Emergency department flow performance: a waiting time analysis.

(1)

Emergency department flow performance:

a waiting time analysis.

20-02-2018

University of Groningen

Faculty Economics and Business

MSc. Thesis Supply Chain Management

J. C. Kuiken

S2372401

(2)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 2 Abstract

An emergency department (ED) of a hospital is influenced by many factors and mechanisms that affect the flow performance of patients. Researchers have identified multiple causes, factors and solutions related to crowding, which occurs when the need for emergency services temporarily exceeds the capacity available at the ED. This causes a reduced flow of patients, which results in longer waiting times. Although waiting times are perceived as challenging in the healthcare sector, there still exists a lack of knowledge in current literature regarding the composition of waiting times at distinctive stages of an ED. A work sampling study is conducted in order to gain more insight into the composition of waiting times at distinctive stages of an ED. With this technique, which is new in this context, waiting times are sampled instead of work. This paper found several recurring patterns inflow of patients and found also factors and mechanisms (e.g. time of the day, crowdedness, patient characteristics) that influence the composition of waiting times. Patient characteristics are identified based on specialisms appointed, types of additional research required, and the need for a consult of a specialist and appear to be an important factor that influences the flow performance of patients. The insights gained with this work sampling study can be used to reduce waiting times and improve flow at the ED.

Keywords: emergency department, flow performance, patient length of stay, waiting time, work sampling

(3)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 3 Table of contents Abstract ... 2 Table of contents ... 3 1. Introduction ... 5 2. Theoretical background ... 7

2.1 Emergency department characteristics ... 7

2.2 Flow performance at the ED ... 8

2.3 Waiting times at the ED ... 9

2.3.1 Explanatory perspective ... 11

2.3.2 Solution-oriented perspective ... 11

2.4 Research designs used in literature ... 12

2.5 Summary of literature findings ... 14

3. Methodology ... 15

3.1 Case description ... 15

3.2 Case study design ... 15

3.3 Research phases... 16

3.4 Sources and analysis... 18

4. Results ... 21

4.1 General analysis ... 21

4.2 Flow performance analysis... 22

4.2.1 Patient arrivals ... 22

4.2.2 Inflow and outflow of patients... 24

4.3 Waiting time analysis ... 27

4.3.1 Overall analysis ... 27

4.3.2 Waiting times split at 15:00 ... 29

4.3.3 Waiting times at busy and calm moments ... 30

4.4 Waiting time analysis for homogenous sets of patients ... 31

4.4.1 Patient specialisms ... 32

4.4.2 Patients that require different types of additional research ... 33

4.4.3 Patients with and without consult specialist ... 35

4.5 Waiting time components throughout the day ... 36

(4)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 4

6. Limitations of this research ... 42

6.1 Limitations of this research ... 42

6.2 Methodical limitations... 42

7. Conclusion ... 44

7.1 Recommendations for practice ... 45

7.2 Further research ... 46

8. References ... 47

Appendix A – Work sampling form ... 51

Appendix B – ED characteristics ... 52

(5)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 5 1. Introduction

Patient waiting time has been recognized by hospital managers as a challenge in the healthcare sector (Harper & Gamlin, 2003; Cayirli & Veral, 2003). Especially waiting time at an emergency department (ED) is challenging as a result of its complex characteristics (Hoot & Aronsky, 2008) and can cause patient dissatisfaction (Taylor & Benger, 2004), higher costs for treatment and even higher mortality rates (Plunkett et al. 2011). By analyzing proportions of waiting time at distinctive stages of an ED, this paper aims to gain more insight in patient waiting times in each phase of the patient pathway.

(6)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 6 ED. This means that there is a lack of knowledge regarding the time a patient spends in each phase of the patient pathway.

Various solutions for decreasing waiting time are mentioned in literature (Hoot & Aronsky, 2008; Paul et al. 2010). However, in order to apply the right solution to a specific situation, it is essential to have clear insights into what exactly patients are waiting for at each distinctive stage. Hoot & Aronsky (2008) mention for example adding extra personnel as a possible solution, which is needless in case of long waiting times due to a coordination problem of medical personnel. In order to gain more insight into waiting times, waiting times will be sampled using a ‘work sampling method’. This method has proven to be effective in sampling fractions of work time (Robinson, 2010), therefore it might seem appropriate to sample fractions of waiting time as well. This method will be used to quantify proportions of waiting times at distinctive stages at the ED. The research question of this research is therefore:

How do diverse factors and mechanisms translate into the composition of waiting times at distinctive stages of an emergency department?

By sampling waiting times the proportions of time a patient spends in each phase of the patient pathway will be measured. Next to that, with the identification of factors and mechanisms that cause waiting times at the ED, waiting times can be decreased and negative effects of crowding are reduced.

(7)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 7 2. Theoretical background

This chapter summarizes the main findings retrieved from literature regarding patient waiting times at the ED. First, specific characteristics of the ED are outlined. Subsequently, the flow of patients at an ED is addressed. After that literature on ED crowding and waiting times are discussed.

2.1 Emergency department characteristics

The ED is the first contact point for patients who need acute care. Acute care is by definition unscheduled (Asplin et al. 2003) and in combination with a broad range of different injuries and illnesses difficult to manage. In general, an ED consists of a waiting room, a triage desk, a trauma room, a plaster room and a certain number of treatment rooms. There are several ways a patient can arrive at the ED. This is by medical transportation (ambulance), self-referral or by general practitioner (GP) referral. After arrival, a patient waits until triage. According to van der Vaart et al. (2011) triage is “a brief clinical assessment that determines the time and sequence in which patients should be seen in the ED”. A triage nurse performs this step in order to determine the urgency and sequence in which patients should be treated. The measurement tool that is often used for triage is the Manchester Triage System (MTS) (Anneveld et al. 2013). The MTS scale has five levels and ranges from level one, i.e. immediate care to level five, i.e. related to non-urgent patients that can wait maximum four hours before being seen by a physician. Next, if capacity is available, the diagnosis based on physician examination is executed. Extra lab tests and medical imaging can be necessary for the diagnosis process. After the diagnosis, the patient gets treated and finally discharged to either home or another department of the hospital.

(8)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 8 occur. It is important to state that this is a static picture. In reality, phases overlap regularly. There are waiting times that are partly inevitable, such as waiting for results of lab tests or medical imaging. However, waiting times related to inadequate staffing or lack of capacity for example are partly preventable in some situations (Sinreich et al. 2012; McCarthy et al., 2014).

