The impact of COVID-19 on emergency department crowding

(1)

The impact of COVID-19 on emergency

department crowding

Author

Tamara Kamp

Student Number:

S2554879

MSc Business Administration - Health

Faculty of Economics and Business

University of Groningen

1

st

_supervisor

Dr. M.J. Land

University of Groningen

2

nd

_supervisor

Prof. Dr. T.J. van der Vaart

(2)

2

List of abbreviations

Acute Medical Unit AMU

Electronic Patients Information Chart EPIC

Emergency Department ED

Intensive Care Unit ICU

Inter Quartile Range IQR

Length Of Stay LOS

Manchester Triage System MTS

Severe Acute Respiratory Syndrome SARS University Medical Centre Groningen UMCG

(5)

5

Abstract

Background: Crowding problems at the emergency department (ED) were known to challenge healthcare systems in many countries before COVID-19. Crowding has the potential to exponentially increase during a viral outbreak. The growing pressure on healthcare makes it increasingly important to understand the impact of COVID-19 on crowding, however up to now little is known about the impact of COVID-19 on patient flow in the ED. Aim: This study aimed to examine the impact of COVID-19 on ED crowding. We aimed to contribute to realizing improvements of ED patient flow in future pandemics and in normal situations. Methods: A single-case study at the University Medical Centre Groningen (UMCG), a large tertiary teaching hospital in the north of the Netherlands, was conducted. Patient flow was analysed by the use of quantitative historical data during two periods (pre-COVID-19, 01-03-2018 until 01-03-2020 and COVID-19, 01-04-2020 until 01-07-2020). Data was extracted from the electronic patients information chart of the ED of the case hospital. Analyses on factors causing crowding were divided into three interdependent components inflow, throughput and outflow. Results: 50,246 patients (Pre-COVID-19; median age 55, 55% male ) and 4,565 patients (COVID-19; median age 58.5, 57% male) were included. During COVID-19 there was less crowding despite the challenges posed by the pandemic. This was reflected in a lower work-in-progress, a small reduction in length-of-stay and less patient arrivals. Furthermore, COVID-19 led to an altered

(6)

6

1. Introduction

The rapid spread of the COVID-19 virus has resulted in a worldwide pandemic. The impact of the COVID-19 pandemic goes beyond individual mortality rates, it highly affects hospitals and in

particular the emergency department (ED) (Cavallo, Donoho, & Forman, 2020). Healthcare systems in many countries were challenged by crowding problems at the ED before COVID-19 (Pines et al., 2011). Van der Linden et al. (2013) showed that 68% of Dutch EDs experience crowding regularly, several times a week or even daily.Crowding has the potential to exponentially increase during a viral outbreak (Schanzer & Schwartz, 2013). Currently, it has been estimated that there are 1.7 million undiscovered viruses in both birds and mammals of which up to 49 percent could infect humans, possibly leading to a future pandemic (Daszak et al., 2020). Research can and should help to investigate and understand the impact of COVID-19 on crowding at the ED to improve patient flow during both normal situations and future pandemics.

The ED is an important component of the hospital. Patients who are treated in the ED are in need of care as soon as possible with the highest quality. The quality of the services provided in the ED depends on a variety of characteristics, for example variability of patient flow, workload and diversity of case complexity (Venkat et al., 2015). The main cause of patient flow impairments is described to be crowding. Causes of crowding at the ED are multifactorial. Crowding at the ED results in patient dissatisfaction, inability of staff to adhere to guideline-recommended treatment and higher patient risks, such as complications and mortality (Morley et al., 2018). The formation of queues is therefore very undesirable (Rosmulder & Luitse, 2011).

Hospitals needed to implement policy changes, that affected EDs, in order to diminish the consequences of COVID-19. The following precaution measures were taken, aimed at reducing the pressure on the ED: Proactive steps to manage the increased patient inflow; Different staffing schedules to influence throughput; Stricter throughput regulations to transfer COVID-19 suspected patients from the ED as soon as possible; Rearrangement of hospital wards, to influence the outflow at the ED (Boyle & Henderson, 2020; Health System Response Monitor, 2020; Lang, 2020). Hence, COVID-19 will highly affect the ED resulting in reorganization and reformation of resources and patient flow. What was the effect of changes that were made to the ED with regard to e.g. staffing schedules, altered regulations and rearrangement of wards to influence patient flow? The impact of this reorganization on crowding and length of stay (LOS) is not clear and can go one of two ways. One could expect a higher process efficiency and thus a shorter LOS. However, from for example

Queueing Theory it is known that splitting a single queue and generating two separate sets of resources (‘infected’ and ‘non-infected’) will lead to a substantial increase in average waiting time (Bocharov, D’Apice, & Pechinkin, 2011).

It is important to understand the impact of COVID-19, because it will enable EDs in the future to manage and improve patient flow. This understanding will enable them to use their resources efficiently and provide timely care to patients during both a future pandemic and the normal situation and thus improve overall performance. It will help us to understand which strategic adjustments should be reused and which should be altered. This becomes increasingly important due to the growing pressure on healthcare caused by increasing demand and declining financial resources. The research question therefore is:

How did COVID-19 impact ED crowding?

Aim of this study is twofold, since we want to better understand and improve patient flow in the ED during both normal situations and future pandemics. We aim to contribute to realizing

(7)

7

2. Theoretical background

This chapter will summarize the main findings regarding crowding. First, literature on crowding and its consequences will be discussed, followed by the causes. Thereafter, queueing and the relation with ED crowding will be discussed. Lastly, literature about the (expected) impact of COVID-19 on the ED and the subsequent sub-questions will be discussed.

2.1 Crowding: definition and consequences.

Crowding is widely known to have a negative influence on patient flow, however there is no universal description of crowding. Crowding and overcrowding are often used interchangeably, referring to the same phenomenon (Rasouli et al., 2019). One of the most cited definitions of crowding in literature is the definition from the American College of Emergency Physicians: “crowding occurs when the identified need for emergency services exceeds available resources for patient care in the emergency department, hospital, or both” (ACEP, 2019). To ease the operationalization of this broad definition, crowding can be described as a mismatch. Crowding occurs when there is a mismatch between supply and demand of emergency services within a reasonable amount of time which hinders the operations of the ED (Venkat et al., 2015). This imbalance negatively affects waiting times for treatment. Besides the reduction in patients’ perceived quality of care, long waiting times will also increase crowding which can negatively affect patient outcomes (Davis et al., 2019; Morley et al., 2018; Rosmulder & Luitse, 2011). Crowding affects not only patients but hospital-staff as well. It contributes to stress and increases the incidence of interruptions and multitasking. Stress and interruptions are known factors of productivity reduction and decrease in effectiveness (Bellow & Gillespie, 2014; Kc & Terwiesch, 2009). Furthermore, higher absenteeism, lower staff morale and satisfaction as a result of high workload are associated with crowding (Weissman et al., 2007). Overall, crowding can be seen as a system-wide problem since it influences the entire chain of care in the hospital and can be reduced by implementation of effective operational management (Davis et al., 2019; Langabeer, 2008).

2.2 Crowding: causality

According to the conceptual model of ED crowding by Asplin et al. (2003), the factors causing crowding in an ED system can be divided into three interdependent components: in flow, throughput and outflow.

