Taking Care of Patients - Avoiding Crowding in an Emergency Department in the USA

(1)

Emergency Department in the USA

August 31, 2016

Master Thesis Supply Chain Management

Master Thesis Technology & Operations Management University of Groningen, Faculty of Economics and Business

Author David Buitelaar

Student number s2057425

Adress Leeuwarderstraat 48a

9718HZ Groningen

E-mail david.buitelaar@gmail.com

Supervisor Dr. M. J. Land Dr. J. T. van der Vaart Dr. L. D. Fredendall

Institution University of University of University of

Groningen, Groningen, Clemson,

Faculty of Faculty of Department of

Economics and Economics and Management

(2)

Thesis TOM/SCM David Buitelaar

ABSTRACT

Emergency Department crowding is a severe problem in the United States. Crowding results in an increased Length-of-Stay for patients, delays in critical treatments and even increased mortality rates. This research builds on the vast amount of literature available on ED crowding by utilizing an Input/Output analysis to identify causes of crowding in an American ED. This research focuses on confirmation or rejection of causes found in literature by using a more in-depth analysis that takes into account the dynamic patient inflow. A dynamic analysis is helpful, since this will create insight of an event earlier in the day and its effect later in the day. It is important to identify the correct causes in order to solve the right problem. In order to measure the number of times waiting occurs, different waiting categories are identified and measured with the use of Work Sampling. Work Sampling is combined with the Input/Output analysis to determine causes of crowding. A single case study is performed at Oconee Memorial Hospital in Seneca, South Carolina. Quantitative and qualitative analysis are used to gain insights for avoiding crowding of an American ED.

(3)

INTRODUCTION

Emergency Department (ED) crowding is a problem affecting most of the hospitals in the United States (Handel et al., 2010). Since the 1990s ED visits steadily increased in the United States, while the number of EDs have declined (Tang, Stein, Hsia, Maselli, & Gonzales, 2010). Bazzoli, Brewster, Liu, & Kuo (2003) recognize that a large number of American EDs cope with capacity constraints. EDs in the United States have serious problems with crowding, leading to all sorts of negative outcomes such as delays in critical treatments, increased waiting times and even increased mortality rates (Pines et al., 2011). In the United States in 2006, the average waiting time of emergent patients to be seen by an ED physician was 37 minutes, where the recommended maximum is 15 minutes (Horwitz, Green, & Bradley, 2010). This requires improvements in the operational performance of American EDs. In order to give suggestions for avoiding ED crowding, this research conducts an analysis such that the dynamic variable patient inflow and patient outflow of an ED are understood. These dynamics need to be understood in order to understand crowding. Problems in the afternoon can be the result of events that hinder flow in the morning.

In order to find ways to improve the operational performance of an ED, causes need to be identified first. The literature identifies several causes for ED crowding: (i) patients competing for the same resources, (ii) insufficient amount of treatment rooms of an ED, (iii) the inability to move patients to other places after their care is completed, e.g. to home, an inpatient clinic or another hospital, (iv) inadequate staffing and (v) non-urgent visits (Handel et al., 2010; Hoot & Aronsky, 2008). In the search for solutions to reduce the waiting time, scholars found that the following actions would decrease crowding of an ED: (i) ED expansion, (ii) additional personnel and (iii) increased inpatient clinic capacity to decrease the access block in the patient outflow (Han et al., 2007; Hoot & Aronsky, 2008; Miró et al., 2003; Olshaker & Rathlev, 2006). Sinreich, Jabali, & Dellaert (2012) recognize that a solution for crowding must be found by adjusting work shifts of the personnel. They state that an appropriate staffing level needs to be determined, to match the patient arrival pattern and use an algorithm to come up with an appropriate staffing level that reduces the waiting time of patients. However, there exists indistinctness in literature about the real effect of causes and solutions. In order to select the proper solution for an ED, an adequate analysis must be conducted first. This research builds on a quantitative analysis, while searching for confirmation of current insights and new insights on causes of waiting times in an American ED. Based on these insights, practical implications for the case ED and similar EDs can be given.

(5)

performing this research more insights can be gained in reasons for crowding specific for an American ED.

The research is based on a case study performed at Oconee Memorial Hospital in Seneca, South Carolina. Adapting and extending the current Input/Output analysis to an American ED is useful in gaining new insights in mechanisms that lead to a crowded ED and waiting times for patients in an American ED. The research answers the following research question:

“How can crowding in the American Emergency Department setting be explained from a dynamic viewpoint?”

The vast body of literature on American EDs uses a research based on averages and therefore implicitly assumes stationary patient arrivals. However, differences in inflow or capacity during specific times of the day can lead to crowding. These events in a narrow timeframe cannot be identified if stationary patient arrivals are assumed. The Input/Output analysis takes a dynamic viewpoint in identifying causes of crowding, which makes it possible to identify events earlier in the day and their effect on crowding. The dynamic viewpoint distinguishes itself from the implicitly assumed stationary viewpoints by being able to see the effect of events at specific moments of a day on the crowding later in the day. This analysis has not been used in an American ED, but has been proved to be successful in Dutch EDs. It is interesting to see if a new analysis is able to confirm or disprove causes of crowding found in American literature. This research will have a two-fold theoretical contribution. The first part of the theoretical contribution is a confirmation or addition of causes of crowding in an American ED, while the second part of the theoretical contribution is to create insights in differences in causes of crowding between Dutch and American EDs. The practical contribution will be implications for the case ED on how to avoid crowding while taking into account case-specific causes of crowding. The methodological contribution is an extension to the current Input/Output analysis by combining the current Input/Output analysis with Work Sampling and Workload analysis.

(6)

THEORETICAL BACKGROUND

In this theoretical background literature on the following topics will be discussed: (i) characteristics and process of an ED, (ii) Crowding in EDs and its causes and solutions and (iii) an overview of findings of causes of a delayed outflow pattern in EDs. The section ends with a conclusion to assess validity of the research question.

Emergency Department – the Process

The ED process starts with the patient inflow. Patient inflow has several sources: (i) self-referrals, (ii) referred by other care provider or (iii) brought in by ambulance. Patients enter with different urgency levels and require different responses of ED personnel with respect to diagnosis and treatment. After arrival, patients are registered and triaged.A triage nurse performs the triage process to provide the physician with an indication of the acuity and ensures that patients who need care, receive it within appropriate time. The triage results in a color code, which resembles the acuteness of the patient and the time limit for the patient to be seen by an ED doctor (Mackway-Jones, Marsden, Windle, & Harris, 2013). Often used triage systems are the Manchester Triage System (MTS) or the Canadian Triage and Acuity Scale (Mackway-Jones et al., 2013; Ospina et al., 2007). The triage system provides clarity about the maximum waiting time before the patient’s 1st medical screening by a physician for the different levels of urgency. The MTS has five levels of urgency: (i) emergent (red), (ii) very urgent (orange), (iii) urgent (yellow), (iv) standard (green) and (v) non-urgent (blue) patients (Roukema et al., 2006). After being registered and triaged, the patients get their 1st medical screening and is tested by a physician. This is followed by a decision on the medical treatment. Based on this decision, patients are being discharged or admitted to an inpatient care facility. The outflow of patients can be divided in: (i) patients who leave without treatment, (ii) patients disposed to an outpatient care facility, (iii) patients admitted to the hospital and (iv) patients who leave after their treatment (Asplin et al., 2003; Lane et al., 2000).