Figure 1.1 Emergency department processes and distinctive stages involving waiting

2.2 Flow performance at the ED

Analyzing the patient flow at an ED is necessary and important in order to improve efficiency, reduce crowding and minimize the average length of stay (Wang & Howard, 2013). Litvak (2009) confirms that improving the patient flow is crucial for increasing patient satisfaction and hospital productivity. Three metrics in order to measure flow are inflow, throughput and outflow. Inflow measures the number of patients entering the ED per time unit, also known as the arrival rate. The outflow measures the number of patients leaving the ED per time unit. The throughput time measures the speed a patient moves along the pathway. This represents the time between the moment a patient enters the ED and the moment a patient is discharged from the ED (Qureshi et al. 2011) and is the total time a patient spends at the ED, also known as the length of stay (LOS). Patient flow is defined by Drupsteen et al. (2013) as “the speed at which patients are transferred from one step in the care process to the next“. In this context, flow at an ED is defined as the speed patient moves from a distinctive stage (Figure 1.1) to another. This definition implies that an increase in the speed of the transfer of patients to a subsequent care process leads to a better flow performance. Waiting times decrease the speed of the transfer between stages along the patient pathway. For this reason, it is important to identify the waiting time components between different stages as hindrance of flow. Waiting times are therefore a key performance measure for flow

(9)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 9 since this significantly reduces the speed by which patients move along the patient pathway.

2.3 Waiting times at the ED

Waiting times occur due to various reasons, such as variable demand, limited capacity and inefficient operations management (Meng et al., 2015). Figure 1.1 shows the stages at which in between waiting times could occur. These lines between stages represent possible moments of waiting, but do not tell anything about the causes. This section reviews existing literature on causes of waiting times at the ED.

Because of an ED’s multifaceted and complex characteristics, numerous factors related to waiting times can be distinguished. According to Miro et al. (2003) factors related to waiting times can be divided into four different categories:

- Factors related to the ED itself (e.g. waiting for a doctor or lab results)

- Factors related to ED-hospital interrelations (e.g. waiting for processes performed outside the ED)

- Factors related to the hospital itself (e.g. waiting for a hospital bed)

- Factors related neither to the ED or hospital (e.g. waiting for patient relatives)

All mentioned factors are related to resources (e.g. waiting times due to a lack of capacity). Note that waiting times could occur due to a lack of synchronization as well. Asplin et al. (2003) developed a conceptual model that applies operations management principles to patient flow in order to help understand crowding at an ED and distinguishes input, throughput and output of an acute care system. Input factors relate to any events, conditions, or systems characteristics that contribute to the demand for care. The throughput component identified all the activities that are performed within the border of an ED. Output factors relate to any events, conditions or system characteristics that influence the ability of an ED to dispose patients.

(10)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 10 times on causes and effects, do solution-oriented studies focus on possible solutions in order to decrease waiting times. Commonly studied factors from both perspectives can be found in Table 2.1 and are discussed in the following two sections. The distinction is made between input, throughput and output factors, which is in line with the conceptual model of Asplin et al. (2003).

Table 2.1. Factors related to waiting times at the ED

Explanatory perspective Solution-oriented perspective

Causes Solutions

Input Unscheduled care (Asplin et al. 2003) Increased responsiveness

(Koomen, 2015; van Manen 2016) Demand management (Hoot & Aronsky, 2008)

Throughput Inadequate staffing (Hoot & Aronsky, 2008)

Insufficient physician capacity during peak hours (Paul & Lin, 2012) Confusing medical staff role definitions and inappropriate ED layout structure (Regan, 2000)

Adding extra personnel or observation units (Hoot & Aronsky, 2008)

Scheduling work shifts (Sinreich et al. 2012)

Improvement of internal factors, such as the layout of the work environment (Miro et al. 2003) Output Hospital bed shortages (Hoot &

Aronsky, 2008; Derlet & Richards, 2000)

The process of admitting patients to the hospital (Paul & Lin, 2012)

ED expansion and improved hospital bed access (Miro et al. 2003; Dunn, 2003; Derlet & Richards, 2000)

(11)

2.3.1 Explanatory perspective

According to Hoot & Aronsky (2008), main causes for long waiting times are inadequate staffing, inpatient boarding and hospital bed shortages. Regan (2000) confirms that inappropriate staffing leads to crowding, and adds confusing medical staff role definitions, and inappropriate ED layout structures as possible causes. In addition, Paul & Lin (2012) state that insufficient physician capacity during peak hours and the process of admitting a patient to the hospital increase waiting times at the ED. When looking at external factors, Lambe et al. (2003) conducted research on average waiting times in California’s Emergency departments. They show that waiting times were longer at hospitals in poorer neighborhoods. With an ordinary least squares regression analysis they show that per $10,000 decrease in per capita income, the average waiting time increased with 10,1 minutes.

Next to identifying causes, several simulation studies aimed to analyze the patient flow and waiting times at the ED (Sinreich & Marmor 2005; Meng et al. 2015). Sinreich & Marmor (2005) developed a detailed model as simulation tool for modeling EDs in Israel describing the operational structure of an ED. They conclude that over 50% of total patient turnover time consists of waiting time. Major components for waiting based on the simulation results are: waiting for the first physician’s examination, waiting times for blood work and time away for an x-ray examination. Also, McCarthy et al. (2014) analyzed the patient flow through the ED and conclude that approximately 75% of the time a patient spends at the ED, it is not interacting with care providers. These researchers analyzed average waiting times, predicted waiting times and mentioned possible types of waiting. However, they did not provide us with any insight on how waiting times are distributed along the patient pathway.

2.3.2 Solution-oriented perspective

(12)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 12 However, literature regarding solutions seems rather contradictive. Miro et al. (2003) state for example that ED expansion leads to lower waiting times, while other studies do not agree with the effectiveness of ED expansion (Hoot & Aronsky, 2008; Han et al. 2007). Also adding extra capacity or personnel does not necessarily result in lower waiting times (Mumma et al. 2014). This illustrates the shortcomings of the mentioned solutions and shows the need for an analysis that takes into account the time specific events that cause crowding.

2.4 Research designs used in literature

Based on articles discussed from both the explanatory and solution-oriented perspective, literature shows that causes, effects and solutions in order to decrease waiting times are studied extensively. However, except for the studies of Gelissen (2016), Buitelaar (2016) and Ten Have (2016), no research method has been able yet to gain insight into factors and mechanisms that translate into the composition of waiting times. This section provides an overview of the research design used in literature on ED waiting times. An overview of different research designs used in literature on waiting times at the ED is shown in Table 2.2. The table displays articles with research methods used in existing literature, their research aim and the perspective to which it belongs.

