The Influence of On-site Medical Specialist Staffing on Crowding Patterns at the Emergency Department

__________________________________________________________________________________________________________________________________

The Influence of On-site Medical Specialist Staffing on

Crowding Patterns at the Emergency Department

Master Thesis Supply Chain Management S.L. van der Meulen1

June 22, 2018

Supervisors:

Dr. M.J. Land Dr. J.T. van der Vaart

1 _{University of Groningen, Faculty of Economics and Business.}

2

Abstract

While medical specialists are needed for several tasks at the emergency department (ED) that are crucial for the patient flow, they are not physically present at the ED. As a result, waiting for medical specialists at the ED was found to be a significant impeding factor to crowding at an ED in the Netherlands. Literature stressed the importance of investigating the patterns of inflow and outflow of patients throughout the day, as it reveals responsiveness problems of the ED and other causes of crowding. This research investigates how the physical presence of medical specialists at the ED influences the crowding patterns at the ED. A single case study is executed at a large sized ED in the Netherlands. During a pilot, this hospital staffed one medical specialist of surgery, cardiology, internal medicines, neurology and radiology to be physically present at the ED between 12:00 and 20:00, in addition to regular staffing levels. By using input/output analysis, it is investigated how the presence of specialists affects the responsiveness of the ED throughout the day. Results show that the output of patients responded faster to peaks in the inflow with the presence of medical specialists at the ED. The ED was capable to treat more patients in a shorter time; the average length of stay of patients was shortened with ten minutes during the presence of the medical specialists. Patients of surgery were found to benefit the most from the presence of the medical specialists in terms of their length of stay at the ED. It was found that multiple factors could have contributed to these results.

3

Table of Content

2.3 Causes of crowding at the ED ... 9

2.3.1 The role of medical specialists ... 10

2.3.2 Responsiveness of ED personnel ... 10

2.4 Solutions for crowing at the ED ... 10

2.4.1 Scheduling additional ED personnel ... 11

2.4.2 Scheduling medical specialists at the ED ... 11

2.5 Research implications ... 11

3 Methodology ... 12

3.1 The case hospital ... 12

3.1.1 The processes at the ED of the case hospital ... 12

3.1.2 The pilot ... 13

3.2 Data ... 14

3.3 Data analysis ... 15

3.4 Conclusion ... 16

4 Results ... 17

4.1 General dataset analysis... 17

4.1.1 Dynamics of inflow and outflow of patients ... 18

4.1.2 Length of stay and work in progress ... 19

4.2 Dynamics per specialism ... 20

4.2.1. Surgery ... 20

4.2.2. Cardiology ... 20

4.2.3. Internal medicines ... 21

4.3 Set of similar days ... 22

4.3.1. General analysis set of similar days ... 22

4.3.2 Dynamics of inflow and outflow of patients ... 23

4.3.3. Length of stay and work in progress ... 23

4.3.4. Analysis per specialism ... 24

4.4 Comparison to how the pilot is experienced ... 25

4.4.1. Outcomes of the survey ... 25

4.4.2. Comparison of the outcomes of the survey to the outcomes of the diagrams... 25

4

6 Conclusion ... 29

Bibliography ... 31

Appendix ... 34

A. Descriptive statistics total dataset ... 34

B. Dynamics per specialty ... 37

C. Selection set of similar days ... 38

D. Descriptive statistics set of similar days ... 40

E. Analysis set of similar days ... 41

F. Survey questions and respondents ... 44

5

1 Introduction

Medical specialists play a substantial part in the acute care process at the emergency department (ED) as they are expected to supervise their residents and advise emergency physicians or other specialists at the ED when needed. Despite these tasks, medical specialists are not physically present at the ED. They work at their own departments in the hospital, which makes it hard to reach them. Consequently, waiting for the specialists to perform their tasks at the ED has found to be a contributing factor of crowding (Derlet & Richards, 2000; Have, 2016; Hoeben, 2017; Magid et al., 2009). This study examines the effect of the physical attendance of medical specialists on dynamics of crowding in the ED. Note that in this article, medical specialists are non-emergency physician medical specialists.

Crowding at EDs is a growing international problem (Moskop, Sklar, Geiderman, Schears, & Bookman, 2009; Pines & Griffey, 2015). Crowding is defined as the situation in which there are inadequate resources to meet patient care demands which leads to a reduction in the quality of care (Pines, 2007). It can result in longer waiting times for patients and in the worst case the ED may no longer be able to treat new patients, being forced to refer the patients to other EDs in the area. Numerous studies have shown that crowding at the ED is a significant risk for patient safety; it is even associated with a higher mortality rate (Bernstein et al., 2009; Carter, Pouch, & Larson, 2014). Hence, resolving crowding at the ED could make a difference between life and death.

6

(2009) shows that only 41% of the staff of an ED in the United States could indicate that specialists arrive within 30 minutes after being contacted for the examination of a critically ill patient. This is alarming as critically ill patients need to be treated as quickly as possible. Moreover, delays in consultation were mentioned most frequently (80% of the respondents) as a contributing factor to crowding at EDs in the Netherlands (van der Linden, 2013). As a consequence of specialists being difficult to reach, residents batch the patients needed to be seen by a specialist, in order to discuss them consecutively instead of bothering them with every single question and thus waiting for every single answer (Hoeben, 2017). This causes patients to wait longer for the needed care which hinders patient flow. So on the one hand, the supervision of medical specialists improves patient flow and their knowledge is needed to make important decisions. But on the other hand, asking for supervision or advice of specialists is discouraged as it is hard to reach them which consequently slows down the process at the ED.

Multiple studies showed that that improved responsiveness of ED staff to patient arrivals can be the key to improved patient flow in the Netherlands, as it was often found that output responds too slowly to peaks in the patient inflow (Koomen, 2015; Mulder, 2017; van Manen, 2015). The importance of analysing the inflow and outflow patterns of patients at the ED throughout the day is emphasized by Ter Avest, Onnes, Van der Vaart, and Land (2018), as it reveals these responsiveness problems of the ED.

In order to investigate whether presence of medical specialists at the ED influences the responsiveness of the ED, this study examines the inflow, outflow and throughput patterns of patients at the ED. Consequently this paper addresses the question: How does the physical attendance of medical

specialists influence the crowding patterns at the ED?

7

knowledge on resolving crowding at the ED. The study is executed by performing a single-case study at the ED of the Haaglanden Medical Centre, a large sized hospital in The Netherlands.

8

2 Theoretical background

This chapter provides an overview of relevant literature for this study. First, the processes at the ED are explained. Hereafter a summary of literature on crowding at the ED is provided, followed by an overview of causes and solutions of crowding related to the scope of this research. The fifth paragraph explains the research framework and the last paragraph concludes.

2.1 Process at the ED

The ED of a hospital is responsible for treating patients arriving at the ED that are in need of immediate care. This implies that the ED provides unscheduled care, which makes it hard to predict the necessary resources to provide the patients with acute care (Asplin et al., 2003; Luscombe & Kozan, 2016). The staffing and organization of EDs may vary across countries and hospitals. Generally, in the Netherlands nurses, nurse practitioners, emergency physicians (EPs) and residents work at the ED (Reiring, 2018). In most hospitals the residents of cardiology, internal medicine, neurology, paediatrics, radiology, surgery and residents of the ED work at the ED.