Emergency Department – Characteristics

(7)

symptoms. As a result of the sudden onset of symptoms, the arrival pattern of patients at the ED is dynamic. The dynamic arrival pattern and higher complexity of patients make it difficult to plan the necessary staff capacity at an ED (Higginson, Whyatt, & Silvester, 2011).

As can be concluded from the process characteristics, an ED copes with diverse input, diverse diseases, diverse treatment times, different treatment time limits and an unknown arrival rate of patients. This combination of characteristics leads to difficulties in planning necessary staff capacity and realizing a proper patient flow, which leads to the possibility of crowding. The next section discusses the concept of crowding.

Crowding of the Emergency Department

Crowding is defined as “the situation where ED function is impeded by the number of patients waiting to be seen, undergoing assessment and treatment, or waiting for departure, exceeding the physical or staffing capacity of the department” (Forero et al., 2010, p. 120). This section discusses the factors, related solutions and their advantages and shortcomings found in literature.

Commonly studied factors and related solutions of ED crowding from literature can be found in table 1. The distinction between input, throughput and output is made, which is in line with the conceptual model of Asplin et al. (2003). In this conceptual model the ED is seen as a system. The input component of the system is any condition, event or system characteristic that contributes to the demand for ED care. The actual process within the borders of the system (i.e. the ED) is called the throughput component and focuses on activities that take place at an ED. The output component of the system is any condition, event or system characteristic that impacts the ability of the ED to disposition patients (Asplin et al., 2003).

Studied factors Sources of ED crowding

Solutions from literature

Input Emergency care,

unscheduled urgent care and safety net care (Asplin et al., 2003)

High demand for care Increased responsiveness (Koomen, 2016; Saghafian, Hopp, Van Oyen, Desmond, & Kronick, 2012; van Achteren, 2014)

Throughput Inadequate staffing Insufficient capacity of personnel available

Additional personnel or observation units (Hoot & Aronsky, 2008)

Output Inability to dispose patients (Hoot & Aronsky, 2008)

Lack of available staffed patient beds (Asplin et al., 2003; Hoot & Aronsky, 2008)

ED expansion in number of beds and personnel (Han et al., 2007; Miró et al., 2003)

Inpatient boarding (Hoot & Aronsky, 2008)

Access block at inpatient clinics (Forero et al., 2010; Hoot & Aronsky, 2008)

Increase inpatient bed capacity such that inpatient clinic access is increased (Hoot & Aronsky, 2008; Olshaker & Rathlev, 2006).

(8)

It is important to state that a patient’s waiting time is affected by the patient’s acuity and the degree of crowdedness of the ED. A longer Length-of-Stay (LOS) of patients in the morning, results in buildup of Work-in-Process (WIP) leading to a crowded ED, which increases the waiting time for subsequent patients. A dynamic Input/Output analysis will enable it to analyze these patterns for both Length-of-Stay and WIP over the course of the day.

The next part discusses several solutions found in literature for avoiding crowding. This part focuses on advantages and shortcomings of these solutions to emphasize the importance of a good diagnosis in order to find the correct cause of ED crowding.

ED expansion

ED expansion implies adding extra treatment rooms, without adding staff capacity. Miró et al. (2003) state that ED expansion leads to lower waiting times of patients and a decrease in the percentage of days that had to cope with crowding. However, other scholars do not agree on the effectiveness of ED expansion (Hoot & Aronsky, 2008; Koomen, 2016; Olshaker & Rathlev, 2006; Saghafian et al., 2012; van Achteren, 2014). Han et al. (2007) found that ED expansion did not lead to a decrease in LOS, but even leads to an increase of the inability to provide safe care to patients.

Additional capacity of personnel

Additional personnel is used to decrease ED crowding, however adding expensive resources will lead to further rising costs in health care. Although it is not expected that additional personnel is necessary throughout the complete day, extra staff on critical moments during the day is expected to lower the WIP, resulting in shorter waiting times without a (big) increase in personnel costs. However, Schull, Lazier, Vermeulen, Mawhinney, & Morrison (2003) did not find a link in their specific case between nursing and physician shortages and ambulance diversion due to crowding, but did find that delays in the admission process of patients to inpatient clinics are an important contributor for ambulance diversion due to ED crowding.

Increased hospital bed access

(9)

This discussion shows that there are still contradictory findings on the effect of solutions. Applying the wrong solution to a problem will not solve the problem. These contradictory findings show the need for an analysis that takes into account time-specific events in identifying causes of crowding.

Research Findings in Dutch hospitals

In some Dutch hospitals a diagnosis based on Input/Output analysis has been used and gained new insights in every case ED. The Input/Output analysis distinguishes itself since it takes the dynamic arrival and departure pattern of patients into account. It is important to take into account the dynamic behavior of arrivals and WIP, because crowding in the afternoon could be the result of events in the morning. This analysis has proven itself to be useful in distinguishing these events. Examples of events are a high inflow of (critical) patients or low ED capacity measured in either rooms or staff. Previous research gained insights in the arrival and departure pattern of patients over the course of a day. By analyzing these patterns for different type of days, e.g. high/low inflow during morning, differences in the patterns can occur. Based on this type of analysis several solutions are proposed to avoid ED crowding. Koomen (2016) and Van Achteren (2014) showed that an increased responsiveness of ED personnel is necessary for a shorter patient Length-of-Stay. Responsiveness is defined as “how easily or fast a proposed change can be implemented” (Upton, 1994 IN van Achteren, 2014). Gelissen (2016) found that a WIP reduction up to 25% of the highest WIP can be achieved by a higher responsiveness during the morning. Dutch EDs are characterized by the fact that during the mornings the ED has a lower Work-in-Process (WIP) than in the afternoon. EDs are practically empty early in the morning, while later in the day crowding of the ED occurs. It is expected that American EDs cope with a faster buildup of WIP in the morning.