Inflow factors refer to events, system characteristics or conditions that contribute to the demand for ED services. This includes patient volume, patient mix (case type and acuity) and arrival pattern. A study conducted among 350 ED’s in the United States illustrated that the d emand for emergency services has been rising, additionally patients expectations of ED care are increasing (Pitts et al., 2012). Besides the increasing patient volumes, patient complexity and acuity is increasing as well. It is well-known that patient arrival patterns drive systems for scheduling resources and staff. Previous research showed that these arrival patterns are fairly predictable (Hall, 2006). Arrival to the ED showed to be stochastic, have a time-dependent rate factor and is influenced by external factors such as weekday, temperature and holidays (Vile et al., 2017).

Throughput factors are activities within the ED that could hinder patient flow. A variety of throughput factors are associated with crowding: medical and nursing staff shortages, waiting time for triage, waiting time for physician’s examination, waiting time for medical diagnostics (e.g. lab tests and medical imaging) and ED design limitations resulting in e.g. treatment room scarcity (Morley et al., 2018; Schull, Slaughter, & Redelmeier, 2002).

(8)

8 Figure 2.1 - ED patient flow

2.3 Queueing

Previous researchers have shown that Queueing Theory can be useful in healthcare, especially to study the performance of the ED in terms of patient flow and the availability of resources (De Véricourt & Jennings, 2011; Fomundam & Herrmann, 2007; Mayhew & Smith, 2008; McClain, 1976). Queueing Theory aims to reduce costs through reduction of inefficiencies and delays (Lakshmi & Appa Iyer, 2013). The ED can be seen as a network of queues and different types of services. Patients arrive, wait for service (triage, medical tests, diagnosis) and after the creation of a treatment policy they will be discharged from the ED (to home or to a hospital ward) (Figure 2.1). Triage is the process of assessing patients’ needs for care, which is used for setting queue priorities among patients and mostly done according to the Manchester Triage System (MTS) with 5 levels of urgency (Table 2.1) (Mackway-Jones et al., 2013). Triage is followed by diagnosis, based on physician examination, lab tests and medical imaging. The next step is to formulate the treatment policy. After treatment patient discharge from the ED can follow different routings: 1. Patient is sent home, 2. Patient is admitted to a hospital ward, 3. Patient is referred to another hospital.

Waiting time patterns are an effective method to support and establish management decisions (Vass & Szabo, 2015). Hospitals operating at high levels of utilization showed to have many

operational implications, such as long waiting times (Green, 2006; Smith‐Daniels, Schweikhart, & Smith‐Daniels, 1988). A fundamental assumption in the current queueing literature is that service time, i.e. the time it takes a patient to flow through the ED, is independent of process state and the current workload (Kc & Terwiesch, 2009). However, Kc and Terwiesch (2009) showed that service time in the hospital decreased during an increasing workload and concluded that level of capacity is adaptive to higher workload levels. Thus, uncertainties exist about the applicability of queueing theory on ED crowding during both the normal situation and a pandemic. Therefore, the impact of generating two separate sets of resources during COVID-19 on waiting time at the ED is unclear.

Table 2.1 Manchester Triage System (MTS) levels of urgency (Mackway-Jones et al., 2013)

Level Urgency Colour Time until seen by physician

1 Immediate Red 0 minutes

2 Very urgent Orange Within 10 minutes

3 Urgent Yellow Within 1 hour

4 Standard Green Within 2 hours

5 Non-urgent Blue Within 4 hours

Inflow

• Patient requests care: • GP referral •Ambulance • walk-in Throughput • Triage •Treatmentroom assignment

• Diagnosis; Physician examination, lab tests, medical imaging •Treatment policy

Outflow

(9)

9 2.4 COVID-19 influences on EDs and research sub-questions

The control of the spread of rapidly transmissible respiratory infections, such as COVID-19, is complicated by crowded EDs, especially during outbreaks and pandemics. Little is known about the impact of COVID-19 on crowding and patient flow in the ED. To answer the main research question, ‘How did COVID-19 impact ED crowding’, thorough understanding of the ED performance is needed. This is done according to three sub-questions on inflow, throughput and outflow based on the

conceptual model of ED crowding by Asplin et al. (2003).

It is known that COVID-19 caused an increase in patient inflow with respiratory symptoms. In contrast, a recent study showed a steep decrease in the number of referrals of patients with a stroke in a hospital in Canada (Bres Bullrich et al., 2020). The authors suggest that the lockdown policy may have impacted the willingness of physicians to refer patients.

1. How did COVID-19 impact the inflow pattern and distribution of patients at the ED?

We will investigate the impact of changes made in the ED, in the light of COVID-19, on ED crowding to understand the formation of queues and bottlenecks in the system. As mentioned before, demand exceeding capacity is a key characteristic of crowding resulting in the formation of queues in several elements of the ED system. Based on the limited COVID-19 related research we expect a change in the patient arrival pattern during the pandemic, leading to an improved match between supply and demand.

Strategic changes to resource availability and allocation were made during COVID-19. The COVID-19 pandemic led to a stretch in hospital resources world-wide (Mitchell & Banks, 2020). It is known that resources such as medical diagnostics (e.g. lab tests and medical imaging) and treatment rooms may influence the throughput at the ED both during a normal situation and a pandemic. Waiting time due to limited availability or inefficient allocation of resources are known causes of patient flow hindrance (Morley et al., 2018). Another modification, resulting from the discontinuing regular care, is an increase in staff. Furthermore, queueing theory suggests that dedicated resources (e.g. COVID-19 treatment rooms and COVID-19 staff) should have led to increased lead-times, since the pooling synergy of shared resources evaporates. However, the altered work flow could also have led to a higher process efficiency and thus to shorter lead-times.

2. How did COVID-19 influence throughput at the ED?

We will use data on waiting times for diagnostic tests and the occupation and utilization of treatment rooms to investigate the availability and utilization of resources during COVID-19. This allows us to investigate if COVID-19 led to an impaired resource availability and hampered utilization, leading to crowding. Based on queueing theory we hypothesize that generating two separate sets of resources, i.e. ‘infected’ and ‘non-infected’, will lead to a substantial increase in waiting time and thus in a changed resource availability and utilization or an increase in crowding (Bocharov et al., 2011).

A previous study conducted in an ED in Sweden showed that the LOS decreased during COVID-19 (af Ugglas et al., 2020). A reduction in LOS may be caused by both throughput factors or outflow factors. For example, a reduction in boarding time (the time between the bed request and the moment a patient leaves the ED) will lead to a reduced LOS. During COVID-19 the in-hospital capacity

transformed, which could have changed patient admission from the ED to the hospital wards. Additionally, a more critical evaluation of the need to admit patients or a changed patient mix could have led to a changed outflow pattern. A study conducted in the United States showed that during COVID-19 a lower percentage of patients was discharged from the ED and thus a higher percentage was admitted (Baugh et al., 2020). An increased proportion of patients being admitted can hamper outflow of the ED.

3. How did COVID-19 impact the outflow pattern and mix of patients at the ED?

(10)

(11)

11

3. Methodology

This method section will outline the approach taken to answer the research (sub)question(s) of this research and to support us to better understand how crowding can be addressed.