Patients arrive at the ED through referral of the general practitioner (GP), by emergency medical transport or by self-referral. When the patient enters the ED he or she will be registered first by the receptionist. As patients arrive under different circumstances, their level of urgency must be determined in order to assign treatment priorities. This is done by triage, which is defined as “the process of quickly sorting patients to determine priority of further evaluation of care at the time of patient arrival in the emergency department” (Fernandes et al., 2005, pp. 31). A widely used triage scale is the Manchester Triage System (MTS), introduced in 1997 (Mackway-Jones, 1997). It entails a five-point triage scale in which each colour represents triage category, depicted in table 2.1. Here, the maximum time is the desired maximum time in which the patient must see the treating clinician. Table 2.1 Manchester Triage System (Mackway-Jones, 1997)

After triage the patient either waits in the waiting room or immediately goes to an examination room. In the examination room, a nurse or EP examines the patient and the patient will be assigned to a specialty (Hoeben, 2017). In most cases, the EP or the resident of the specialty performs the initial treatment. However, a specialist of the assigned specialty is in the end responsible for the patient.

Number Colour Triage category Max. Time (minutes)

1 Red Immediate 0

2 Orange Very urgent 10

3 Yellow Urgent 60

4 Green Standard 120

9

Therefore the EPs and residents need to report to the medical specialist in order to supervise their decisions. Moreover, when patients require services or knowledge outside the scope of practice of the EP or specialist, a consultation of the assigned specialist or a specialist of another specialty will take place (Reiring, 2018). Both the consultation and supervision are initially done by phone and when necessary the specialist comes to the ED to examine the patient. Additional laboratory tests, scans or x-rays might be performed to help forming a diagnosis. Subsequently the patient is diagnosed and treated. Finally, the patient is admitted to the hospital, moved to an external facility or discharged home.

2.2 Crowding at the ED

Crowding is defined as a mismatch between supply and demand, which implies that there are insufficient resources at the ED to fulfil the demand of the patients (Pines & Griffey, 2015). This results in prolonged waiting times and reduced quality of care for patients (Pines, 2007). For the past twenty years, crowding at EDs has been an international problem and many have examined the phenomenon (Moskop et al., 2009; Pines & Griffey, 2015). Unless the latter, it is still hard to find universal measures and solutions for crowding as every hospital has its own characteristics (Hoot & Aronsky, 2008; Pines & Griffey, 2015).

Asplin (2003) developed a conceptual model of emergency department crowding which separates crowding into three components: input, output and throughput. The first component, input, consist of every event or condition that contributes to demand for ED services. Secondly, output involves the disposition of ED patients. This is mostly defined as the bottleneck causing crowding, due to a lack of inpatient beds (Boyle, Beniuk, Higginson, & Atkinson, 2012). Finally, throughput is influenced by all processes a patient goes through inside the ED, in other words it identifies the length of stay of patients. Hence, to improve throughput it is needed to improve the processes internally in the ED. This research investigates the latter.

2.3 Causes of crowding at the ED

10

2.3.1 The role of medical specialists

Waiting for specialty consultations is often named as a contributing factor of crowding at the ED (Derlet & Richards, 2000; Have, 2016; Hoeben, 2017; Magid et al., 2009). As residents and EPs should always report to the responsible specialist and often specialist’s knowledge is needed to diagnose or treat the patients, specialists have a significant role in the process at the ED. Since medical specialists are not present at the ED, it often takes too much time to fulfil this role (Brick et al., 2014). The waiting for specialty consultation differs per specialty (Yoon, Steiner, & Reinhardt, 2003). Research by van der Linden (2013) showed that 80% of their respondents mentioned delays in consultation as a contributing factor to crowding at EDs in the Netherlands. As a consequence, ED staff may batch patients that need to be examined by a specialist in order to treat them consecutively (Hoeben, 2017). In that way, residents do not have to call the specialists for every single patient and thus they do not have to wait multiple times for this specialist to arrive. However, this prolongs the waiting time for some of the patients and thus keeps the examination rooms unnecessarily longer occupied which hinders the patient flow. Furthermore, the work of Hoeben (2017) shows that supervision of residents by specialists helps diagnosing the patient in an early stage which enhances the patient flow and thus the responsiveness of the ED, which will be discussed in the following paragraph.

2.3.2 Responsiveness of ED personnel

The way ED staff reacts to new patient arrivals can be the key to improve patient flow (Van Achteren, 2014). Responsiveness problems arise when output responses too slowly to peaks in the patient inflow. When arrivals at the ED are increasing, productivity of the ED should also increase, otherwise the amount of patients in the ED will build up. Hence the lack of responsiveness contributes to crowding at the ED. However, ED staff should not only respond when the ED is getting crowded. This is shown by Ter Avest, Onnes, Van der Vaart and Land (2018) who state that accepting longer throughput times when the ED is relatively quiet results in patient accumulation at the ED later in the day. To detect whether there is a lack of responsiveness causing backlogs of patients, the analysis of the dynamics of inflow and outflow patterns of the ED is necessary (Ter Avest et al., 2018).

2.4 Solutions for crowing at the ED

11

2.4.1 Scheduling additional ED personnel

Many studies have suggested to improve staffing in order to reduce crowding at the ED (Hoot & Aronsky, 2008). However, limited research is done on the effect of improved staffing at the ED. Prior research that has been done shows divergent results. The work of Bucheli and Martina (2004) shows that adding one extra physician during a busy shift in an ED in Switzerland reduced outpatient LOS with 35 minutes. However, other studies show that additional staffing does not necessarily lead to shorter LOS. Kawano, Nishiyama and Hayashi (2014) who conducted their study in Japan, argue that the effect of additional ED personnel is dependent on the personnel’s experience. They found that an extra junior residents prolonged the patient’s LOS at the ED while adding an extra senior residents decreased the LOS.

2.4.2 Scheduling medical specialists at the ED

Van der Linden et al. (2017) investigated the effect of medical specialists being present at the ED or at the hospital during out-of-hours instead of being available on-call outside of the hospital. The LOS of the total amount of patients did not decrease, while the LOS of admitted patients decreased from 197 to 181 minutes. Comparing these results to the work of Bucheli and Martina (2004), it suggests that higher decrease in outpatient LOS is obtained by adding one extra physician instead of adding five extra specialists during busy shifts at the ED.

2.5 Research implications

Literature indicates that crowding in EDs is still an international problem that needs to be resolved. Lack of responsiveness of ED personnel and waiting for specialty consultants have found to be contributing factors to crowding at EDs that put patients at risk. The physical presence of the specialists at the ED would suggest that they are able to fulfil their roles at the ED faster, which might lead to shorter LOS of patients. This improves the outflow of patients which on its turn might enable a higher responsiveness to the inflow of patients.

12

3 Methodology

This study aims to examine the effect of the physical attendance of medical specialists at the ED on its crowding patterns. To thoroughly analyse this effect, a single case study is executed. This enables to observe the phenomenon in its natural setting and describes the effects in a rich and detailed way (Voss, Tsikriktsis, & Frohlich, 2002).

