Explaining lagged output in the ED, a case study on patient waiting.

(1)

Explaining lagged output in the ED, a

case study on patient waiting.

Manon ten Have

University of Groningen

Faculty of Economics and Business MSc Technology operations Management 20 June 2016

Supervisors: Dr. M.J. Land

Dr. J.T. van der Vaart

Saffierstraat 22 9743 LH Groningen

m.ten.have.1@student.rug.nl 0640981791

(2)

Master thesis M. ten Have, University of Groningen, 2016. 2

Abstract

Crowding in the emergency department is a source of concern in many countries. Hindering flow and performance, the phenomenon is yet to be explained fully. Specifically the delay in output occurring in the morning, influencing throughput in the afternoon, is a mystery. This ‘lagged output’ is especially worrying when there are many arrivals in the morning. This research consists of a single case study of a small-sized ED in the north of the Netherlands. Utilizing interviews, historical time-series analysis, work-sampling analysis and direct observation in order to find causes of lagged output in the

emergency department. The main causes found are: low control over patient arrivals, low coordination of work, low amount of communication, high waiting for human resources, low responsiveness and delay in availability of physical resources. The quantification of the influence of the six causes is an opportunity for further research. This case study points out the need for attention to throughput in the ED at any point in time by all ED-personnel. A possible vehicle for improving flow in the ED is to increase patient flow visibility to establish a sense of urgency and induce action by ED-personnel to focus on output. The insights in the phenomenon of lagged output in this study can be used to reduce patient waiting and thus increase both patient satisfaction and flow in the ED.

(3)

1. Introduction

Patient waiting times are perceived as a challenge in healthcare by hospital management (Cayirli & Veral, 2003; Harper & Gamlin, 2003; Green & Savin, 2008). Long waiting times can severely affect patient satisfaction (Gardner, et al., 2007), morbidity, costs for treatment (Krochmal & Riley, 1994) and even higher mortality rates (Plunkett, et al., 2011). Specifically challenging is the Emergency Department (ED). A study of five Emergency departments in Israel showed a significant increase in the number of patients and the length of stay in recent years (Sinreich, et al., 2012).

Long waiting times at the Emergency department (ED) are generally associated with overcrowding (Derlet & Richards, 2000; Sinreich, et al., 2012). This crowding, which is an increase of patient arrivals, increases patient length of stay and delays treatment even with patients that need treatment urgently (McCarthy, et al., 2009). The patient length of stay is the time between entering the ED and departing from the ED. Findings from Krochmal & Riley (1994) show that an increased length of stay generates an increase in costs of ED patient treatment. A systematic literature review on causes, effects and solutions of overcrowding by Hoot & Aronsky (2008) revealed the multifaceted and complex characteristics of crowding in the ED. Co-dependencies, shared resources and arrival variability hinder application of operations research concepts such as queuing theory to optimize flow through the ED. Queuing theory is commonly referred to as a solution for crowding, but no universal model is able to explain and solve crowding in any ED so far (Hoot & Aronsky, 2008). Another common solution to less patient waiting is to increase human resources and beds, but according Kroneman & Siegers (2004) supply of these resources yields no significant effect on the average patient length of stay. Recent case study research in the Netherlands has found that the length of stay for relatively quiet and busy periods in the ED is similar (Koomen, 2016). A possible explanation is found in Diwas & Terwiesch (2009) who argue that hospital workers accelerate their service rate when arrivals increase and thus raise their productivity. However data from EDs previous case studies in the Netherlands (Koomen, 2016) show that this increase in productivity only starts when queuing of patients becomes visible. Physicians are able to keep up with inflow, however in the late mornings, lags in output are experienced. Acceleration in service rate happens only to catch up with inflow again late in the afternoon. Although the daily peak in arrivals is relatively predictable, it is unclear why this lag in output persist and why the acceleration in service occurs too late to free up beds in time to avoid crowding. The research question of this research is therefore:

How can lagged output in the Emergency Department be explained?

The aim of this research is to create more insight in ED-personnel responses to arrivals and patient waiting times. Specific focus will be on the cause of output delay in the morning and accelerated service in the afternoon by investigating patient waiting times during these periods. This research will consist of a case study utilizing both qualitative and quantitative data collected at the Emergency Department of a small-sized hospital in the north of the Netherlands.

(6)

2. Theoretical Background

Specific studies on the phenomenon of lagged output are scarce. However literature of crowding is extensive (Hoot & Aronsky, 2008). This chapter will summarize the main findings regarding crowding and research on hospital performance measures such as output, length of stay, and patient waiting times. This will be preceded by outlining the specific characteristics of the ED. Hereafter the current research regarding the arrival pattern in the ED will be outlined. Consecutively research on

establishing an even patient flow and causes of crowding and hindrances of flow will be discussed. Lastly, types of patient waiting found in literature and causes for output delays in literature will be elaborated upon.

2.1 Emergency Department characteristics

The lay-out of most EDs consists of a waiting room, a triage desk, a trauma room and several treatment rooms. Patients can arrive at the ED through emergency medical transportation, General Practioner (GP) referral or by self-referral. Next the patient waits until triage. Triage is according to van der Vaart et al. (2011) “a brief clinical assessment that determines the time and sequence in which patients should be seen in the ED.” The triage step is performed by a triage nurse and is used to determine the level of urgency of patient treatment. A widely used measurement tool is the

Manchester Triage System developed in 1997 (Mackway-Jones, et al., 2014) (Figure 2.1). The next step is diagnosis on the basis of physician examination, lab tests and medical imaging. After the diagnosis, policy is formed and the patient gets treated at the ED and followed by discharge to either home, the outpatient clinic, the mortuary, or an inpatient department at the hospital.

Level Urgency Color Time until seen by physician

1 Immediate Red 0

2 Very urgent Orange 10 Minutes

3 Urgent Yellow 1 hour

4 Standard Green 2 hours

5 Non-urgent Blue 4 hours

Figure 2.1 - Manchester Triage System levels of urgency (Mackway-Jones, et al., 2014).

Nature of work. Contact with the ill and afflicted, constant learning and meeting standards for quality precisely characterize the nature of healthcare work and make jobs in this sector stressful by definition (Parent-Thirion, 2007). Also professionals in health care are often needed to work overtime and unsocial hours (Messenger et al., 2007). They also face heavy workload, a multitude of tasks and hours that are longer, which can have serious consequences (Gundersen, 2001). This heavy workload is also described by Visser et al. (2003) in a study about Dutch medical specialists and stress. A study of Pilsjar et al. (2011) concludes “that workload and overtime can lead to poorer health among hospital employees across Europe”. Specifically in the ED overtime and high workload are not uncommon, therefore particularly ED-personnel is at risk of poorer health.