Conclusion

(10)

METHODOLOGY

The aim of the study is to gain insights in the dynamics of patient Length-of-Stay at the ED over the course of the day. The Length-of-Stay of patients is subject to events over the day that can increase the patient Length-of-Stay, even if the event has happened earlier in the day. A single case study is used, since it gives the possibility to have a more in-depth analysis. A case-study is perfectly suitable for “how” questions and extensive in-depth analysis (Yin, 2014). This research investigates a situation to understand a phenomenon. According to Meredith (1998) case study research can be used to perform both quantitative and qualitative methodologies to help understand a phenomenon. This research will use and extend an Input/Output analysis in an American ED. This study will contribute by verifying causes of crowding in the United States found in literature by using a dynamic approach. The study also makes it possible to compare with findings of an Input/Output analysis in Dutch EDs. The study derives its practical value from providing the ED with insights in the main causes of waiting times, considering the dynamics of hourly patient inflow and outflow patterns. The study contributes to literature by confirming causes found in American literature on ED crowding. An Input/Output analysis is extended such that it takes the outflow and problems with the flow of patients in the treatment into account. It is important to take these factors into account while identifying causes of crowding, since literature has shown that American EDs cope with discharge and admission problems. It is expected that an extended approach will be able to confirm or disprove causes of crowding in American EDs found in literature. To the best of our knowledge this will be the first application of a combined Input/Output, Work Sampling and Workload analysis in the US-setting.

Oconee Memorial Hospital

The research is performed at the Oconee Memorial Hospital in Seneca, South Carolina. Oconee Memorial Hospital consists of 169 beds and features inpatient and outpatient clinics and an ED that is on duty 24 hours a day. Around 42500 patients visit the ED a year. The ED consists of 20 beds. These are: 4 Fast-track beds, 6 exam rooms, 2 cardiac rooms, 2 trauma rooms, 1 cast room, 2 OB Gyn (Obstetrics and Gynaecology) rooms, 1 ENT (Ear, Nose and Throat) room, 1 Eye room and 1 psych room. The ED has the possibility of 3 extra hallway beds.

(11)

Analysis

In the search for causes of crowding and suggestions for solutions, different types of data gathering and analysis are used. The different sources of data and analysis performed on these data can be found in table 2. The general analysis will give insights in the performance of an average day, while later phases will use this as a base case to build on. This will lead to triangulated findings.

Data source Analysis

1st phase: General analysis Quantitative data from IT system Medhost

Triage division and corresponding Length-of-Stay; Boxplot of arrivals per hour; Input/Output diagrams 2nd Phase: High-Low analysis Quantitative data from IT

system Medhost

Input/Output diagrams; Throughput diagrams; Comparisons between types of days; Workload analysis 3rd Phase: Work Sampling Structured observations

(30-minute interval)

Determine causes of waiting for patient; comparison between morning and afternoon; comparison between admitted and discharged patient 4th Phase: Consulting

physicians and direct observations

Consulting Physicians and direct observations

Discuss observations and findings with physicians to get better insight in flow patterns

5th Phase: Triangulate findings of Work Sampling with Quantitative data

Quantitative data from IT system Medhost and Structured observations (30-minute interval)

Throughput diagrams

Table 2 Data Analysis

Quantitative data from Medhost - The provided data exists of the treating ED physician, determined

acuity level of the patient and date and time of the following activities: (i) arrival, (ii) triage, (iii) registration, (iv) patient moved to room, (v) first medical screening, (vi) ready for departure and (vii) departure. The reliability of the data is checked while shadowing an ED physician by noticing how and when data is entered in Medhost. The raw dataset includes de-identified patient data from 27-05-2015 until 26-05-2016 and accounts for 42463 patients. The dataset is de-identified to guarantee patient privacy. The first step after collection is to check for errors in the data set. Patients who have not used scarce resources, e.g. ED physicians/treatment rooms, will be removed from the data set because these patients have a negligible effect on crowding. After removing errors from the data set, a total of 41619 patients between 27-05-2015 and 26-05-2016 remained. This data is used, since it allows for Throughput and Input/Output analysis.

(12)

Work-in-process (WIP) and average Length-of-Stay can be seen. The vertical distance between the arrival graph and the departure graph will show the WIP at a certain time. The horizontal distance between the arrival graph and the departure graph will show the average Length-of-Stay of the nth patient. The Throughput and Input/Output diagrams are used for a base case for the High-Low analysis, since these diagrams represent long-term average flow patterns. This base case will be used for comparisons with days with a different inflow pattern in the High-Low analysis, which will be discussed next.

During High-Low analysis specific types of days are analyzed. The first comparison is made between the top 25% days with longest average Length-of-Stay and the top 25% days with shortest average Length-of-Stay. This analysis will create insight in differences in flow patterns between two different type of days with respect to Length-of-Stay. These insights are useful in understanding which events result in crowding, without taking into account specific differences in inflow of patients.

Next, a comparison between the four groups presented in table 2 is performed to take into account different inflow patterns throughout the day. This analysis requires a split of the total data set in the four groups such that comparisons can be made. This creates insights on the effect of a high demanding morning in comparison to a high demanding afternoon, together with either a high or low demanding morning/afternoon. Insights are gained in the buildup of the outflow pattern and the resulting WIP at the ED. By comparing the different inflow patterns, it can be identified when crowding is more likely to occur based on the inflow of patients. During these type of days, causes will contribute to the occurrence of crowding. Hence, by splitting the data into these four types of days causes of crowding can be identified for different types of inflow patterns during different moments of the day.

Above median number of visitors in afternoon (High demand afternoon)

Median or below median number of visitors in afternoon (Low demand afternoon) Above median number of visitors

in morning (High demand morning)

Group 1 (High/High); Abbreviation HMHA

Group 2 (High/Low); Abbreviation HMLA Median or below median number

of visitors in morning (Low demand morning)

Group 3(Low/High); Abbreviation LMHA

Group 4 (Low/Low); Abbreviation LMLA Table 2 Group comparison high/low demanding morning/afternoon

(13)

Work Sampling – In general, Work Sampling is used to determine the proportion of time used for

specified categories of work (Ho & Pape, 2001). In this research Work Sampling is used to identify what patients are waiting for during the patient pathway. After shadowing an ED physician the patient pathway is identified, which is presented in figure 1. Note that the presented pathway is a static picture, while in reality it could be that phases overlap. If phases overlap, a physician is consulted to determine which cause is the most important cause for delaying the patient pathway.

Figure 1 Patient Treatment Pathway

Delays in the patient pathway occur between these phases. The Work Sampling gathers the number of times each delay occurs. The categories for delays used are the waiting period in between the activities in the patient treatment journey. These are: (i) waiting for 1st contact with physician, (ii) waiting for CT-scan, Ray, Blood test or medication order, (iii) waiting for results of the CT-scan, X-ray or Blood test, (iv) waiting for effect of treatment, (v) waiting for physician, (vi) waiting for nurse, (vii) waiting for care completion, (viii) waiting for departure. The Work Sampling form used can be found in appendix 1.

The Work Sampling data is analyzed in different ways. First, it is used to see which one of the delays accounts for the highest number of registered delays during the data gathering. This gives an impression on the average patient treatment pathway and where delays in this pathway occur. After this, the data is split at 12:30 to make an equal comparison between morning and afternoon to analyze different patterns during different times of the day. This is performed for both discharged and admitted patients. An admitted patient has two specific waiting categories: (i) waiting for the necessary consult of a hospitalist and (ii) waiting to be moved to an inpatient bed. The discharged patient on its turn has one specific category: (i) waiting for discharge.