Article(s) Perspective Research design Aim Regan (2000) Explanatory Observational,

interviews

Identify operational inefficiencies of ED’s Asplin et al. (2003) Explanatory Theoretical Create a conceptual model

for understanding ED related factors Miro et al. (2003) Solution Observational,

historical data

The effect of ED expansion on ED waiting time

Lambe et al (2003) Explanatory Observational, survey

Investigate if geographic factors influence waiting time

Schneider et al. (2003)

Explanatory Observational, survey

Measure the degree of physical crowding and personnel shortage Plunkett et al.

(2011)

Explanatory Observational, historical data

(13)

Table 2.2. Research methods used in literature on ED waiting times

Only a few recent studies (Gelissen, 2016; Buitelaar, 2016; Ten Have, 2016) used work sampling in this context as a method in order to gain more insight in waiting times. In this context, the work sampling technique is used to sample waiting times, while this technique normally is used to samples fractions of work time (Robinson, 2010). Sampling waiting times, compared to work time, is challenging. This is caused by the fact that it is not always clear what exactly a patient is waiting for. Also, a patient could be waiting due to multiple reasons (e.g. both for an ED physician and a nurse). While when sampling work, most resources will do only one job at the same time. Next to that, the patient pathway differs per patient and is highly variable. Not every patient follows the same steps and processes in the same sequence, which makes it difficult to compare different patients. With sampling work time for a capacity analysis, these factors are more evident.

Even though there are challenges that have to be faced using this technique, Gelissen (2016) states that this work sampling has a great potential in order to measure patient waiting proportions. This method will be elaborated upon more extensively in the method section in order to investigate if this leads to new insights.

Sinreich et al. (2012); Sinreich & Marmor (2005)

Solution Simulation Investigate if scheduling work shifts lead to lower waiting times

Paul & Lin (2012) Solution Simulation Identify causes of reduced ED throughput

Bellow & Gillespie (2014) Explanatory Observational, historical data Provide a historical overview of ED crowding Gelissen (2016), Buitelaar (2016) & Ten Have (2016) Explanatory Observational, work sampling

(14)

2.5 Summary of literature findings

(15)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 15 3. Methodology

To gain more insight in waiting times at an emergency department, a single case study is performed at a hospital in the Netherlands. This method allows for an in-depth analysis and lends itself to exploratory investigations where the variables are unknown and the phenomenon is not understood (Karlsson, 2016). This chapter outlines the framework used in order to answer the research question of this research:

3.1 Case description

The case study takes place in a small-sized hospital in the north of the Netherlands. The ED is open 24 hours per day, 7 days per week and 15.162 patients were treated from 01/01/2017 to 31/12/2017. This means that on average, 42 patients visit the ED per day. The ED consist of six treatment rooms, a family room, a plaster room and a trauma room. There is no triage room since the determination of the urgency of the patient, which is done by a triage nurse, normally happens in the hallway. ED personnel consists of ED physicians, physician assistants, trainees and nurses. Each day a nurse is appointed as care coordinator. Together with the head ED physician of that specific day, the nurse controls the logistic aspects and flow through the department.

3.2 Case study design

This research consists of different phases that are aimed at analyzing the flow performance and indicating factors and mechanism that cause waiting. Even more important is creating a better understanding of what exactly patients are waiting for at each distinctive stage at the ED. First, the overall flow performance of the ED is analyzed. Subsequently, possible factors causing waiting are identified, after which a detailed analysis of waiting times using a work sampling method at the ED is done. Also developing the work sampling tool in this context is part of this research.

(16)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 16 questions this method is valuable (Yin, 2014). In addition, Meredith (1998) states that a case study can be used to collect both qualitative and quantitative in order to understand a phenomenon. This research is categorized as explorative, which means that how fewer the amount of cases used the more opportunity for in-depth observation there is. For this reason, a single case study is performed. A small-sized hospital in the Netherlands was selected to study since a small-sized ED allows for a better overview and process visibility compared to a larger ED. A better overview and visibility of processes of an ED leads to more reliable results.

3.3 Research phases

In order to structure this research on waiting times at the ED, research is conducted in four phases. Table 3.1 shows each individual phase.

Phase Data source Analysis Goal

1. General analysis Semi-structured interviews, direct observations Identify patient pathway, its distinctive stages and possible moments of waiting

Analyze current way of working 2. Flow performance analysis Historical time-series data Weekly pattern patient arrivals/ average LOS, input/output diagrams, throughput diagrams Understand case-specific characteristics 3. Waiting time analysis Structured observations (30-minutes interval)

Work sampling Determine causes of waiting

4. Combine findings Historical time-series data, structured observations, consulting physicians, direct observations

Combine all data and triangulate findings

Gain insight into composition of waiting times and underlying factors and mechanisms

(17)

1st phase: General analysis - In the first phase, the global characteristics of the case

are identified. This phase consists of semi-structured interviews with physicians, nurses and technicians together with direct observations. The observer is present at the case ED several days in order to observe the patient flows and the current way of working. Also, the processes at the ED will be visualized in order to identify all distinctive stages and possible moments of waiting. This is done based on de distinctive stages mentioned in the theoretical background.

2nd phase: Flow performance analysis - In order to understand the case-specific

characteristics a flow performance analysis is performed. This is done based on quantitative time-series data retrieved from the hospital information management system. First, weekly patterns in patient arrivals and average LOS are identified. In addition, the inflow, outflow, average work in progress (WIP) are analyzed.

3rd phase: Waiting time analysis - The third phase of this research aims to quantify

proportions of waiting times at distinctive stages of the ED. Using a work sampling study, there is identified what exactly patients are waiting for at distinctive stages of the ED. Waiting times for different patient specialisms, at different times of the day, and homogenous sets of patients are compared.

4th phase: Combine findings - In the final phase of this research all data is combined.

(18)

3.4 Sources and analysis

Both quantitative and qualitative data is gathered, qualitative by means of observations and consulting ED personnel and quantitative by means of data retrieved from hospital information system and work sampling. The combination of using both qualitative and quantitative is mutually supportive, as qualitative data gives more insights into quantitative data and quantitative data gives more reliability to qualitative statements of practitioners.

Historical time-series data - The hospital uses a hospital information management

system (HiX) to record all patient data. The main purpose of this system is to store medical records, but also quantitative data on logistical aspects are recorded. For purposes of this research, the following information is retrieved: patient origin, specialism, starting time at the ED, movement throughout the ED, departure time and the patient destination. After consultation with ED physicians, it became clear that movement throughout the ED is not recorded accurately. For this reason, this information is not used for this research. For the first analysis, data from 01/01/2017 – 18/11/2017 was used. During a later stage of this research, data until 31/12/2017 was added when available. The first step was to delete errors from the raw data set. Some records did not include a departure time and were for this reason deleted.