(7)

2.2 Arrival pattern

Patient arrival fluctuations are often associated with variables that are temporal (hour, weekday and month) (Wargon, et al., 2010; Batal, et al., 2001; Jones, et al., 2008). Moreover arrivals are associated with holidays (Jones, et al., 2008; Wargon, et al., 2010), the weather (Jones, et al., 2008) and affluenza outbreaks (Xu, et al., 2013). Specific cases in US and the UK find that arrivals are typically elevated on Mondays and in the weekends (Jones, et al., 2008; Morzuch & Allen, 2006). In a case study of five general Israeli hospitals, arrivals peaked between 11:00 and 12:00 and between 20:00 and 21:00 (Sinreich, et al., 2012). A peak in the afternoon was perceived in a study of Draeger (1992). Moreover in a British hospital arrivals between 10:00 and 22:00 were reported as higher than during the rest of the day (Morzuch & Allen, 2006). However the pattern of arrival in Dutch EDs is not yet known to literature. A study of three American hospitals by Jones et al. (2008) concludes that models should incorporate variables that are specific to the site and environment of the hospital for the most accurate way of forecasting daily ED patient numbers. For example applying exponential smoothing of

historical data to the ED in four Belgian hospitals made useful predictions of the number of visits (Bergs, et al., 2014). To establish an accurate response to changing arrivals, it is first necessary to distinguish what the general arrival pattern looks like for the case ED.

2.3 Responsiveness to arriving patients

When the arrival pattern is known, anticipating on this pattern can be key to improved performance. Koomen (2016) and Van Achteren (2014) have found a preliminary indication that increasing responsiveness of ED-personnel can lead to a decreased length of stay. Drawing on operations management literature, responsiveness can be defined as “how fast or easy a proposed change can be implemented” (Upton, 1994). Another definition of responsiveness is “a prompt service willingness” (Fitzsimmons & Fitzsimmons, 2011). In the ED this can be adapted to how fast and easy can be responded to new patient arrivals. So far little research has been conducted on the topic of responsiveness in the ED.

2.4 Measuring flow in the ED

(8)

2.5 Main reasons of overcrowding & patient waiting

There are many reasons for crowding in the ED (Derlet & Richards, 2000; Hoot & Aronsky, 2008). Main reasons for crowding present in literature are mainly inadequate staffing and inpatient bed shortages (Hoot & Aronsky, 2008). Regarding the latter Cooke et al. (2004) also relate hospital bed occupancy to the time a patient spends in the ED. Hospital bed occupancy levels that exceeded 90% caused substantial increases in length of stay in the case of Forster, et al. (2003). Furthermore the process of admitting patients to the hospital has been named as slow and seen as a cause of reduced throughput in the ED (Paul & Lin, 2012). Other causes named by Paul & Lin (2012) are insufficient physician capacity during peak hours, time spent waiting for lab results and turnaround times for radiology tests. According to McCarthy et al. (2014) approximately 75% of the time patients spend in the care area is without interaction with providers. This can be waiting time related to imaging or lab tests but this can also be time patients wait until the physician is ready with another patient. According to McCarthy et al. (2014) this high amount of waiting represents an opportunity for process

improvement

2.6 Possible causes of lagged output

The solution of responsiveness presupposes that personnel needs to perceive urgency before they increase their service rate. Theory about deadline behavior with dynamic deadlines indicates that attention to time increases with the approximation of the deadline rather than around a midpoint (Waller, et al., 2002). The attention to time increases and motivates the pacing effort of the group with the effect of increasing task performance (Waller, et al., 2002). The relationship between the rate of task performance over allotted work time is often perceived as positive linear or curvilinear (Karau & Kelly, 1992; Lim & Murnighan, 1994). Groups become aware of the expiring time and they then increase their task performance to complete their work on time. However such attention to time may be less present in the emergency department as some overview is lacking with the emergency physicians. For example emergency physicians in the study of Nugus & Braithwaite (2010) were relatively unaware of the amount of patients arriving to the ED at any point in time. Also they were unaware of the transfer of patients to inpatient departments and the availability of non-ED physicians for patient review in the ED. Another plausible cause specific to lagged output could be the way the activities in the ED are coordinated. According to Fredendall, et al. (2009) the barriers to swift even flow in the hospital call for the need to incorporate the variable of supply chain coordination. Coordination is “the management of dependencies between activities” (Malone & Crowston, 1994). As there are many actors in the ED, poor coordination of those can account for a part of the lag in the output of patients.

2.7 Summary of literature findings

So far in literature, much research is conducted on the causes of crowding (Hoot & Aronsky, 2008) however little can be found about the necessity of responsiveness (Koomen, 2016; van Achteren, 2014) to cover the lag in output in the morning. The operational theory of swift even flow

recommends the throughput time to find hindrances in flow and thus output delay (Fitzsimmons & Fitzsimmons, 2011). As also the behavioral patient centred care perspective is adopted, patient waiting time can be used as a determinant of patient satisfaction in the case study (Handel, et al., 2010). Possible causes of lagged output found in literature include deadline behavior and lack of

(9)

3. Methodology

To research how ED-personnel responds to workload and how lagged output in the morning starts to arise, affecting the waiting times and the length of stay, a single case study will be conducted at a small-sized hospital in northern Netherlands. This method section will outline the approach taken to answer the research questions of this research.

3.1 Single case study design

The case method used in this research allows for a relatively full understanding of the complexity and nature of the phenomenon (Karlsson, 2009). Especially for examining ‘how’ and ‘why’ questions this method is particularly valuable (Yin, 2014). Moreover by using this method, the processes in the ED can be studied in its natural setting and observing actual practice can amount to a new understanding of the phenomenon of lagged output (Karlsson, 2009). As the origin of this phenomenon is relatively unknown the explorative nature of the case method lends itself for exploring new variables. Because of this explorative nature, the fewer the amount of cases the more opportunity for in-depth observation in the allotted time. Therefore this research will consist out of a single case. To study the phenomenon of lagged output in-depth, a small-sized ED in a hospital in the north of the Netherlands was selected to participate in the study. The size of the department allowed for greater process visibility and overview than in a larger ED would have been the case. This because a small ED is easier to oversee for the observer.

3.2 The case hospital

The case hospital is a small hospital in the North of the Netherlands. The case hospital has an ED that is seen as a basic ED. Relatively complex patients will move to a specialized larger ED in the area. In 2015, 13514 patients were treated by the ED. The case ED consists of seven treatment rooms, a plaster room, a triage room, a waiting room and three hallway places for beds. Additional to several

physician assistants, trainees and nurses, the ED makes employs eight specialized emergency physicians. These emergency physicians have control over their shift and coordinate the team. The triage nurse controls the rooms to establish flow through the department. Furthermore physicians and nurses in the ED are available continuously, as they do not take breaks.

Most patients arrive to the ED by referral of the GP (40%) or specialist (22%). Others arrive by ambulance (16%), come from radiology (10%), or are referring themselves (8%) to the ED (Appendix I: Figure A). These patients are first seen by the secretary and signed in if they are not already signed in by the ambulance. From that moment on the patient is officially present at the ED. The arriving patients that are already signed on are visible on a routing board hung in the central department. Moreover the board displays all patients present in the ED, the room they are in and personnel assigned to them. Any medical personnel therefore can see where any patient is located. Next to that, icons depict tests to come in, triage color and the specialty per patient. After arrival the patient receives a triage color, in the case hospital most patients received yellow (52%) or green (30%) (Appendix I: Figure B). After treatment, most patients leave the ED to get admitted to the hospital (39%), go to the outpatient clinic (36%) or to go home (20%) (Appendix I: Figure C).