Consulting Physicians – Consults with physicians are used to discuss the findings with physicians.

During these consults physicians are asked to give their view on topics as: (i) the most demanding time of the day, (ii) reasons of delays in patient pathways, (iii) maximum level of patients in the ER that is still manageable, (iv) manageable workload per physician and (v) found patterns in the quantitative data.

Direct Observations – According to Meredith (1998) direct observation is required to discover why and how circumstances occur. During direct observations, attention is paid to how data is entered in the IT system and how physicians and nurses organize their work. This is used to understand the ED process and how data is recorded in the IT system, i.e. Medhost.

Triangulate findings – By comparing the Work Sampling findings with physician consults and the

(14)

(15)

RESULTS

The first part of this section will cover the general analysis. Next, the flow variability of the inflow and outflow pattern is evaluated to determine delays in the patient pathway. A high-low analysis based on the extreme days is performed to address the differences in flow patterns between high and low demanding days. The sector ends with results of the physician workload analysis and the Work Sampling. The results are triangulated by combining findings in the Input/Output analysis and Work Sampling with physician consults.

General analysis

General info is extracted to provide an overall impression of the case characteristics. The general analysis is used as a base case for the high-low analysis. The median number of patients was 113 per day. The number of arrivals per day varied from 81 patients per day on 27-08-2015 till 156 patients on 25-04-2016. The median Length-of-Stay was 2 hours and 41 minutes and the Length-of-Stay varied from 10 minutes to 91 hours and 18 minutes. From the 41619 patients in the corrected data file, the disposition location of 1162 patients was not recorded. From the patients of which the disposition location was known, 19% was admitted at a care facility and 81% were discharged. The percentage of patients per triage level is presented in table 4. The biggest group is patients in level 3 or 4. These two patient groups represent 89% of the patients. Only 0.2% of the total number of patients is triaged as level 1.

Triage Level Number of patients Percentage of total

Level 1 88 0.2% Level 2 1966 4.7% Level 3 21216 51.0% Level 4 15938 38.3% Level 5 2303 5.5% No triage 108 0.3%

Table 3 Triage Division

The division of the Length-of-Stay per triage level is presented in table 4. The 95% percentile is 24 hours for triage level 2 patients. This can be explained by the fact that psychological patients are determined as triage level 2. The protocol for treating psychological patients is to observe these patients for at least 24 hours before admission to a psychological care facility. The differences in Length-of-Stay in the 4 other levels are logical, since less acute patients will take less time than critical patients. A big part of the patients is triaged as level 3 patients, thus these patients affect the median Length-of-Stay the most.

Level 1 Level 2 Level 3 Level 4 Level 5 No triage

1st Quartile 2:06:58 2:32:14 2:13:48 1:20:49 0:54:40 1:50:36 Median 2:56:05 3:34:39 3:10:54 2:06:11 1:30:08 2:49:35 3rd Quartile 3:38:05 5:24:03 4:21:16 3:06:43 2:27:41 4:49:44 95% Percentile 6:58:26 24:00:08 6:26:33 5:06:39 4:24:10 7:01:49

(16)

Arrival pattern variability

In order to analyze the variability in the arrival pattern over the day, a year-average Input/Output diagram can be found in figure 2. Figure 2 presents the inflow and outflow as curves in one figure with the number of available physicians as bars at each time of the day. The end time of the hourly measurement intervals is presented on the horizontal axis, while the number of patients is presented on the primary vertical axis. The number of available ED physicians is presented on the secondary vertical axis.

Figure 2 Yearly flow pattern and physician capacity

The average inflow of patients can be divided in three sectors. The arrival rate increases in the first section, between 6:00 and 11:00. The second section is a stable arrival rate of almost 7 patients an hour, between 11:00 and 20:00. The arrival rate declines in the third section, between 20:00 and 6:00. At 15:00 the inflow and outflow cross each other in the graph, indicating that the ED is able to have a corresponding outflow of patients, which in its turn indicates that the average number of patients in the ED does no longer increase at that point.

The average inflow of patients increases from 1.3 to 6.6 patients per hour between 6:00 and 11:00, while the outflow of patients is only at 3.8 outflowing patients per hour at 11:00. The delay between the increase in the inflow and the outflow to reach the same level results in buildup of WIP and an increased Length-of-Stay of patients.

(17)

Figure 2 shows that the ED is capable of taking care of the maximum inflow per hour on an average day, i.e. the outflow level per hour reaches the same or even higher levels of patients per hour. The inflow during the morning however, increases faster than the outflow is increasing.

In order to compare the current WIP over the course of a day with the WIP if the outflow is at a desirable rate, a new curve is added. This is the desirable outflow per hour. The desirable outflow follows the average inflow 2.5 hours later, i.e. the desirable outflow at 9:30 is equal to the inflow at 7:00. 2.5 hours is chosen, since the median Length-of-Stay in the complete data set is just over 2.5 hours. Figure 3 compares the current average outflow and the desirable average outflow in the period of 7:00 and 21:30.

Figure 3 Implications of current and desirable outflow for Work-in-Process

The WIP in the current state builds up to an average close to 18 patients between 13:00 and 21:00, after which the WIP decreases to an average WIP of 2 patients at 7:00. During the morning (7:00-13:00) the WIP builds up to an average of close to 18 patients, due to a delayed outflow pattern in the morning. This delayed outflow seems to indicate that the ED is coping with capacity problems during the morning, resulting in buildup of WIP and longer patient Length-of-Stay.

(18)

High-Low Analysis

This section presents the results of the high-low analysis. First, the day with the most visiting patients is discussed by means of a throughput diagram as an example of possible delays in the patient pathway. Next, a comparison between days with low and high arrivals between 7:00 and 12:00 is made. The sector ends with a comparison between the four types of days as presented earlier and a workload analysis.

Figure 4 is a throughput diagram of the most demanding day of the data set with 156 visiting patients. The vertical axis presents the number of patients while the horizontal axis presents the time of the day. Note that at midnight, 19 patients that arrived the preceding day were still in the ED. Hence the arrival curve starts at a count of 19 patients. The other curves start at 18, 16, 14, 4 and 0 for respectively registration, in bed, first medical screening, ready for departure and departure. This means that there are 14 patients who have had their first medical screening at midnight, but not yet their actual treatment.

Figure 4 Throughput diagram 25 April 2016

(19)

Over the course of the day, delays occur in the patient pathway resulting in a long Length-of-Stays of patients. The delays in the patient pathway that occur are: (i) patient moved to treatment room and first medical screening, (ii) patient moved from waiting room to treatment room and (iii) ready for departure and actual departure.

The first delay occurs after the patients are moved to a treatment room and before their first medical screening, between 03:00 and 08:30. The physician is not able to see patients directly after their move to a treatment room during this time, resulting in a delay in the patient pathway. This seems to occur as a result of the high WIP in the night.

