have happened if patients were treated in exactly the same sequence as they arrived. Considering the actual lateness, 266 of the 355 cases (75%) are labelled as ‘Negative lateness’, patients that receive their service on time or early. Thus, the majority of patients are finished early or right on time. On the other hand, 89 out of 355 (25%) are labelled as ‘Positive lateness’, implying the patients were served too late. Compared to the distribution of lateness for FCFS, the distribution for the actual situation shows more variation. If FCFS would have been applied 270 of the 355 (76%) cases would have finished on time. The average throughput time for FCFS is 8,89 days and the median is 8 days. .
Figure 6: Distribution of lateness Step 1
Figure 5: Throughput diagram Step 1 0 50 100 150 200 250 300 350 18-1-2010 15-3-2010 10-5-2010 5-7-2010 30-8-2010 25-10-2010 20-12-2010 14-2-2011 11-4-2011 6-6-2011 1-8-2011 26-9-2011 21-11-2011 Dates A m ou nt of pa ti e nt s
The scatter diagram for Step 1 is displayed in Figure 7. The horizontal axis shows the actual date at which Results Review 1 was finished and the vertical axis shows the lateness of an order. The scale of the horizontal axis is set to 56 days (8 weeks). No trend can be found in the scatter diagram so there is ought to be no seasonality in the distribution of lateness for Step 1.
Figure 7: Scatter diagram Step 1
-8 -6 -4 -2 0 2 4 6 8 10 12 18-1-2010 15-3-2010 10-5-2010 5-7-2010 30-8-2010 25-10-2010 20-12-2010 14-2-2011 11-4-2011 6-6-2011 1-8-2011 26-9-2011 21-11-2011
Date Results Review 1
La te ne s s Lateness
4.2 Step 2: from Results Review 1 to Scans
Also, in most cases the third scan is performed more than a year later than the second scan. These cases are considered to be exceptional. Therefore, only the cases for which the first or second scan is performed at a later date than Results Review 1 are considered. This means that, in some cases, the time between the Results Review 1 and the first scan, as well as the time between the Results Review 1 and the second scan are considered. In total, 124 scans remain for 94 patients. For these scans a throughput diagram is created, which is displayed in Figure 8 on page 16. The scale of the horizontal axis is set to represent 8 weeks (56 days). The blue line represents the amount of patients for Results Review 1, the cumulative input, the red line represents the amount of patients for the Scans, the cumulative output, and the green line represents the amount of patients that should have received a scan considering the time standard of 10 days. The (red) curve for the Scans drops below the (green) curve for the time standard a few times, indicating that these scans were finished late. However no major deviations are visible. Also, the red and green lines tend to run on top of each other indicating that, in most cases, the scan was performed within the time standard of 10 days.
Based on the same calculation that was used for Step 1, an overview of the distribution of lateness for Step 2 was made (Figure 9). Notably, the formulated time standard for this step is 7 to 10 days. Since this thesis focuses on whether the time standards are met, the maximum limit of 10 days was used for calculation of lateness. In comparison, the actual average throughput time is 8.48 days and the median is 7 days. Note that the average as well as the median is lower than the time standard. From the 124 scans 92 were finished on time (74%). However, in the remaining 32 the cases the patient received the scan too late. In case FCFS would have been applied, 94 patients (76%) would have been served on time with an average of 8.48 days and a median of 8 days. The planner from the radiology department stated that a distinction is made between routine and urgent patients. The time slots for the scanners are reserved for urgent patients. This implies that normal patients are waiting for time slots occupied by urgent patients. However, when these time slots tend to remain unused, normal patients are assigned at the last moment.
Figure 9: Distribution of lateness Step 2
Figure 8: Throughput diagram Step 2 0 20 40 60 80 100 120 140 29-1-2010 26-3-2010 21-5-2010 16-7-2010 10-9-2010 5-11-2010 31-12-2010 25-2-2011 22-4-2011 17-6-2011 12-8-2011 7-10-2011 2-12-2011 Date A m o u n t o f p a ti e n ts
The scatter diagram for Step 2 is displayed in Figure 10. The horizontal axis shows the actual date at which the Scan was finished. There is a slight increase in the dispersion of lateness between 8-5-2010 and 16-8-2010. This could be due to vacations from patients as well as from the medical personnel. This corresponds with the remark made by the planner from the radiology department that the availability of scanners decreases slightly during holidays but that this does not result in major delays. However, the distribution increases even more later on so there is no clear influence of seasonality.