3.3 Research phases

(10)

Master thesis M. ten Have, University of Groningen, 2016. 10 Phase 1: Lagged output in the case ED. To find explanations for lagged output, it must be made sure that this hospital also deals with the phenomenon of lagged output. Moreover patterns of arrival, throughput over the course of the day, and differences between mornings and afternoons will be investigated in this phase to establish a first look on flow patterns in the case ED.

Phase 2: Responses to increases in arrival. The second phase of the research will outline the main qualitative findings regarding responses of medical personnel on the arrival pattern within the case ED. Attention will also be paid to recognizability of flow patterns and what physicians perceive to impede flow and affect output negatively. In this way the responsiveness of the ED and the influences on output can be made visible. Hereby probable causes of delays can be identified.

Phase 3: Waiting in the ED. Consecutively phase three will investigate the main causes of delays in patient flow by use of waiting times in the case ED. Furthermore as to investigate the concept of responsiveness as a solution to lagged output (Koomen, 2016), causes of waiting in the morning are compared to causes of waiting in the afternoon.

Phase 4: Main causes of lagged output. Lastly, phase four will then consist of an attempt to summarize findings and establish cause-effect relations found in phase 2 and 3 to the concept of lagged output.

3.4 Sources and Analysis

To ensure validity of the case study, different methods and types of data collection will be used for data triangulation purposes. Table 3.1 gives an overview of the methods of analysis employed in this study.

Table 3.1: methods of data collection

Next the methods and techniques of analysis to answer these questions that will be used in this research will be explained further.

Historical time-series data: The dataset consisted out of all patients that arrived on the 1st of June 2015 until the 31st of May 2016. First all patients that were discharged as ‘expired’ and ‘not processed’ will be removed as these patients have not entered the ED. The ‘raw data’ will then be analysed with Excel. The dataset listed 13514 patients arriving to the ED, 50% of whom were male and the other 50% being female. The median age was 51 years old. Analysis of the data will be done on length of How can lagged output in the ED be explained?

Phase Source Analysis

1. Lagged output Historical time-series data Input-output diagram, Throughput diagram, weekly arrival boxplots. 2. Responses Interviews / Direct observation Coding of interviews

3. Waiting Structured observations Work-sampling

(11)

Master thesis M. ten Have, University of Groningen, 2016. 11 stay over time, work-in-progress and input-output diagrams will be drawn. The latter will be made to analyse if the lag in output also persists in this ED. Moreover an accumulated throughput diagram will be used to see how patients flow through the ED.

Semi-structured interviews: Semi-structured interviews will be used to examine physicians views on throughput, attention to time, main sources of delay in output and their thoughts on process

improvement. Two specialized emergency physicians and two assistant physicians will be

interviewed. An interview protocol will be used to structure these interviews (Appendix II). The input-output diagram, throughput diagram, work-in progress diagram and length of stay will be used as input for these interviews. Interviews are conducted in Dutch and recorded and transcribed (Appendix III). Analysis of these interviews is done as outlined in Karlsson (2009): The first step is open coding, making categories and identifying concepts in terms of dimensions and properties. The second step is axial coding, regrouping the data in a new way to link categories rationally. The third and last step is be selecting a core category and its relation to other categories. All interviews will be coded using Atlas.ti.

Direct observations. Direct observations of the processes in the ED and especially responses to increasing arrivals of patients will be written down. These notes will be used to illustrate findings in the data. For observing directly, a notebook and Excel will be employed. The observer will spend several days in the case ED to observe flows and the current way of working.

Work-sampling. The last method used to identify sources of waiting is the work-sampling technique as outlined in Robinson (2010). This method is used as an estimating approach for the percentage of time that employees spent on predefined work tasks (Kirwan & Ainsworth, 1988). Data will be collected at a number of sample points through an observer in the ED. This observer uses half hour intervals to observe the status of each patient present in the ED. Each observation consist of a patient status, time of treatment, start time, treatment room number and a binary option seen/not seen by a physician. With this technique the observer focuses on patient waiting times as opposed to what will be perceived in the interviews by physicians as long waiting times. The observer will ask ED-personnel about the status of every patient they are seeing at the moment at half hour intervals.

3.5 Lagged output

(12)

4. Results

This chapter describes the summarized outcomes of the case study about lagged output in the ED in the North of the Netherlands. The gap between the real output rate and the ‘ideal’ output rate is described by the four phases: first the lagged output in the case ED, second the responses to arrivals in the ED, thirdly the main causes of waiting and fourth the main causes of lagged output.

4.1 Lagged output in the case ED

This section discusses the phenomenon of lagged output in the case ED. All graphs in this section are based on time-series data from June 2015 until May 2016. To determine if the lag in output as seen in other cases also occurs in the case hospital, an input-output diagram is drawn. Figure 4.1 shows the rate at which patients arrive at the ED and the rate at which they are discharged. Also the ideal output rate is depicted (Figure 4.1). This is the output rate that would occur if the ED is able to keep up with the arrival rate. This means that after an arrival, on average this patients is discharged after the average value of the length of stay, which is 137 minutes, roughly two hours.

Figure 4.1: Input and output diagram of an average day in the ED.

In this graph it is clear that at on average in the beginning of the day, the ED is capable of keeping up with incoming patients. However later in the day the output starts lagging. This means specifically that there is a gap between the ideal output rate and the realized output rate. In this ED, this is between 10:30 and 16:00. Until 16:00 the realized speed is slower than is the ideal speed to keep up with arrivals. This gap was also found in three medium-sized hospital in the north of the Netherlands (Koomen, 2016; Manen, 2016; Gelissen, 2016), moreover this is found in the west of the Netherlands (Westerhof, 2016) and a large regional teaching hospital in the North of the Netherlands (Dijk, 2013). The lag in output causes patients to occupy rooms needed for new patients during peak hours.

0 0,5 1 1,5 2 2,5 3 3,5 P a tient s

Input-Output diagram ED

(13)

Master thesis M. ten Have, University of Groningen, 2016. 13 Weekly arrivals. To find patterns in arrival of patients a boxplot graph is drawn. When looking at daily arrivals of patients (Figure 4.2), it is visible that Mondays (Median = 39) and Fridays (Median = 41) are generally the days with the most arriving patients. This was also perceived by physicians (Appendix III, Interview 1). With a total median of 37 patients and an interquartile range of 11 patients, the amount of patients seems relatively predictable. However as this is a small ED, a difference of a few patients can be the cause of crowding when they arrive at the same time. The maximum number of patients can go up to 61 patients on the busiest day(May/13/2016) and down to 18 patients on the least busy day (June/11/2016).