3.1 Study design

To investigate the impact of COVID-19 on ED crowding a single case study will be conducted at a large tertiary university hospital (University Medical Centre Groningen, UMCG) in the Netherlands. A case study enables to study a phenomenon in its natural setting (Eisenhardt, 1989), focusing on the embeddedness of a phenomenon in real-life context (Blumberg, Cooper, & Schindler, 2011). It is a way to get more understanding of the complexity and nature of the investigated phenomenon (Karlsson, 2009), in this case the impact of COVID-19 on ED crowding. It is important to take into account that a case study limits the generalizability. Interval validity and reliability are guaranteed by the use of multiple sources of data (i.e. triangulation) to provide a more complete, holistic and stronger confirmation of the observed phenomena, leading to an increase of internal validity and reliability (Harvey & MacDonald, 1993; Karlsson, 2009). In this study both quantitative as well as qualitative data is used.

3.2 Hospital case description

The study is performed at the UMCG, a tertiary university hospital with approximately 34,000 ED visits each year. Investigation of this specific ED is chosen because UMCG is the largest hospital in the northern region of the Netherlands and fulfilled an important role in providing care to COVID-19 patients. Most ED’s have a similar lay-out: a waiting room, a triage room, a trauma room and various treatment rooms. The UMCG has 22 treatment rooms, which can be expanded with 6 hallway beds and 5 observatory beds (see Table 1, Appendix A for an overview of the treatment rooms). If a

condition requires more investigation, diagnostics or longer observation, a patient can be transferred to the observatory for a few hours. Furthermore, a patient can temporarily be sent back to the waiting room or can be moved to a different department such as the radiology department for additional diagnostics. Patient flow encompasses the movement of patients throughout the ED processes (Figure. 2.1). The process of patients at the ED of the UMCG is similar to the process discussed in section 2.3, which is tracked using the patient information system EPIC (electronic patients information chart). 3.3 Data collection

The data collected consists of different types, namely quantitative data (two sets of historical data on patient flow measured during two periods) and qualitative data (professionals’ view on observed effects).

For analysing patient flow, two EPIC datasets of the UMCG were used. The analyses of these datasets were subdivided based on the research question and sub questions into crowding level, inflow, throughput, and outflow. The first dataset included information on crowding level, inflow, throughput and outflow of patients. The second dataset included detailed information on throughput factors i.e. treatment rooms and diagnostics. These datasets consist of timestamps for all patients arriving at the ED.

Two time periods were compared to each other, namely pre-19 and during COVID-19. In this study pre-COVID-19 was defined as the period from 01-03-2018 until 01-03-2020 and COVID-19 was defined as the period from 01-04-2020 until 01-07-2020. The decision of the pre-COVID-19 period was based on the availability of data until 01-03-2018 and lockdown regulations starting from half March 2020. Consequently, the start of the COVID-19 period was marked by the start of lock-down regulations (half March) and the relaxation of the regulations starting the first of July. Both datasets consisted of timestamps for all patients arriving at the ED.

(12)

12 outcomes in terms of inflow, throughput and outflow. Furthermore, the observed phenomena were discussed in dept with two ED physicians (E. ter Avest and T.M. Stevens-Stolmeijer).

3.4 Data cleaning

The first dataset (on crowding level, inflow, throughput and outflow) was divided into pre-COVID-19 and COVID-19. The pre-COVID-19 period consisted of 50,513 patients and the COVID-19 period of 4,601 patients. In both periods patients were excluded if they had a LOS of more than 18 hours (validated by the involved ED physician), because these measurements can be seen as incorrect. Secondly, patients with an arrival time later than their departure time were excluded as well. Lastly, patients with an abnormal age (>120yrs) were excluded. In the end the pre-COVID-19 period consisted of 50,247 patients and the COVID-19 period of 4,565 patients (Figure 1, Appendix B).

The second dataset (on throughput i.e. treatment rooms and diagnostics) was also divided into pre-COVID-19 and COVID-19 as mentioned under 3.2. Both periods started with a similar number of patients, however exclusion was based on different criteria. Treatment room occupation time stamps shorter than 5 minutes were excluded, since it was expected that this was the result of an

administrative error. For the pre-19 period this were 13,726 registrations and for the COVID-19 this were 1,802 registrations. Furthermore, negative diagnostics registrations, with a results

timestamp before request time, were excluded. Pre-COVID-19 this encompassed 8,667 measurements and during COVID-19 this were 878 registrations. The overall dataset was checked for accurateness by the involved physician. Nonetheless, minor administrative errors could still be present in the datasets.

3.5 Data analysis

All analysis described in this part were conducted for both periods, i.e. pre-19 and COVID-19. The observed outcomes from both periods were compared to analyse differences, to investigate the impact of COVID-19 on crowding and how this can be used during both normal situations and future pandemics. The total amount of patients differed between the periods, hence the majority of the analyses were carried out relative to the total number of patients in the dataset. To gain insight into the baseline characteristics of the studied population the age and gender distribution were collected and examined. Furthermore, when analysing patterns per hour of the day, patients were grouped. For instance, a patient arriving (or departing) between 13:30 and 14:30 was coded as 14:00. The rest of the data analysis will be discussed according the main research question and the according sub questions (as mentioned under 2.4). First, I will discuss the analyses on crowding level in general. Thereafter, the analysis on factors which could explain the differences in crowding levels. These analyses are divided in inflow, throughput and outflow.

3.5.1 Crowding levels

Analyses on crowding levels where conducted using the following metrics: work in progress (WIP), length of stay (LOS), patients within the national norm and the overall inflow-outflow pattern. WIP was measured as the number of patients already present at the ED at the moment a particular patient arrived. This was determined by counting the patients of whom the time of arrival was earlier than a particular patient arrival and the count of patients of whom discharge was earlier than the particular patient arrived. WIP and the distributions of WIP were graphed to analyse crowded moments. Crowding at the ED was defined as 80% or more treatment rooms (≥17 beds) being occupied, based on the perception of crowding by the involved ED physician. LOS was defined as the time between arrival and departure of a particular patient. The national norm for LOS is four hours. Average LOS was displayed for all patients per triage colour and per age category. Average LOS was used as performance indictor to visualize the patients seen within the national norm. Furthermore, the

(13)

13 3.5.2 Inflow

To investigate how COVID-19 influenced the inflow pattern and distribution of patients, comparative analyses were conducted on arrival specialism, patient mix and arrival patterns. An altered inflow pattern and distribution is known as a cause of crowding, leading to longer waiting times and the formation of queues further down the ED system. To compare patient complexity and urge ncy the way of arrival (e.g. via ambulance or via helicopter) was displayed. Arrival specialty was analysed to further investigate patient mix. The average arrival rate per hour was calculated to investigate

differences in arrival patterns between the two periods. An arrival pattern diagram was made showing the number of patients arriving per hour of the day. To detect noteworthy arrival patterns, this was plotted per day of the week. Additionally, we conducted analyses with regard to number of patients arriving per weekday and per month to observe patterns. A high number of patients arriving within a 1-hour timeframe puts pressure on the availability of capacity and resources and influences the throughput and outflow, resulting in the formation of queues in several elements of the ED system. The consulted ED physician indicated that 8 or more patients arriving per hour is considered as a high inflow. To investigate the frequency of occurrence of high patient arrivals within an 1-hour timeframe, constraining the availability of capacity and resources, graphs were made of 8 or more patient arrivals per hour. This shows the frequency of occurrence of a high inflow, resulting in high pressure on the ED system. Lastly, we analysed triage colour to gain insight into urgency. The required care a patient needs is important and can influence the LOS. Patients with high urgency require more staff and specific equipment, resulting in less available capacity for other patients.