More importantly is the second delay that occurs between patient registration and move to a treatment room. This delay happens between 10:00-12:00 and after 15:00. The ED is not able to move all the patients in the waiting room to a treatment room directly after their arrival, which suggests a lack of available staffed rooms during this period of time. Direct observations indicated that several rooms empty are during this time period, but patients are not moved to a room due to a lack of staff capacity during this period of time, which leads to an increasing number of patients in the waiting room during the morning. An increase in waiting patients, results in an increased Length-of-Stay of these patients and subsequent patients of equal acuity. These events early in the day can result in crowding later in the day, which emphasizes the importance of a dynamic analysis.

The third delay occurs for patients who are ready for departure, but have not left the treatment room yet. This can happen for three reasons: (i) patients cannot be admitted to a hospital bed, since there are no empty hospital beds available, (ii) patients cannot leave the ED, since there is no relative to pick them up and (iii) a nurse is required to administrate the discharge paper before the patient can leave. These patients keep using a treatment room, which thus cannot be used by a subsequent patient. This not only increases the Stay of these patients, but also increases the Length-of-Stay of subsequent patients, since there is no treatment room available for them at this time. Nurses cope with an increasing workload over the day. This higher workload prevents them from discharging patients since other more acute activities will go first, however it will result in a more packed department.

This single day gives a clear impression of the capacity problems that cause long Length-of-Stay for patients. To be able to analyze ED capacity issues in a more general way, the following graph compares the top 25% days with the bottom 25% days based on the Length-of-Stay of the patients. Note that the average number of patients that arrives is just 7.1% lower on days with short Length-of-Stay compared to days with long Length-Length-of-Stay. A group size of 93 days is expected to be sufficient in overcoming clinical variability between patients. The explanation for a difference in the flow performance has thus to be explained by a different cause. This is presented in figure 5.

(20)

Thesis TOM/SCM David Buitelaar Figure 5 Comparison long versus short TH time days

The outflow patterns of the two types of days are quite similar during the morning, except for the hurdle between 11:00 and 12:00 for days with a high LOS. From figure 5 we see that this hurdle results in a lower outflow for long LOS days in comparison to short LOS days until 17:00. The delay between 11:00 and 12:00 for days with a long LOS seems to go back to partial unavailability of lab readers on critical days, which results in a longer processing time for test results if a high number of lab tests is necessary for determining the treatment of patients.

The resulting WIP shows the differences in inflow and outflow patterns between the two types of days. The WIP on days with a high LOS builds up to a level of over 20 between 14:00 and 22:00, while on days with a low LOS the WIP has a maximum of 13 patients in process.

Figure 5 indicates that days with a long LOS are characterized by a high inflow during the morning and a hurdle in the outflow. This is the reason why the WIP on days with long LOS builds up to an average of close to 24 patients, while days with short LOS do not exceed an average of 13 patients in process. As a result of the fast buildup of WIP in the morning for days with a long LOS, the waiting time for subsequent patients increases.

(21)

Comparison Where to find?

1 – LMLA and HMLA Appendix 2, 4

2 – LMLA and LMHA Appendix 3, 5

3 – LMLA and HMHA Figure 6, 7

Table 5 Comparisons different days

The third comparison is shown here as an example for the comparisons between the four type of days. This example is chosen, since the differences are more noticeable in this comparison than in comparison 1 or 2. Comparison 1 and 2 can be found in Appendix 2 and 3.

Figure 6 Comparison Flow Performance LMLA -LMHA

Figure 6 (and Appendix 2 and 3) not only presents the inflow and outflow rate of patients but also the ready for departure rate as curves. Interesting is that in every type of day, a drop in the ready for departure rate between 18:00 and 20:00 is noticeable. As a physician states it during an consult:

“I think it goes back to the hospitalist. Their shift ends at 19:00 and often at 17:30/18:00 they stop taking new admissions, to not run late in their shift. At 19:00 one hospitalist replaces five hospitalists and has to cope with the workload left by the prior hospitalists.”

If a patient needs to be admitted, the patient first needs to be seen by the hospitalist during a consult. After this consult, the hospitalist admits the patient. Before this consult takes place, a patient cannot be ready for departure, hence the drop in the ready for departure rate.

(22)

(comparison 1 and 3). It seems to be caused by a lack of capacity at either the ED or supporting services during the morning, since these hurdles cannot be found to this extent on a low demanding morning.

To see how the Length-of-Stay behaves over the course of a day, the same comparison is made for the Expected Length-of-Stay of patients. The same comparisons are made as the ones mentioned in table 4. The most extreme comparison (LMLA-HMHA) is shown in figure 7, comparison 2 and 3 are shown in Appendix 4 and 5.

Figure 7 presents the Expected Length-of-Stay with the WIP over the course of the day. The expected Length-of-Stay is related to the WIP, as can be seen in figure 7 (and Appendix 4 and 5). An increase in WIP, results in an increase in the expected Length-of-Stay. The expected Length-of-Stay of the HMHA days is close to 3 hours and 30 minutes during the afternoon, while the expected Length-of-Stay for the LMLA is around 2 hours and 45 minutes during the afternoon.

Figure 7 Comparison Length-of-Stay LMLA-HMHA

(23)

the delay in the outflow in the morning. This delayed outflow leads to a buildup of WIP. From the analysis so far, the delayed outflow seems to be the result of: (i) lack of physicians in 70% of the days and (ii) a hick-up in the ready for departure rate due to wrongly coordinated supporting services. To be able to draw conclusions on causes of crowding at critical moments, the Work Sampling will be added later on. First we will take a look at the workload of physicians, to be able to create insights in the workload over the day and the possibility of responsiveness problems.

Workload analysis

Part of the in-depth analysis is the workload analysis per physician shift to see if shifts are overburdened. Figure 8 presents the workload per physician shift over time. Shift 3 is the shift of the physician assistant, which could be a reason for the lower workload for this shift. Physician assistants treat less acute patients and are able to create a shorter Length-of-Stay for these patients. On the other hand, ED physicians are responsible for more acute patients, which have a longer Length-of-Stay. It is important to stress that figure 9 does not take the patients in the waiting room into account, since a physician is not yet processing them.

Figure 8 Workload per physician shift

(24)

shift is because physicians are less likely to take on new patients at the end of their shift is to prevent hand-overs. Physicians made clear that hand-overs are prevented as much as possible to avoid liability issues. Levin et al. (2007) acknowledge the finding of a decreased workload at the end of a shift.

Assuming that the patient acuity mix does not differ per shift, each dayshift has a comparable workload. Consults with physicians made clear that physicians are capable of seeing an average of 3 patients per hour, since administration (e.g. ordering laboratory/imaging tests, read patient files etc.) takes time as well. In the patient-physician contact points three phases can be distinguished: (i) Start work-up of patient, (ii) review results and make clinical decision and (iii) discharge/admit patient. Regarding the amount of time every phase takes, a physician states:

“In most cases I think my biggest holdup is to find a moment to sit down and review all the results of

a patient. Try to put all the data together into one clinical picture. Decision making is time consuming, while if you are being pressured to see new patients that are coming in, it is hard to find the time to sit down and make a clinical decision.”