Figure 4.2: Boxplots of patient arrivals per weekday in the case hospital

The build-up of work in progress in the ED is depicted in Figure E as a throughput diagram (Appendix I). As input and output are relatively slow in the beginning of the day, later it is clear that the amount of work-in-progress is increasing and that length of stay also increases. This trend persists until around 18:00 and then the length of stay and the amount of work-in-progress decreases again.

Work in progress and length of stay. The work in progress is also visible in figure 4.3. This figure also represents the length of stay given per minute of arrival. The average length of stay of patients arriving in this minute was calculated from all patients in the year. During the nights, the length of stay is relatively similar, around 140 minutes. However as soon as the day starts, the length of stay

decreases rapidly to around 9:00. From 9:00, as work in progress builds, the average length of stay increases until a peak on 13:00. After 13:00 the length of stay decreases to form a small valley around 16:00 with another small peak on 18:00. This is followed by a decrease in length of stay in the

evening. 0 10 20 30 40 50 60 70

Sunday Monday Tuesday Wednesday Thursday Friday Saturday All

(14)

Figure 4.3: Length of stay and Work in progress in the ED

Figure 4.4 indicates a correlation between the number of arrivals in the ED and the length of stay of patients. So the length of stay is partly dependent on the amount of patients visiting the ED. And thus the amount of capacity is an influence on the length of stay. However as there is a lot of variability (Figure 4.4) this effect is small. As with the same capacity more arrivals do not necessarily mean an increase in length of stay, something other than the amount of capacity must also be explaining the delay in output.

Figure 4.4: The relation between length of stay and amount of patients visiting the ED.

0 20 40 60 80 100 120 140 160 180 0 1 2 3 4 5 6 7 8 9 10 0 :0 0 1 :0 0 2 :0 0 3 :0 0 4 :0 0 5 :0 0 6 :0 0 7 :0 0 8 :0 0 9 :0 0 1 0 :0 0 1 1 :0 0 1 2 :0 0 1 3 :0 0 1 4 :0 0 1 5 :0 0 1 6 :0 0 1 7 :0 0 1 8 :0 0 1 9 :0 0 2 0 :0 0 2 1 :0 0 2 2 :0 0 2 3 :0 0 L eng th o f st a y ( m in) Am o un t o f pa tient s in t he E D

Length of stay and Work in progress

Work in progress Length of stay

0 50 100 150 200 250 0 10 20 30 40 50 60 70 L eng th o f st a y ( L O S) Amount of patients

(15)

Master thesis M. ten Have, University of Groningen, 2016. 15 This seasonal variation is also seen in a weekly pattern (Appendix I, Figure F). The median length of stay in the case hospital was 123 minutes. The deviations of the median length of stay per weekday differ 27 minutes between the day with the highest arrivals (Monday) and the day with the lowest arrivals (Saturday) (Appendix I, Figure F). Slightly lower lengths of stay were mainly visible during the weekend days, this is explainable as the number of arrivals is lower. The pattern during the week is comparable with the number of patients during the week. As weekends have less arrivals and a peak that starts later (Appendix III, Interview 4), the week is most representative to study output and patient waiting with the work-sampling technique.

Effect of the amount of arrivals in the morning. Because Figure 4.1 indicated a delay in output starting in the morning, a split was made in the data on the amount of arrivals during the morning. As visible in figure 4.3 the work in progress starts building from 9:00 is at a max at around 13:00. The median amount of arrivals between 9:00 and 13:00 is eleven patients. The data was split between low arrivals (ten or less arrivals) and high arrivals (eleven or more arrivals). To see if the amount of

arrivals in the mornings affected the output rate in the afternoon, two input-output diagrams are drawn. Comparing low arrivals in the morning (Figure 4.6) and high arrivals in the morning (Figure 4.7) in both cases the delay in output is visible. On the days with the lowest arrivals in the morning, ED-personnel is able to keep up with demand until around 11:00. For the high arrival days, output starts to lag around 10:30. Next on an average day with high arrivals, around 12:00 an output rate of 2.6 is generated whereas on a day with low arrivals, only 1.7 is reached. As the ideal output rate in the case of low arrivals should be 2.2, it is curious why this is not generated especially because on days with high arrivals even 2.6 can be reached around that time. This indicates a slower pace of working during the morning with low arrivals than during a day with higher arrivals. The slower pace could be due to the low amount of arrivals influencing perceived urgency. This perceived urgency may then indicate low responsiveness.

(16)

Figure 4.5: Input-output diagram of the average of all days with low arrivals

Figure 4.6: Input-output diagram of the average of all days with high arrivals

0 0,5 1 1,5 2 2,5 3 3,5 P a tient s

Input/Output low arrivals morning

Arrival rate Output rate Ideal Output rate

0,00 0,50 1,00 1,50 2,00 2,50 3,00 3,50 4,00 P a tient s

Input/Output High arrivals morning

(17)

4.2 Responses to arrivals

With the use of interviews and direct observation, the current way of working was researched. Appendix IV gives the coding tree used in the process. The main categories tied to the codes in the interviews were: the arrival pattern in the ED, capacity, physician methods of working, ED visibility, external resources and physician ideas for improvement. When investigating current responses to arrivals and ways of working in the ED, the first question in the interviews was to what extent the arrival pattern was known to the physicians.

Arrival pattern. With the use of different graphs this question was asked to four physicians in the ED. It was found that both the ED-physicians and the assistant physicians were aware of the daily pattern in the ED. One of the ED-physicians remarked when looking at a graph of the daily work in progress:

“Yes this is compatible with the busiest moments of the day. It differs per day, some days the peak of arrivals is a bit earlier than other days, but on average these are the busiest times (Appendix III, Interview 1)”.

Physicians were also asked to comment on the peaks in length of stay during the day to identify probable causes of output delay. The increased length of stay the early morning (Figure 4.4) was explained as patients who are either complex and wait on the radiology shift to start at eight or having an alcohol intoxication (Appendix III, Interview 2). Another delay was that specialists were reported to be more difficult to reach in the morning (Appendix III, Interview 3). The small valley in the afternoon around four was mainly explained by increased capacity.

Capacity. The valley in length of stay in the afternoons (Figure 4.4) was explained in all interviews as the extra capacity by means of an overlap in two shifts in the afternoon.

“You can really see a peak in the output between 4 and 5 PM and maybe this is because on that moment the late shift arrives and you can wrap up old patients and new patients will be seen by someone else (Appendix III, Interview 1)”.

Extra capacity in the form of an extra physician during peak moments in the day was also reported to be utilized to anticipate on arrivals by all physicians. But not only capacity was reported as methods of anticipations, physicians had different ways of anticipating on arrivals.

Anticipation of arrivals. One physician said not to plan any meetings during the day and tried to have meetings more in the mornings (Appendix III, Interview 2). Another physician said for example that they prepare patients in advance:

“When we get a notice that a new patient is signed on, the ED-physician would estimate what types of research are necessary for this patient (Appendix III, Interview 1)”.