Figure 10: Scatter diagram Step 2
-10 -5 0 5 10 15 20-10-2009 28-1-2010 8-5-2010 16-8-2010 24-11-2010 4-3-2011 12-6-2011 20-9-2011 29-12-2011 Date Scan L a te n e s s Lateness
4.3 Step 3: from Scans to Results Review 2
In total 225 patients have received 309 scans, which means that some patients had received more than one scan. For step 3 the scans are followed by an consultation regarding the results of these scans. This consultation after the scan is defined as Results Review 2. The care path description does not specify which scan should be considered as an end point for the time standard. Therefore 2 different analyses were made. One analyses starting from the first scan that was registered and one analysis focusing on the time after the second scan, if a second scan took place.
Figure 11: Throughput diagram Step 3 – first scan 0 20 40 60 80 100 120 140 160 180 12-1-2010 9-3-2010 4-5-2010 29-6-2010 24-8-2010 19-10-2010 14-12-2010 8-2-2011 5-4-2011 31-5-2011 26-7-2011 20-9-2011 15-11-2011 Date A m o u n t o f p a ti e n ts
The distribution of lateness for the analysis of the first scan is displayed in Figure 12. In summary, 68 of the 166 patients (41%) received a consultation early or right on time. This means that the majority, the remaining 98 patients is served late. Furthermore, the actual average throughput time is 9.19 days and the median is 8 days. Both are greater than the time standard so this acknowledges that the most of the patients were served late. In comparison, for FCFS 62 of the 166 patients (37%) would have been on time with an average throughput time of 9.19 days and a median of 8 days.
Figure 12: Distribution of lateness Step 3 - first scan
3 2 1 3 10 15 34 26 8 17 9 5 3 10 8 4 1 2 1 2 2 0 1 1 1 5 16 38 25 8 22 14 8 11 5 5 4 0 2 0 0 0 0 5 10 15 20 25 30 35 40 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Lateness in days A m o u n t o f p a ti e n ts Actual FCFS
The distribution of lateness for the second scan is provided in Figure 14 below. In summary, 24 of the 55 patients (44%) were served on time. On the other hand, more than half of the patients (56%) were served late. The actual average throughput time is 8.78 days and the median is 8 days. Both are greater than the time standard of 7 days. If FCFS would have been applied, these results would only slightly improve. In that case 25 of the 55 patients (46%) would have received a consultation on time with an average throughput time of 8.78 days and a median of 8 days.
Figure 14: Distribution of lateness Step 3 – second scan
1 1 2 3 2 6 9 9 6 3 3 1 2 1 0 2 1 3 1 1 2 3 2 4 12 5 7 2 2 2 3 3 3 2 1 0 0 2 4 6 8 10 12 14 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 14 Lateness in days A m o u n t o f p a ti e n ts Actual FCFS
Both distributions of lateness (for the first and for the second scan) show that the delivery reliability of this step is low. This is supported by statements made by the planner of the radiology department. One of the statements was that the meetings between the specialists of the radiology department happen on a weekly basis. This could explain the lateness because Results Review 2 cannot happen before the specialists meet to discuss the results and decide the follow up treatment. Furthermore, it was state that, in general, the documentation of the scans is available on the same day that the scans are finished. Yet, these documentations have to be ‘authorized’ by a specialist, which is not likely to happen on the same day. Also,
radiology distinguishes between routine and urgent patients. This urgency is assigned by specialists from their respective department (e.g. urology). In case of urgent patients the medical specialists deliberate by phone, the same applies to multidisciplinary consultation between medical specialists for different departments. It is likely that these urgent patients are also prioritized when it comes to consultation. Some patients are thus prioritized and served in a quicker fashion, thereby reducing the amount of time that is available to serve normal
patients.