Another physician adds:

“We are hitting our maximum capacity on most days, which means you hit the stage where you do

not have time to see new patients.”

It can be concluded that the peak of each shift leads to a high workload for physicians, which retains physicians from decision-making with respect to the treatment of a patient. Burdened periods for physicians will lead to: (i) worse patient outcomes in terms of Length-of-Stay, mortality rate and readmission percentages (Levin et al., 2007), (ii) an increased WIP over the day, which results in (iii) longer burdened periods and cognitive overload of ED physicians.

Work Sampling

This section presents the Work Sampling results. In total 1693 data points are collected, both from patients in the waiting room as from patients in a treatment room. These 1693 data points result in an absolute error rate between 0.98% and 1.50% for a confidence level of 95%. The data set of the Work Sampling was split to compare the morning and the afternoon. The morning has an absolute error rate between 0.90% and 2.54%, while the afternoon has an absolute error rate between 1.41% and 2.61%. Both morning and afternoon had these error rates for a confidence level of 95%.

(25)

Thesis TOM/SCM David Buitelaar Figure 9 Comparison morning/afternoon Work Sampling

In order to see what the percentage of measurements per category is for both admitted and discharged patients, figure 10 presents a comparison between admitted and discharged patients both in morning and afternoon. Important to state is that although certain categories are measured more, a translation to the size of the actual waiting time cannot be made, since long events will be measured with a higher chance than events that take less time, due to the 30-minute measurement interval.

(26)

The number of measurements for waiting for consult and waiting for admittance, accounts for roughly 23% of the measurements for admitted patients. These measurements indicate that patients in need of admission are waiting for their consult and their admittance. This not only lengthens their Length-of-Stay but also withholds new patients to be moved to a bed and be treated by a physician. A physician stated:

“There is a big delay in getting a patient admitted. When we are finished and ready to admit a patient, a hospitalist takes quite some time to come to the ED. The longer patients sit here and wait to be admitted, the longer we cannot see new patients.”

Physicians cannot see new patients, since admitted patients that are not yet admitted occupy a treatment room while waiting for the hospitalist. Later in the afternoon, the ED is packed with patients waiting for admission resulting in a lack of available treatment rooms.

Figure 11 presents the number of repetitions the same patient has to wait for the same waiting category.

Figure 21 Number of repetitions per waiting category

(27)

Combining findings

In this section the findings from quantitative data and Work Sampling are combined. By combining the flow pattern with the cause of delays for patients a more detailed picture of the flow at an ED can be obtained.

Figure 12 Flow and delays in the treatment pathway

Figure 12 combines the cumulative inflow and outflow of days on which Work Sampling data was gathered with the delays in the treatment pathway according to the Work Sampling data. The Work Sampling data indicates how the WIP was built up. Note that each bar represents the average across the days that included work samples at this time. However, on some of the times only data was recorded for one day, which explains the difference with the actual average WIP from the throughput diagram.

Waiting for first contact and discharge are measured the most as the cause of waiting at that time. Waiting for first contact is measured the most during the morning, it declines throughout the late afternoon. This indicates that later in the afternoon, the waiting room starts to run empty which indicates that there is no lack of staffed beds late in the afternoon.

(28)

Combining the findings of the quantitative analysis and the Work Sampling indicates the same issues, but creates more insight in specific moments of the day, e.g. between 10:30 and 11:30 the average number of patients waiting for lab results increases from 2 to 3.5 while the number of patients waiting for 1st contact increases from 4.5 to 6. Both methodologies indicate the same delays in patient flow, but combining them will create a better insight. Consults with physicians are used to triangulate these findings. The conclusion can be drawn that the case ED copes with:

Lack of staff capacity – the ED has too few physicians available during the morning in 3 out of 4 type

of days. As a result of this, the WIP increases throughout these days. Due to this increase in WIP the Length-of-Stay of patients is longer, both in morning and afternoon.

Lack of supporting services – the ED is dependent on certain supporting services that help the ED

(29)

DISCUSSION

This study found a delayed outflow pattern in the morning, resulting in a WIP that is too high during the rest of the day. This degree of crowdedness leads to longer waiting times for patients in different stages in the patient pathway, e.g. waiting for 1st medical screening, waiting for available room and waiting for departure. Two causes were found to explain this delayed pattern: (i) lack of staff capacity at the ED, (ii) lack of reconciliation with supporting services, e.g. lab and hospitalists. The next section discusses main findings of literature and compares these with findings found in this research.

Main Findings

Main findings in American literature on causes of ED crowding that will be discussed here are: (i) inadequate staffing, (ii) problems with inpatient boarding and (iii) a high inflow of (non-urgent and urgent) patients. This research aimed to confirm or disprove causes found in literature and adding new causes while also comparing with Dutch findings for ED crowding.

Inadequate staffing – inadequate staffing is found in literature to affect crowding of EDs. Hoot &

Aronsky (2008) their literature review discusses the average workload per physician and nurse. The average workload per day was determined, while this Input/Output analysis distinguishes itself with its dynamic perspective, leading to insights in events at specific times of the day. It was found that in 70% of the days, physicians lacked in capacity during the morning. The WIP of patients during the morning builds up to a level that leads to problems with patient flow in the afternoon. However, adding physician capacity during the morning will also result in idle physicians in 30% of the days and the chance that physicians will feel less need to work during the morning, since more physicians are available. This could lead to responsiveness problems during the morning. Currently, the workload is balanced over the six different shifts, which indicates that there is no responsiveness issue. It is important to recognize that if an ED wants to solve this cause of crowding, it needs to use a time-series analysis in order to identify a lack of staff capacity. EDs need to consider the extra costs that additional staff capacity during critical moments when taking this decision.

Problems with inpatient boarding – literature on American EDs recognize the problems with

inpatient boarding as an important factor influencing the level of crowding, with access blocks exceeding eight hours. The literature mainly focuses on the number of inpatient beds available, but does not take into account the process of admission. This research found that the number of inpatient beds is not a determinant for ED crowding, but found that the admission process requires staff which is inadequately scheduled during critical times. Hospitalists are required for admission processes, but are (partly) unavailable for admission consults between 17:00 and 19:00, leading to patients occupying ED rooms after their care completion instead of being moved to an inpatient bed. It seems to be that the hospitalist staffing is inadequate during shift changes, leading to this access block.

Lack of reconciliation with supporting services – literature has not paid much attention to the

(30)

However, supporting services cannot only schedule to the demand of the ED but also have to take into account the demand of other inpatient departments of these supporting services.