The data is used to analyze weekly patterns and understand case-specific characteristics. Also, the flow performance analysis is done based on this data. Input/output diagrams, graphs with the average LOS and average WIP during the day are drawn in order to recognize flow patterns, crowding and causes of crowding.

(19)

Work sampling - In the third phase of this research, data is collected by means of work

sampling. Data retrieved from ED systems does not comprise reliable information on what each patient is waiting for between distinctive stages in the care process. Therefore, this research aims to identify proportions of waiting time at each distinctive stage at the ED. Based on the analysis done in the first phase of this research, nine possible moments of waiting are identified. These are inflow, 1st contact physician, (additional) research, results (additional) research, ED physician, nurse, consult specialist, supervision and outflow. The work sampling technique is normally used to sample fractions of work time (Robinson, 2010), and will be used for sampling fractions of waiting time in this context. A few preliminary studies have used this tool (Gelissen, 2016; Buitelaar, 2016; Ten Have, 2016). Although they faced problems regarding the sample size, the number of rooms measured and interval of measurement, they showed that this tool has true potential in this context but also needs further development. Using a work sampling form (Appendix A) the following aspects are recorded per patient with a 30 minutes interval on weekdays from 11:00 till 19:00:

- Cause of waiting - Time under treatment - Patient specialism

- ED physician that treats the patient

In preliminary research (Gelissen, 2016; Buitelaar, 2016; Ten Have, 2016) the work sampling tool was used primarily to sample waiting times at a 30 minutes interval. All samples were combined, without taking into account that there are homogenous sets of patients that can be distinguished. In this research, waiting times are sampled with a 30 minutes interval, but also the following aspects are noted in order to distinguish homogenous sets of patients.

- Additional research required (lab, x-ray, CT scan, echo) - Consult from specialist required (yes/no)

- Possibility for supervision trainee (yes/no)

(20)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 20 times for homogenous sets of patients can be drawn and compared. This data is retrieved from the hospital information system the day after the waiting times are sampled.

The samples gathered with this tool will be analyzed in multiple ways. First, factors (e.g. time of the day, specialisms) that influence components of waiting are analyzed. Subsequently, the dataset is split up and a more in-depth analysis is conducted. Homogenous sets of patients are identified in order to create distributions of waiting times that are representable for specific groups of patients.

Consulting physicians - In addition to the quantitative data collected, consults with

physicians are used in this research to discuss findings. Physicians are asked for their opinion the following topics: what they experience as the busiest moments of the day, how they anticipate on this, causes of delays in the patient pathway and patterns found in the quantitative data.

Direct observations - In order to analyze how and why circumstances occur, direct

observations are useful (Meredith, 1998). Prior to and during the work sampling study the researcher is present at the ED to observe. Attention is paid to the way physicians and nurses communicate, the way physicians and nurses organize their work and the way data is recorded in the hospital information system. This is used to understand procedures and the current way of working at the ED.

Combined findings - The findings from both the work sampling study and historical

(21)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 21 4. Results

This chapter discusses the outcomes of the case study on waiting times at the ED. First, a general analysis gives a global overview of the case characteristics. Furthermore, an ED flow performance analysis is done in order to give insight into the ED’s logistics performances. Subsequently, the results of the work sampling study are discussed in the following sections. Factors as time of the day, crowdedness and patient characteristics are investigated in relation to the composition of waiting times.

4.1 General analysis

In order to give an overall impression of the case characteristics, first general info is extracted. The dataset retrieved from the hospital information systems is based on time-series data from 01/01/2017 – 31/12/2017. Patients with a throughput time higher than 1000 minutes were deleted from the dataset, which was the case for 2 patients. This resulted in a dataset that consists of 15162 patients that visited the ED distributed over 365 days. The average number of patients visiting the ED per day is therefore 42 patients and the average LOS is 129.92 minutes with a median LOS of 119 minutes. With a standard deviation of 69.6 minutes, it is evident the average LOS is rather variable.

(22)

4.2 Flow performance analysis 4.2.1 Patient arrivals

Emergency care is per definition unpredictable and unscheduled. Even though patient arrivals are highly variable, several patterns can be identified. The graph in Figure 4.1 shows the weekly pattern of patient arrivals.

Figure 4.1 Weekly pattern patient arrivals

Figure 4.1 shows boxplots with the 25th and 75th percentile that display the number of patients that arrive each day. On average, there are more patients visiting the ED during weekdays compared to the weekend. It also becomes clear that Monday and Friday are the busiest days, which is confirmed by ED personnel. The average number of patients that visit the ED each day is on average 42. Patient arrivals are therefore predictable to a certain extent, but still vary a lot taking the maximum and minimum number arriving at one day into account, respectively 60 and 19 patients.

15,0 20,0 25,0 30,0 35,0 40,0 45,0 50,0 55,0 60,0 65,0

Monday Tuesday Wednesday Thursday Friday Saturday Sunday

N u m b er o f p at ie n ts

(23)

Figure 4.2 Weekly pattern average LOS (minutes)

Figure 4.2 shows the average, the median, the 25th_{quartile and 75}th_{quartile LOS per}

day. The graph shows that the average LOS does not differ a lot between days. However, compared to the weekly patterns of patient arrivals (Figure 4.1), the outcomes on Friday are remarkable. Friday is a relatively busy day compared to other weekdays, but the average LOS is shorter compared with other weekdays. On Thursday this the other way around, with a relatively low number of patient arrivals and a relatively high LOS. The graph of figure 4.2 also tells that the average LOS per patient on the same day differs quite a lot. This is confirmed by the fact the LOS has a standard deviation of 70.5 minutes, which results in a 0.55 coefficient of variation for the whole dataset.

Figure 4.3 Number of patients vs. average LOS 50 70 90 110 130 150 170 190

Monday Tuesday Wednesday Thursday Friday Saturday Sunday

Av era ge LOS (m in u tes )

Weekly pattern average LOS

25th quartile Median Average LOS 75th quartile

0,00 50,00 100,00 150,00 200,00 0 10 20 30 40 50 60 70 Av era ge LOS (m in u tes ) Number of patients

(24)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 24 Figure 4.3 presents a scatterplot of the number of patients arriving per day compared to the average LOS. This figure indicates that there is a correlation between both factors. This means that the LOS is slightly related to the number of patients visiting the ED. It can also be observed from the graph that the dots are considerable spread among the axes. This indicates that a high number of patient arrivals per day does not necessarily mean the average LOS increases as well.