Also this anticipation after sign on of a patient was reported by another physician:

(18)

Master thesis M. ten Have, University of Groningen, 2016. 18 Here the physician reported to take action as many arrivals are perceived, this imposes a sudden ‘deadline’ on present patients consistent with deadline behavior (Waller, et al., 2002). When many patients arrived, less urgent patients were for example placed in the waiting room or discharged (Appendix III, Interview 3). This was also directly observed by the author. When many patients arrived, suddenly the throughput of patients was discussed and attention to throughput started:

“I actually only start to discuss freeing up rooms when patients are signed on, and I expect them within half an hour (Appendix III, Interview 3)”.

An increase in communication of patient throughput was also perceived during direct observation in the ED on busy days (3rd , 10th , 13th of May 2016). Next to this during direct observations ambulance arrivals were also prepared for. ED-personnel was ready for very urgent patients directly when they arrived by ambulance.

Responses to arrivals by physicians. When then the patients actually arrived, different mechanisms were reported in the interviews and perceived in the direct observations. Coordination of work consisted of division of patients between physicians by either the coordinating ED-physician (Appendix III, Interview 4) or patients were seen by priorities set by the physician. The former was mostly observed when it was busy in the ED or at a transfer of patients during a switch of shifts. The decision of which patient to see next was based on how many and what kind of patients were currently seen by the physician. Also this was based which physician was idle and who was waiting for

additional research next to triaged urgency and length of stay (Appendix III, Interview 1). Another physician based this decision also on how much time it would take to help a patient or to prevent idleness of nurses (Appendix III, Interview 2). Lastly when a shift almost ends, physicians are less likely to take on new patients (Appendix III, Interview 4).

When taking on new patients, three of the four physicians interviewed reported a multitasking effect

(Appendix III, Interview 1,3,4)”. As between 10:00 and 11:00 patients start to pile up and

work-in-progress starts to increase, physicians are required to take on more patients.

“In the beginning you can keep up, but at some point you have reached your maximum capacity (Appendix III, Interview 3)”.

When presented with figure 4.1 a physician elaborated extensively on this phenomenon, remarking that at first you feel in control but then you suddenly spend time catching up with the peak arrivals.

“During a shift, on a given moment you notice some kind of turning point wherein you have less visibility and you have the idea a patient spends more time in the ED because multiple patients and things pile up and not all is finished…..Figure 4.1 illustrates this nicely, at the busiest moments, it is a bit more messy (Appendix III, Interview 4)”.

Also this phenomenon was reported to be strengthened by the influence of complex patients

(19)

Master thesis M. ten Have, University of Groningen, 2016. 19 Visibility of patients. When seeing how many patients were present in the ED, the most used and beneficial tool was the routing board and seen as the way to keep oversight on patient flow, care providers assigned, attention to time and the room a patient was in (Appendix III, Interview, 1, 2,3,4). However when in a treatment room, the board was not visible and this was seen as a problem

especially during busy times (Appendix III, Interview 3). Also during direct observations it was clear that arrival of patients or patients in the waiting room were not always noticed by physicians.

The second key factor influencing visibility of patient flow is lay-out, a physician elaborated on the difference between EDs when looking at the lay-out of the department and how that influences visibility of patients. About the case hospital the physician said:

“You see it happening, you see the ambulances, patients coming by, it is transparent and clear what goes on.” and “Also to me it is important that the nurses are close by and the secretary, that everything is more a whole (Appendix III, Interview 4)”.

All actors are easier to communicate to as they are close by. Also mentioned before, when increasing arrivals are visible, urgency is perceived and attention is drawn to time. This can be time a patient is waiting (Appendix III, Interview 2), a time a physician waits for a resource or the general time of day. This attention to time is not always present according to physicians.

“I don’t really watch time a lot, at a given moment it is time to go home and I think that was fast”(Appendix III, Interview 1)

At a length of stay of 3 hours and 40 minutes a physician said during direct observation on the 3rd of may: “I did not know the patient is here that long!”. Or sometimes patients are not visible when in the room, for example on the 13th of May a physician forgot to see a patient before discharge and another was forgotten to be admitted to elsewhere in the hospital. On the 26th of May a patient was forgotten to be signed on by a nurse for a CT-scan and thus was waiting until nurses and physicians noticed. Waiting and delay. The reported issues with visibility can result easily in patient waiting and thus delay in patient flow. Complexity of patients was clearly perceived as having an influence on patient length of stay. For example patients wait longer when the pattern of complaints was a-typical (Appendix III, Interview 1 & 4). The relatively easier surgical patients can be processed much faster (Appendix III, Interview 1 & 2). A major source of waiting is related to lab results that was reported to take between half an hour up to two hours by different physicians, secretaries and nurses.

Another often perceived source of waiting is the radiology department that is in charge of a lot of the additional research done. During lunch only one radiologist is available (Appendix III, Interview 2) and it can even take up to 1,5 hours when radiology is busy (Appendix III, Interview 4). Another physician elaborated on specialists as a cause of delay in patient flow:

“Some specialist would like to come by themselves to examine the patient, they say they will be right there, but in the mean time they finish their consultation hour, then the patient here occupies a room, because you expect the specialist soon (Appendix III, Interview 3)”.

(20)

Master thesis M. ten Have, University of Groningen, 2016. 20 working with specialists fast and that admittance to the hospital was fast. The latter was also again reported as taking both a long (Appendix III, Interview 4) and a short (Appendix III, Interview 1) amount of time. This related to lunchbreaks of departments and delay in picking up patients (Appendix III, Interview 4). The last delay was found in disturbances for ED-physicians.

“Our work is already unpredictable, continuously you are disturbed by everyone, the trainee has questions, the assistant physicians have questions, you have all GP-phone calls, the nurses want to know something… the continuous interruption make you slower, because every time you need to try to focus on the piece you were working on (Appendix III, Interview 3).

In short flow of patients in the ED has many actors and co-dependencies. The effects of multitasking and the way of coordinating the process influences the length of stay and hindrances of flow take all shapes and sizes. The next section will describe in detail the frequency of causes of waiting times that impede flow.

4.3 Waiting in the ED.

In the interviews different causes of delay in output were mentioned. To investigate what hinders patient flow most, a work-sampling technique was utilized. In total the observer gathered 1030 data points containing patient statuses per half hour. In total 274 patients were observed, 134 arrived in the morning and 135 arrived during the afternoon. Five patients were only observed in the afternoon, but arrived in the morning. Appendix V provides the frequencies of the gathered data. Appendix VI elaborates on the meaning on each of the observed patient statuses.

It was found that 26,5% of patient statuses were ‘under treatment’. This translates to approximately 1/4th of the patient time. This means that 73,5% was patient idle time and thus the patient was waiting. Figure 4.6 depicts a diagram of all found causes of waiting and their frequency of occurrence. The causes of waiting are all related to the main hindrance of flow perceived at a data point. The resources depicted in figure 4.7 are necessary for a patient to flow to a next step in their treatment.

First, from the figure it is clear that patients wait the most for a physician to see them. This can be for the first time or after initial examination done by the nurse. Also this can be after additional research or for treatment. Physicians also work on multiple patients and thus they might be busy working on another patient or they are making policy or write up a report of a patient while other patients wait. Moreover, the main coordinating ED-physician often has extra tasks such as supervision, coordinating tasks, and handling calls from other parts of the hospital.