3.5.3 Throughput

Throughput encompasses everything that happens within the ED system. It is known that resources influence the throughput at the ED and limited availability or inefficient allocation can lead to patient flow hindrance (Morley et al., 2018). Analyses were conducted on triage time, the occupation and utilization of treatment rooms and waiting times for diagnostic tests, to analyse differences due to the generation of two separate sets of resources (i.e. ‘infected’ and ‘non-infected’). The national norm for triage encounters 15 minutes. Average triage times were displayed for all patients per triage colour and were used as performance indictors to visualize the patients seen within the national norm. According to the ED physicians at the UMCG treatment room availability is one of the main bottlenecks.

Therefore, to investigate treatment room availability the average occupation time per room and utilization of the rooms were plotted. Utilization was calculated as a percentage of the total hours a treatment room is occupied divided by the total amount of hours in the dataset. Results on treatment room occupation time and utilization will help to understand how COVID-19 impacted treatment room availability. Moreover, the amount a patient changed treatment room was plotted, which aided in visualizing if healthcare was more efficient during COVID-19. Lastly, the median waiting time for diagnostics (laboratory, imaging, CT) and their utilization were displayed to investigate if COVID-19 led to impaired availability.

3.5.4 Outflow

Outflow aspects can be a cause of crowding and influence throughput aspects. Departure patterns were plotted per hour of the day and per weekday to detect particularities in outflow patterns. Furthermore, outflow patient mix was analysed by the proportion of patients admitted to an in-hospital ward. Outflow mix can be an explanation for altered outflow patterns. Previous studies showed that high boarding times are associated with decreased ED efficiency (Napoli, Ali, Lawrence, & Baird, 2020).Therefore, the admittance process was further analysed by boarding time. We determined boarding time by the time between the bed request and the moment a patient leaves the ED. Besides boarding time distribution, we will gain insight at which moment boarding times increase by means of a combined graph of boarding time and WIP.

3.5.5 Qualitative data

(14)

(15)

15

4. Results

This chapter describes the outcomes of the case study conducted at UMCG about the impact of COVID-19 on ED crowding. The results are divided in five subsections, according to the research- and sub question(s), to analyse and understand the impact of COVID-19 on ED crowding. First, we will provide an overview of the baseline characteristics of the studied population. Secondly, results on crowding levels will be shown. Thereafter, we will present which factors could explain the observed crowding levels divided in inflow, throughput and outflow. Qualitative data, gathered during and after the presentations for the ED department and discussions with the ED physicians to gain better

understanding of the impact of COVID-19 on ED crowding, will be described immediately after the relevant items. Per subsection a short conclusion will be provided. Within this section the pre -COVID-19 period will be referred to as the normal situation.

4.1 Baseline study population characteristics.

To create a baseline reference for the analyses described in section 3 we determined the baseline characteristics of the studied population (Table 4.1). From the 50,513 patients who were admitted to the ED during the normal situation (1-3-2018 until 1-3-2020), 50,247 were included in the study. The sample consisted of 27,746 (55.2%) males and 22,501 (44.8%) females. The median age of the total patient group was 55 years (interquartile range, IQR 30 - 69). From the 4,601 patients who were admitted to the ED during the COVID-19 situation (1-4-2020 until 1-7-2020), 4,565 were included in the study. During the COVID-19 period the proportion of males and the median age slightly increased. The sample consisted of 2,600 (57.0 %) males and 1,965 (43.0 %) females. The median age of the total patient group was 58.5 years (IQR 38 – 71).

The distribution of age categories showed a decrease in the proportion of patients aged between 0-19 and 20-39 years during COVID-19 compared to the normal situation. An increase is observed in the proportion of patients in the 40-59, 60-79 and 80+ age group, with the highest increase in the 60-79 years age group (+5.1%) (Table 4.1). The increase in age (median and age categories 40-59, 60-79 and 80+) can be explained by older patients being more susceptible to the COVID-19 virus. Furthermore, it is known that patients with multiple comorbidities, often connected to a higher age, are associated with a severity of a COVID-19 infection leading to an ED presentation. In addition, males infected with COVID-19 are more at risk for worse outcomes, explaining the higher proportion of presentations at the ED (Jin et al., 2020).

Table 4.1: Baseline study population characteristics.

Pre-COVID-19 COVID-19

Variables Total (N=50,247) Total (N=4,565)

Male/female (n (%)) 27,745 (55.2) / 22,501 (44.8) 2,600 (57.0) / 1,966 (43.0)

Age (median, IQR) 55 (30 – 69) 58.5 (38 -71)

Age categories (n (%)) 0-19 5,321 (10.6) 390 (8.5) 20-39 11,195 (22.3) 824 (18.1) 40-59 12,572 (25.0) 1,158 (25.4) 60-79 17,277 (34.4) 1,802 (39.5) 80+ 3,882 (7.7) 391 (8.6)

(16)

16 4.2 Crowding levels

In this section analyses on crowding levels are presented using the following metrics: WIP, LOS, patients seen within the national norm and inflow/outflow pattern.

The maximum WIP measured during both periods exceeded 22 patients, the amount of treatment rooms available at the ED. This high maximum number of patients in the ED is possible due to the use of extra hallway beds, patients being sent back to the waiting room and patients undergoing extra diagnostic tests outside of the ED (see Table 1, Appendix A for an overview of the treatment rooms). All those patients were still being counted as present in the ED system during these activities. Crowding at the ED is assumed with 80% or more treatment rooms (≥17 beds) occupied, according to the consulted physician. Figure 4.1 shows the WIP and cumulative WIP, with the dashed black line resembling crowding (≥17 patients at the ED). The figure shows that during COVID-19 6.7% it occurred that a patient arrived at the ED when already 17 or more patients were at the ED, in contrast to 31.9% during the normal situation.

The ED physicians were not aware that the WIP levels during the normal situation sometimes exceeded 30. Furthermore, they feel that the lower WIP during COVID-19 resulted from more focus on throughput times and efficient working manners.

Figure 4.1 – Percentage and cumulative percentage of work in process. The black dashed line represents crowding (≥80% treatment rooms occupied)

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 0% 1% 2% 3% 4% 5% 6% 7% 8% 9% 10% 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 C u m u la ti v e p e rc e n ta g e o f p a ti e n ts F re q u e n c y o f o c c u re n c e Number of patients

(17)

17 Figure 4.2 shows the distributions of WIP for each hour of the day as boxplots. All the boxplots in this thesis show the median as the s olid middle line of the box. The lower edge of the boxplot represents the first quartile and the upper edge represents the third quartile. The whiskers represent the tails of the distribution, which is equal or less than 1.5 times the interquartile range. Observations which are not within this range are considered as outliers and

represented as dots. The ‘x’ represents the mean. The distribution of WIP for each hour of the day shows a similar pattern among both periods, but the WIP is lower during the COVID-19 period. There is a rise in WIP starting in the morning, with a peak during the afternoon hours (between 14:00 and 18:00). The decrease in WIP during the night enables the ED to begin with a nearly empty start at 7:00.