3.1 The case hospital

This single case study is performed at the ED of a large sized teaching hospital called ‘’Haaglanden Medical Centre’’ (HMC), located at the western part of the Netherlands. The hospital has three locations, this study is performed at the ED of the location ‘’Westeinde’’. The ED contains of 24 treatment rooms and around 52.000 patients visited this ED last year.

3.1.1 The processes at the ED of the case hospital

Figure 3.1 illustrates the process a patient goes through when visiting the ED of this specific hospital. When the patient arrives at the ED, he or she is registered in the system and triaged by a nurse. The triage nurse assigns every patient to a specialty. When a multidisciplinary patient comes in, it is unclear to which specialism the patient should be assigned. Multiple specialists must see the patient before he or she will be assigned to a specialty. The specialists see the patient in sequential order, as often the specialists are not simultaneously available to see the patient. When the patient is assigned to a specialty, a medical specialist of this particular specialty is responsible for the patient which implies that the specialist should always be involved in the decisions about the patient. But normally specialists are not physically present at the ED. The EPs or the residents diagnose and treat the patients. They may contact the specialist for advice and if needed the specialist comes to the ED to see the patient. It is possible that even this specialist is unable to diagnose or treat the patient. Then the patient will be reassigned to another specialty. Thereafter, radiology and lab requests may be executed. Finally, after treating the patient he or she is discharged, admitted to a department or moved to an external facility.

13

Registration Triage Assigned to a

specialty Treated by EP or resident of assigned specialty Treated by specialist or EP Discharged Admitted to ward Moved to external facility Specialists see patients in sequential order If multidisciplinary patient If resident unable If specialist unable

3.1.2 The pilot

During the pilot, specialists of the five main specialties were physically present at the ED from 12:00 till 20:00, in addition to the regular staff occupation. These specialties were surgery, cardiology, neurology, internal medicines and radiology. In this way, the hospital hoped to reduce the patient’s LOS, the number of diagnostics, the admissions to the departments, and subsequently improve the quality of care.

From 3 October 2016 until 27 November 2016 a similar pilot study was carried out by van der Linden et al. (2017). However, in this pilot the specialists were not obliged to be physically present at the ED. During shifts where the specialists would initially have been on-call at their homes, they now had to be present at the hospital and available for ED staff when needed. Outcomes showed that the average LOS of all patients remained unchanged (van der Linden et al., 2017). Only the LOS of admitted patients decreased from 197 min to 181 min. They suggested that the positive effect of the pilot would have been more pronounced when the specialists would haves been present at the ED continuously. To gain more knowledge the pilot was repeated over a longer period (1 November 2017 until 31 May 2018). During this pilot the five specialists had to be physically present at the ED from 12:00 until 20:00, fully committed to the ED. This timeslot was chosen as the hospital found these to be the peak hours at the ED. The process during the pilot is depicted in figure 3.2. Still, the patient is registered, triaged and assigned to a specialty first. However, when a multidisciplinary patient comes in, multiple specialists will see the patient simultaneously and jointly decide to which specialty the patient will be assigned.The patient has to explain his or her complaints only once to all the specialists together instead of speaking to each specialist separately. After this, a resident or EP will diagnose and treat the patient. When assistance is required they may contact the specialist. However, in the case that a complex patient arrives, the specialists might assist and treat the patient immediately. Lab and radiology requests are executed when needed. After treating the patient, he or she is discharged, admitted to a department within the hospital or moved to an external facility.

14

During the pilot, agreements on the availability of specialists were similar to control period for all specialisms. Still one specialists of every specialism should be on-call available and should be at the ED within 15 or 30 minutes. But as the specialists of surgery, cardiology, neurology, internal medicines and radiology were physically present at the ED between 12:00 and 20:00, they had to be immediately available for advice and supervision during this time slot.

Registration Triage Assigned to a specialty Treated by EP or resident of assigned specialty Treated by specialist(s) and resident(s) together EP/resident/ specialists help simultaneously Discharged Admitted to ward Moved to external facility If multidisciplinary patient If resident unable

3.2 Data

The case hospital uses an information system to record patient data, which includes the arrival number, patient’s age, type of referral, time of arrival, time of triage, triage code, specialty assigned, diagnose code, time the medical specialists first sees the patient, number of lab requests, number of radiology requests and time of discharge. The hospital provided this data for the period of 01-11-2017 until 30-03-2018, which is the pilot period, and 01-11-2016 until 30-03-2017, which is the control period. The time a medical specialist first sees the patient was determined by the time the medical file was opened first. Unfortunately, this was found to be an unreliable measure as the data was incomplete and inconsistent. Therefore this data could not be considered for the analysis. Patients that were directly send to the GP after being triaged were deleted from the dataset, as these patients have a very short LOS that would distort the average LOS. For the control period 3874 records and for the pilot period 4837 records were removed. The descriptive statistics for both periods can be found in appendix A.

Next to the time-series data, the hospital provided outcomes of a survey conducted with the ED staff and specialists to investigate how the pilot was experienced. The survey was fully completed by 245 participants of which 98 medical specialist and 12 emergency physicians. The other 135 respondents were nurses, interns, residents, chefs de Clinique and X-ray technicians. For this research, six questions were used to analyse the timeslots at which the presence of specialists was considered most useful. These survey questions and information of the respondents of the survey can be found in appendix F. In addition to using above described secondary data, the ED was observed and field notes were made in order to analyse the processes at the ED and the role of the specialists. This enabled the writer to interpret and analyse the data properly.

15

3.3 Data analysis

The input-output modelling technique of Soepenberg, Land, & Gaalman (2012) was used to analyse the effect of the presence of specialists on the patient flow at the ED. This technique enables to analyse the pattern of inflow and outflow of patients throughout the day. The framework was actually designed for improving throughput times of make-to-order companies, but has also been used successfully in in-depth analyses of the daily dynamics at the ED (Verbree, 2013). Especially the input-output modelling technique is useful as it provides insight into the patterns of the three components that Asplin et al. (2003) used to describe crowding at the ED: input, throughput and output. It reveals the pattern of the patients entering and leaving the ED and as such provides insight into the patients’ throughput times. In this research the patient throughput time is called the length of stay (LOS), which is defined as the time that the patient spends inside the ED, from the moment of arrival until the departure from the ED. The LOS is determined by input- and output control decisions (Kingsman & Hendry, 2002). At the ED, input is hard to control as patient just enter with established priorities. However output is controllable as it entails capacity control decisions (Soepenberg, Land, & Gaalman, 2008). In the scope of this research the output control decision is the staffing of specialists at the ED. Hence the input-output modelling technique is applicable to analyse the effect of this output control decision on the flow of patients. It allows to analyse the in- and outflow patterns of patients at the ED and detects whether backlogs of patients building up throughout the day.

For this technique a throughput diagram is used (Wiendahl, 1988). This diagram shows the average cumulative in- and outflow of patients throughout the day. This is calculated by accumulating the total amount of patients that arrived per minute, divided by the total amount of days in the dataset. In this study, the horizontal axis of the throughput diagram represents the time in hours of the day and the vertical axis represents the cumulative number of patients. The horizontal distance between the arrivals and departures curves in the throughput diagram shows the time a patient spends on average in the ED, which is the LOS. The vertical distance between the two curves is equal to the work-in-progress (WIP), which represents the amount of patients that are at the ED at that time. In order to calculate the WIP, it is essential to determine the average amount of patients that were in the ED at 00:00. This was done by summing the amount of patients that arrived before and departed after 00:00 during the period, divided by the total amount of days.