Figure 15: Scatter diagram Step 3 combined -10 -5 0 5 10 15 19-1-2010 29-4-2010 7-8-2010 15-11-2010 23-2-2011 3-6-2011 11-9-2011 20-12-2011
Date Results Review 2
L a te n e s s
4.4 Step 4: from Results Review 2 to Treatment
For the analysis of the final step in the carcinoma of the prostate care path, an additional database was provided by the hospital. This database contains data for 118 treatments for 78 unique patients. However, not all cases described in the data file were useable. First, the data had to be filtered based on whether the patients have had a consultation before the treatment, labelled as Results Review 2. Also, some patients had received multiple treatments, ranging from two to four treatments per patient. All these treatments for one patient took place on one and the same date, however. Therefore, in agreement with the CNS, one date per patient was used for the analysis of the step. Because this is the final step in the care path and the amount of patients that received treatment the extreme values are not deleted. In total 39 patients remain, 8 of which had received a Lymph Node Dissection (LND). The remaining 31 patients had received either a lymphadenectomy or a prostatectomy and are labelled as having
Figure 16: Throughput diagram Step 4
0
5
10
15
20
25
30
35
40
31-5-2010 26-7-2010 20-9-2010 15-11-2010 10-1-2011 7-3-2011 2-5-2011 27-6-2011 22-8-2011 17-10-2011 12-12-2011 DateA
m
o
u
n
t
o
f
p
a
ti
e
n
ts
The distribution of lateness for the 8 patients that received a LND is depicted in Figure 17. Because of the limited amount of data regarding the patients that received an LND all patients are used in the analysis. Note that not one of the patients has a negative value for lateness. This means that every patient was treated too late. Also note that for both the actual situation and the FCFS simulation the distribution of lateness is the same. The average throughput time for both cases is 38.50 days and the median is 28.50 days. Both are more than twice the size of the time standard of 14 days.
Figure 17: Distribution of lateness Step 4 - LND
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 1 2 5 6 12 13 16 18 23 103 Lateness in days A m ou nt of pa ti e nt s Actual FCFS
Figure 18 shows the distribution of lateness in days for the remaining 31 patients that received Surgery. Because of the limited amount of patients that received Surgery, all patients are considered. In the actual situation only 3 patients (10%) were treated on time. This means that the remaining 28 patients (90%) were all treated too late. The actual average throughput time is 32.84 days and the median 28. Both are (more than) twice the time standard for this step. In case FCFS would have been applied the average throughput time remains 32.84 days but the median becomes 31 days. Furthermore, the amount of patients treated on time decreases even further: 2 of the 31 patients (6%) are treated on time. When the treatment of patients is structurally running behind schedule it is logical that the amount of patients that is treated on time decreases even further for FCFS.
Figure 18: Variance of lateness Step 4 - Surgery
The combined results for Step 4 indicate that the delivery reliability of this step of the care pathway is low. This is supported by statements made by the planner of the urology department (where the surgery or LND is performed). Like for the radiology department, priority or urgency is assigned to patients by the medical specialist, who also determines the time span of the care pathway for that patient. In this case the formulated time standards are not always taken into consideration. The planners need to reserve time slots for surgery for urgent patients, thereby also risking that these slots remain unused in the end. However, they try to fill these empty slots at the last moment with normal patients. Still, these urgent patients prevent routine patients to be treated and therefore tend to delay these patients. The planner also stated that most delays are a result of a lack of capacity. Another reason for delays, according to the planner, is that the anaesthesiologist needs to consult the patients about their use of medication approximately one week before their treatment.
The distribution of lateness over time for both treatments is displayed in Figure 19. From 22-6-2010 to 30-9-2010, around 27-7-2011 and near the end of the year 2011, the lateness tends to increase. There seems to be some effect of the holiday seasons but not conclusive,
however. According to the planner of the urology department, the time available for surgery decreases by 50% during the holiday seasons. This could be deliberate because fewer patients are expected to demand a treatment during the summer holiday. It could also be due to the vacation of surgeons. One reason for the delay of patients could therefore be the lack of capacity. However, another explanation could be that patients require their treatment to be postponed to a date after their planned vacation.
Figure 19: Scatter diagram Step 4 combined
4.5 Summary of results
For Step 1 it was found that from the 355 patients that were analysed, 266 patients (75%) was treated early or right on time. The mean throughput time for the first step was found to be 8,90 days and the median 8 days. Compared to the formulated time standard of 10 days, both statistics explain why most of the patients were on time. Both the input and output curve in the throughput diagram show no major fluctuations and tend to run parallel to each other. According to Soepenberg et al. this implies that the amount of patients waiting for their first results review as well as the throughput times are constant and controllable. If FCFS would have been applied in step 1 only an additional 1% of the patients would have been treated early or on time. No conclusive influence of seasonality was found for step 1.
