Buffers at Emergency Departments: Explaining the Influence of Buffer Mechanisms on Emergency Departments’ Responsiveness to Volume Variability

(1)

Buffers at Emergency Departments: Explaining the Influence of Buffer

Mechanisms on Emergency Departments’ Responsiveness to Volume Variability

Ilse Grasmeijer

S2365200

Faculty of Economics and Business

University of Groningen

Master Thesis Supply Chain Management

Supervisors:

Dr. M.J. Land

Dr. J.T. van der Vaart

(2)

Abstract

Emergency Departments have to provide acute and unpredictable care. Because of this unpredictability, Emergency Departments should be able to respond to variability in patient inflow and make sure that they can deliver the needed care within limited time. Different theories have proposed the existence of buffers in healthcare in order to catch up with variability. A single case study is conducted in order to investigate those buffers in the unpredictable setting of an Emergency Department. The focus will be on the capacity, processing time and time buffer. This research has found evidence of the existence of all proposed buffers at the investigated Emergency Department and further investigated how these mechanisms work. The capacity buffer and processing time buffer are positively influencing the responsive capability to volume variability. On top of that, the importance of a flexible capacity buffer is pointed out. According to this research, the time buffer can be seen as a lack of responsiveness to volume variability by having insufficient buffers or misusing the available buffers. Findings are supported with both quantitative data from measurements and qualitative data from experienced ED-physicians in order to better understand the influence of buffers on the responsiveness. Better understanding of the buffer mechanism will help to increase this responsiveness, which will positively influence the total patient flow and consequently the patient satisfaction.

(3)

1. Introduction

This paper investigates which mechanisms Emergency Departments use to absorb variability in inflow volume. When looking at patient inflow in Emergency Departments, this daily patient inflow can vary from 50% less than average to 100% above average patient inflow (Gelissen, 2016). During some days, less patients come to visit the Emergency Department, while other days are overcrowded and have a larger inflow compared to quiet days. According to Higginson et al. (2011) acute care is unplanned and unpredictable, which makes it hard to determine the necessary resources for providing care to patients. However, investigated Emergency Departments in the Netherlands (Gelissen, 2016; Koomen, 2015; Van Achteren, 2014) all seem to have somehow the ability to respond to this varying patient inflow. This responsiveness is up to a certain height, since research of Ten Have (2016) showed that a delay in output occurs, especially during mornings with a high number of patient arrivals. This indicates that Emergency Department’s responsiveness is not sufficient during these moments. What mechanisms are influencing Emergency Department’s responsiveness and what factors determine limitations to this responsiveness, will be investigated in this research. A better understanding of these mechanisms can help to improve the responsiveness to volume variability and can help to reduce waiting times during busy moments. A crowded Emergency Department, where patients have to wait for their care, negatively influences the clinical outcomes of patients via several manners, including higher complication rates and mortality (Pines et al., 2011). Better understanding and further optimizing Emergency Department’s ability to respond to variability in volume will therefore positively influence patients’ outcomes.

(6)

to a new type of buffer, which could be used by service providers, adjusting service processing time in order to cope with variability (Hopp et al., 2007 and Roemeling et al., 2017). This buffer in processing time is proposed as being the third buffer in a healthcare environment. A recent study of Roemeling et al. (2017) suggested the definition of processing time buffering as “the mechanism through which processing times are reduced by removing critical and/or non-critical tasks and/or increasing task speed in response to variability”. Despite the strong theoretical evidence for the phenomenon ‘processing time buffering’, this buffer is not empirically established yet.

This study aims to 1) generate better understanding of the (buffer) mechanisms as a solution to inflow variability at the Emergency Department and II) contribute to literature on buffers, and thereby focus in specific on the existence of the proposed phenomenon ‘processing time buffer’. For the first purpose, this study aims to get more insight in capacity, time and processing time buffers in Emergency Departments. Literature has shown the existence of these buffers in healthcare separately, but we do not know how to use these buffers, in what situation and whether they can be combined or not in this specific acute care setting. Better understanding of these buffers can help to explain why Emergency Departments can keep up with an increased inflow at some days, while at less busy days long waiting times may exist. Next to the buffer mechanisms, this research will investigate whether there are other mechanisms which influence the responsive capability as well. For the second purpose, this research will strive for a better understanding of the buffers used at the Emergency Department. This insight can subsequently help in order to better understand buffers in other circumstances. The processing time buffer will be empirically investigated in order to find out whether this phenomenon actually exists and if so, how this buffer functions in achieving an increase in output rate at the Emergency Department. In order to fulfill both purposes, this research aims to answer the research question:

What (buffer) mechanisms influence the ability to respond to variability in patient volume in Emergency Departments?

(7)

2. Theoretical background

Several authors (Andersen et al. 2014; Radnor, 2010; Mazzocato et al., 2010) are convinced that dealing with variability in healthcare is underexposed as research theme. More attention has been paid towards tools and techniques in order to improve healthcare performance. This chapter will summarize most important findings regarding variability and buffers in hospital environments. First, some general characteristics of Emergency Departments are given. Thereafter, variation at the Emergency Department will be discussed, with in particular volume variability. Lastly, the new proposed processing time buffer will be explained, followed by a research model in which literature findings are combined.

2.1 Characteristics Emergency Department

If acute care is required, a patient often goes to the Emergency Department. According to Higginson et al. (2011) we can qualify acute care as unplanned and unpredictable, which makes it hard to determine necessary resources to treat Emergency Department patients. Emergency Departments often have a similar general lay-out, consisting of a waiting room, triage desk, trauma room and several treatment rooms. After arrival, patients are assigned to a specific triage level at the triage desk, where often the Manchester triage system is used as standard (Mackway-Jones et al., 2013). The Manchester triage system has five levels of acuity, and each level represents a certain level of priority for the patient and the resources which should be allocated (Anneveld et al., 2013).

Table 1.1 - Manchester Triage System levels of urgency

Depending on the triage level, a patient should be examined by a physician within established norms. Every urgency level has a time target until patient has to be seen by physician, see table 1.1. After first seeing the patient, the physician can decide to take some additional tests in order to come up with the diagnosis, for example medical imagining or lab tests. After examination, the treatment policy is formed and patient gets treated at the Emergency Department. After treatment discharge of the patient follows, which can be an admittance to the hospital, a referral to the outpatient clinic, a discharge with no check-up, a discharge with a check-up from a general practitioner, a referral to another hospital or, in case a patient dies, a mortuary (Ten Have, 2016). Specific characteristics of the case hospital are given in Appendix I.