16

Microsoft Excel is used to develop the diagrams and tables for analysing the data. By comparing the graphs of the control period to the graphs of the pilot period, it can be investigated whether there are changes in the daily patterns of inflow, outflow, LOS and WIP during the pilot period. First, the average daily patterns of in- and outflow of the control and pilot period will be analysed and compared. To investigate whether the effect of the physical presence of specialists at the ED differs among the specialties, separate analyses per specialty were carried out for surgery, cardiology, internal medicines and neurology. Unfortunately the data provided did not contain information on the time of a radiology request and the time when the results were available. For that reason it was not possible to investigate the patterns for radiology. To reduce the influence of the difference in the total amount of patients and the distribution of patients per specialty per period on the patterns of in- and outflow, similar days of both periods were analysed. This data set is constructed according to three criteria. Firstly, to form a pair of similar days, the day from the control period and the day from the pilot period should be the same day of the week. Secondly, the total amount of patients that visited the ED on the two days may differ at most two patients. Out of the pairs that satisfied the first two criteria, pairs that were also similar in terms of the distribution among the four different specialties were selected. Ultimately, this resulted in 28 pairs of days, for every day of the week 4 days. An overview of the selected days can be found in appendix C.

After analysing the dynamics of both periods as described above, the findings will be compared to the outcomes of the survey that show at what time the respondents found the specialists the most useful. Respondent could indicate whether they found the presence of specialists the most useful between 12:00-15:00, 15:00-18:00, 18:00-20:00 or they had no opinion. This comparison shows whether the outcomes of the graphs that indicate if the presence of specialists influences the patient flow matches with how ED staff and the specialists experience the usefulness of the presence of the specialists throughout the day. This might help explaining the results found using the diagrams. Besides, the survey gives more information on the usefulness of the radiologist, as it is not possible to analyse its influence on patient flow using the diagrams, as explained earlier.

3.4 Conclusion

17

4 Results

This paragraph presents and summarizes the results with respect to the influence of the presence of medical specialists at the ED on the crowding patterns at the ED. First, we focus on the comparison between the control and the pilot period. Then we present the results for the four separate specialties. Third, the results of the analysis of the pairs of similar days are presented. Lastly, these outcomes are compared to outcomes of the survey that show when the specialists were experienced as most useful.

4.1 General dataset analysis

To determine the main differences in the inflow of patients and their LOS between both periods, table 4.1 provides an overview of the number of patients that arrived between 12:00 and 19:30 and the corresponding LOS per specialty for both periods. Patients that arrived after 19:30 were excluded as they were expected not to benefit from the physical presence of the specialists that stay until 20:00. The table shows that in the pilot period 477 (+6%) more patients visited the ED, while the average LOS decreased with almost 10 minutes (-5%). The specialty surgery contributes the most to this effect. Surgery treated 12% more patients during the pilot period, while the LOS of these patients decreased with more than 13 minutes. The descriptive statistics in table A1 in the appendix shows that surgery admitted more than 4000 radiology request in both periods, while the other specialties admitted between 1400 and 2000 radiology requests per specialty. This indicates that surgery benefitted more from the presence of the extra radiologist during the pilot. Besides, it is notable that the LOS of the other specialties has in absolute terms the highest decrease, namely 14 minutes. This could be caused by the extra radiologist as well. It was found that the decrease in LOS does not hold for patients with the highest urgency level for surgery, cardiology and neurology. This is illustrated in table A3 in the appendix. On average patients with triage level U1 stayed more than 32 minutes longer at the ED. Since the amount of patients assigned to U1 is very small, the average LOS is sensitive to outliers. The next paragraphs shows how the figures below are established, by analysing the inflow and outflow pattern of patients throughout the day for both periods.

Control period Pilot period

Patients LOS Patients (% difference control period) LOS (% difference control period) Surgery 2308 02:38:32 2586 (+12%) 02:25:03 (-9%) Cardiology 1141 03:08:32 1168 (+2%) 03:00:16 (-4%) Internal medicines 1152 03:40:20 1244 (+8%) 03:43:29 (+1%) Neurology 986 03:30:43 1095 (+11%) 03:25:08 (-3%) Others 2076 03:02:08 2047 (-1%) 02:48:00 (-8%) Total 7663 03:06:09 8140 (+6%) 02:56:13 (-5%)

18

4.1.1 Dynamics of inflow and outflow of patients

In figure 4.1 the throughput diagram of the control and the pilot period are depicted. These graphs show the patterns of the average accumulated inflow and outflow of patients during the day. Since the five medical specialist were present from 12:00, it is expected that outflow of patients in the pilot period would be higher from 12:00 on. This should be illustrated by a steeper slope of the outflow curve during the pilot period from 12:00.

As can be seen from this graph, during the pilot period arrivals are rising more quickly from 13:00 onwards compared to the control period, which automatically raises the output curve of the pilot period. This makes it difficult to see whether the output curve is steeper due to the higher amount of inflow, or due to the fact that patients stay shorter at the ED. To investigate if the latter is the case, figure 4.2 shows the ‘’expected outflow’’, which is calculated by adding the corresponding LOS of the control period to the inflow during the pilot period. This graph illustrates that the outflow of patients during the pilot period does not show a change in the pattern from 12:00 onwards. The actual outflow is only slightly above the outflow curve calculated by using the throughput times of the control period. To analyse the hourly inflow and outflow patterns, the average rates at which patients arrive at the ED and the rate at which they are discharged for both periods are depicted in figure 4.3 and 4.4. These averages are calculated by taking the average amount of patients that arrived between half an hour before and half an hour after the full hour. Besides that, the ‘’desired outflow’’ is added, which is the outflow rate when the ED is able to keep up with the arrival rate. This is calculated by adding the median LOS (155 minutes) of patients during the control period to the time of arrival. For both periods this implies that in the desired situation each patient is discharged after 155 minutes.

19

Figure 4.3 Input/output diagram: Control period

Figure 4.5 Length of stay and work in progress: Control and pilot period

The figures 4.4 and 4.5 show that inflow and outflow rate roughly follow the same patterns for both periods. For both periods, the outflow peaks one hour after the inflow peaks which is indicated with an arrow. The peak in outflow is 7.3 patients during the control period and 7.9 patients during the pilot period. The outflow of patients is higher than the inflow of patients from 16:45 for both periods, which implicates that the amount of patients in the ED is decreasing from this point onwards. Both outflow curves show that in both periods the ability of the ED to keep up with the inflow rate is comparable, as the distance between the inflow and outflow curve does not substantially differ.

4.1.2 Length of stay and work in progress

From the throughput diagram shown in figure 4.1, the average LOS and the average amount of patients in the ED (WIP) can be derived, which are depicted in figure 4.5 below. Here it can be seen that during the pilot period the ED was able to treat more patients and simultaneously shorten their LOS. The WIP in the pilot period is around one patient higher during 12:00 and 18:00, while the LOS during the pilot period is below the LOS of the control period. But what is particularly striking is that this graph reveals that the LOS is shortened from 08:30 in the morning, while the specialists are only physically present between 12:00 and 20:00. This suggests that there are more factors during the pilot period that have shortened the LOS.