(21)

Figure 4.7: Pareto diagram of causes of patient waiting during the day.

Comparison between morning and afternoon. To compare if during the morning waiting for patients was different than during the afternoon, the work-sampling data is split into causes of waiting observed in the morning and causes observed in the afternoon. Observations between 9:00 and 13:00 were labelled as morning and all observations between 13:00 and 17:00 were labelled as afternoon. The average waiting time of the observed patients arriving in the morning was 152 minutes, for the afternoon this was 149 minutes. As the difference is negligible it is assumed is that the percentages of waiting for the causes are comparable. The causes of waiting are aggregated using the main categories of waiting in Appendix VI. The aggregated types of waiting are displayed in figure 4.8.

Wait for discharge. In the graph 17% and 20% of the hindrances in flow are happening in the end of the of the process, the outflow from the ED. The small three percent point increase in the afternoon, can be related to the fact that during the day beds elsewhere in the hospital become full because of admittances. It will then take longer to find an appropriate bed in the right department. Moreover nurses have less time per patient as afternoons are busier than in the morning. Priority during busy times can then not lie at discharge, as urgent patients need help immediately.

Wait for physician. Similarly waiting for physicians hinder patients flow a little more in the afternoon (19%) than in the morning (16%). In the afternoon the amount of work in progress is higher and capacity is similar to the mornings, therefore there are more patients per physicians. Although the increase is not shocking, this could have something to do with an acceleration in the output (Figure 4.1). If this service rate was also achieved during mornings, waiting for physicians could be less as physicians have less patients per person. It also offers the possibility that time for making reports and policy could have an higher influence than multitasking with multiple patients.

0% 20% 40% 60% 80% 100% 120% 0 20 40 60 80 100 120 140 160 F re uq ncy

(22)

Figure 4.8: The build-up of types of patient waiting in the morning and afternoon in the ED.

Wait for outcomes. A 4 percent point decrease in the afternoon is visible in waiting for results. The proportion that was waited by patients for human resources such as physician (+3), specialist (+1) and nurse (+4) was bigger. In the work-sampling method the observer registered the cause of waiting that was hindering flow. For example an outcome could already be in, but the physician was so busy that he/she was not able to look at the patient yet and thus the patient waited for a physician rather than for the results. In this case waiting for outcomes could be observed less as these human resources were more busy during the afternoon. Thus in that way waiting for results did no longer hinder flow.

Wait for room. Waiting for rooms increased by 3 percent point to a total of 15% of all observations in the afternoon. These are additional patients in the waiting room. In the afternoon, most patients that are not urgent (triaged green) are in the waiting room for some time in the ED. In the morning only some patients spent a lot of time in the waiting room. However from the patients that were in the waiting room in the morning, a much larger percentage waited for the plaster room (45% ) than in the afternoons (22%). It was preferred to do plaster for fractures in the designated room, but during busy times, any room was good to help patients. Even the triage room was used by physicians instead of the triage nurse, who opted for the waiting room to do small proceedings.

Waiting for specialist. Waiting for specialists is similar during mornings (12%) and afternoons (13%). Specialists act independent of knowledge of ED-arrivals and therefore perceive little urgency over the phone as they do not know the status of the ED, only that of the urgency of patient-health. Some specialists wait until they have finished their consulting hours as interrupting this would cause

17% 20% 16% 19% 17% 13% 12% 15% 12% 13% 17% 6% 7% 11% 2% 3% 0% 20% 40% 60% 80% 100% Morning Afternoon

Types of patient waitng

Wait for assistant consulting physician

Wait for nurse

Wait for additional research (CT, photo, echo, ECG, MRI) Wait for specialist

Wait for (plaster) room Wait for outcome (lab, CT,photo,echo) Wait for physician Wait for discharge

(23)

Master thesis M. ten Have, University of Groningen, 2016. 23 delay in their own department. Figure G outlines the amount of occurrence of different specialisms (Appendix I). As most low-complex surgical patients are fractures, they are handled by the ED and move fairly quickly through the ED. Other surgical patients and specialties have average lengths of stay as high as 180 minutes or as low as less than 120 minutes, this is can be due to complexity of patients, time to arrival of the specialists and how easily the specialists can be reached by phone. The latter depends on the requirement of the specialist to see the patient at the ED or the ability to discuss the patient over the phone.

Wait for additional research. Waiting for research also decreased with 11 percent points, the cause of this could be similar to the cause of waiting for results, it is simply not hindering flow but is simultaneous to waiting for something that hinders flow. But it is also likely that because of the busyness in the afternoon in the hospital, other resources such as echo and CT prioritized ED patients more because of the perceived increase in urgency pointed out by the physicians of the ED as their work in progress is rising. And when the resource such as the ECG is at the ED, multiple patients can be handled sequential.

Wait for nurse. During the morning, patients were seen to wait less frequent for nurses (7%) than in the afternoon (11%). As work in progress in the ED during the afternoon was higher than in the mornings, nurses needed to handle more patients at the same time. This multitasking effect could well speak to planning skills of the nurses and makes it difficult to determine what to prioritize. As the nurses often take the first step of the process (taking patients to the room, first examination and taking blood/urine) a delay in this part of the process can have serious consequences.

Wait for assistant consulting ED-physician. Lastly, during the days of observations mostly one trainee was present in the ED during the day. The trainee needed supervision for mostly one or two patients that he treated together with the ED-physician. When making policy, the assistant needed to consult the ED-physician who was also busy with calls, other patients, coordinating tasks or giving a consult to another physician. One ED-physician in the interview remarked that supervision takes a lot of time. The 1 percent point increase during afternoon can be due to a decrease in availability of the ED-physician during busy times.

Waiting times in the ED are the result of a complex process depended on multiple actors who together deliver the end result. Both physical and human resources in and outside the ED are interrelated to an extent that keeping up with flow in the ED is challenging. The causes of waiting times that differ during the mornings and afternoons will be related to the phenomenon of lagged output in the next section of this chapter.

4.4 Main causes of lagged output

(24)

Figure 4.9: Ishikawa diagram of causes of lagged output 1. Low control over patient arrivals

The amount of control the ED has over patient arrivals influences the amount of lagged output in the ED.

 High volume during the late morning results in more congestion of the ED. As found in the data-analysis when arrivals are higher during the morning (Figure 4.8), it is more difficult to keep up with the arrival rate. As there is a relation between the amount of arrivals and length of stay (Figure 4.5), this could lead to lagged output when capacity is then not increased.  Out of the control of physicians is the mix of arriving patients. Found is that some higher

complex patients have higher length of stays than others (Appendix I, Figure G).

 The effect of low control over patient arrivals causing lagged output is also enhanced because the spread of patients can only partly be regulated (patients are diverted to the GP and re-evaluations of patients happen during mornings). Utilizing Queuing theory principles, increases in these types of variability increase the likelihood of system congestion at a fixed capacity.