2.2 Variability at the Emergency Department

We can identify three different types of variability, which are relevant for healthcare: (1) flow variability, which is related to the arrival of patients, (2) clinical variability, which is related to the degree of illness or difference in response to a particular treatment; and (3) professional variability, which is related to the different capabilities of medical professionals (Litvak and Long, 2000). Next to these so-called types of natural variability, artificial variability exists (Litvak et al., 2005). Artificial variability is created by own actions, and is therefore controllable (Roemeling et al., 2017). Emergency Departments are dealing with variability in patient inflow, since it is concerned with acute care, which is unplanned and unpredictable (Higginson et al., 2011). The abovementioned

Level Urgency Code Target until seen by physician

Level 1 Immediate Red 0 Minutes

Level 2 Very urgent Orange 10 Minutes

Level 3 Urgent Yellow 1 hour

Level 4 Standard Green 2 hours

(8)

types of variability make it hard to predict patient volume and patient mix at the Emergency Department. Because of this, it could be that Emergency Departments become overcrowded and consequently, patients are queuing for care. This is called the phenomenon of crowding, and worldwide Emergency Departments experience an increase of this problem (Schiff, 2011). Crowding at Emergency Departments can be defined as ‘a state where care demand exceeds available resources, resulting in long waits for tests and treatments’ (Moskop et al., 2009). Research of Stang et al. (2015) has concluded that patients in a crowded Emergency Department encounter important delay in care, sometimes for life-saving interventions, higher rates of errors and adverse events, and poorer survival. For hospitals, it is highly important to find a way to prevent crowding and thus find a way to deal with the earlier mentioned forms of variability.

2.3 Variability in patient inflow

Whether it is a busy moment or not is mainly influenced by the incoming patients. Their time of arrival, degree of illness and personal characteristics can vary. As mentioned by Asplin et al., (2006), the variability in patient inflow influences the degree of crowding most at the Emergency Department. Therefore, this research will focus on this type of variability. However, investigating this type of variability as a whole, exposes immediately a challenge of research in healthcare, namely that fact that besides arrival pattern, also patient mix, i.e. degree of illness or personal characteristics, differs. This gives implications for the generalizability of such research. As mentioned in the introduction part, the volume variability can be strengthened or weakened by patient mix, which is the variability in type and degree of illness. On quiet days, a low volume of highly complex patients, could ask for a larger capacity of resources compared to a busier day with relatively simple patients. An attempt to quantify a patient’s degree of illness is done by applying triage-codes, as described above. However, similar codes can still differ a lot and the perfect quantification is not identified in literature yet. Higginson et al., (2011) have conclude that the combination of the dynamic arrival pattern, diversity and high complexity of patients make acute care unpredictable and therefore difficult to plan necessary staff capacity. An attempt to describe the dynamic arrival pattern at the Emergency Department is given below.

(9)

2.4 Responsiveness to variability

In order to be able to deliver care on time, Emergency Departments should somehow be able to respond to this variability in demand. Responsiveness can be regarded as ‘the behavior or result of the system with respect to task being performed in a timely fashion’ (Gläßer et al., 2009). When applying this concept towards healthcare settings, Jones et al. (2008) defined responsiveness as ‘to what extent health care providers are strategically driven to respond to fluctuating levels of patient demand for their services’. Research of Ten Have (2016) suggests a difference in the responsiveness in the morning versus the responsiveness in the afternoon. Whereas in the morning the studied Emergency Department faced a delay in output rate, an accelerated service in the afternoon was possible. Suggested causes of this higher responsiveness were increased capacity, perceived urgency among physicians that can cause an incentive to speed up service rate and an increased attention to throughput in the form of ‘freeing up rooms’ (Ten Have, 2016). Other possible causes, such as other types of tasks in the morning compared to the afternoon and having a different way of work coordination, are also conceivable.

2.5 Buffers in healthcare

(10)

2.6 Buffer in processing time

Although literature supports the existence of capacity and time buffers, recent literature has argued that these two buffers are not the only two buffers used to respond to volume variability in Emergency Departments. Hopp et al. (2007) concluded that next to time and capacity buffers, quality buffers could be considered as additional buffer, which is also applicable in service environments. When actors have discretion over their task, they can increase or decrease the quality of delivered service in response to potential idle capacity or imminent congestion (Hopp et al., 2007). This phenomenon of increasing working speed among employees in production environments was already investigated in 1998, in which was concluded that workers in production systems tend to work faster or slower when they were aware of the actual WIP (work in process) inventory (Schultz et al., 1998). More recently research of Diwas and Terwiesch (2009) has investigated whether this phenomenon also exists in a service environment. They concluded that care providers are able to increase their productivity at busy moments, by speeding up their service rate. Batt and Terwiesch (2016) also describe the possibility of speeding-up work pace via several mechanisms to absorb an increase in patient inflow, ranging from working harder, to truncating the work done for the customer, and multitasking. Thus, the presence of another mechanism that can catch up variability by speeding up the work speed of care providers is supported by multiple studies. Where Hopp et al. (2007) has mentioned that actors can increase or decrease quality of service, more recent research of Roemeling et al. (2017) has shown that reducing processing time does not necessarily lead to lower perceived quality. These authors argue that the quality of a service can be seen as a possible consequence of processing time adjustment rather than being directly affected, which make them define the third buffers as follows: ‘Processing time buffering is the mechanism through which processing times are reduced by removing critical and/or non-critical tasks and/or increasing task speed in response to variability.

2.7 Summary

In this review of literature, we have seen that individual buffers are described as mechanism in order to respond to volume variability at the Emergency Department. However, specifications of how these buffers are used, in what scenarios and whether it is possible to combine the buffers, are not mentioned. We can summarize this in the following research model, whereby volume variability forces Emergency Department to use certain buffer mechanisms in order to be able to respond to volume variability. How these buffer mechanisms are used and their particular influence on Emergency Department’s responsiveness, is not known yet and will be further investigated in this research.

Figure 1.2 - Research model

Increase in incoming patient volume Responsive capability of Emergency Departments Capacity buffer

Processing time buffer

Emergency Department’s Responsiveness

(11)

(12)

3. Methodology

In order to investigate the role of buffers in Emergency Departments’ responsiveness to volume variability, a single case study will be conducted. This case study will be conducted in Nij Smellinghe, a small-sized hospital in the north of the Netherlands.

3.1 Case description

In the following paragraphs, the explanation for a single case study and the investigated hospital is given.