20

4.2 Dynamics per specialism

The descriptive statistics in table 4.1 showed varying outcomes for the change in LOS per specialism during the pilot period. This paragraph zooms into the dynamics of surgery, cardiology and internal medicines patients. As the dynamics of neurology patients showed similar results to surgery, these graphs as well as the throughput diagrams of the separate specialties are presented in appendix B.

4.2.1. Surgery

The descriptive statistics showed that the greatest progress in LOS was made by surgery. The patterns of the in- and outflow rate of surgery patients are illustrated in figure 4.6. The figure shows that during the pilot period the outflow curve lies closer to the inflow curve, which implies that the ED is more responsive to the inflow of patients. This is clearly illustrated by the fact that during the control period the outflow curve peaks three hours after the peak in the inflow, while in the pilot period the outflow curves peaks one hour after the peak in the inflow (indicated with the arrows). Besides, also for surgery patients the LOS is shortened from 08:00. This implicates that a factor beside the physical presence of the surgeon at the ED influenced the LOS of surgery patients. Nonetheless, the difference between the LOS of both periods increases relatively throughout the afternoon, while during that timeslot the WIP of the pilot period is higher as well.

4.2.2. Cardiology

The graphs for cardiology patients show remarkable patterns for the outflow of patients during the pilot period. In figure 4.8 it can be seen that the inflow of patients for both periods is similar and peaks at 12:00. Although the outflow of patients in the pilot period lags more between 12:00 and 14:00 compared to the control period, the outflow in the pilot period is better able to keep up with the peak in the inflow. The arrows in graph 4.8 indicate the peaks in the outflow for both periods. In the pilot period the outflow rate is able to reach the level of inflow, while the outflow rate of the control period is not able to reach this level. This results in a backlog of patients and consequently a longer LOS throughout the afternoon in the control period. This is visible in figure 4.9; until 16:00 the LOS of the Figure 4.6 Input/output diagram: Surgery

Figure 4.7 Input/output diagram: Surgery

21

pilot period is slightly higher than the LOS of the control period, but after this moment the LOS of the control period increases while the LOS of the pilot period decreases. Moreover, this graph reveals that the earlier found shortened LOS of 6 minutes during the pilot period is due to the shorter LOS between 16:00 and 20:00. The higher LOS in the beginning of the afternoon may be caused by the higher amount of patients that was in the ED between 13:00 and 17:00 during the pilot period.

4.2.3. Internal medicines

Internal medicines stands out as it is the only specialty that had an increase in the LOS between 12:00 and 19:30 during the pilot period. However, figure 4.11 reveals that the LOS is actually lower between 10:00 and 13:00, which was also found for the surgery and neurology patients. Figure 4.10 shows that from 14:00 in the pilot period, output in the responds slower to the inflow of patients. This lack of responsiveness affects the LOS in the pilot period which is depicted in figure 4.11. The LOS of patients in the pilot period is higher from 13:00 onwards. This suggests that the attendance of the extra internist has limited contribution compared to the other specialists. This might be caused by the fact that internal medicines submits the most lab request (see table A1 in the appendix). Waiting for these lab results covers a large part of the LOS of internal medicines patients. This is a process on which the medical specialists have no influence.

Figure 4.10 Input/output diagram: Internal medicines

Figure 4.7 Input/output diagram: Surgery

Figure 4.11 Length of stay and work in progress: Internal medicines

Figure 4.8 Input/output diagram: Cardiology

Figure 4.7 Input/output diagram: Surgery

22

4.3 Set of similar days

The previous paragraphs showed that the inflow of patients was higher during the pilot period, which is also illustrated in table 4.2. The average daily dynamics found for both periods might be influenced by this difference. Therefore, this paragraph analyses a set of days of the control and pilot period that are similar regarding the total number of patients arrived per day and the distribution of patients among the four specialties.

Table 4.3 shows that the average daily inflow of the selected days for both periods are equal, except for neurology which had on average one patient more per day during the control period. An overview of the set of days selected and the descriptive statistics can be found in appendix C and D.

Total Surgery Cardiology Internal medicines Neurology

Control period 113 33 17 16 14

Pilot period 113 33 17 16 13

4.3.1. General analysis set of similar days

First the main differences in the inflow of patients and their LOS between the control and pilot periods for this set of days are determined. Table 4.4 provides an overview of the number of patients that arrived between 12:00 and 19:30 and the corresponding LOS for both periods. The table shows that within the data set of similar days, the LOS of cardiology and internal medicines are different compared to those within the entire data set. Internal medicines had overall a larger average LOS in the pilot period. Within the specific data set, cardiology has a larger LOS and internal medicines a shorter LOS during the pilot period. For surgery, neurology and the other specialties the effect on the LOS found for the pilot period in the entire data set is enhanced for this specific set of similar days.

Total Surgery Cardiology Internal medicines Neurology

Control period 112 32 16 16 14

Pilot period 116 35 17 17 14

Control period Pilot period

Patients LOS Patients (% difference control period) LOS (% difference control period) Surgery 458 02:44:29 443 (-3%) 02:22:18 (-13%) Cardiology 222 03:12:45 234 (+5%) 03:14:34 (+1%) Internal medicines 223 03:46:33 233 (+4%) 03:34:35 (-5%) Neurology 179 03:41:14 179 (+0%) 03:20:11 (-10%) Others 361 03:03:41 382 (+6%) 02:37:33 (-14%) Total 1443 03:10:16 1471 (+2%) 02:58:31 (-6%)

Table 4.4 Descriptive statistics set similar days 12:00-19:30: Number of patients and average length of stay Table 4.3 Average amount of patient arrivals per day: Set of similar days

23

Figure 4.14 Length of stay and work in progress similar set of days

Figure 4.7 Input/output diagram: Surgery

4.3.2 Dynamics of inflow and outflow of patients

Figures 4.12 and 4.13 illustrate the inflow, outflow and desired outflow curves of the control and pilot period for the set of similar days. Despite the fact that the dataset includes similar days, the inflow patterns still differ substantially.

Nonetheless, the graphs illustrate that the ED is more responsive to the arrival of patients in the pilot period. This is shown by two phenomena. First, the outflow curve tightly follows the pattern of the inflow curve between 12:00 and 20:00 in the pilot period, whereas the outflow curve of the control period starts lagging from 12:00. Second, during the control period the outflow was not able to reach the level of inflow, while during the pilot period the level of outflow exceeds the level of inflow (both indicated with an arrow in graphs 4.12 and 4.13).

4.3.3. Length of stay and work in progress

The higher responsiveness during the pilot period has its effect on the LOS, which is depicted in figure 4.14. Here it can be seen that the LOS gradually declines between 12:00 and 20:00 and the difference between the LOS of both periods increases. However, again this graph shows that the LOS during the pilot period is already lower from 07:00, while the specialist are only scheduled between 12:00 and 20:00 which suggests that other factors lower the LOS during the pilot period.

Figure 4.12 Input/output diagram similar set of days: control period

Figure 4.7 Input/output diagram: Surgery

Figure 4.13 Input/output diagram similar set of days: pilot period

24

4.3.4. Analysis per specialism

This paragraph analyses the difference in the patterns of in- and outflow of patients between both periods within this specific dataset. It should be noted that discussing the separate specialisms must be done carefully, since the curves of in- and outflow show jagged patterns. When the dataset included more days, the curves would have been more smooth.