2. Delay in availability of physical resources

In the work-sampling it was already visible, demand for physical resources increases during the afternoon as more patients arrive (figure 4.3). The availability of these physical resources can cause delays in output and therefore increase likelihood of lagged output.

 Co-dependency exist between departments exist because they make sequential use of the resources.

(25)

Master thesis M. ten Have, University of Groningen, 2016. 25 as lab and radiology are shared by the whole hospital. This would explain the increased amount of observations during the afternoon of waiting for ‘additional research’.  Delay in pick-up of patients caused was found in the interviews and observation to be

influential to discharge of patients. 19% of all patient waiting was related to the discharge of patients. Waiting for pick-up of patients could take long around 12:00 – 13:00 as nurses of departments were on lunch break. Also hospital occupancy increased during the day and therefore it became less easy to find rooms for a patient, as observed in the study of Cooke et al. (2004) & Forster et al. (2003). Also the 3 percent point increase in the work-sampling data is an indication of this phenomenon (Figure 4.8).

 The decrease in availability of resources is also visible when looking at the (plaster) room. Due to crowding, waiting for rooms was perceived more during afternoons (Figure 4.8).

3. Low responsiveness

In the morning ED-personnel is unable to keep up with the ideal output rate. But during the afternoon an increase in output rate can be realized, which is not reached during the morning. As therefore explained capacity is not the only factor in the existence of lagged output (Figure 4.1). Low responsiveness consists of multiple components:

 In the interviews physicians said that visibility of patient flow influenced their perceived urgency (Appendix III, Interview 1-4).

 Perceiving urgency can cause incentive to speed up the service rate. The fact that they are unable to create this output earlier in the ED can be caused by physicians not perceiving urgency to treat patients.

 Also attention to time (time of patient stay, time of day) was found as influencing on how one thought of the flow of patients and the amount of urgency perceived.

 At moments of crowding, attention shifted to throughput in the form of ‘freeing up rooms’ rather than only on medical treatment and processes (Appendix III, Interview 3).

4. High waiting for human resources

Output starts lagging when due to crowding, patients need to wait longer for human resources such as physicians and specialists. At this point in time capacity is not sufficient to prevent delays caused by human resources.

 Professional variability causes variation between days in amount of congestion as some do better than others in coordination, planning and attention to time (Appendix I, Figure G). The difference between average length of stay generated per physician differs 26 minutes between the fastest and slowest physician.

 As nurses, specialists and physicians are increasingly busy during the day they might miss out on consulting each other sooner causing waiting for each other more often. In one interview a physician explained:

(26)

Master thesis M. ten Have, University of Groningen, 2016. 26 The likelihood of missing out on each other increases as the ED gets crowded as ED-personnel is increasingly busy. Especially availability of the ED-physician can be impediment as this is the main gatekeeper for decisions regarding patients.

 Last it was found both in the work-sampling analysis and the interviews that some specialist respond slower to phone calls and have less time to visit the ED in the afternoon. The lower the urgency they perceive, the more likely they finish other business first. Incentivizing by communicating urgency to specialists only occurred by physicians in the ED during very busy times.

5. Low amount of communication.

Flow through the ED is dependent on how well actors work together and communication facilitates the process of delivering care. Several signs of hindrances in communication were found in the case hospital.

 Firstly, physicians get no notice when results are in, they have to hit F5 and wait by the computer to be receive them. Therefore there is a delay in the time results are in and the time when they are actually noticed. Noticing these can be take longer when physicians are busy.  Secondly, in the case hospital ED there are no rules for when to communicate changes in

patient status. The lack of formalization and forgetting to signal that additional research must be initiated, causes increased waiting times. Especially as nurses and physicians are

increasingly busy during the day (Figure 4.7) more strain is put on these resources and this induces stress and forgetfulness.

 Thirdly, as mentioned before, ED-physicians start to communicate extra to triage about freeing up rooms when crowding already starts and arrivals are visible on the board (Appendix III).

These hiccups in communication can have an increased hindering effect on output and flow during busy times. Forgetting to sign up a patient for CT in the early morning has less severe consequences for new patients regarding room availability then in the peak of arrivals.

.

6. Low coordination of work

In the case ED the way work was divided was seen as influencing the amount of output in the ED.  Especially when multitasking it was reported that length of stay increased (Appendix III). As

arrivals increase, with the same capacity, multitasking and thus the lag in output increased. The arrival of the second shift that overlapped the first decreased multitasking and thus length of stay (See 16:00 – 17:00 in Figure 4.3). Sometimes physicians even arrived earlier than their shift started to help out.

 The increased attention to throughput that happens when the ED gets crowded caused ED-physicians to start dividing work when multiple patients had no physician assigned.  Additionally before crowding happens, it was perceived that physicians take on any patient

(27)

Master thesis M. ten Have, University of Groningen, 2016. 27 some physicians had only one patient, another had four patients. Coming back to the

multitasking-effect, this unbalanced divide can increase throughput and patient waiting.

(28)

5. Discussion

This study found multiple explanations for the difference between the target output and the actual output in the morning. Six causes have been identified to have an influence on the size of this lagged output in this case study. These include low control over patient arrivals, delay in availability of physical resources, increased waiting for human resources, low responsiveness, low amount of communication and low coordination of work. The following section will discuss the main findings of this research.

5.1 Limits to generalizability of the case

The case study consisted out of one ED, in the north of the Netherlands. Therefore little comparison can be drawn between different hospitals, causes present at the case study ED might not be present elsewhere, or in a lesser form. Nevertheless in multiple case studies lagged output was found in the time-series data (Gelissen, 2016; Westerhof, 2016). Therefore the phenomenon studied is not unique and main causes found in this case study are potentially also present in other EDs. However the influence of some main causes can differ because of different characteristics of other EDs. For example an ED in a research hospital has a higher variety of patients and thus patient complexity is higher, affecting control over arrivals and thus the output. It also must be taken into account that the case study ED was a small-sized ED. According to Morton & Bevan (2008) smaller EDs tend to have more patients per physician than larger ones. Therefore the effect of multitasking on output lag can be smaller or even negligible in a larger ED. However adversely this could affect perceived urgency in larger EDs as physicians have less patients and are generally with more they might be less inclined to service promptly.

The work-sampling method that was utilized to research what exactly patient waiting consists of. In the study 73,5% of patient time consisted of waiting, this is similar to the finding of McCarthy et al. (2014) that 75% of patient time is not spend on treatment. Also similar to Cooke et al. (2004) & Forster et al. (2003) this study found a relation between waiting for admittance, time of day and hospital bed occupancy. Next to waiting for a physician, waiting for admittance is the second most frequent measured waiting time. Also delays in availability of resources are insuperable because both human and physical resources are often shared with other departments in the ED. Optimization could include heuristics employing priority for the ED or the use of resources that are only meant for the ED. For example; usage of dedicated resources decreased length of stay in Hyer et al. (2009). However this can mean idle physicians or suboptimal use of machines such as the CT-scan. As these are expensive resources, this can be inefficient for smaller hospitals.