3.1.1 Single case study

According to Meredith (1998) a case study enables the researcher to study a phenomenon in its natural setting, which makes it possible to generate a relevant theory from the understanding gained through observing actual practice. Observation of natural settings is important, since experiments usually deliberately separate the phenomenon from the context and isolate it from the natural context, while case study research is focused on the embeddedness of a phenomenon in real-life context (Blumberg et al., 2011). Furthermore, Karlsson (2009) mentioned that the case method is a way to get full understanding of the complexity and nature of a phenomenon. Although we (partly) know that the buffers exist separately, we do not know what their role is, what their effects are and how they interact in healthcare setting. Since a case study is desirable in early, exploratory investigations where the variables are still unknown and the phenomenon not at all understood (Meredith, 1998), a case study is conducted. We could say that this case study has theory building on the buffering mechanism in hospitals as purpose, since we want to identify the role of the variables and also the linkages between those. According to the research purpose overview of Handfield and Melnyk (1998), research with theory building as purpose, needs to have a research structure with few focused and in-depth field studies. Karlsson (2009) mentioned that the fewer the case studies, the greater the opportunity for depth of observation. In order to be able to investigate in-depth the role of an Emergency Department’s buffering mechanisms and their interdependence, a single case study has been chosen as a research method. However, we have to take into account that a case study limits the generalizability of conclusions. There is a risk of misjudging the representativeness of a single event or exaggerating easily available data (Karlsson, 2009). In order to overcome this risk, multiple sources of data are used to reduce the effect of different interpretations or viewpoints (Karlsson, 2009).

3.1.2 Case study – description hospital

(13)

hospital to assist at the Emergency Department. This additional capacity is hard to observe and therefore other hospitals are less suitable for this exploratory research compared to the selected case hospital.

3.2 Data collection

In order to describe the data collection method, we first distinguish the different stages of this research. As mentioned by Karlsson (2009), the collection of data in case research is characterized by triangulation, which means the use and combination of different methods to study a particular phenomenon. A description of the data collection methods in each research stage will be given in the next paragraph. In paragraph 3.2.2 will be explained how data is collected and analyzed per method.

3.2.1 Research stages

Research stage I – Understanding and Developing

During this stage, the current course of action at the Emergency Department will be observed in order to understand the flows and look for the existence of the capacity, time and processing time buffers. The focus will be on the differences between quiet and busy moments with regards to these buffers by observing physicians’ activities, patient waiting time and time spent on a patient treatment. During this stage, several data collection templates will be developed in Excel. First drafts are based on previous studies of Ten Have (2016), Gelissen (2016), and Van Achteren (2014), and will be adjusted in order to be able to clarify differences between busy and quiet moments. The data collection method for processing times is based on descriptions of the phenomenon by Roemeling et al. (2017). After making some drafts, the methods will be tested. This is done in order to investigate how buffers can be measured most precise, to avoid mistakes during data collection and to gain some experience with data collection. Next to gaining an understanding by observing, historical data will be analyzed. The dataset is provided by the hospital itself and contains data of all incoming patients at the Emergency Department in 2016. Which analyses are done, will be described in section 3.2.2. In conclusion, this stage aims to understand the current course of action by observing the Emergency Department, and developing the most suitable data collection template, by testing several developed templates.

Research stage II – Measuring

After the measurement instruments are developed, specific measurements will be made in order to investigate the different buffers. To measure these buffers, an operationalization of each buffer is given below.

Capacity buffers – From literature, a capacity buffer is defined as resources waiting for a job (Roemeling et al., 2017). In order to measure when physicians are waiting for a job, the capacity buffer is operationalized by measuring the time spent on non-work-related activities. This will be measured by a work sampling study. Since physicians will always need some time for personal activities, regardless of the number of patients at the Emergency Department, this research will look at the deviation of the average amount of time spent on non-work-related activities. In case there is an increase in time spent on these non-work-non-work-related activities, we can speak of a capacity buffer. Additional capacity is not considered as capacity buffer, since there is no flexible capacity in the observed hospital, so capacity is according to schedule.

(14)

removing tasks or increase task speed. The existence of this buffer can therefore be proven in case the processing times as described above, are reduced during busy moments. Measurements will be done on treatments executed by the physicians, and a more detailed description will be given in the coming paragraph. Patient treatments are chosen as activity to measure, since this activity has most visible start and end, whereas other activities, such as administrative tasks, preparing patient reports or diagnosing a patient, are often executed simultaneously and are hard to distinguish.

Time buffers – Patients spend often a part of their time on waiting for diagnosis or treatment. This specific time will be measured by making use of patient arrival and departure time using historical data, i.e. a patient’s LOS. From this dataset can be concluded whether patient’s LOS increases in case more patients are present at the Emergency Department. This increase in LOS will be considered as additional time spent on waiting, and thus as time buffer. Hereby is assumed that it is unlikely that a longer LOS is caused by longer treatment times, since historical data of more than 13,000 patient visits is used for those measurements.

After collection, the data of the work sampling method and processing time measurements will be analyzed and combined with findings from historical data. Data analyses can be done by making plots and (pivot) tables and conducting regression analyses and a T-test, which will be further elaborated on in section 3.2.2. Aim of this stage is analyzing results in order to interpret the collected data and come up with results related to the research question.

Research stage III – Validating

During this stage, the results will be discussed with physicians by conducting several interviews. This stage aims to validate results and to get a better understanding why things happen, for example the reason for an increase or decrease in a certain activity measured by the work sampling method. In the next paragraph will be explained how those interviews will be conducted.

3.2.2 Data collection method per research stage

As mentioned by Voss et al. (2012), reliability of data increases if multiple sources of data on the same phenomenon are used. In this section, a more extended description of the data collection methods that will be used during the research stages is given, followed by the tools that will be used to analyze the data. Figure III gives an overview of the data collection method per research stage.

Table 3.1 - Overview data collection and analysis methods per research stage

Research stage Data source Data analysis

Understanding and Developing

Direct observations and historical data from IT system

Input/output diagrams, Boxplots on patient arrival patterns, regression on WIP and corresponding LOS.

Measuring and analyzing

Structured observations by using work sampling method and measuring processing times for each physician

Determine the proportion of time spent on activities, comparison between busy and quiet moments and between different physician types. Processing time will be analyzed by boxplot on average time spent on a patient’s treatment, plus regression on number of patients a physician is responsible for and corresponding length of treatment.