(18)

18 To investigate whether the reduction of WIP during COVID-19 affected throughput times, LOS was calculated (Table 4.2). The average LOS for all patients was shorter during COVID-19 (3:18 vs 3:12, Table 4.2). Furthermore, although the complexity of cases was higher during COVID-19, reflected by the increased proportion of patients with a red, orange or yellow triage colour, the according average LOS was lower. Despite that the average LOS for both periods was within the 4-hour target (the national norm in which a patient needs to be discharged from the ED), still 31% and 28% did not meet the target during the normal and COVID-19 period, respectively (Figure 2a and 2b, appendix C). Since the distribution of age categories of the baseline study population showed

differences between the periods we conducted analysis on LOS per age category. However, this did not reveal differences between the periods (Figure 3, appendix C). ED staff acknowledged that during COVID-19 they focused on throughput times by aiming to adhere more to protocols. This led to an increased perceived work efficiency. However, they perceived that it was not a sustainable way of working in the long-term. Furthermore, they were surprised that a significant percentage of patients did not meet the 4 hour target.

Table 4.2: Descriptive overview of LOS per triage colour

Triage colour Patients (n (%)) (Pre-COVID-19) Average LOS (Pre-COVID-19) Patients (n (%)) (COVID-19) Average LOS (COVID-19) Red 1,035 (2.1) 2:26 142 (3.1) 2:23 Orange 6,992 (13.9) 3:52 651 (14.3) 3:35 Yellow 29,127 (58.0) 3:46 2,898 (63.4) 3:34 Green 8,150 (16.2) 2:31 574 (12.6) 2:07 Blue 1,104 (2.2) 1:24 56 (1.2) 1:25 NULL 3,839 (7.6) 1:09 245 (5.4) 1:18 All patients 50,247 3:18 4,566 3:12

All numbers are represented as numbers (n) and percentages (%) or as time (hh:mm).

The observed decrease of WIP should instinctively lead to a similar decrease of LOS,

therefore a combined graph of WIP and LOS over time is displayed (Figure 4.3). The graph shows that although the average WIP is lower during COVID-19 for every hour of the day, the average LOS remains similar to the normal situation. A similar pattern is visible for both periods. Noteworthy is that the WIP increases during the morning and afternoon, but the according LOS is more or less stable for both periods. Between 7:00 and 8:00 there is a spike in LOS, thereafter the LOS stabilizes during the day. This is important information for the ED in general because the capacity at the ED can be influenced by the number of patients present at a certain moment.

Figure 4.3 – Average LOS (minutes) combined with the average number of patients in the ED (WIP) per hour of the day 0 2 4 6 8 10 12 14 16 18 20 22 24 0 40 80 120 160 200 240 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 W IP ( # p a ti e n ts ) L O S (m in u te s )

Hour of the day

(19)

19 The consulted ED physician acknowledged that the sense of urgency with a high number of patients present at the ED increases, resulting in little change of LOS when WIP rises. They speculated that when the ED is quiet physicians tend to spend more time on administrative tasks.

Based on the previous results, showing similar patterns of WIP and LOS, it is not surprising that the inflow-outflow diagram also shows a similar pattern during both periods (Figure 4.4). The solid curves depict the normal situation and the dashed curves depict COVID-19. The desired outflow curve is based on the outflow rate that would correspond to the actual arrival rate after a constant delay, implying that the ED is capable of creating a balance between demand and supply. The desired outflow curve is the arrival rate delayed with the average LOS (198 and 192 minutes during the normal situation and COVID-19, respectively). The graphs of both periods show that during the morning the ED is capable of maintaining an outflow rate corresponding to the arrivals, however it is important to notice that the difference between the desired outflow rate and the actual outflow rate increases during the day. There is a maximum inflow of 5.3 patients at 11:00 during the normal situation and a maximum inflow of 3.8 patients at 14:00 during the COVID-19 period, which is almost a 30% decrease of inflow rate. The outflow patterns are similar. Building up of backlog (when outflow matches inflow) stops for the first time at 17:00 during both periods. Analyses on specific inflow and outflow factors and patterns follows in section 4.3 and 4.5, respectively.

Figure 4.4 – Inflow-outflow diagram including the desired outflow.

The ED physician argued that during COVID-19 staff focused on throughput times, which was reflected in the perceived LOS and led to perceived higher process efficiency. To check this statement, Little’s Law is used. Based on Little’s law (Hopp & Spearman, 2008) the argument of the ED

physician regarding extra focus on throughput resulting in higher efficiency during COVID-19 can be questioned. The reduction in LOS could also be explained by the reduction in WIP. The average LOS decreased during COVID-19 with only 3.0% (198 minutes to 192 minutes) and the average WIP decreased with 26.7% (from 13.33 patients to 9.77 patients). The decrease in LOS during COVID-19 was accompanied by a larger decrease in WIP, which contradicts the perceived higher e fficiency if we were comparing two equal periods. The theoretical explanation of Little’s Law and its application can

0 1 2 3 4 5 6 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 N u mb e r of p ati e n ts

Hour of the day

(20)

20 be found in appendix D. In-depth analyses of the LOS during COVID-19 showed that the LOS for the red triage colour decreased significantly over the months while the proportion of patients with that colour remained the same (Figure 4.5). The ED physicians argued that if pre-hospital they thought that a patient would be admitted to the ICU, the ICU staff would be present at the ED. After initial

assessment the ED and ICU physicians decided whether the patient had to be admitted to the ICU. This altered workflow led to increased multidisciplinary work and faster throughput times.

Figure 4.5 – LOS (hh:mm) during the COVID-19 period per triage colour per month

To explore why the LOS did not decrease as much as the WIP during COVID-19, analyses were conducted to investigate if patient characteristics and/or WIP distribution could explain this phenomenon. Population characteristics of patients who did not meet the 4-hour target are shown in Table 4.3. Gender and age distribution was similar during both periods. The proportion of patients with a red, orange and yellow triage colour and arrival via ambulance was higher during COVID-19. This indicates that patients not meeting the 4-hour target had a higher urgency during COVID-19. Furthermore, in line with the increased urgency, the proportion of patients requiring a n admission was higher during COVID-19.

April May June

(21)

21 Table 4.3 Population characteristics of patients not meeting the 4-hour LOS target

Pre-COVID-19 COVID-19

Variables Total (N=15508) Total (N=1257)

Ma le/fema le (n (%)) 8,214 (53.0) / 7291 (47.0) 693 (55.1) / 564 (44.9)

Age (media n, IQR) 57 (45 – 72) 59 (47 -73) Tria ge colours (n (%)) Red 171 (1.10) 17 (1.35) Ora nge 2,716 (17.51) 223 (17.74) Yellow 10,992 (70.88) 937 (74.54) Green 1,327 (8.56) 60 (4.77) Blue 70 (0.45) 4 (0.32) NULL 232 (1.50) 16 (1.27) Wa y of a rriva l (n (%)) Ambula nce 8,033 (51.80) 718 (57.12) Other 7,207 (46.47) 519 (41.29) Police 9 (0.06) 2 (0.16) NULL 243 (1.57) 18 (1.43) Admission required (n (%)) 9,374 (60.45) 836 (66.5)

All numbers are represented as median with interquartile range (IQR), or numbers (n) and percentages (%).