4.2.2 Inflow and outflow of patients

Figure 4.4 Input/output diagram ED

In Figure 4.4 the daily pattern of inflow and outflow of patients at the ED is shown. This is the rate at which patients arrive and the rate at which they are discharged. The x-axis represents the time and the y-axis represents the number of patients per hour. Previous research has shown (Ten Have, 2016; Koomen 2015; van Manen 2016) that similar hospitals have a comparable daily pattern.

In addition, the virtual unlagged output rate is displayed in Figure 4.4. This is the output rate at which the patients should be discharged if the ED would be able to keep up with the arrival rate. Given the fact that the average LOS is 130 minutes, this means in an

0,0 0,5 1,0 1,5 2,0 2,5 3,0 3,5 Pa tient s

Input/output diagram ED

(25)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 25 ideal situation the patients are on average discharged 130 minutes after an arrival. It becomes clear that at around 10:00 the ED is able to keep up with the incoming patients. However, later during the day the output starts lagging as the output rate slightly decreases, which results in longer patient waiting times at the ED. There is no explanation or clear insights yet of the factors that cause waiting times at this point

The graph can be divided into roughly three stages. The first stage is from 00:00 to 7:00 where the inflow and outflow are low and more or less the same. The second stage is from 7:00 to 17:00. At this stage, the WIP increases all the time since the inflow is higher than the outflow. This also can be observed in Figure 4.5, where the average WIP has added to the graph. The third stage ranges from 17:00 to 00:00 where the point is reached that both inflow and outflow decrease again. Also at this point the output rate kept up with the arrival rate, which means that from this point, the WIP slowly decreases again.

Figure 4.5 Input/output diagram + average WIP

In Figure 4.5 the average WIP is added, on the secondary vertical axes, measured as the number of patients present at the ED. The graph is in line with the statements made earlier, and also can be divided into the three stages. In the first stage, the WIP stays

0,0 1,1 2,3 3,4 4,6 5,7 6,8 8,0 0,0 0,5 1,0 1,5 2,0 2,5 3,0 3,5 1:00:00 2:00:00 3:00:00 4:00:00 5:00:00 6:00:00 7:00:00 8:00:00 9:00:00 _10:00:00 _11:00:00 _12:00:00 _13:00:00 _14:00:00 _15:00:00 _16:00:00 _17:00:00 _18:00:00 _19:00:00 _20:00:00 _21:00:00 _22:00:00 _23:00:00 0:00:00 N u m b er o f p at ie n ts Pa tient s p er h o u r

Input/output diagram + average WIP

(26)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 26 more or less the same. The WIP increases in the second phase and decreases again in the third phase.

Figure 4.6 Average LOS + WIP

Figure 4.6 shows the daily pattern of the average LOS is compared with the average WIP. The two things that stand out from this graph are the peak LOS around 7:00 AM and a similar pattern in the average LOS during the day and the arrival rate of patients. The average LOS increases as the arrival rate increases (Figure 4.5)

In previous section there is stated that there is only a weak relation between the number of patients visiting the ED per day and the average LOS per day. However, looking at this graph it can be obtained that there is an interaction between the number of patients visiting the ED per hour (see Figure 4.4) and the average LOS per hour. Given the fact the average LOS per day is not influenced by the number of patients, and the average LOS per hour is influenced by arrival rate of patients, this indicates a possibility for improvements.

The peak in average LOS around 7:00 AM is explained by two factors. The first reason is that ED physicians change work shifts at 8:00 AM. A physician who almost finished its night shift does not want to take a (non-urgent) patient with the risk to work over

0,0 1,0 2,0 3,0 4,0 5,0 6,0 7,0 8,0 80,00 90,00 100,00 110,00 120,00 130,00 140,00 150,00 160,00 1:00:00 2:00:00 3:00:00 4:00:00 5:00:00 6:00:00 7:00:00 8:00:00 9:00:00 _10:00:00 _11:00:00 _12:00:00 _13:00:00 _14:00:00 _15:00:00 _16:00:00 _17:00:00 _18:00:00 _19:00:00 _20:00:00 _21:00:00 _22:00:00 _23:00:00 0:00:00 Pa tient s Min u tes

Average LOS + WIP

(27)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 27 hours. Also, the X-ray department opens at 8:00, which means that patients at the ED sometimes have to wait at the ED until they can go there for additional research.

4.3 Waiting time analysis

In order to gain more insight into waiting times and hindrances of flow, a work sampling study is performed. With this method, it is determined what percentages of time patients are waiting and at what stage of the patient pathway. Samples are taken between the 5th of December 2017 and the 9th of January 2018. The samples are taken with a 30 minutes interval on weekdays from 11:00 to 19:00. In case it was impossible to measure within a time frame of 5 minutes from the planned time of measurement, the sample is seen as an invalid data point and is deleted from the dataset. This resulted in 1213 samples taken, from which 288 appeared to be under treatment when observations took place. This means the patient was seen at this moment by a physician or nurse. This is roughly 24%, which means that over 76% of the time the patient was present at the ED, the patient was not interacting with a healthcare provider.

4.3.1 Overall analysis

(28)

Figure 4.7 Components of waiting

Figure 4.7 provides an overview of the proportions of time a patient is waiting for along the patient pathway. If a patient stays 130 minutes at the ED for example, this patient waits on average 16.6% of this time (21,6 minutes) to be seen by a physician (1st contact). It is important to state that this is an average of all patients recorded. It should be taken into account that patients’ characteristics and complexity are highly variable. For this reason, the patient pathway could vary and is difficult to compare. Nevertheless, this distribution is useful in order to give a global overview of waiting times at the ED.

It can be observed that there are six major components of waiting. These are waiting for: 1st contact physician, additional research, results additional research, ED physician, consult of a specialist and the outflow. The differences between waiting time components are explained by the fact that nearly all patients require these steps in all conditions. Waiting for inflow (e.g. waiting for a room) is only measured at busy moments at the ED for example. The same applies to waiting for supervision (e.g. supervision for trainee). This is only measured if a physician in training actually treats a patient, which is only the case for 13% of all patients (see Table 4.1). In section 4.4, homogenous sets of patients are distinguished and patients that all require certain steps along the patient pathway. This creates a distribution of waiting that is more representable for specific groups of patients.