(15)

Direct observations - During the start of this research, direct observations will be made of the processes at the

Emergency Department. According to Meredith (1998), direct observations are important in order to investigate the role of the context in which the phenomenon occurs, and therefore helps to understand the how and why elements of the phenomenon. During these direct observations, the observant will get familiar with the processes at the Emergency Department. Attention will be paid to differences between quiet and busy moments. The observant will make notes, in case an observation may contribute to the explanation of the different phenomena.

Historical data from the IT System - For analyzing patient inflow and outflow, a dataset of the hospital will be

used. This dataset consists of timestamps for all patients arriving at the Emergency Department between the 1st_of

January 2016 until the 31st_{of December 2016. Of each patient’s visit the following data will be collected: patient}

number, triage color, patient’s origin (i.e. by whom the patient is referred), the responsible physician, specialty of diagnosis and several times: registration, arrival, triage, first seen by physician, and discharge time. After observations and explanations of how data is collected by the physicians and nurses, it became clear that time of registration, arrival and discharge were filled in correctly. Other registrations, such as time of triage and first seen by physician were filled in inaccurately, and will therefore not be taken into consideration. When checking the dataset on errors, 10 patient-visits are removed. For those patients, time of arrival was later than time of discharge, which indicates an error in measurements. After deleting those items, the dataset consists of 13,888 measurements. Some analyses will be done on this dataset. First, an input/output diagram will be made, in order to observe the flow pattern at the Emergency Department. Input/output diagrams give the average number of patients arriving or leaving at a certain moment in time. The average number of patients arriving and leaving will be plotted for every hour of the day. Data consists of half an hour before the certain moment in time until half an hour after this moment in time (for example, 8.00 AM consists of data from 07.30AM until 08.30AM). In specific, attention will be paid to the angle of inclination of both graphs, in order to see if the inflow diagram can ‘follow’ the outflow diagram by having a similar steep graph.

Besides input/output diagrams, some analyses on the distribution of patient volume will be performed. A bar chart with the frequency for every patient volume gives a graphically display of this distribution. More in-depth analyses on the distribution of patient volumes related to weekdays and months could give an indication whether monthly or daily patterns exists. Hereby, we have to make the remark that monthly patterns are hard to indicate, since the dataset consists of one year, i.e. all months once, which makes it hard to distinguish a monthly pattern.

(16)

This relationship between WIP and LOS will also be used in order to make a distinction between busy and quiet moments, that could be used during the work sampling method. Different dummy variables will be created for ‘busy’ and ‘quiet’. This will be done multiple times, with different number of patients in the process as split for ‘busy’ and ‘quiet’. After creating the dummies, T-tests will be conducted in order to see whether there are significant differences between the groups. The split for which the T-test showed most significant differences between busy and quiet, will be used as distinction in this research. Historical data will be used to support this choice, since we are looking for the most natural distinction between busy and quiet. If work sampling data would be used, we would already influence the differences between busy and quiet in the collected work sampling data. By using historical data, a natural distinction can be made, without influencing work sampling results on forehand. Lastly, from historical data will be checked whether a day consists of ‘easy’ or ‘difficult’ patients. As already mentioned in the literature section, a day with few patient arrivals can be experienced as busy by the physicians, since the few patients that arrive are relatively ‘difficult patients’. As described, the triage code tries to make a classification in difficulty of a patient. However, after discussing this with the case hospital, it became clear that this is not a perfect classification, since there is still a lot of variation within the triage codes. Because there is no better classification designed in literature yet, this research will make use of these codes, but not as the focus of this research. The focus of this research is the variation in patient volume. In order to prevent that patient-mix influences the results, triage codes will be checked to see whether that particular day can be considered as a “normal day” or it should be excluded from analysis, for example by extremely rare or difficult patients. This decision will be made in consultation with the physician on duty.

Work sampling method - Work sampling is a method which makes an estimation on the percentage of time an

employee spend on predefined work tasks (Kirwan and Ainsworth, 1988). A large number of observations of the physicians will be made by the observant at the Emergency Department. The observant will ask what activity each physician is working on every time-interval. For these observations, a time-interval of 30 minutes will be used, which means that after every 30 minutes the observant observes the tasks that all physicians are working on at that particular moment. Each observation consists of the category to which the observed physician belongs (ED-physician, physician-assistant or physician-in-training), the observed task, and information which will help to determine whether it was a busy moment or not, consisting of number of rooms that are occupied, triage codes of those patients and the number of patients in the waiting room. Measurements will be conducted until 15:30, since this is the end time of the day-shift. Measuring during the handover would deliver unreliable results, since there is additional capacity during the handover. After the handover patients were transferred to another physician, who can decide to (partly) redo some tests, which was depending on previous physician and transition. This would also influence the results.

(17)

Measuring processing times for each physician - Processing times will be measured by the time between the start

and the end of a patient treatment. The start of a treatment will be measured by the time a physician leaves the coordination room. The end of a treatment will be measured by the time a physician enters the treatment room again. As discussed with the physicians of the case hospital, after every treatment a physician returns to the coordination room, either to write a patient report or to gain information of a new patient. There is explicitly chosen for the coordination room as location of observations, since it is for the observant possible to follow the conversations of the physicians, which will provide insight into the treatment a physician is going to execute. On top of that, the coordination room provides the best overview of the corridors, and in case the observant missed a physician that leaves the room, security cameras with a delay of half a minute give the opportunity to follow the physician. Measurements will be done during several days, at which different physicians are on duty. This is done in order to increase generalizability. Each measurement will consist of the time leaving and entering the treatment room, physician’s name, patient’s diagnosis, type of treatment (for example first physical examination or discussing results) and the number of patients a physician is responsible for at that specific moment. Last mentioned information will be used in order to determine whether processing time, i.e. time spent on a specific patient treatment, decreases in case a physician becomes responsible for more patients.

Data will be analyzed by plotting the duration of the first physical examination per number of patients a physician is responsible for. Next to that, a regression analysis will be conducted in order to see whether a significant relationship exists. There is explicitly chosen for first physical examination as the focus of the research on processing times. As mentioned by Panella et al. (2003), variability has a critical impact on healthcare processes. Every patient, but also every physician differs, and this influences the duration of the treatment. In order to minimize the variability during the measurements of processing times, the decision to measure one specific treatment as the focus of this measurement has been made, namely the first physical examination by the physician. According to the physicians in the case hospital, the first physical examination is done regarding protocols and predefined work-standards. These protocols stimulate physicians to work in a similar way. Because of this, professional variability (i.e. variability caused by the physician) can be reduced. For other treatments, less protocols are available and the professional variability increases, which would influence variation in treatment times. As mentioned by Roemeling et al. (2017), clinical variability is a purely natural affair, and therefore impossible to influence. This research will try to reduce the influence of variability caused by a patient’s characteristics and degree of illness, clinical variability, by comparing processing times of patients with similar diagnoses.