The in- and output diagrams show the same phenomenon for all four specialisms. In the control period, the four specialisms were not able to keep up with the inflow rates as outflow rates did not get to the level of the highest inflow rate. While during the pilot period, all four specialties showed that the outflow rates got at least to the level of the highest inflow rate, and in the case of neurology and cardiology the outflow exceeded the level of the highest inflow rate. This is illustrated in the input/output diagrams for internal medicines in figure 4.15 and 4.16. During the control period, there is a peak in the inflow at 16:00. Unless the fact that outflow follows the desired outflow curve closely, the ED is not able reach the desired output peak at 19:15. Figure 4.16 shows that the ED was able to keep up with these peaks in the inflow in the pilot period. The peak in the outflow at 20:00, indicated with the arrow, exceeds the peak in the inflow. The in- and output diagrams for surgery, cardiology and neurology show similar results, these graphs are presented in appendix E.

Regarding the LOS for the four separate specialties, except for cardiology, the LOS in the pilot period is already lower from at least 10:00 in the morning. This was found in the analysis for the entire data set as well. The graphs illustrating the patterns of the LOS and WIP throughout the day for the separate specialties can be found in appendix E.

Figure 4.15 Input/output diagram similar set of days control period: Internal medicines

Figure 4.7 Input/output diagram: Surgery

Figure 4.16 Input/output diagram similar set of days pilot period: Internal medicines

25

4.4 Comparison to how the pilot is experienced

In the case hospital a survey was conducted with the ED staff to investigate how the pilot was experienced. This survey contained questions about at which time slots the presence of the surgeon, cardiologist, neurologist, internist and radiologist were experienced as most useful. This is especially helpful to get insight in the influence of radiology, as it was not possible to analyse the influence of radiology on the dynamics in the pilot period in the previous analysis. The outcomes of these survey questions will be presented and compared to the findings of the patterns showing the influence of specialists throughout the day during the pilot period.

4.4.1. Outcomes of the survey

The results indicate that all five medical specialists were found the most useful between 15:00 and 18:00, which is depicted in figure 4.17. It can be seen that radiology stands out: for all three timeslots, the highest percentage of respondents indicated that the radiologist is the most useful. It should be noted that the questions only indicate at which timeslots the specialists were experienced as most useful and not whether the specialists were experienced useful at all. However, almost twice as much respondents found the presence of radiology specialist useful in all timeslots compared to the other four specialties. This outcome suggest that the presence of the radiologist was over all experienced as most useful. The outcomes for surgery, cardiology, neurology and internal medicines show comparable results. Appendix G gives an overview of the type of respondents per question. Here it can be seen that overall, EPs found the presence of the medical specialists the most useful while specialists found their presence the least useful.

4.4.2. Comparison of the outcomes of the survey to the outcomes of the diagrams

According to the analysis done in paragraph 4.1 where it was found that surgery benefitted the most from the pilot regarding the LOS of their patients, it was expected that the presence of the surgeon was experienced as most useful. However, the results of the survey show that the presence of the surgeon, cardiologist, internist and neurologist was experienced equally useful. This supports the Figure 4.17 Timeslots at which the specialists are experienced as most useful

26

earlier thought that the effect found for surgery is due to the extra radiologist, as surgery submits almost four times as much radiology requests compared to the other three specialisms.

27

5 Discussion

In this paper, the patterns of average inflow, outflow, LOS and WIP throughout the day were analysed for a pilot period in which medical specialists were physically present at the ED between 12:00 and 20:00. It was investigated how the patterns of crowding at the ED in the pilot period differed from the period in which there were no medical specialists present on-site at the ED.

The first analysis showed that the pattern of the outflow of patients throughout the day was not substantially different when the five specialists were present at the ED. The curves of in- and outflow of patients set against the time of the day were found to be relatively similar. Although on average the amount of patients in the ED was uniformly higher between 12:00 and 20:00 during the pilot period, these patients had a shorter LOS. This implicates that the ED was able to treat more patients in shorter time during the pilot period.

More strikingly is the fact that the patterns of the LOS showed that the average LOS in the pilot period was already shortened from 08:30 onwards. Only a small part of the patients that arrive at the ED at 08:30 could have benefitted from the specialist that were present between 12:00 and 20:00. This implicates that other factors have influenced the LOS during the pilot period. A possible explanation could be that during the control period, the ED of the case hospital was renovated. The amount of rooms at the ED was the same for both periods, but the lay-out of the rooms was revised. The ED staff might had to get used to this new lay-out, which could have lengthened the LOS in the control period. In addition, during the pilot period there was a CT-scan at the ED while this was not the case during the control period. This might have shortened the LOS of patients during the pilot period. Therefore, differences found between the control and pilot period cannot be solely dedicated to the presence of the medical specialists.

Also noteworthy is the fact that the patients with the highest urgency level did not experience shorter LOS during the pilot, despite specialists being able to see the patients immediately. Especially cardiology patients with the highest urgency level experienced an increase in LOS of 58% between 12:00 and 20:00 compared to the control period. This might suggest that specialists slow down the process for patients with high urgency. However, as the group of patients with the highest emergency level is small, results are sensitive to outliers.

28

for all three timeslots. Radiology stood out, as the amount of respondents indicating the radiologist to be most useful was more than 30% higher for every timeslot compared to the other specialties. This suggests that the respondents experienced the presence of the radiologist as most useful compared to the other four specialties. It was found that the specialty surgery submitted 50% more radiology requests compared to cardiology, neurology and internal medicines. This implicates that surgery benefits the most from the presence of the extra radiologist. Therefore, it is likely that the outstanding effect found for surgery was mostly caused by the extra radiology specialists instead of the extra surgeon. The extra radiologist affects the outcomes for the other three specialties as well, but to a lesser extent.

An analysis of a set of days of the control and pilot period of which the amount of visitors per day and the distribution of patients among the specialties was similar showed a consistent change in the patterns of outflow throughout the day for all specialties in the pilot period. Surgery, cardiology, internal medicines and neurology were better able to keep up with the inflow of patients. The patterns of inflow and outflow throughout the day showed that in the pilot period, the peaks in the outflow followed shortly after the peaks in the inflow. In addition, in the pilot period the peak levels of outflow were at least the same height as the peak levels of inflow. While in the control period, the peaks in the outflow followed with a delay and after the peaks in the inflow. Furthermore, the peak level of the outflow did not reach the peak level of inflow, causing output to lag throughout the day. This finding is likely to be caused by the extra capacity that the medical specialists provided during the pilot period. As other studies found that staffing one extra physician at the ED shortened the LOS of discharged patients by 35 minutes (Bucheli & Martina, 2004), it is questionable whether the same or even better results could not be achieved by staffing for instance additional nurses or emergency physicians at the ED.