0,0% 10,0% 20,0% 30,0% 40,0% 50,0% 60,0% 70,0% 80,0% 90,0% 100,0%

Components of waiting

Inflow 1st contact physician Additional research Results additional research ED physician Nurse

(29)

4.3.2 Waiting times split at 15:00

Figure 4.8 Distribution waiting times split at 15:00

The dataset is split at 15:00 in order to see if there are differences between the composition of waiting times before and after 15:00. Based on the flow performance analysis, the WIP increases from approximately 8:00 till 15:00, then remains the same until 17:00, and decreases again after 17:00. This would suggest that waiting times earlier the day are still developing, which could lead to differences compared to waiting times later during the day. Also, the flow patterns and average LOS differ throughout the day. Due to the fact the dataset is split, the absolute error increases to +/- 3.14% for the largest category.

A few issues emerge from Figure 4.8. First, the waiting time components inflow and outflow is higher after 15:00. This is explained by the fact that it is busier in terms of patients present at the ED after 15:00. This means patients have to wait more often until a room is available (inflow) and nurses are busy and do not have time to transport patients to another department at the hospital (outflow). Also at the end of the day, there is a higher chance that all beds are occupied, which causes a delay in output. Furthermore, the percentage waiting time for consult of a specialist is higher during the morning. This is caused by the fact that specialists need to run an outpatient clinic till

0,0% 2,0% 4,0% 6,0% 8,0% 10,0% 12,0% 14,0% 16,0% 18,0% 20,0%

Waiting times split at 15:00

(30)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 30 17:00 and have less time during the day to help ED physicians. After 17:00 specialists finished their outpatient clinic and are more accessible for this reason, which results in less waiting time for this category.

4.3.3 Waiting times at busy and calm moments

During the work sampling study, also the degree of crowdedness was measured. At every time of measurement, the number of occupied rooms was recorded. This gives insight into the composition of waiting times during calm and busy moments, shown in Figure 4.9.

Figure 4.9 Distribution waiting times at busy vs. calm moments

Figure 4.9 shows differences between the composition of waiting times at busy and calm moments. Busy is in this context defined as a situation where over 7 treatment rooms (of in total 9) are occupied. Calm moments are therefore situations where 7 or fewer rooms are occupied. The waiting time components inflow and outflow are higher at busy moments, which is in line with the results shown in Figure 4.8. This is explained by the fact that in general, it is busier after 15:00 in terms of WIP compared to earlier during the day. Next to the factors mentioned earlier, does the outflow component increase at busy moments due to a coordination problem. It was observed that patients sometimes have to wait at ED, while they are ready to be admitted to the hospital already. This is probably caused by a lack of communication between ED physicians and nurses.

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Calm Busy

Busy vs. calm moments

Inflow 1st contact physician Additional research Results additional research ED physician Nurse

(31)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 31 The degree of crowdedness is measured by taking the number occupied treatment rooms into account, which means that the patients in the waiting room are ignored. It could be that there were many patients waiting in the waiting room, while only a few treatment rooms were occupied. Although this is unlikely, it should be taken into account when interpreting the results.

Previous waiting time components are presented as a percentage of the total time a patient is waiting at the ED. However, these results can be misinterpreted since they only give an overview of the waiting time components, not taking the total throughput time (average LOS) of patients into account. The flow performance analysis showed that the throughput time is rather variable. Therefore, section 4.4 will adjust the waiting time components to the average LOS of homogeneous sets of patients.

4.4 Waiting time analysis for homogenous sets of patients

The quantitative data retrieved from the hospital information system did not provide information about throughput times of specific groups of patients. However, using both the data collected by means of work sampling and the historical time-series data, this could be combined in order determine the average LOS for homogenous sets of patients. Also based on the combination of this data, the distribution of waiting times for homogenous sets of patients is determined. First, this is done for different patient specialisms. Subsequently, waiting times for patients that required different types of additional research are analyzed. Finally, there is a distinction made between patients that need consult and those who do need a consult.

(32)

4.4.1 Patient specialisms

Figure 4.10 Distribution waiting times per specialism

Figure 4.10 shows the components of waiting distributed over the average LOS for different patient specialisms. The components of waiting are now adjusted to the average LOS, instead of a percentage of the total waiting time that is shown in section 4.3. For this reason, the component ‘under treatment’ is added. The components of waiting are now expressed in minutes (of the total LOS), instead of a percentage of the total time a patient is waiting at the ED.

Most of the patients that were handled during the work sampling study, were categorized as surgical, lung diseases or cardiology. Therefore these three specialisms are compared. The average LOS for surgical patients is 140 minutes and the waiting time component inflow is 6.3%, which means that surgical patients wait on average 9 minutes to enter a treatment room. The component under treatment counts for 24%, which means that surgical patients are 33 minutes under treatment while being 140 minutes present at the ED.

Between patient specialisms, there are differences for both the average LOS and the distributions of waiting times. The average LOS for patients with lung diseases is higher compared to other patient specialisms. This is probably due to the complex

0,0 20,0 40,0 60,0 80,0 100,0 120,0 140,0 160,0 180,0 Total

Cardiology Lung diseases Surgical

Average LOS (minutes)

Waiting times per specialism

(33)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 33 characteristics of patients with lung diseases. For surgical patients, this is the other way around. These patients are less complex, which results in a lower average LOS. The waiting time components inflow, 1st contact physician and consults from a specialist represent are higher for patients with lung diseases, which indicates that these patient wait longer at these stages compared to other specialisms. Furthermore, the inflow component for cardiology patients is negligible. This is because in general, cardiology patients need to enter a treatment room for cardiac monitoring. It should be noted that due the fact that the dataset is split up again, the maximum absolute error increases to 6,63% for the largest category now. Therefore there is a chance some differences are not significant anymore.

4.4.2 Patients that require different types of additional research

As mentioned earlier in section 2.1, the stages every patient goes through vary per patient. There are several stages every patient goes through, but there are also stages certain patients pass. The first distinction between patients was made based on specialisms. In this section, also a distinction is made between patients that require different types of additional research.

With the work sampling technique, not only the cause of patient waiting was recorded. Also for every patient it is recorded if additional research is required, what kind of additional research, if the physician needs consult from a specialist (yes/no) and if the patient is treated by a trainee (yes/no). This last factor indicates if there is a possibility for supervision or not.

Additional research Consult specialist Supervision

Number without 74 132 1057

Number with 1139 1081 156

Total 1213 1213 1213

Percentage (%) 93,90% 89,12% 12,86%

Table 4.1 Percentages of patients that require resources

(34)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 34 There are different types of additional research at the ED. Table 4.1 assumes that a patient needs at least one of these types, which means that 94,25% of patients need at least 1 type of additional research (e.g. X-ray examination, CT-scan, lab, echo) However, different types of additional research lead to a change in the composition of waiting time. Table 4.2 shows the percentage of different types of additional research and the percentages of patients that require each of them.