Semi-structured interviews - In order to validate results, semi-structured interviews will be held among

(18)

4. Results

In this chapter, the results of the case study on buffers at Emergency Departments will are explained. First, information about the flow patterns of the case hospital is given. This is done in order to sketch a frame on volume variability at this hospital. Thereafter, the different buffers are discussed, following the same structure as presented in the research model. We will start with the buffers that belong to the responsive capability of the hospital, the capacity buffer and the processing time buffer. Possible other mechanisms that contribute to the responsive capability of the Emergency Department will be described here as well. Lastly, findings on the time buffer will be discussed.

4.1. Patterns in patient arrivals

Based on the historical dataset, some patterns can already be recognized. In order to give an overview of the daily pattern of this specific hospital, an input/output diagram has been made.

Figure 4.1 - Input and output-diagram case hospital, based on 2016

Figure 4.1 shows the inflow and outflow rate of patients, which means the average number of patients arriving per hour. Measurements are done on whole hours, which means that for example 08:00 consists of data from 07:30 until 08:30. For this analysis we looked at the possible acceleration that can occur in the inflow and outflow rate. This acceleration can be determined by looking at the angle of inclination. In case the angle is the same for blue and red, a similar acceleration is achieved. Next to the inflow and outflow rate, the figure displays the desired outflow rate. This rate is based on the average LOS, which was 137 minutes. We could say that in case the Emergency Department is able to keep up with the inflow rate, the outflow rate follows the same trend as the inflow rate, but roughly two hours later. We see similar angles of inclination between the blue and red dotted line. Until 10:30, the actual outflow rate is higher than the desired outflow rate. After 10:30, we see that the outflow rate stays behind compared with the desired outflow rate and the Emergency Department is not able to keep up with the inflow rate. This lagged outflow rate lasts until 15:00, after which the actual outflow rate becomes higher

(19)

than the desired outflow rate. Between 18:30 and 19:30 and between 23:30 and 00:30 we see again that the outflow rate cannot follow the desired outflow rate.

The daily pattern in inflow/outflow rates is based on data collected over a year. Earlier studies of Ten Have (2016), Koomen (2016) and Gelissen (2016), showed a similar pattern. However, the daily pattern and quantity of patients can vary a lot. In order to show this variation graphically, the following graphs are made.

Figure 4.2 - Distribution of number of patients arriving per day

Figure 4.2 shows the distribution of patients arriving per day. Most patients (62) visited the Emergency Department on the 5th_{of January, fewest patients (21) visited the Emergency Department on the 24}th_{of November and 25}th_of

December. On average, 37.9 patients visit the Emergency Department per day.

Figure 4.3 - Distribution of patient inflow related to weekdays

In order to further understand the patterns at the Emergency Department, figure 4.3 shows the distribution of patient inflow related to weekdays. On average, most patients visit the Emergency Department on Friday (median is 43) and Monday (median is 39). However, this pattern is hard to use as a forecast, since the graph shows large variation for the same day of the week. For example, on Thursdays 26 patients can visit the Emergency Department, but a patient inflow of 46 also occurs.

0 5 10 15 20 25 30 35 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 54 55 56 57 60 61 62 Ab so lu te f re qu en cy

Number of patients (total per day)

Distribution of number of patients arriving per day

20 25 30 35 40 45 50 55 60

Sunday Monday Tuesday Wednesday Thursday Friday Saturday

Nu m be r of p at ie nt s pe r da y

Distribution of patient inflow related to weekdays

(20)

In figure 4.4 the variation between months is graphically presented. A larger patient inflow is visible for March and September, whereas July and August had a smaller patient inflow. However, this graph shows the variation between months based on 2016. Therefore, conclusions on monthly patterns are not drawn yet.

Figure 4.4 - Distribution of patient inflow related to months

4.2 Capacity buffer

From the previous paragraph can be concluded that the observed Emergency Department has to deal with variation in patient inflow. Emergency Departments need to be capable to deal with this variability. In this thesis, we are specifically interested in the role of buffers as mechanisms that respond to the variability.

The first buffer that will be discussed is the capacity buffer. In order to see how the capacity of physicians is allocated, a work sampling study is conducted, as discussed in the Methodology chapter. The data is collected during 140 measurement moments, with an interval of 30 minutes. In total, the observer gained 535 observations, related to 19 different physicians. The percentages of time which were devoted to a specific activity, are graphically represented in figure 4.6. Table 4.5 is confined to the activities with the largest percentages.

Table 4.5 – Percentages five largest activities

The absolute error rate per activity is calculated for a confidence level of 90%, based on calculations of Robinson (2010). For both busy and quiet moments, and all physicians included, the absolute error rate lies between 1.9 and 3.1 percent. For the different physicians, we give an overview of the minimal and maximal error rate during busy and quiet moments. ED-physicians had an error rate between 3.2 and 7.1 for quiet moments, and between 3.5 and

20 25 30 35 40 45 50 55 60 Janua ry Febr uary _March _April May June July Augu st Septe mbe r Octobe r Nove mbe r Dece mbe r Nu m be r of p at ie nt s pe r da y

Distribution of patient inflow related to months

5th percentile 25th percentile Median 75th percentile 95th percentile

Activity Time share (%)

All types Time share (%) ED-physician Time share (%) Physician-assistant Time share (%) Physician in training Patient treatment 25.9 23.2 26.7 26.9

Prepare patient reports 20.3 11.6 19.4 27.6

Non-work-related activities 14.2 12.6 17.2 11.2

Consultation 14.2 16.8 10.0 17.0

(21)

6.9 for busy moments. Physician-assistants had an error rate between 2.7 and 5.4 for quiet moments and 3.3 and 5.7 for busy moments. Physicians-in-training had an error rate between 3.3 and 6.2 during quiet moments and 3.5 and 6.5 during busy moments. Differences occur since there were more quiet moments, which resulted in more data for quiet moments. Between physicians occur differences because most of the day two physician-assistants were working, which makes it possible for them to collect double data.