5.2 Country limitations

(29)

5.3 Methodical limitations

The methods used in these case study included semi-structured interviews, work-sampling, time-series analysis, and direct observation. Below the main limitations of these methods in the case study are discussed.

Firstly the recording of the interviews could have had an influence on the interviewed physician, in the form of selective provision of information or giving favorable answers to the interviewer. Secondly, the semi-structured nature of the interview allowed for different questions than prepared in the protocol. Some questions asked by the interviewer could be interpreted as confirmation questions instead of open questions. Thirdly, when asking consecutive questions interviewer bias could have occurred, by asking question directly related to the phenomenon. Lastly as time of physicians was limited, interviews had a maximum of 30 minutes, this hindered more in-depth questioning. This was obviated by many conversations with nurses, physicians, and the secretaries during direct observation. The use of direct observation to gather data was limited as they were subject to interpretation and time limits. However direct observation resulted in an understanding of the processes in the ED necessary to put the findings in perspective. However the presence of an observer can have an effect on output. For example the workers in the Hawthorne experiments showed an increase in output when being watched (Gillespie, 1993; Roethlisberger & Dickinson, 2003). This effect could also have been stimulated by showing the four interviewed physicians graphs of output performance (Appendix II) halfway through the research. Similarly the presence of the observer during the collection of work-sampling data and during direct observation could have had an influence on work speed and methods. The 1030 data points of work-sampling were mainly collected between 11:00 and 17:00. The amount of collected data-points collected in the morning was lower than in the afternoon, making morning findings less reliable than findings of the afternoon. This could have impacted the representativeness of the study. Additionally, the work-sampling relied on reporting physicians and nurses giving account of the current situation. This was obviated by using multiple sources: in case of doubt another nurse or physician was asked what the status of the patient was. If this was not achieved, the status was not written down.

The time-series analysis used to draw comparisons between mornings and afternoons split the data similarly as the work-sampling data was split. The graph with low arrivals in the morning can possibly be distorted as on weekends patients arrive later in the day. The peak in patient arrival tends to arrive later in the day during weekends (Appendix II). Possibly these days can be overrepresented in the graph with low arrivals.

5.4 Recommendations for practice

(30)

Master thesis M. ten Have, University of Groningen, 2016. 30 Low responsiveness, low coordination of work, and low communication are all related to perceived urgency. The point at which is this perceived is different for everyone. For practice is recommended to install a way of signaling increasing arrivals to ensure that ED-personnel perceives this in time. Then as soon as patients are signed on they can be evenly divided between personnel to even up the workload. In perceiving this crowding, visibility of patient flow is key. This study would like to draw attention to output factors and flow in the department on a daily basis to establish a sense of urgency to service. Next as an increase in waiting for discharge in the afternoon was measured, this can increase lagged output. Therefore the author recommends to prioritize outflow of patients at any point in time. In that way there is less need to ‘free up rooms’ when suddenly many patients arrive..

Another way to enhance flow of patients is signaling of results. Waiting for lab, CT and other

additional research must be noticed by physicians when they arrive. In practice this can mean a pop-up notice on screen when results are visible. Lastly, as there are different ways of dividing work and coordinating flow, recommended is to employ the best practices principle. Physicians can learn from each other to find the best way of handling flow and use standard heuristics to formalize the best way of coordinating flow. For example a rule can be established that the coordinating ED-physician only takes on a certain amount of low-complexity patients. This can result in increased availability of the ED-physician for assistants/trainees/nurses/general practitioners/specialists to consult.

5.5 Further research

(31)

6. Conclusion

This case study researched the additional waiting times caused by lagged output in the ED. This conclusion summarizes the main results of the case study.

The arrival pattern of the case ED is found subject to weekly and daily seasonality. The phenomenon of lagged output is found to be present in the case ED. By utilizing interviews it was found that additional capacity is speeding up output in the form of overlapping shifts and an additional physician during peak hours. Low responsiveness in the case ED is tied to low flow visibility causing little incentive to increase service rate. Urgency must be perceived for ED-personnel to start coordination of work, increase communication and to establish focus on throughput. As physicians have more patients during the day, multitasking multiple patients can delay output and thus increase lagging by

unchanged capacity. The work-sampling method identified waiting for physician, waiting for hospital admittance and waiting for lab results as the top three causes of patient waiting. The main increase in frequency between observed causes of waiting in mornings and afternoons is waiting for human resources. With the increase of waiting for human resources, waiting for physical resources no longer hindered output and decreased. In the afternoon, waiting for discharge and rooms was also observed to be more frequent.

(32)

References

Batal, H. et al., 2001. Predicting patient visits to an urgent care clinic using calendar variables.

Academic Emergency Medicine, 8(1), pp. 48-53.

Bergs, J., Heerinckx, P. & Verelst, S., 2014. Knowing what to expect, forecasting monthly emergency department visits: A time-series analysis.. International emergency nursing, 22(2), pp. 112-115. Bloor, M., 1976. Bishop Berkely and the adenotonsillectomy enigma: An exploration of variation in the social construction of medical disposals. Sociology, 10(1), pp. 43-61.

Cayirli, T. & Veral, E., 2003. Outpatient scheduling in healthcare: a review of literature. Production

and Operations Management, 12(4), pp. 519-549.

Chin, L. & Fleisher, G., 1998. Planning model of resource utilization in an academic pediatric emergency department. Pediatric Emergency care, 14(1), pp. 4-9.

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), pp. 1177-1185. Cooke, M. W., Wilson, S., Halsall, J. & Roalfe, A., 2004. Total time in English accident and

emergency departments is related to bed occupancy. Emergency Medicine Journal, 21(5), pp. 575-576. 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), pp. 63-68.

Dijk, T., 2013. How can the throughput time of patients in the Emergency Department be Improved, Groningen: University of Groningen.

Diwas, S. & Terwiesch, C., 2009. Impact of Workload on Service Time and Patient Safety: An Econometric Analysis of Hospital Operations. Management Science, 55(9), pp. 1486 - 1498.

Draeger, M. A., 1992. An emergency department simulation model used to evaluate alternative nurse

staffing and patient population scenarios. s.l., ACM, pp. 1057-1064.

Fitzsimmons, J. A. & Fitzsimmons, M. J., 2011. Service Management: Operations, Strategy,

Information Technology. 7 ed. New York: McGraw-Hill.

Forster, A. J. et al., 2003. The effect of hospital occupancy on emergency department length of stay and patient disposition. Academic Emergency Medicine, 10(2), pp. 127-133.

Fredendall, L. D., Craig, J. B., Fowler, P. J. & Damali, U., 2009. Barriers to Swift, Even Flow in the Internal Supply Chain of Perioperative Surgival Services Department: A Case Study. Decision

Sciences, 40(2), pp. 327-349.

Gardner, R. L., Sarkar, U., Maselli, J. H. & Gonzales, R., 2007. Factors associated with longer ED lengths of stay. The American journal of emergency medicine, 25(6), pp. 643-650.

Explaining lagged output in the ED, a case study on patient waiting.