Lab X-ray CT ECG Echo

Number without 289 645 908 457 1093

Number with 924 568 305 756 120

Total 1213 1213 1213 1213 1213

Percentage (%) 76,17% 46,83% 25,14% 62,32% 9,89% Table 4.2 Percentage of patients that require types of additional research specified

Based on this information, a more representative distribution of waiting times for specific patients can be composed (see Figure 4.11). On the y-axis, a distinction is made between patients that require different types of additional research. The first row shows the components of waiting in minutes for all patients that needed a CT-scan. The x-axis represents the average LOS in minutes for each set of patients.

Figure 4.11 Composition of waiting times for patients with different types of additional research 0,00 20,00 40,00 60,00 80,00 100,00 120,00 140,00 160,00 180,00 200,00 All

Lab X-Ray CT

(35)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 35 Figure 4.11 shows that the average LOS for patients with different types of additional research varies. All patients that require a CT-scan have an average LOS of 189 minutes. Also patients that need lab research have a longer average LOS than average, namely 163 minutes. This influences the composition of waiting times as well. The components wait for additional research and wait for results additional research are clearly higher for patients that need a CT-scan. Also, the waiting time components are slightly higher for this specific group of patients. Another thing that emerges from Figure 4.11 is the comparison between the waiting times of all patients and those who require X-ray examination. The composition of waiting times, including waiting for additional research and results are almost equal. Given the fact that only 47% of all patients need X-ray examination, the X-ray examination does not influence the composition of waiting times a lot. Next to making a distinction between patients with different types of research, the distinction can also be made between patients that need consult from a specialist and those who do not. Section 4.4.3 analysis the composition of waiting times for this set of patients.

4.4.3 Patients with and without consult specialist

Figure 4.12 Waiting times components for patients with and without consult specialist

Figure 4.12 shows that there is a considerable difference between the average LOS of patients that require consult from a specialist and those who do not, which are 100 and 157 minutes respectively. Comparing the waiting time components, all components are less for patients that do not require a consult from a specialist. Especially the

0,00 20,00 40,00 60,00 80,00 100,00 120,00 140,00 160,00 All patients

With consult Without consult

(36)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 36 components wait for additional research and wait for results additional research are significantly lower. Patient that need consult from a specialist wait on average 19 and 17 minutes for additional research and results, while patients without a consult wait on average respectively 5 and 2 minutes. This is explained by the fact that patients without consult from a specialist have less complex characteristics in general, and do not require additional tests. Also, the composition of waiting times for patients that do not require consults from a specialist differs a lot. This set of patients wait on average 70 minutes in total while being present at the ED, and wait 35 minutes to be seen by a physician.

4.5 Waiting time components throughout the day

Based on the flow performance analysis in section 4.2 and the waiting time analysis in section 4.3 and 4.4, the waiting time components throughout the day are discussed. In order to gain more insight into waiting times throughout the day, Figure 4.12 and 4.14 show the waiting times components per half an hour.

Figure 4.13 Components of waiting vs. average LOS 0,00 20,00 40,00 60,00 80,00 100,00 120,00 140,00 160,00 Av era ge LOS (m in u ts ) Minutes

Components of waiting vs. average LOS

Outflow Supervision Consult specialist Nurse

ED physician

Results additional research Additional research 1st contact physician Inflow

(37)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 37 Figure 4.13 shows the components of waiting as a percentage of the average LOS per half an hour. The average LOS of patients that arrive 11:00 at the ED is 135 minutes. During these 135 minutes, the patient is on average 39 minutes under treatment, waits 11 minutes for 1st contact physician, waits 22 minutes for additional research and so on.

Figure 4.14 Components of waiting vs. average WIP

Figure 4.14 shows the components of waiting as a percentage of the average WIP per half an hour. From 11:00 to 17:00 the average WIP ranges from six to eight patients. At 15:00 the average WIP in terms of patients present at the ED is seven and the waiting time component 1st contact physicians is approximately two. This means that two of in total even patients present at the ED are under treatment at 15:00. The component outflow is approximately one at 15:00, which means that one patient is waiting to leave the ED at this moment.

There are several things that can be obtained from the graphs in Figure 4.13 and 4.14. First, the inflow component occurs more frequently as the WIP increases. The opposite

0,00 1,00 2,00 3,00 4,00 5,00 6,00 7,00 8,00 9,00 Av era ge WIP (# p at ie n ts ) Time

Components of waiting vs. average WIP

Outflow Supervision Consult specialist Nurse

ED physician

Results additional research Additional research 1st contact physician Inflow

(38)

(39)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 39 5. Discussion

This research aimed to gain more insight into the composition waiting times at the ED. Factors and mechanisms influencing the composition of waiting times are investigated and the work sampling tool to investigate this has been developed further. This chapter gives an overview and discusses the results presented in chapter 4.

Due to an ED its complex characteristics, there are many variables that influence the logistical performances. In this research, several factors and mechanisms that influence the flow and hindrances of flow are analyzed. Despite the fact that emergency care is per definition unscheduled and unpredictable, daily and weekly patterns are identified. Therefore both patient arrivals and inflow/outflow patterns are predictable to a certain extent. The comparison between the number of patient arrivals and average (LOS) on Friday and Thursday are remarkable. On Friday the number patient arrivals is higher compared with other weekdays, while the average LOS is relatively short. This is explained by the fact that there is more ED personnel employed on Friday, which results in more capacity and lower throughput times. Thursday shows contrary results as the number of patients arrivals is relatively low, while the average throughput time is relatively high.

(40)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 40 adjust the components of waiting to the average LOS of specific groups of patients. This resulted in components of waiting in terms of minutes of the average LOS.

With the development of the work sampling tool, clear insights are gained into the composition of waiting times for specific groups of patients. This research has shown that factors such as time of the day, crowdedness at the ED and patient characteristics influence the components of waiting. Patient characteristics are identified based on specialisms appointed, different types of additional research required and the need for a consult from a specialist.

Both the comparison of waiting times before and after 15:00 and waiting times throughout the day show that time of the day is an important factor. The waiting times are developing until 15: 00, as the WIP decreases. Comparing this with the composition of waiting times after 15:00, there are considerable differences regarding the inflow and outflow components. Same results are shown when comparing calm and busy moments. Based on consults with physicians, the increase of the waiting component outflow can be explained by the fact nurses are too busy to arrange hospital admittance and transport to another department. This implies that there is a coordination problem. Also, due to a lack of communication between physicians and nurses, patients sometimes wait at the ED while they are ready for departure already. This mechanisms cause hindrances of flow and therefore result in increasing waiting times. Moreover, the way of working of ED physicians influence the composition of waiting times as well. While observing at the ED, it became clear that ED physicians organize their work in different ways. However, the exact differences between how ED physicians organize their work are not analyzed and are an opportunity for further research.