Figure 4.6 - Work-sampling results per physician category

25,8% 23,2% 26,7% 26,9% 20,2% 11,6% 19,4% 27,6% 14,1% 12,6% 17,2% 11,2% 14,1% 16,8% 10,0% 17,9% 8,0% 16,8% 5,0% 6,0% 4,4% 1,1% _5,6% 5,2% 3,6% 10,5% 2,8% 2,4% 2,1% 3,9% 0,7% 1,9% 2,1% 2,8% 0,7% 1,7% 1,1% 2,2% 2,9% 1,1% _3,3% 3,7% 0,0% 10,0% 20,0% 30,0% 40,0% 50,0% 60,0% 70,0% 80,0% 90,0% 100,0%

All types ED-physician Assistant physician Physician in training

Pe rc en ta ge o f tim e di vi de d to s pe ci fi c ta sk Category of physician

Work sampling results

Gaining patient information Coordination

(22)

Figure 4.6 gives a schematic overview of the time share allocated to each activity. The first bar shows the results for all physicians, the remaining three bars give the results per category of physician. We speak of three categories of physicians:

1) ED-physicians: These physicians have followed specialized education on Emergency Departments and are most experienced. At all time, at least one ED-physician is working, and he or she is seen as coordinator of the physicians and nurses.

2) Physician-assistants: Two types of physicians belong to this category. First, the physicians who are following specialized education in order to become ED-physician. Second, physicians who are graduated as basic-physician, but are not following specialized education yet. Both physicians can work autonomous.

3) Physicians-in-training: These physicians are not graduated as basic-physician yet, but are gaining experience during an internship at the Emergency Department. Physicians-in-training are always supervised by an ED-physician, and in some cases by a physician-assistant as well.

From figure 4.6 we can conclude that most time is devoted to the treatment of patients, with exception of the physicians-in-training. This category spends most of its time on preparing patient reports. For ED-physicians, the time share for preparing patient reports is lower. Furthermore, we see a larger time share for consultation for ED-physicians and ED-physicians-in-training. Lastly, we see that the time share for non-work-related activities is on average 14.1% for all physicians. In order to find out how capacity is influenced by the number of patients at the Emergency Department and better understand their responsive capability, a distinction between busy and quiet moment has been made in the work sampling data. Before showing the differences between busy and quiet moments per physician category, we will first explain why we consider an occupation of more than five rooms as “busy”.

As mentioned in the Methodology section, historical data is used in order to see from which number of patients a delay in LOS is best visible. Probably from that number of patients, processes are delayed and patients have to wait longer, resulting in a longer LOS. Several categorizations are made, with a varying split between “busy” and “quiet”. Every time, the differences between “busy” and “quiet” were tested with a T-test, in order to find out whether the two groups are significantly different from each other. Below, the categorization with highest significant difference between the two groups, is described.

Table 4.7 - Results T-test, differences in LOS between “quiet” and “busy”

Variable Obs. Mean Std. Dev.

Quiet (0-5 rooms) 7,718 126.33 0.85

Busy (6-8+ rooms) 6,165 150.10 0.94

T-statistic -18.63 ***

*** p < 0.01, ** p < 0.05, * p< 0.10

(23)

considered as most natural distinction between busy and quiet. For “quiet” the average LOS is 126 minutes, while for “busy” the average LOS is 150 minutes. This distinction will be used in upcoming sections.

4.2.1 ED-physicians

From figure 4.8, we can see the following differences between “quiet” and “busy” for ED-physicians: I) preparing patient reports and consultation by phone increases in case more rooms are occupied, II) patient treatment, non-work-related activities and consultation remains (approximately) the same, and III) supervision decreases in case more rooms are occupied. We will discuss the main findings per activity. These findings were also presented to four physicians during the interviews in order to validate results and gain deeper understanding why above-mentioned differences would occur.

Figure 4.8 - Overview time shares for ED-physicians.

Patient treatment – The time share spent on patient treatment remains the same. However, ED-physicians do treat more patients during busy moments. During those moments, ED-physicians often take multiple ‘easy’ patients, which means that they can treat them relatively fast. In paragraph 4.3.2 and 4.3 we will further elaborate on this, since this trend is also visible by measuring processing times. Multiple easy patients with a shorter LOS instead of one difficult patient who requires longer and more treatments, can result in a similar amount of time spent on patient treatment in total.

Preparing patient reports - A possible explanation for a similar time share spent on preparing patient reports while it becomes busier, could be that the ED-physician becomes responsible for more patients him- or herself, instead of supervising a physician-in-training. As a consequence, the ED-physician have to prepare more patient reports of these patients him- or herself.

Consultation by phone – At busy moments, ED-physicians spent a larger time share on consultation by phone. This includes acceptance of new patients and consultation with specialists or general practitioners. In case more patients are at the Emergency Department, specialists call more often for a consult or test-results are given by phone. Just before it becomes busy, consultation by phone already increases since patients are registered by their general practitioner, ambulance or other referral, which is discussed by phone. The increase in phone consultation is also confirmed during the interviews, see Appendix III, interview I and III. ED-physician: I can already

5+ rooms occupied Patient treatment 23% Prepare patient reports 15% Non-work related 12% Consultation 15% Consultation by phone 23% Supervision 8% Take-over ambulance 4% 0-5 rooms occupied Patient treatment 23% Prepare patient reports 10% Non-work related 13% Consultation 17% Consultation by phone 15% Diagnose patient 2% Supervision 12% Gaining patient information 1% Administrative tasks 3%

Consultation with specialist

(24)

experience busy moments, when there are no patients actually at the Emergency Department, but we receive their registrations by phone, like now. Currently it is not busy, but I know that it will be busy within a few minutes, so than I am already taking some consequences. Physician-assistant: “In case you have the 37-phone (coordination phone), which is often the ED-physician, you are continuously disturbed for registrations and consultations.” Non-work-related activities – The time share spent on non-work-related activities is approximately the same for busy and quiet moments. It is likely that an ED-physician has more additional administrative or coordinative tasks, which can be done during busy moments. Those tasks are not patient-related and can therefore be done during quiet moments. As mentioned during the interviews (Appendix III, Interview I), “ED-physicians have so many additional tasks, which you can’t finish during your management-day only. So during quiet moments, ED-physicians can already do some of those additional tasks.” This statement is also visible in the time share of administrative tasks and coordination during quiet moments.

Supervision – During busy moments, the time share spent on supervision decreases. During these moments, an ED-physician sees more patients him- or herself, and consequently, the ED-physician will be less available for supervision. In case the ED-physician is available for supervision, supervision is done faster. Physician-in-training: “In case it is busy, I try to consult my patients quicker with my supervisor. Then, there is less time for an ‘education-mode’. Otherwise, an ED-physician asks more on my opinion, what I think the diagnosis is and which additional tests I would like to take.”