Comparison with Dutch findings – Input/Output research performed in Dutch EDs found the delayed

outflow pattern in the morning to be caused by a lack of responsiveness during the morning and focused on increasing this responsiveness by e.g. adding flexible capacity and emphasizing the need of a quick response of ED staff. This research in an American ED however, did not find a lack of responsiveness to be a determinant in the level of crowding. The average workload of the five day-shifts, shows a comparable pattern in the buildup of workload. This shows that the Input/Output analysis continues to find new causes in different settings and contributes to the usability of the analysis.

Methodology

This study used the following methods: (i) quantitative analysis, (ii) Work Sampling, (iii) direct observations and (iv) consultation of physicians.

Quantitative analysis – the quantitative analysis was used to create insight in the patient flow

pattern at the ED. After initial analysis the data was split into four types of days, each containing a number of days. This split was based on the number of visitors per morning/afternoon. The data set was split at the median value to minimize the effect of outliers. However, splitting it at the median results in a small difference between a low demanding morning and a high demanding morning, i.e. one more patient can make the difference between low demanding and high demanding. The quantitative analysis however, distinguishes itself from other analysis with its dynamic viewpoint. This creates better insights than analysis based on averages, which implicitly assume stationary patient arrival. The workload analysis gave insight in the buildup of physician workload, which made it possible to conclude on the average responsiveness of ED physicians.

Work Sampling – 1693 valid data points are collected with Work Sampling on nine days. These are

collected between 7:00 and 19:00. However, a majority of the data points was collected between 9:00 and 17:00. This can have an impact on the validity of findings for the periods between 7:00-9:00 and 17:00-19:00. Work sampling contributed to the research by making it possible to find delays in the treatment pathway, creating insights in the collaboration between the ED and supporting services.

Consulting physicians – During the consults with physicians several questions in relation to their

work organization and the flow patterns are posed. Physicians could give biased answers during these consults. However, physicians will give their opinion where crowding can occur. These opinions are used in the analysis of the quantitative data, leading to triangulated findings. This contributes to the quality of the research.

Direct observations – Direct observation was used to understand how an American ED operates and

(31)

Limitations to the generalizability

The research is conducted at one American ED in a rural area in South Carolina. The size, type and location of a case ED can have an effect on causes of waiting times and degree of crowdedness in an ED. A comparison with other EDs with respect to the delayed outflow pattern in the morning and causes must therefore be made with caution. Other case studies recognize the lagged outflow pattern but focus on different solutions (Gelissen, 2016; Ten Have, 2016). Although the delayed outflow pattern is recognized in different settings, the causes can be different due to different characteristics of case EDs.

An ED in a rural area is expected to have different type of patients than a teaching or specialized hospital in a city with respect to number of visitors and their acuity. The number of visitors and their acuity have an expected big impact on the flow, since lower acuity patients are easier to treat, hence a shorter Length-of-Stay.

(32)

CONCLUSION

In this case study an analysis of the crowding of an ED has been performed. In this conclusion an answer to the research question is given in order to answer the research question:

“How can crowding in the American Emergency Department setting be explained from a dynamic viewpoint?”

The arrival pattern of patients can show undesirable variability, but an ED cannot fully manage its demand for care. In order to avoid crowding the ED needs to manage its supply of care. To avoid crowding, the crowding must first be explained and analyzed. In this research an in-depth Input/Output analysis is used to explain crowding in an American ED. Input/Output analysis was successfully used in Dutch EDs, but has not been combined with Work Sampling and Workload analysis to create triangulation and a more in-depth analysis of the case ED.

The study contributes to literature in identifying causes of crowding for an ED in a rural area in the United States. Crowding in the afternoon is often the result of a delayed patient outflow in the morning. Four different types of daily inflow patterns are found. Three out of the four inflow patterns lead to higher WIP, especially in the afternoon due to a delayed outflow pattern in the morning in combination with higher inflow. The delayed patient outflow in the morning seems to be the result of: (i) a lack of available physicians at critical times and (ii) a lack of capacity at supporting services, i.e. lab and hospitalists, on critical times over the course of the day.

The study stresses the importance of an aligned organization in order to create an even flow of patients, such that crowding and waiting times can be avoided as much as possible. The Work Sampling showed that (i) waiting for 1st contact, (ii) waiting for ordered test to be drawn and (iii) waiting for results of tests are the top three causes of patient waiting. Waiting for 1st contact was measured less during afternoon in comparison with the morning, while waiting for ordered test to be drawn was measured more in the afternoon. These results indicate that a lack of physicians is measured more during the morning, while lab cannot process all the test results within a reasonable amount of time in the afternoon. Even small delays in a patient pathway have an effect on the WIP level at an ED and can lead to crowding and longer waiting times.

(33)

Recommendations for the researched ED

In order to cope with the delayed outflow of patients, additional capacity is needed during the morning in 70% of the days. The ED needs to weigh the importance of a shorter Length-of-Stay in 70% of the days with the spilled capacity during 30% of the days. Urgency must be paid to the possibility of a decreased responsiveness if physician capacity is added during the morning.

It is important to recognize that the ED is dependent on supporting services as hospitalists and the lab. A delay in supporting services can lead to delays in treatments for patients and crowding of the ED. Waiting patients are occupying an ED bed such that new patients cannot be moved to a bed yet. This will specifically contribute to crowding if many patients arrive. The patient flow is dependent on a proper collaboration of all ED personnel and supporting services. It is important to stress that the complete hospital has to cooperate in order to have a properly working ED with a low level of crowding and short waiting times.

Recommendations for further research

In order to understand the delayed outflow pattern better and to see if this flow pattern is similar in other EDs, further research is needed. This research needs to take into account different sizes and types of EDs to define causes of crowding better for different types of EDs.

The Work Sampling can be combined with a time-and-motion study, such that more insight is created in the exact time a patient spends in each stage of the patient pathway. There still exists a lack of knowledge regarding the time a patient spends in each phase of the patient pathway. A time-and-motion study uses continuous observation and is therefore much more labor-intensive (Ho & Pape, 2001), since it requires an observer to stay with one patient during the complete patient pathway. Time-and-motion study on its own will probably not result in representativeness.

(34)

REFERENCES

Asplin, B. R., Magid, D. J., Rhodes, K. V., Solberg, L. I., Lurie, N., & Camargo, C. a. (2003). A conceptual model of emergency department crowding. Annals of Emergency Medicine, 42(August), 173– 180. doi:10.1067/mem.2003.302

Bazzoli, G. J., Brewster, L. R., Liu, G., & Kuo, S. (2003). Does U.S. Hospital Capacity Need To Be Expanded? Health Affairs, 6(6), 40–54. doi:10.1377/hlthaff.22.6.40

Forero, R., Hillman, K. M., Mccarthy, S., Fatovich, D. M., Joseph, A. P., & Richardson, D. B. (2010). Access block and ED overcrowding. Emergency Medicine Australasia, 22, 119–135.

doi:10.1111/j.1742-6723.2010.01270.x

Frank, I. C. (2001). ED crowding and diversion: Strategies and concerns from across the United States.