(41)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 41 require at least a CT-scan. The results show that an increase in LOS is many situations is caused by an increase in the waiting time components related to additional research. However, not every type of additional research changes the composition of waiting times. X-ray examination does for example not influence the composition of waiting times.

(42)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 42 6. Limitations of this research

6.1 Limitations of this research

Each ED of a hospital has its own characteristics and therefore different causes related to waiting times. The same goes for the case study used in this research, an emergency department of a small-sized hospital in the north of the Netherlands. Causes identified in this case might not be present at another ED, which limits the generalizability of this case. The data gathered in this research is not representable for other EDs. However, several factors and mechanism found in this research that influence the compositions of waiting time are found in similar studies using a work sampling technique (Gelissen, 2016; Ten Have, 2016; Buitelaar, 2016). Nevertheless, the mechanisms did not result in the same change in compositions in waiting time.

By further developing the work sampling tool a distribution of waiting times for homogenous sets of patients is created. This does not mean this distribution is completely accurate for this group of patients since there are still differences due to the high variability in patient pathways. However, the differences in composition of waiting time between patient groups shows the influence of patient characteristics on the flow performance.

6.2 Methodical limitations

The following sources are used in this study to collect data: direct observations, quantitative analysis, work sampling and the consultation of ED physicians. This section discusses the limitations of each individual method. Due to the fact work sampling is a new method in this context, there are several methodical limitations.

Direct observations – the researcher has been physically present at the ED several days

(43)

Quantitative analysis – quantitative time-series data between 01/01/2017 - 31/12/2017

and 05/12/2017 – 09/01/2018 is retrieved from the hospital information system. The first dataset is used for the flow performance analysis and the second dataset is used in combination with the work sampling data. The only reliable data related to the EDs operational performances were the arrival time and departure time of patients. Extreme values were deleted from the dataset since they would influence the results significantly, which is not representative for a usual weekday at the ED.

Work sampling – this research used a further developed work sampling tool in order to

sample waiting times. Data was gathered on weekdays from 11:00 to 19:00. The tool relied on physicians and nurses giving account of the current situation. In case the cause of waiting was not sure, another physicians or nurse was asked what the patient was waiting for. Most of the data was gathered from 11:00 to 17:00, which could influence the validity of finding after 17:00. Also as the level of in-depth analysis increased, the maximum error rate increased as well, which made the results less reliable.

Consultation of ED physicians – based on the consultations with ED physicians, it

(44)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 44 7. Conclusion

In this single case study, an analysis of waiting times at an ED has been done. This chapter summarizes the main findings of this research in order to answer the research question:

From the total amount of samples taken with the work sampling study, 24% of the patients appeared under treatment when observations took place. This is in line with the statement of McCarthy et al. (2014), who states that that about 75% of the time a patient spends at an ED, it is not interacting with healthcare providers. The work sampling tool has been developed further by identifying homogenous sets of patients based on patient characteristics and also the composition of waiting has been adjusted to the average LOS. This made it possible to express to components of waiting in terms minutes of the LOS, instead of as a percentage waiting time of the total waiting time (Gelissen, 2016; Ten Have, 2016; Buitelaar, 2016).

(45)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 45 research. Nevertheless, by identifying different patient groups, the distribution of waiting times at distinctive stage of the ED are investigated.

This paper gained more insight into waiting times at the ED of a small-sized hospital in the north of the Netherlands. Factors and mechanisms are identified, but it should be noted that there are no simple solutions in order to decrease ED waiting times. They would have been implemented if they would have been there. This research generated tools in order to make the ED more oriented at the logistics aspects.

7.1 Recommendations for practice

Logistics performance is not the main priority of an ED, and will never be since patient safety is the most important aspect. However, it could be argued that better operational performances lead to higher patient satisfaction (Taylor & Benger, 2004), lower costs for treatment and even lower mortality rates (Plunkett et al. 2011). It should be noted that while keeping the quality of healthcare at the same level, there is an opportunity for improvement regarding logistics performances. By conducting a work sampling study, more awareness at the personnel of the ED was created. ED personnel can use the waiting time distribution displayed is sections 4.3 and 4.4 in order to determine what waiting time components could be decreased. Also, the factors identified as for example time of the day, different specialisms and different types of patients should be taken into account. ED personnel should be aware that these factors affect the composition of waiting times.

(46)

7.2 Further research

In this research, the work sampling technique has proven to be an effective tool in order to gain insight at waiting times at the ED. Though several limitations were encountered while conducting the study, as described in previous sections. The next step in this field of research comprises improving the generalizability of findings. This can be done by performing a sample work sampling study at different EDs. Regional differences can be analyzed when the results of different EDs are compared. Lambe et al (2003) state for example that waiting times are influenced by the neighborhood characteristics, e.g. the location of a hospital. The same work sampling technique could be used at different EDs, through which can be analyzed if factors like hospital size, location and ED characteristics influence the composition of waiting times as well.

(47)

MSc. Thesis J. C. Kuiken, University of Groningen, 2018. 47 8. References

Anneveld, M., Van Der Linden, C., Grootendorst, D., & Galli-Leslie, M. (2013). Measuring emergency department crowding in an inner city hospital in The Netherlands. International journal of emergency medicine, 6(1), 21.

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.

Bellow, A. A., & Gillespie, G. L. (2014). The evolution of ED crowding. Journal of

Emergency Nursing, 40(2), 153-160.

Boyle, A., Beniuk, K., Higginson, I., & Atkinson, P. (2012). Emergency department crowding: time for interventions and policy evaluations. Emergency medicine

international.

Buitelaar, D. (2016). Taking Care of Patients - Avoiding Crowding in an Emergency Department in the USA. University of Groningen.

Cayirli, T., & Veral, E. (2003). Outpatient scheduling in health care: a review of literature. Production and operations management, 12(4), 519-549.

Connelly, L. G., & Bair, A. E. (2004). Discrete event simulation of emergency department activity: A platform for system‐level operations research. Academic

Emergency Medicine, 11(11), 1177-1185.

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.

Dunn, R. (2003). Reduced access block causes shorter emergency department waiting times: an historical control observational study. Emerg Med (Fremantle), 15:232-238. Drupsteen, J., van der Vaart, T., & Pieter van Donk, D. (2013). Integrative practices in hospitals and their impact on patient flow. International Journal of Operations &

Production Management, 33(7), 912-933.

Gelissen, J. H. (2016). Emergency Department Flow Performance, What are you waiting for? University of Groningen.