29

6 Conclusion

This study aimed at investigating the influence of the physical presence of medical specialists at the ED on the patterns of crowding at the ED. Theory states that waiting for medical specialists to perform their tasks at the ED is frequently found to hinder patient flow (Derlet & Richards, 2000; Have, 2016; Hoeben, 2017; Magid et al., 2009). On the other hand, the supervision of residents by medical specialists enhances patient flow as the patients are diagnosed in an early stage Hoeben (2017). This study investigated the influence of the physical presence of medical specialists at the ED on the patterns of the inflow and outflow of patients throughout the day. Hence, the research question that was to be answered in this paper is: How does the physical attendance of medical specialists influence

the crowding patterns at the ED? Prior research on the staffing of medical specialists at the ED only

showed the effect on the median LOS (van der Linden, 2017). This research contributed to literature by thoroughly analysing the patterns of average inflow and outflow of patients throughout the day. The in-depth analysis reveals how the physical presence of specialists influenced the responsiveness of the ED throughout the day. It provides a more complete picture of the effect of staffing medical specialists at the ED as a measure contributing to resolving crowding at the ED.

The research was conducted using a single case study at an ED in the Netherlands. For the period of 1-11-2017 until 31-05-2018 a pilot was carried out in which a surgeon, cardiologist, internist, neurologist and radiologist were physically present at the ED between 12:00 and 20:00, in addition to regular staffing levels. Patterns of inflow and outflow of patients of this period were compared to the period one year before the pilot, which is the control period (1-11-2016/31-05-2017).

30

In order to get more insight into the effect of the physical presence of the five specialties on the patterns of inflow and outflow of patients throughout the day, further research on the influence of the radiologist on the flow of patients at the ED is recommended. This enables to investigate which specialism is the most useful at the ED in terms of patient flow. It is recommended to further analyse the influence of the presence of the medical specialists on the different stages between inflow and outflow of patients at the ED. This would provide a better overview of where the medical specialist influence the patient flow and what possible other stages influence the responsiveness of the ED. Lastly, it is recommended to investigate whether other types of additional staffing could not achieve the same or even better results.

Factor of influence Explanation Shortened LOS

from 10 a.m.

The projections of LOS over time revealed that the LOS of patients was already shorter from 10:00 on, while the specialists were only present from 12:00. This implicates that other factors have influenced the LOS of patients during the pilot.

Radiologist The radiologist might be needed in the process of every patient, which causes that the

effect of the extra radiologist is included in the results found for surgery, cardiology, neurology and internal medicines. A survey on the usefulness of the presence of the medical specialists between 12:00 and 20:00 suggested that the radiologist was experienced as most useful. Therefore, the presence of the radiologist might have positively influenced the outcomes for the separate specialties of this research.

Extra capacity This research did not distinguish whether the patterns of inflow and outflow of patients

in the pilot period were influenced by the tasks (responsibility, supervision, expertise) medical specialist have at the ED, or through the extra capacity they provided.

31

Bibliography

Asplin, B. R., Magid, D. J., Rhodes, K. V., Solberg, L. I., Lurie, N., & Camargo, C. A. (2003). A conceptual model of emergency department crowding. Annals of Emergency Medicine, 42(2), 173–180. Bernstein, S. L., Aronsky, D., Duseja, R., Epstein, S., Handel, D., Hwang, U., … Asplin, B. R. (2009). The

effect of emergency department crowding on clinically oriented outcomes. Academic

Emergency Medicine, 16(1), 1–10.

Boyle, A., Beniuk, K., Higginson, I., & Atkinson, P. (2012). Emergency Department Crowding: Time for Interventions and Policy Evaluations. Emergency Medicine International, 2012, 1–8.

Brick, C., Lowes, J., Lovstrom, L., Kokotilo, A., Villa-Roel, C., Lee, P., … Rowe, B. H. (2014). The impact of consultation on length of stay in tertiary care emergency departments. Emergency Medicine

Journal, 31(2), 134–138.

Bucheli, B., Martina B. (2004). Reduced length of stay in medical emergency department patients: a prospective controlled study on emergency physician staffing. European Journal of Emergency

Medicines, 11, 29-34.

Carter, E. J., Pouch, S. M., & Larson, E. L. (2014). The relationship between emergency department crowding and patient outcomes: A systematic review. Journal of Nursing Scholarship, 46(2), 106–115.

Derlet, R. W., & Richards, J. R. (2000). Overcrowding in the nation’s emergency departments: Complex causes and disturbing effects. Annals of Emergency Medicine, 35(1), 63–68.

Fernandes, C. M. B., Tanabe, P., Gilboy, N., Johnson, L. A., McNair, R. S., Rosenau, A. M., … Suter, R. E. (2005). Five-level triage: A report from the ACEP/ENA five-level triage task force. Journal of

Emergency Nursing, 31(1), 39–50.

Have, M. ten. (2016). Explaining lagged output in the ED, a case study on patient waiting . Hoeben, L. (2017). The impact of the supervision structure on Emergency Department flow . Hoot, N. R., & Aronsky, D. (2008). Systematic Review of Emergency Department Crowding: Causes,

Effects, and Solutions. Annals of Emergency Medicine, 52(2).

32

Kingsman, B., & Hendry, L. (2002). The relative contributions of input and output controls on the performance of a workload control system in make-to-order companies. Production Planning

and Control, 13(7), 579–590.

Koomen, A. (2015). The influence of emergency department responsiveness on throughput times. Lee, R. S., Woods, R., Bullard, M., Holroyd, B. R., & Rowe, B. H. (2008). Consultations in the

emergency department: A systematic review of the literature. Emergency Medicine Journal,

25(1), 4–9.

Luscombe, R., & Kozan, E. (2016). Dynamic resource allocation to improve emergency department efficiency in real time. European Journal of Operational Research, 255(2), 593–603.

Mackway-Jones, K. (1997) Emergency triage. London: BMJ Publishing Group

Magid, D. J., Sullivan, A. F., Cleary, P. D., Rao, S. R., Gordon, J. A., Kaushal, R., … Blumenthal, D. (2009). The Safety of Emergency Care Systems: Results of a Survey of Clinicians in 65 US Emergency Departments. Annals of Emergency Medicine, 53(6), 715–723.e1.

Moskop, J. C., Sklar, D. P., Geiderman, J. M., Schears, R. M., & Bookman, K. J. (2009). Emergency Department Crowding, Part 1-Concept, Causes, and Moral Consequences. Annals of Emergency

Medicine, 53(5), 605–611.

Mulder, C. F. (2017). The more patients in the morning, the more patience in the afternoon: An

Emergency Department Case Study.

Pines, J. M. (2007). Moving Closer to an Operational Definition for ED Crowding. Academic

Emergency Medicine, 14(4), 381–382.

Pines, J. M., & Griffey, R. T. (2015). What We Have Learned from a Decade of ED Crowding Research.

Academic Emergency Medicine, 22(8), 985–987.

Reiring, S. (2018). Case study : on the relational integration between emergency physicians and

medical specialists of other departments and the influence on the patient flow at the ED .

Soepenberg, G. D., Land, M., & 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), 495–503.

Soepenberg, G. D., Land, M. J., & Gaalman, G. J. C. (2012). A framework for diagnosing the delivery reliability performance of make-to-order companies. International Journal of Production

33

Ter Avest, E., Onnes, B. T., Van der Vaart, T., & Land, M. J. (2018). Hurry up, it’s quiet in the emergency department. The Netherlands Journal of Medicine, 62(3), 32–35.