Differences between ED-physicians - However, multiple interviewees argued that there are large differences between ED-physicians. Also from the processing time measurements (see paragraph 4.3), we can conclude that the time spent on patient treatment as percentage of the total work shift, differs per ED-physician. Figure 4.9 shows the difference between the ED-physician of that day and their percentage of time of their work shift spent on patient treatment compared to the average percentage of all physicians that day. In order to increase generalizability, five different ED-physicians are observed. The differences between ED-physicians are also recognized during the interviews, Appendix III, interview I and III. ED-physician: “For myself, I think that coordination and supervision is a larger part of the time, but I know that I am working more on that part compared to colleagues.” Physician-assistant: “I think the degree of seeing patients is depending on the physician who is coordinating that day. Some physicians coordinate more, while others continue with seeing patients.”

Figure 4.9 – Overview of time spent on patient treatment per ED-physician 0,00% 5,00% 10,00% 15,00% 20,00% 25,00% 30,00% 35,00%

14-Nov 15-Nov 17-Nov 23-Nov 24-Nov 29-Nov 30-Nov 6-Dec 7-Dec

Pe rc en ta ge o f w or k sh if t s pe nt o n pa ti en t tre atm en t

Time spent on patient treatment: ED-physicians compared to average

Average time at patient ED-physician's time at patient

-0,75% -0,84% -4,52%

-1,19% _-3,83%

-3,85% _-7,72%

-4,87%

(25)

0-5 rooms occupied 5+ rooms occupied

Conclusions ED-physicians – When looking at the shifts in time share spent on several activities, we can conclude that ED-physicians focus more on the process during busy moments, by seeing more patients him- or herself and do more other patient-related activities. During quiet moments, ED-physicians are more available for education and supervision. We can consider this as flexibility in capacity. However, we also see that the time share of non-work-related activities did not change. ED-physicians have additional tasks which they can do, so that they don’t have to wait for a job during quiet moments, which means that the capacity buffer is minimal for this category.

4.2.2 Physician-assistants

From figure 4.10, we can see the following differences between “quiet” and “busy” for physician-assistants: I) patient treatment, consultation by phone, administrative tasks and diagnose patients increase during busy moments, II) non-work-related activities drops drastically.

Figure 4.10 - Overview work sampling outcomes for physician-assistants

Patient treatment – For physician-assistants a large difference is visible between patient treatment during busy and quiet moments. As mentioned by multiple interviewees, this is because physician-assistant more often take multiple “difficult” patients. Whereas an ED-physician has to be available for coordinating, a physician-assistant can spend more time on patients related tasks during busy moments. This was also visible in the data collected with regard to processing times. During measurements at busy moments, each physician-assistant was seven times responsible for three or more patients. In comparison with physicians, it is significantly more, since ED-physician were on average two times responsible for three or more patients. Next to more patients, another explanation for the increase in time share during busy moments was given by a physician-assistant, see Appendix III, interview III: “During busy moments, it could be that I am not able to do the first physical examination completely. In that case, I have to go back to one patient multiple times. At the end, this could result in longer time at that one patient compared to finishing the examination in once.”

Consultation by phone – We have already explained the increase in consultation by phone for ED-physicians. However, also for the physicians without the coordination phone, the time share spent on this activity increases. Physician-assistant, Appendix III, interview III: “In case the ED-physician is busy, the physician-assistant gets more phone-calls himself, especially from a specialist or for receiving results from additional tests.”

Patient treatment 31% Prepare patient reports 18% Non-work related 7% Consultation 12% Consultation by phone 9% Diagnose patient 7% Supervision 1% Gaining patient information 1% Administrative tasks 7% Consultation with specialist 3% Coordination 3% Take-over ambulance 1% Patient treatment 23% Prepare patient reports 21% Non-work related 24% Consultation 8% Consultation by phone 2% Diagnose patient 5% Supervision 4% Gaining patient information 5% Administrative tasks 2%

(26)

Non-work-related activities – Probably the largest difference between busy and quiet is caused by non-work-related activities. Whereas for ED-physicians not much changed, physician-assistants decrease a large amount of time spent on work-related activities. Although physicians were amazed about the large time share on non-work-related activities at first, after overthinking they were convinced that this percentage was a correct representation. Main cause for this decrease is that during quiet moments, physician-assistant don’t have much additional tasks, like ED-physicians, and are therefore waiting for the next job.

Conclusions physician-assistants – For physician-assistants, the largest differences occur between busy and quiet moments. The time-share spent on non-work-related activities can drop from a quart of their time to less than 10 percent. The possibility to drop this time share confirms the existence of a capacity buffer during quiet moments.

4.2.3 Physicians-in-training

From figure 4.11, we can see the following differences between “quiet” and “busy” for physicians-in-training: I) prepare patient reports and consultation by phone increases, II) patient treatment, prepare patient reports and consultation remain (approximately) the same, and III) non-work-related activities and supervision decreases. Again, the findings are explained per activity and complemented with additional findings from the interviews.

Figure 4.11- Overview work sampling outcomes for physicians-in-training

Patient treatment – In comparison to physician-assistants, the time share for patient treatment does not differ much. This is mainly because physicians-in-training do not postpone activities or have multiple patients like ED-physicians or physician-assistants do. This results in less change in the time share for patient treatment. Physician-in-training, Appendix III, interview IV: “I often do the same, I complete patient reports totally before visiting a new patient. So in that case, the number of patients isn’t influencing me.”

Prepare patient reports – Although time share for patient treatment remains the same for physician-in-training, they did spend more time on preparing patient reports. This is not caused by more patients, as it was by ED-physicians, since physicians-in-training always see one patient at the same time. This difference is explained during the interview with the physician-assistant, see Appendix III, Interview IV: “I have to consult my patients with my

0-5 rooms occupied 5+ rooms occupied

Patient treatment 28% Prepare patient reports 30% Non-work related 8% Consultation 13% Consultation by phone 9% Diagnose patient 6% Gaining patient information 4% Administrative tasks 2% Patient treatment 26% Prepare patient reports 26% Non-work related 21% Consultation 13% Consultation by phone 5% Diagnose patient 4% Supervision

(27)

supervisor and in case it is very busy, you have to wait for supervision. During waiting, I have time to make a patient report, for which I can take some more time than.”

Non-work-related activities – Similarly to physician-assistants, we see a large decrease in non-work-related activities during busy moments. This indicates that physician-assistants also do not have many additional tasks which they can do during quiet moments.