Previous results showed that WIP and LOS patterns were similar between the two periods. However, looking at the proportion of patients not meeting the 4-hour target per WIP, some differences were observed (Figure 4.6). This figure shows for every amount of WIP at the ED the proportion of patients that does not meet the 4-hour target. During COVID-19 the proportion of patients not meeting the 4-hour target for the lower WIP (2 through 6 patients; A) and the high WIP (17 through 22 patients; B) was lower.

(22)

22 Figure 4.6– Proportion of patients not meeting the 4-hour target per WIP. A (2 through 6) and B (17 through 22) represent the WIP where the proportion of patients not meeting the 4-hour target was lower during COVID-19.

4.2.1 Conclusion on crowding levels

The results have revealed 3 relevant general insights for the COVID-19 period: i) A lower WIP, ii) a small reduction in LOS and iii) less patient arrivals.

The reduction in WIP during COVID-19 led to a reduction in crowding. The reduction in LOS (3%) could be explained by a reduction in WIP (26.7%). Moreover, the proportion of patients not meeting the 4-hour target decreased with 3% during COVID-19, while the WIP decreased with 26.7%. Patients not meeting the 4-hour target during COVID-19 had a higher urgency, as seen from the higher proportion of: i) triage colour red, orange and yellow, ii) arrival per ambulance, and iii) admissions. For certain WIP levels (2 through 6 and 17 through 22) the ED physicians were able to work more efficient (expressed in terms of the 4-hour target) during COVID-19.

(23)

23 4.3 Inflow

In this section analyses on inflow are presented to have a closer look at the arrival specialisms, patient mix and arrival pattern, these results can support the understanding of h ow COVID-19 influenced crowding levels.

Differences were observed in the way of arrival (Figure 4.7). In our dataset patients could arrive at the ED in four different ways, via ambulance, via helicopter, via the police or via another way (coded as ‘other’). The arrival way ‘other’ is used for patients arriving at the ED with their own transport. In the normal situation 2.75% of the patients and during the COVID-19 period 1.69% of the patients had an unknown way of arrival and were therefore excluded from this analysis. During both periods most patients arrived using their own transport or via ambulance. During COVID-19 a higher percentage of ambulances were used to bring patients to the ED (40.9% vs 45.9%). According to the ED staff the increase of arrivals per ambulance during COVID-19 could be explained by take-overs of COVID-19 patients from other hospitals in the country or from neighbouring countries.

Figure 4.7 – Way of arrival

In addition, differences were observed in the top ten arrival specialisms (Figure 4.8). The bars in the figure are ranked according to the largest relative change between the periods, showing that the largest relative change was among trauma and orthopaedic patients. Furthermore, during COVID-19 a higher proportion of cardiology and neurology patients was registered compared to the normal situation. The category ‘other’ encompasses the proportion of patients arriving for specialisms outside the top-10.

Figure 4.8 - Overview of the top 10 arrival specialisms, ranked according the largest relative change. Ambulance Other Helicopter Police

Pre-COVID-19 40,9% 56,1% 0,2% 0,1% COVID-19 45,9% 52,2% 0,2% 0,0% 10,0% 20,0% 30,0% 40,0% 50,0% 60,0% Pre-COVID-19 COVID-19 0% 5% 10% 15% 20% 25% 30% 35% Other Internal medicine Abdominal surgery Gastroenterology and hepatology

(24)

24 The ED staff could explain the differences in the top-10 specialisms. They anticipated that cardiology and neurology patients have more comorbidities, making them more susceptible for COVID-19. Furthermore, COVID-19 suspected individuals were not only registered for the specialism lung diseases, but also for the internal medicine and emergency medicine. This explains the decreased percentage of patients registered for the lung diseases specialism during COVID-19. In their opinion, the decrease in orthopaedic and trauma patients resulted from the downscaling of regular care (e.g. orthopaedic surgery). This probably led to less post-surgery complications. Furthermore, as a result of an altered activity pattern during COVID-19 people stayed- and worked inside more often leading to a decrease in (occupational) accidents and vehicle injuries.

Based on our previous results, we showed that the overall arrival patterns of both periods are similar, with a decreased average amount of patients arriving per hour during the COVID-19 period. To gain further understanding of the arrival patterns, which is the average number of patients arriving per hour, they were plotted as a function of weekday (Figure 4.9a and 4.9b). As a result of the

(25)

25 Figure 4.9a – Arrival pattern per hour of the day per weekday pre-COVID-19

Figure 4.9b – Arrival pattern per hour of the day per weekday during COVID-19.

The ED staff argued that scaling down of healthcare during COVID-19 explained the lower amount of patient arrival per hour. Furthermore, it is known that the number of patients arriving on Fridays is higher compared to other weekdays, however the ED staff was surprised by the noteworthy increase during COVID-19. They did not have an explanation for this phenomenon. Therefore, we conducted additional analyses for the Fridays during COVID-19. Number of patient arrivals on Fridays during COVID-19 (in total 13 Fridays) ranged between 34 and 67 with a median of 56 (IQR 46 - 60.75). The patient characteristics (age, triage colour and way of arrival) on Fridays did not remarkably differ from the total group during COVID-19 (Table 2, appendix E). Hence, the observed noteworthy increase of patients on Fridays could not be explained based on patient characteristics. 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 N U M BE R O F P A T IE N T S

HOUR OF THE DAY

Monday Tuesday Wednesday Thursday Friday Saturday Sunday

0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 N U M BE R O F P A T IE T N S

HOUR OF THE DAY

(26)

26 Analyses on separate days revealed some variation. The average number of patients arriving on one day during the normal situation was 68.7 patients, with a minimum of 39 patients (28th_of December 2019) and a maximum of 107 patients (3rd_{of December 2018). The average number of} patients arriving on one day during COVID-19 was 50.2 patients, with a minimum of 27 patients (5th of April 2020) and a maximum of 72 patients (29th_{of June 2020) (Figure 4, appendix E). To attain} deeper understanding of the observed arrival variability during both periods, the number of patients arriving at the ED has been plotted per weekday (Figure 4.10) and per month (Figure 4.11). The line connecting the boxplots represents the mean trend line. The arrivals per weekday (Figure 4.10) show variation between the two periods, with a smoother trend during COVID-19. During COVID-19 there is an increase of patient arrival on Mondays and Fridays in contrast to the increase of p atient arrivals on Tuesday and Friday during the normal situation.

Figure 4.10 – Number of patient arrivals per weekday

The daily arrivals per month (Figure 4.11) also show variation between the two periods. During COVID-19 there is an increase of patient arrival as the year progresses. During the normal situation this increase of patient arrival during the same months (April to June) was not observed.

(27)

27 The ED staff argued that the increased trend over the months during COVID-19 could be explained by postponement of medical care seeking by patients due to fear of COVID-19 transmission. However, it could also be the result of the changing activity patterns towards the end of the lockdown lead ing to more small accidents and traumas. This would lead to an increase in patients with triage colour blue and green. Figure 4.12 shows the distribution of triage colours, with the column colours resembling the according triage colour. During May there indeed was a slight increase in blue and green triage colours, however the distribution of triage colours in June is similar to April. Therefore, our data suggests that the increased arrivals over the months were not the result of the increase of small accidents which would have led to an increase of blue and green triage colours.