Based the throughput diagram for Step 2 it was found that in most cases the scans were performed on time. No major deviations between the amount of patients that received Results Review 1 and the amount of patients that were waiting for their scan were found. From the 124 scans (for 94 patients) that were analysed, 92 (74%) were finished on time. The average throughput time for this step is 8,48 days and the median 7 days. Considering the time standard of 10 days both statistics explain why the majority of the patients were served on time. The simulation of FCFS showed only small improvement. In comparison with the actual situation, 94 patients (76%) would have received their scan on time. No trend or seasonality was found in the distribution of lateness over time.
For the third step both throughput diagrams started to show more deviations between the amount of patients that received a scan and the amount of patients that was waiting for their second results review. For the first scan it was found that 68 patients from total of 168 analysed patients (41%) received Results Review 2 early or on time. This is explained by a mean throughput time of 9,19 days and a median of 8 days, both of which are greater than the time standard of 7 days. For the second scan it was found that from the total of 55 patients that were analysed, 24 patients (44%) received their consultation on time. In this case, both the average throughput time of 8,78 days and the median of 8 days are also greater then the time standard of 7 days. The simulation of FCFS for the first scan showed that the amount of patients that would have been treated on time decreases from 68 to 63 (37%). For the second scan this simulation showed only a very small improvement, from 24 to 25 (46%). Based on comments made by the planner from the radiology department this could be due to the distinction between normal and urgent patients and their inherent priority. However, this is not found in the data. Neither was a conclusive influence of seasonality was found in the scatter diagram for Step 3.
5. CONCLUSION
The main objective of this thesis was to determine delivery reliability of the carcinoma of the prostate care pathway, and to determine the causes for delays so improvements can be made. Both priority setting of patients and capacity adjustments were expected to positively
influence the delivery reliability. The delivery reliability for every step in the care pathway was assessed by calculating the degree to which the formulated time standards were met. Even though no clear goals were stated in relation to the realisation of time standards, the findings of this thesis suggest that the delivery reliability can be improved for all steps. For the first step, 75 percent of the patients was served early or on time. For Step 2, 74% of the scans was performed on time. Only 39 percent of the patients was consulted on time for Step 3. Finally for Step 4, less than 8 percent of the patients was treated on time. Therefore
improvements should be mainly aimed at Step 3 and 4.
Capacity adjustments are very likely to increase the delivery reliability of these final two steps. The average throughput times as well as the medians for both step 3 and 4 are considerably higher than the formulated time standard. This indicates that the majority of patients is likely to be late and by adding capacity these mean throughput times could be reduced. Also, during the holiday seasons the time available for surgery decreases by 50 percent. However, this is based on comments made by the planners from the radiology department and the urology department and is not apparent from the data. Therefore, additional research is required to determine how capacity adjustments can improve the delivery reliability of the care pathway. Priority setting of patients could also increase the delivery reliability. The hospital distinguishes between normal and urgent patients. This last group of patients is given priority over normal patients. Time slots for scanners and surgery are reserved for this type of patients, hereby limiting the availability of these shared resources for normal patients. According to Vissers and Beech (2005 p.265), this philosophy of
maximum resource use is increasingly viewed as unacceptable because priority is given to optimization of resource use without considering the consequences for the service level. Further investigation of the care pathway needed to identify alternative methods of patient priority setting and their influence on delivery reliability. Conclusively, it became clear that the time standards for the care path are considered by planners and medical specialist only in some cases. Without consideration of these time standards it is likely that these are not met. Adding surveillance, registration and responsibility regarding the realisation of time standards can thus be a starting point for improving the delivery reliability.
Two key future research directions emerge from this study. First, the concept of delivery reliability was applied to a healthcare context. Both the influence of capacity adjustments and priority setting on delivery reliability were investigated. However, additional research is needed to verify both relationships as well as the possible influence of the other input control decisions. Second, the applicability of business diagnosis methods for MTO companies in a healthcare context was tested. It was found that these methods are applicable to some extend, depending on the degree to which the care pathway or process can be divided in measurable, consecutive steps. However, more research is needed to increase the knowledge about the differences and similarities between MTO companies and hospitals as well as the degree to which patients and orders are interchangeable.