Journal of Emergency Nursing, 27(December), 559–565. doi:10.1067/men.2001.120244

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

Han, J. H., Zhou, C., France, D. J., Zhong, S., Jones, I., Storrow, A. B., & Aronsky, D. (2007). The Effect of Emergency Department Expansion on Emergency Department Overcrowding. Academic

Emergency Medicine, 14, 338–343. doi:10.1197/j.aem.2006.12.005

Handel, D. a., Hilton, J. a., Ward, M. J., Rabin, E., Zwemer, F. L., & Pines, J. M. (2010). Emergency department throughput, crowding, and financial outcomes for hospitals. Academic Emergency

Medicine, 17, 840–847. doi:10.1111/j.1553-2712.2010.00814.x

Higginson, I., Whyatt, J., & Silvester, K. (2011). Demand and capacity planning in the emergency department : how to do it. Emergency Medicine Journal : EMJ, 28, 128–135.

doi:10.1136/emj.2009.087411

Ho, C., & Pape, E. S. (2001). Continuous observation work sampling and its verification. Work Study,

50, 23–30. doi:10.1108/00438020110363757

Hoot, N. R., & Aronsky, D. (2008). Systematic Review of Emergency Department Crowding: Causes, Effects, and Solutions. Annals of Emergency Medicine, 52.

doi:10.1016/j.annemergmed.2008.03.014

Horwitz, L. I., Green, J., & Bradley, E. H. (2010). US Emergency Department Performance on Wait Time and Length of Visit. Annals of Emergency Medicine, 55(2), 133–141.

doi:10.1016/j.annemergmed.2009.07.023

Koomen, A. (2016). The influence of emergency department responsiveness on throughput times. MSc Thesis University of Groningen.

Lane, D. C., Monefeldt, C., & Rosenhead, J. (2000). Looking in the wrong place for healthcare

improvements: a system dynamics study of an accident and emergency department. Journal of

Operational Research Society, 51(5), 518–531. doi:10.1057/palgrave.jors.2600892

Levin, S., Aronsky, D., Hemphill, R., Han, J., Slagle, J., & France, D. J. (2007). Shifting Toward Balance : Measuring the Distribution of Workload Among Emergency Physician Teams. Annals of

Emergency Medicine, 50(4), 419–423. doi:10.1016/j.annemergmed.2007.04.007

Mackway-Jones, K., Marsden, J., Windle, J., & Harris, N. (2013). Advanced Life Support Group:

Emergency Triage. Somerset, NJ: John Wiley & Sons.

Meredith, J. (1998). Building operations management theory through case and field research. Journal

(35)

Miró, O., Sánchez, M., Espinosa, G., Coll-Vinent, B., Bragulat, E., & Millá, J. (2003). Analysis of patient flow in the emergency department and the effect of an extensive reorganisation. Emergency

Medicine Journal : EMJ, 20, 143–148; discussion 148. doi:10.1136/emj.20.2.143

Olshaker, J. S., & Rathlev, N. K. (2006). Emergency Department overcrowding and ambulance diversion: The impact and potential solutions of extended boarding of admitted patients in the Emergency Department. Journal of Emergency Medicine, 30(3), 351–356.

doi:10.1016/j.jemermed.2005.05.023

Ospina, M. B., Bond, K., Schull, M., Innes, G., Blitz, S., & Rowe, B. H. (2007). Key indicators of overcrowding in Canadian emergency departments: A Delphi study. Canadian Journal of

Emergency Medicine, 9(5), 339–346. doi:3d08c1e6d8874781b07835eaae873420 [pii]

Pines, J. M., Hilton, J. a., Weber, E. J., Alkemade, A. J., Al Shabanah, H., Anderson, P. D., … Schull, M. J. (2011). International perspectives on emergency department crowding. Academic Emergency

Medicine, 18, 1358–1370. doi:10.1111/j.1553-2712.2011.01235.x

Redlener, I., & Grant, R. (2009). America’s Safety Net and Health Care Reform — What Lies Ahead?

The New England Journal of Medicine, 361(23), 2201–2204.

Roukema, J., Steyerberg, E. W., van Meurs, a, Ruige, M., van der Lei, J., & Moll, H. a. (2006). Validity of the Manchester Triage System in paediatric emergency care. Emergency Medicine Journal :

EMJ, 23, 906–910. doi:10.1136/emj.2006.038877

Saghafian, S., Hopp, W. J., Van Oyen, M. P., Desmond, J. S., & Kronick, S. L. (2012). Patient Streaming As a Mechanism for Improving Responsiveness in Emergency Departments. Operations

Research, 60(February 2016), 1080–1097. doi:10.1287/opre.1120.1096

Schull, M. J., Lazier, K., Vermeulen, M., Mawhinney, S., & Morrison, L. J. (2003). Emergency

department contributors to ambulance diversion: A quantitative analysis. Annals of Emergency

Medicine, 41(4), 467–476. doi:10.1067/mem.2003.23

Sinreich, D., Jabali, O., & Dellaert, N. P. (2012). Reducing emergency department waiting times by adjusting work shifts considering patient visits to multiple care providers. IIE Transactions, 44, 163–180. doi:10.1080/0740817X.2011.609875

Tang, N., Stein, J., Hsia, R. Y., Maselli, J. H., & Gonzales, R. (2010). Trends and characteristics of US emergency department visits, 1997-2007. JAMA : The Journal of the American Medical

Association, 304(6), 664–670. doi:10.1001/jama.2010.1112

Ten Have, M. (2016). Lagged output in ED. MSc Thesis University of Groningen.

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

patient waiting times. MSc Thesis University of Groningen.

(36)

APPENDIX

Treatment Room Start Time Patient1

Patient waiting for3 Additional info for waiting

FAST TRACK 1 FAST TRACK 2 FAST TRACK 3 FAST TRACK 4 EXAM ROOM 5 EXAM ROOM 6 EXAM ROOM 7 EXAM ROOM 8 CARDIAC ROOM 9 CARDIAC ROOM 10 TRAUMA ROOM 11 TRAUMA ROOM 12 CAST ROOM 13 OB GYN/EXAM 14 OB GYN/EXAM 15 ENT ROOM 16 EYE ROOM 17 EXAM ROOM 18 EXAM ROOM 19 PSYCH ROOM 20 HALLWAY BED 1 HALLWAY BED 2

Appendix 1 Work Sampling Form

1

Start time treatment patient; used for finding back patient in data set.

3

(1) 1st contact, (2) Lab/Med order, (3) Results Lab/Med, (4) Effect treatment, (5) Departure, (6) Physician/nurse

(37)

Thesis TOM/SCM David Buitelaar Appendix 2 Comparison Flow LMLA-LMHA

(38)

Thesis TOM/SCM David Buitelaar Appendix 4 Comparison Expected Length-of-Stay LMLA-LMHA