Figure 4.12 – Proportion of patients per triage colour during COVID-19. The grey column represents triage code ‘null’. Previous results showed that the average arrival rate differed between the two periods (2.86 and 2.09 patients per hour for the normal- and COVID-19 period, respectively). To investigate high patient arrivals within a short time interval Figure 4.13 was made. A high amount of patients arriving within a 1-hour timeframe puts pressure on the availability of capacity and resources and influences the throughput and outflow, resulting in the formation of queues in several elements of the ED system. The consulted ED physician indicated that 8 or more patients arriving per hour is considered as a high inflow. Figure 4.13 shows the patterns of high arrival numbers within a 1-hour timeframe corrected for the days in the period, with the dashed lines resembling the COVID-19 period. During the normal situation the maximum amount of patients arriving within a 1-hour timeframe is 16, compared to 11 in the COVID-19 period. There is a lower maximum amount of patient arrivals within a 1-hour

timeframe during COVID-19, but it occurred more often that a high number of patients arrived within a 1-hour timeframe. During both situations high arrival numbers occurred between 9:00 and 20:00. Although previous figures showed that during COVID-19 there was less pressure on the system with regard to WIP and patients numbers, here we showed that with regard to patients arriving within 1-hour there sometimes was more pressure on the ED during COVID-19. However, it cannot fully explain the small difference in LOS since it rarely occurred .

4% 1% 12% 66% 14% 3% 6% 1% 15% 61% 14% 3% 6% 2% 11% 64% 15% 3% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

(28)

28 Figure 4.13 High patient arrival numbers in 1-hour per hour of the day. The dashed lines represent the COVID-19 period.

Thus far, the small difference in LOS during COVID-19 could not fully be explained by high patient arrival within a short time interval. Besides number of patient arrivals, the required care a patient needs is important and can influence the LOS. Patients with high urgency require more staff and specific equipment, resulting in less available capacity for other patients. To gain further insight into patient urgency a graph was made displaying the triage colour (Figure 4.14). The colours of the columns in the figure represent the according triage colour. During the COVID-19 period there was a higher percentage of urgent cases (triage colour red, orange and yellow). In total 80.9% of the cases where urgent during COVID-19, in comparison to only 74% during the normal situation. A significant amount of patients did not receive a triage colour, in the dataset coded as ‘Null’ and visualized as the grey column in the figure. According to the involved ED physician, this is against the case hospital’s policy since it requires that every patients receives a triage colour. The increase of urgent cases during COVID-19 could explain the small difference in LOS with a lower WIP, according to the ED staff. They indicated that the higher

percentage of complex cases could be explained in one of two ways. Either there were more critically ill patients as result of COVID-19 infections or patients waited longer until going to the doctor, resulting in further developed diseases with higher urgencies.

0% 1% 2% 3% 4% 5% 6% 7% 8% 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 F re q u e n c y o f o c c u re n c e

Hour of the day

(29)

29 Figure 4.14 – Proportion of patients per triage colour. The grey column represents triage code ‘null’.

4.3.1 Conclusion on inflow analyses

Based on the inflow analyses COVID-19 led to an altered distribution of patients and an altered patient mix (higher complexity, higher urgency, other arrival specialisms and increased proportion with an urgent triage colour). Furthermore, during COVID-19 patient volume increased as the year progressed.

Our quantitative data of the inflow analyses can explain the reduced WIP with the small reduction in LOS in one of two ways: 1) The increased patient complexity led to the relatively small decrease in LOS or 2) as a result of the increased complexity more staff and specific equipment was needed resulting in less available capacity for other patients. Our qualitative data suggests that the relatively small reduction in LOS was the result of the increased patient complexity. Moreover, the increased occurrence of high patient arrivals within a 1-hour timeframe during COVID-19 showed that sometimes there was more pressure on the ED.

(30)

30 4.4 Throughput

In this section analyses on throughput aspects are shown, i.e. triage time, the occupation and utilization of treatment rooms and waiting times for diagnostic tests. It is known that resources influence the throughput at the ED and limited availability or inefficient allocation can lead to patient flow hindrance (Morley et al., 2018). These analyses will help us to understand how the generation of two separate sets of resources (i.e. ‘infected’ and ‘non-infected’) during COVID-19 influenced crowding levels.

The LOS encompasses several processes at the ED such as triage. Triage time (from start to end) can help us to determine the difference in LOS. Average triage time during both periods did not meet the national norm (within 15 minutes after arrival) for any triage colour (Table 3 – appendix F). Triage times were seen as not reliable, based on the experience of the consulted ED physician. Quite often it happens that entering triage information into the system occurs when nurses have time and not immediately when they are finished with triage, resulting in a high and unreliable average triage time. Therefore, further analyses with triage times were not conducted.

To gain insight into resource availability the average time that treatment rooms were occupied by a patient was plotted (Figure 4.15). The average time that treatment rooms were occupied was longer for the majority of the rooms during COVID-19. Insights on treatment room utilization is needed to explore whether this affected crowding levels.

Figure 4.15 – Average occupation time (hh:mm) per treatment room.

Figure 4.16 was made to explore if the increased treatment room occupation time led to an altered utilization during COVID-19. The figure shows that the utilization of the majority of treatment rooms lowered during COVID-19, indicating that treatment room availability was not a cause of the small decrease in LOS. The ED physician indicated that the low utilization could be overestimated. Since during COVID-19 a room without patient was not immediately available until after thorough

disinfection. Treatment room 1 (K1) and treatment room 2 (K2) were specifically reserved for unstable COVID-19 patients, however the utilization was lower compared to the normal situation. Furthermore, during COVID-19 the utilization of all the observatory beds was higher compared to the normal situation. The medical staff indicated that the high utilization of the observatory during COVID-19 resulted from these beds being used as the ‘clean’ (non-COVID-19) beds. The function of the

observatory was, thus, changed during COVID-19 resulting in higher utilization and a longer average time per bed. Moreover, the lower utilization of K1 and K2 was the result of keeping those available in case an unstable COVID-19 patient would be registered. Patients on K1 or K2 were stabilized and admitted to a ward as soon as possible to prevent spread of the virus.

(31)

31 Figure 4.16 – Utilization per treatment room.

The observatory beds were used for ‘regular’ patients during COVID-19 instead of their intended function to conduct more investigation, diagnostics or longer observation. Therefore, the average occupation time of the observatory beds during COVID-19 were compared to the average occupation time of the regular treatment rooms (K1 through K14) during the normal situation (Table 4.4). The analysis showed that during COVID-19 the median occupation time of the treatment rooms for regular patients was higher compared to the normal situation. Consequently, it can be concluded that during COVID-19 it took physicians longer to treat the regular patients. The ED physician indicated that they used a similar nurse apportionment as during the normal period. This implies that 1 to 2 nurses were in charge of the observatory beds, not taking into account that the function altered. Normally a nurse is in charge of 1 patient instead of 5, therefore this led to increased workload. Furthermore, the observatory beds are out of the doctors’ sight, causing them to feel less urgency to hurry. The aforementioned are two explanations for the increased occupation time of the regular patients during COVID-19.