REFERENCES
Literature:
Ekelund, U., Kurland, L., Eklund, F., Torkki, P., Letterstål, A., Lindmarker, P., and Castrén, M., 2011. Patient throughput times and inflow patters in Swedish emergency departments. A Basis for ANSWER, A National Swedish Emergency Registry.
Scandinavian Journal of Trauma, Resuscitation and Emergency Medicine, 19 (37).
Goldberg, A.J., and Robbins, S.B., 2011. Portion Control Opportunities: Real Time Gains for Hospital Patient Throughput. Journal of Healthcare Management, 56 (5): p.293-297.
Henrich, P., Land, M., and Gaalman, G., 2004. Exploring applicability of the workload control concept. International Journal of Production Economics, 90 (1): p.187-198.
Kingsman, B.G., and Hendry, L.C., 2002. The relative contributions of input and output control on the performance of a workload control system in make-to-order companies. Production Planning & Control, 13 (7): p.579-590.
Kusters, R.J., and Groot, P.M.A., 1996. Modelling resource availability in general hospitals. Design and implementation of a decision support model. European Journal of Operations Research, 88: p.428-445.
Luc, de K., 2000. Are different models of pathways being developed? International Journal of Health Care Quality Assurance, 13 (2): p.80-86.
Oers, van, J.A.M, 2002. Gezondheid op koers? Volksgezondheid Toekomst
Verkenning 2002, RIVM rapport nummer 270551001.
Reerink, A., and Stokmans, D.. Klink: de wachtlijsten in de zorg komen weer terug, in: NRC Handelsblad, november 14th, 2011.
Schönsleben, P., 2007. Integral Logistics Management: Operations and Supply Chain
Management in Comprehensive Value-Added Networks. Florida, USA: Auerbach Publications.
Slack, N., Chambers, S. and Johnston, R., 2007. Operations Management. Harlow,
UK: Prentice Hall.
Sobolev, B.G., Fradet, G., Hayden, R., Kuramoto, L., Levy, A.R., and Fitzgerald, M.J., 2008. Delay in admission for elective coronary bypass grafting is associated with increased mortality. BMC Health Services Research, 19 (8): p.185.
Soepenberg, G.D., 2010. ‘Workload control under diagnosis’. Groningen, The
Netherlands: University of Groningen, PhD thesis.
Soepenberg, G.D., Land, M. And Gaalman, G., 2008. The order progress diagram: A
supportive tool for diagnosing delivery reliability performance in make-to-order companies. International Journal of Production Economics, 112 (1): p.495-503.
Stoop, P.P.M., and Bertrand, J.W.M., 1997. Performance prediction and diagnosis in
two production departments. Integrated Manufacturing Systems, 8 (2): p.103-109. Stoop, A.P., Vrangbæk, K., and Berg, M., 2005. Theory and practice of waiting time
Wiendahl, H.P., 1988. The throughput diagram – An universal model for the
illustration, control and supervision of logistic processes. Annals of the CIRP, 31 (1): p.6-9.
Websites:
http://www.martiniziekenhuis.nl/.
APPENDICES
Appendix 2: Example lay out of datasets
Example lay out of the dataset describing all finished (DTC’s) started in 2010/2011 for patient 00002089
Column Description Comment Example
A Patientcode 00002089
B Date 19-3-2010
C Activity 190013
D Description of activity Repetition polyclinic visit
with an in progress DTC. E Costing unit 3770 F Description of costing unit Polyclini Urology G Amount 1 H DTC number 1000182704
I Dataset Describes department, care type, demand for care, diagnosis and treatment. In this case Urology Department.Column K.column M. column.O.column Q.
06.21.11.41.323
J Description Sequential, raised prostate
specific antigen (PSA)
K Care type Sequential
L Description of care type Sequential
M Demand for care 11
N Description of demand for care
Raised PSA
O Diagnosis 41
P Description of diagnosis Benign prostatic hyperplasia
(BPH).
Q Treatment 323
R Description of treatment Endoscopy
S Starting date DTC 7-3-2010
T End date DTC 9-7-2010
U Performer of activity 00355
V Performing specialist Urologist
W Applicant of activity 00355
X Applicant specialist Urologist
Y Type of applicant Internal specialist
Z Care trajectory number 1000393460
AA Starting date care trajectory
9-9-2011
Example lay out of the dataset describing patient appointment dates for patient 00002089 Column Description Example