Supervision – The decrease in supervision, which was already visible by the ED-physicians and physicians-assistants, is also visible for physician-in-training. This is logic, since their supervisor is less available for supervision. In case ED-physicians are available for supervision, this is done faster, as mentioned in Interview IV, Appendix III.

Conclusion physicians-in-training – We can conclude that the decrease in time spent on non-work-related activities also confirms a capacity buffer for physicians-in-training. Since this category also has less additional tasks, we can also consider this buffer as less flexible compared to ED-physicians.

4.2.4. Conclusions capacity buffer

(28)

4.3 Processing time buffer

The previous paragraph has already argued the existence and use of the capacity buffers in Emergency Departments. In order to investigate the second buffer, the processing time buffer, processing times of patient treatments are measured and analyzed.

4.3.1 Relationship between more patients and the duration of treatment

In order to investigate whether the relationship between the number of patients a physician is responsible for and the average duration of the first physical examination is significant, a regression has been conducted. First a graphical representation is given in figure 4.12. Thereafter the results of the OLS regression are given, see table 4.13 and 4.14. Five extreme values have been deleted, since their processing time was longer than 30 minutes (with maximum of 80 minutes). These outliers influenced the average duration a lot, while it is unlikely that those long durations were situations in which protocols and work standards were used. We assumed that in case a physician examination was longer than 30 minutes, it was an extraordinary case.

Figure 4.12 - Overview duration first physical examination per quantity of patients within the responsibility of the physician

Table 4.13 – Descriptive statistics number of patients within the responsibility of a physician

Variable Obs. Mean Std. Dev. Min. Max.

Dependent: Duration of physical examination 99 10.63 6.19 2 30

Independent: Number of patients a physician is responsible for 99 1.87 0.80 1 4

Table 4.14 – Relationship between number of patients within the responsibility of a physician and the duration of first physical examination

Duration of physical examination

Number of patients a physician is responsible for -2.08 *** (0.75) F-statistic 7.65 *** R-squared 0.073 *** p < 0.01, ** p < 0.05, * p< 0.10 0 5 10 15 20 25 30

Responsible for 1 Responsible for 2 Responsible for 3

Du ra ti on ( in m in ut es )

Number of patients within responsibility of physician

Duration first physical examination

(29)

From this regression analysis can be concluded that there is a significant relationship between the number of patients the physician is responsible for and the duration of physical examination. If a physician is responsible for one additional patient, the duration of the physical examination will on average decrease with 2.08 minutes. An overview per physician-category is given in figure 4.15. Results of physician-assistants were most reliable, since their average is based on 22, 19 and 18 measurements for respectively one, two and three patients. Only five measurements could be done for situations in which ED-physicians were responsible for three patients simultaneously and the third column for ED-physicians in figure 4.15 is therefore less reliable. Since physicians-in-training are most of their time responsible for one patients, they are left out of this overview.

Figure 4.15 - Average duration first physical examination for physician-assistants and ED-physicians

4.3.2 Causes of the decrease in processing times

During the interviews, more depth information was asked about this shortening in average duration of an examination. When posing the following question, what the physician would do different in case he or she became responsible for more patients, the following answers were given: ED-physician: “I ask more closed questions, so the patient only can answer yes or no. This helps to decrease the time the patient is talking. In that case I am less focused on building a relationship with my patient. I want to get my information and then leave the patient.” Physician-in-training: “I recognize that in case there are less patients at the Emergency Department, I take more time for the patient. In case it is busy, you ask the required questions in a structured way and pay less attention to other things which are not relevant at that moment.” Combining findings from the interviews with the observations, we can consider this ‘relational aspect’ as a non-critical task, which is not executed during busy moments. It can be said that the processing time buffer exists in hospital environments.

Less attention is thus paid on relation-building during busy moments, which could enable the physician to decrease the duration of the treatment. Another activity that was skipped in case a physician becomes responsible for more patients, was informing how a patient feels. By responsibility for multiple patients, the total amount of times a physician informs on a patient’s status decreased, see table 4.17. This could mean that informing on patient’s status is a non-critical task for a physician, since nurses are responsible for monitoring patient’s status. This was also confirmed in the interviews, Appendix III, interview IV. Physician-in-training: “In case it is busy, you trust the nurses. They will inform you when status is going backwards. In case it is quiet, I walk by myself, to see how patients is doing.” 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Du ra tio n (i n m in tu es ) Physician-assistant ED-physician

Duration first examination per physician category

(30)

Table 4.16 - Frequencies for informing how patient feels

Responsible for (number of patients) Total amount of times informed on patient’s status

1 9

2 2

3 2

As we saw in figure 4.15, a difference is visible between the ED-physicians and physician-assistants. As mentioned in the introduction of the graph, it could be that this is partly caused by less measurements for ED-physicians, which make their measurement for three patients less reliable. However, also between one and two patients, we see differences between these two physician categories. For physician-assistants, the average duration decreases gradually, while for the ED-physicians the average duration ‘drops’ in case the physician became responsible for multiple patients. An explanation for this phenomenon is that ED-physicians take often fractures or other ‘easy’ patients as second or third patients. This is done explicitly, since the ED-physician needs to be available for supervision or other coordinating tasks (see interviews and results section 4.2).

Because every patient differs, it is hard to make homogeneous patient groups. To see whether abovementioned phenomenon of decreased processing time also occurs when looking only at similar diagnosis, the following categorization is made. In the following table an overview of the average duration of the first examination per diagnosis is given.

Table 4.17 - Average duration per number of patients per diagnosis

Also in this overview, a decrease in processing time is visible when a physician becomes responsible for multiple patients. However, because of the uniqueness of patient’s diagnosis, fewer measurements could be included, which increases the influence of extra-ordinary patients and make results less reliable.

4.3.2 Conclusions processing time buffer

In general, we can conclude that there are clear indications that the Emergency Department uses a processing time buffer. This processing time buffer is used when a physician becomes responsible for multiple patients, and leads to a decrease of time spend on first physician examination. Tasks which are related to relation building with a patient and winning empathy of a patient are decreased during busy moments, which make it possible to decrease the duration of the treatment and consequently response to variability. The possibility to decrease processing times differs per physician-category.

Diagnosis Average duration

responsible for 1 patient (minutes) Average duration responsible for 2 patients (minutes) Average duration responsible for 3 patients (minutes) Number of measurements included Examination fractures 0:09 0:07 0:08 24 Abdominal complaints 0:11 0:09 0:10 32 Stroke 0:22 - 0:09 5

Buffers at Emergency Departments: Explaining the Influence of Buffer Mechanisms on Emergency Departments’ Responsiveness to Volume Variability