Van Achteren, G. (2014). Care for waiting: the effect of emergency department responsiveness on

patient waiting times.

van der Linden. (2013). Emergency department crowding in The Netherlands: managers’ experciences, 2–9.

van der Linden, C. C., de Beaufort, R. A. Y., Meylaerts, S. A. G., van den Brand, C. L., & van der Linden, N. (2017). The impact of medical specialist staffing on emergency department patient flow and satisfaction. European Journal of Emergency Medicine, 1–6.

van Manen, V. (2015). Emergency Care Waiting Times; The Impact of Dynamics and Room Utilization. Verbree, J. M. (2013). Diagnosing and improving the patient throughput time in emergency

departments.

Voss, C., Tsikriktsis, N., & Frohlich, M. (2002). Case research in operations management. International

Journal of Operations & Production Management, 22(2), 195–219.

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), 6–9.

Yoon, P., Steiner, I., & Reinhardt, G. (2003). ED A DMINISTRATION • L ’ ADMINISTRATION DE LA MU Analysis of factors influencing length of stay in the emergency department. Journal of

34

Appendix

A. Descriptive statistics total dataset

Control period Pilot period % Difference

Total number of patients 16938 17480 +3%

Average age 45.42 46.09 +1%

Patients per specialism

Surgery 4831 5232 +8%

Cardiology 2467 2499 +1%

Internal medicines 2433 2591 +6%

Neurology 2064 2083 +1%

Patients per triage level

U1: Immediate 254 248 -2%

U2: Very urgent 4277 4149 -3%

U3: Urgent 7483 8257 +10% U4: Standard 4565 4149 -9% U5: Non-urgent 95 88 -7% No triage 264 589 +123% Radiology requests Total 11284 11187 -1% Surgery 4291 4688 +9% Cardiology 1441 1480 +3% Internal medicines 1447 1645 +14% Neurology 1903 1638 -14% Others 2202 1736 -21% Lab requests Total 14164 14565 +3% Surgery 2366 2472 +4% Cardiology 2815 2833 +1% Internal medicines 3468 3664 +6% Neurology 2197 2248 +2% Others 3318 3348 +1%

35

Figure A1 Descriptive statistics total period: Distribution of number of patients arriving per day

Figure A2 Descriptive statistics control period: Patient arrivals per weekday

36

Table A2 Descriptive statistics: Control period 12:00-19:30

Surgery Internal medicines Cardiology Neurology Total

Patients LOS Patients LOS Patients LOS Patients LOS Patients LOS

U1: Immediate 9 02:32:13 22 02:44:25 25 02:45:24 16 02:19:49 89 02:36:18 U2: Very urgent 179 03:22:37 249 03:55:55 629 03:00:59 312 03:09:09 1936 03:14:22 U3: Urgent 1043 02:52:04 606 03:45:37 415 03:20:30 517 03:46:55 3519 03:19:57 U4: Standard 1009 02:19:00 251 03:40:17 66 03:15:55 133 03:30:58 1945 02:39:22 U5: Non-urgent 14 00:47:51 7 01:22:34 1 02:45:00 0 37 01:06:52 No triage 54 02:25:24 17 04:33:04 5 02:46:24 8 02:22:15 137 02:27:22 Total 2308 02:38:32 1152 03:45:21 1141 03:08:32 986 03:30:43 7663 03:06:09

Table A3 Descriptive statistics: Pilot period 12:00-19:30

(% difference control period)

Surgery Internal medicines Cardiology Neurology Total

Patients LOS Patients LOS Patients LOS Patients LOS Patients LOS

37

B. Dynamics per specialty

Figure B5 Input/output diagram: Neurology

Figure 4.7 Input/output diagram: Surgery

Figure B6 Length of stay and work in progress: Neurology Figure B1 Throughput diagram: Surgery Figure B2 Throughput diagram: Cardiology

38

C. Selection set of similar days

Date* Day Number of

patients

Surgery Cardiology Internal medicines Neurology 27-2-2017 Monday 128 40 22 12 15 29-1-2018 Monday 127 42 23 14 15 9-1-2017 Monday 121 33 21 16 18 8-1-2018 Monday 119 31 23 18 17 12-12-2016 Monday 115 34 17 15 14 27-11-2017 Monday 115 33 19 21 17 6-3-2017 Monday 96 30 18 11 11 11-12-2017 Monday 98 25 20 13 14 7-2-2017 Tuesday 107 32 17 11 16 16-1-2018 Tuesday 108 29 23 14 15 14-2-2017 Tuesday 116 26 20 20 12 27-2-2018 Tuesday 116 31 19 17 12 1-11-2016 Tuesday 117 35 14 21 12 20-2-2018 Tuesday 119 35 17 21 15 14-3-2017 Tuesday 99 31 14 15 9 5-12-2017 Tuesday 101 30 18 13 9 14-12-2016 Wednesday 136 44 19 13 17 27-12-2017 Wednesday 136 38 21 17 17 1-2-2017 Wednesday 119 34 23 17 17 15-11-2017 Wednesday 119 37 19 14 17 30-11-2016 Wednesday 108 27 14 20 12 31-1-2018 Wednesday 109 29 14 16 10 28-12-2016 Wednesday 98 20 16 12 15 14-3-2018 Wednesday 98 23 13 15 13 5-1-2017 Thursday 123 30 17 21 19 21-12-2017 Thursday 123 34 22 23 14 9-3-2017 Thursday 117 31 20 20 13 14-12-2017 Thursday 116 34 21 22 12 3-11-2016 Thursday 114 37 18 16 16 18-1-2018 Thursday 114 30 16 16 15 12-1-2017 Thursday 94 24 16 12 15 11-1-2018 Thursday 90 24 17 9 15 3-3-2017 Friday 108 33 18 15 12 9-2-2018 Friday 108 30 12 18 14 23-12-2016 Friday 117 42 15 19 14 23-2-2018 Friday 117 45 18 15 13

* Blue = Control period White = Pilot period

39

Date* Day Number of

patients

Surgery Cardiology Internal medicines Neurology 10-2-2017 Friday 124 36 18 19 9 8-12-2017 Friday 124 36 20 17 13 9-12-2016 Friday 139 42 18 15 16 16-2-2018 Friday 137 42 21 19 16 4-3-2017 Saturday 98 22 11 11 12 17-3-2018 Saturday 98 24 12 16 10 11-2-2017 Saturday 105 33 12 11 14 2-12-2017 Saturday 107 34 11 14 14 21-1-2017 Saturday 111 33 17 15 12 16-12-2017 Saturday 111 35 13 18 12 10-12-2016 Saturday 124 37 12 19 13 30-12-2017 Saturday 124 39 16 14 8 19-3-2017 Sunday 94 33 13 16 7 3-12-2017 Sunday 92 31 14 12 10 15-1-2017 Sunday 106 29 17 18 17 11-2-2018 Sunday 107 30 10 19 12 27-11-2016 Sunday 113 37 12 19 12 14-1-2018 Sunday 112 35 13 16 12 26-3-2017 Sunday 118 41 15 16 12 5-11-2017 Sunday 117 42 12 16 9 Daily average control period - 113 33 17 16 14 Daily average pilot period - 113 33 17 16 13

* Blue = Control period White = Pilot period