Table 4.4 Overview of analyses on resources

Variables Pre-COVID-19 COVID-19

Occupation time ‘regular treatment rooms’

(median, IQR)

1:57 (00:58 – 03:10) 2:27 (1:18 – 3:35)

Waiting time diagnostics (median) Lab

00:26 (00:18 – 00:38) 00:31 (00:21 – 00:46) Imaging _{00:45 (00:25 – 01:18)} _{00:45 (00:24 – 01:18)} Imaging (request – start)

00:15 (00:06 – 00:28) 00:18 (00:08 – 00:35) Imaging (end – results) _{00:24 (00:10 – 00:52)} _{00:23 (00:09 – 00:50)}

CT 02:46 (02:11 – 03:55) 02:35 (2:19 – 03:13) Utilization diagnostics (n (%)) Laboratory 35,447 (70.18) 3,555 (77.25) Imaging 19,771 (39.14) 1,795 (39.00) CT 570 (1.15) 64 (1.39)

All times (hh:mm) are represented as median with interquartile range (IQR), or numbers (n) and percentages (%). Median waiting time diagnostics encompasses time between request and results.

(32)

32 The medical staff was extremely focussed on efficiency during COVID-19. Therefore, we investigated the amount of treatment room changes a patient underwent to demonstrate whether this c ould be quantitatively objectified (Figure 4.17). A treatment room change can for example consist of a patient moving from a treatment room to the CT, back to the waiting room and finally to another treatment room. This flow is hampered by COVID-19 infected patients, since rooms need to be disinfected and staff has to wear and change special protective equipment every time they enter and leave the room. During the COVID-19 period the maximum of treatment room changes a patient underwent was 6, compared to 11 in the normal situation. Furthermore, the percentage of patients with zero (31.6% vs 36.4%) or one (37.0% vs 42.2%) changes of treatment rooms was high er during the COVID-19 period. The high amount of treatment room changes surprised the ED staff, they were unaware of this phenomenon. The decrease of changes during COVID-19 was recognizable by the staff, since they tried to critically assess in which room a patient would be seen. However, they experienced it as a hampering of care quality, since it limited their freedom to switch rooms and sometimes led to an uncomfortable situation. Therefore, they would not implement this altered working methods in a normal situation.

Figure 4.17 – Frequency of treatment room changes.

Besides treatment rooms, other resources such as diagnostics are used in the ED. Analysis on waiting time for diagnostics between the two periods showed differences for lab and CT results (Table 4.4). During COVID-19 the median time between the request of laboratory investigations and the results was longer (00:26 vs 00:31) and the median time between the request and results for CT was shorter (02:46 vs 2:35). To investigate whether these differences could be explained by altered

utilization, additional analysis were conducted (Table 4.4). During COVID-19 an increased proportion of patients required lab tests (+ 7.1%), which could explain the increased waiting time for the test. We found a similar proportion of patients requiring imaging or a CT, although patient complexity

increased. The ED physician stated that they aimed to conduct additional diagnostics at the wards, instead of at the ED. Previous results showed that during COVID-19 there was a decrease in treatment room changes, which was thus not the result of decreased imaging or CT utilisation.

4.4.1 Conclusion on throughput analyses

Analyses on throughput factors showed that COVID-19 did not lead to an impaired resource availability and hampered utilization. COVID-19 led to an increased average treatment room

(33)

33 4.5 Outflow

In this section analyses on outflow are presented to have a closer look at the outflow pattern, patient mix and boarding times.

Our previous results showed that the overall departure patterns of both periods were similar, with a decreased average amount of patients per hour during the COVID-19 period (figure 4.4, subsection 4.2). To gain further understanding of the departure patterns, the average number of patients departing per hour were plotted as a function of weekday (Figure 4.18a and 4.18b). The departure patterns per weekday show similarities between the periods among the majority. The curves for averages over time during COVID-19 are less smooth as a result of the lower amount of patients. As mentioned earlier there is a noteworthy increase on Fridays during COVID-19 between 18:00 and 21:00, which could not be explained based on patient characteristics or interpretation of th e physicians (subsection 4.3) .

Figure 4.18a – Departure patterns per hour of the day per weekday pre-COVID-19.

Figure 4.18b – Departure patterns per hour of the day per weekday during COVID-19. 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 N u m b e r o f P a ti e n ts

Hour of the day

Monday Tuesday Wednesday Thursday Friday Saturday Sunday

0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 N u m b e r o f p a ti e n ts

Hour of the day

(34)

34 To investigate if we could explain the phenomenon of a small difference in LOS during COVID-19, the outflow patient mix was analysed. Previous literature showed that, during the normal situation, boarding time is associated with decreased ED efficiency (Napoli et al., 2020). Boarding times can be influenced by the proportion admitted patients. Our analyses showed that during COVID-19 a larger proportion of patients was admitted to an in-hospital ward (39.1% and 47.6% for the normal and COVID-19 period, respectively). However, boarding times were similar between the two time periods, consisting of a median of 1 hour and 5 minutes for both periods (Figure 5, appendix G). These boarding times account for almost one third of the LOS. While the boarding times were not influenced by the increased proportion of admitted patients during COVID-19, it is interesting to investigate if differences in average boarding times existed in relation to the number of admittance requests per day. However, this analysis showed a similar pattern between the two periods and did not reveal new insights (Figure 6, appendix G).

The ED staff was not aware of the long boarding times. They thought that a possible

explanation for the long boarding time could result from the discrepancy in urgency of a f ast outflow. The ED aims for a boarding time as short as possible. In contrast, the various departments do not focus on the admission of ED patients. Their focus is on care for patients present at their ward, for example during the morning they do ward rounds. Moreover, the ED physician gave the example that

sometimes they have to wait until a patient can be admitted to a ward until the break of the nurses is over. This example clearly demonstrates the discrepancy in urgency. The fact that boarding times did not differ between the two periods, was according to the staff, the result of unchanged working methods.

Analyses were conducted to investigate if a specific patient group had a long boarding time (≥ third quartile) and if there existed differences in population characteristics between the two periods. These results show that there were minor differences in characteristics between the two periods (Table 4, appendix G). To gain more insight into factors that influence waiting time components and

potentially causing delays in the admittance process, boarding time distribution was investigated. Analysis of the distribution of boarding time per 15 minute showed that there were minor differences in boarding time distribution between both periods, although less patients were present during COVID-19 (Figure 7, appendix G).

(35)

35 Figure 4.19 – Number of patients in the ED (WIP) combined with the average boarding time (minutes) per hour of the day. 4.5.1 Conclusion on outflow analyses

Based on the outflow analyses we concluded that although the WIP was lower during COVID-19, a higher proportion of patients was admitted to the in-hospital wards, which is in line with the observed higher urgency (inflow analyses). This did not affect the overall outflow patterns or boarding times during COVID-19. Furthermore, we showed that boarding times were not affected by the WIP. This may be important information for future ED and hospital operations planning. Further research should investigate in-hospital activities during the moments boarding times rose to identify possible causes or bottlenecks and solutions towards integrated care.

0 2 4 6 8 10 12 14 16 18 20 0 20 40 60 80 100 120 140 160 180 200 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 W IP ( # p a ti e n ts ) B o a rd in g t im e ( m in u te s )

Hour of the day

The impact of COVID-19 on emergency department crowding